MERGE ... RETURNING

diff --git a/doc/src/sgml/ref/merge.sgml b/doc/src/sgml/ref/merge.sgml
new file mode 100644
index 0995fe0..bedb2c8
--- a/doc/src/sgml/ref/merge.sgml
+++ b/doc/src/sgml/ref/merge.sgml
@@ -25,6 +25,7 @@ PostgreSQL documentation
 MERGE INTO [ ONLY ] <replaceable class="parameter">target_table_name</replaceable> [ * ] [ [ AS ] <replaceable class="parameter">target_alias</replaceable> ]
 USING <replaceable class="parameter">data_source</replaceable> ON <replaceable class="parameter">join_condition</replaceable>
 <replaceable class="parameter">when_clause</replaceable> [...]
+[ RETURNING * | <replaceable class="parameter">output_expression</replaceable> [ [ AS ] <replaceable class="parameter">output_name</replaceable> ] [, ...] ]
 
 <phrase>where <replaceable class="parameter">data_source</replaceable> is:</phrase>
 
diff --git a/src/backend/commands/copy.c b/src/backend/commands/copy.c
new file mode 100644
index e34f583..aa3cca0
--- a/src/backend/commands/copy.c
+++ b/src/backend/commands/copy.c
@@ -274,12 +274,6 @@ DoCopy(ParseState *pstate, const CopyStm
 	{
 		Assert(stmt->query);
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(stmt->query, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in COPY"));
-
 		query = makeNode(RawStmt);
 		query->stmt = stmt->query;
 		query->stmt_location = stmt_location;
diff --git a/src/backend/commands/copyto.c b/src/backend/commands/copyto.c
new file mode 100644
index 8043b4e..e02d7d0
--- a/src/backend/commands/copyto.c
+++ b/src/backend/commands/copyto.c
@@ -510,7 +510,8 @@ BeginCopyTo(ParseState *pstate,
 		{
 			Assert(query->commandType == CMD_INSERT ||
 				   query->commandType == CMD_UPDATE ||
-				   query->commandType == CMD_DELETE);
+				   query->commandType == CMD_DELETE ||
+				   query->commandType == CMD_MERGE);
 
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
diff --git a/src/backend/executor/execExpr.c b/src/backend/executor/execExpr.c
new file mode 100644
index 812ead9..8572b01
--- a/src/backend/executor/execExpr.c
+++ b/src/backend/executor/execExpr.c
@@ -48,6 +48,7 @@
 #include "utils/array.h"
 #include "utils/builtins.h"
 #include "utils/datum.h"
+#include "utils/fmgroids.h"
 #include "utils/lsyscache.h"
 #include "utils/typcache.h"
 
@@ -2485,6 +2486,22 @@ ExecInitFunc(ExprEvalStep *scratch, Expr
 	InitFunctionCallInfoData(*fcinfo, flinfo,
 							 nargs, inputcollid, NULL, NULL);
 
+	/*
+	 * Merge support functions should only be called directly from a MERGE
+	 * command, and need access to the parent ModifyTableState.  The parser
+	 * should have checked that such functions only appear in the RETURNING
+	 * list of a MERGE, so this should never fail.
+	 */
+	if (IsMergeSupportFunction(funcid))
+	{
+		if (!state->parent ||
+			!IsA(state->parent, ModifyTableState) ||
+			((ModifyTableState *) state->parent)->operation != CMD_MERGE)
+			elog(ERROR, "merge support function called in non-merge context");
+
+		fcinfo->context = (Node *) state->parent;
+	}
+
 	/* Keep extra copies of this info to save an indirection at runtime */
 	scratch->d.func.fn_addr = flinfo->fn_addr;
 	scratch->d.func.nargs = nargs;
diff --git a/src/backend/executor/execPartition.c b/src/backend/executor/execPartition.c
new file mode 100644
index 651ad24..3391269
--- a/src/backend/executor/execPartition.c
+++ b/src/backend/executor/execPartition.c
@@ -612,6 +612,9 @@ ExecInitPartitionInfo(ModifyTableState *
 	 * case or in the case of UPDATE tuple routing where we didn't find a
 	 * result rel to reuse.
 	 */
+
+	/* XXX: What about the MERGE case ??? */
+
 	if (node && node->returningLists != NIL)
 	{
 		TupleTableSlot *slot;
@@ -877,6 +880,7 @@ ExecInitPartitionInfo(ModifyTableState *
 		List	   *firstMergeActionList = linitial(node->mergeActionLists);
 		ListCell   *lc;
 		ExprContext *econtext = mtstate->ps.ps_ExprContext;
+		int			action_idx = 1;
 
 		if (part_attmap == NULL)
 			part_attmap =
@@ -897,6 +901,7 @@ ExecInitPartitionInfo(ModifyTableState *
 			/* Generate the action's state for this relation */
 			action_state = makeNode(MergeActionState);
 			action_state->mas_action = action;
+			action_state->mas_action_idx = action_idx++;
 
 			/* And put the action in the appropriate list */
 			if (action->matched)
diff --git a/src/backend/executor/functions.c b/src/backend/executor/functions.c
new file mode 100644
index 50e06ec..7e4717a
--- a/src/backend/executor/functions.c
+++ b/src/backend/executor/functions.c
@@ -1665,8 +1665,8 @@ check_sql_fn_retval(List *queryTreeLists
 
 	/*
 	 * If it's a plain SELECT, it returns whatever the targetlist says.
-	 * Otherwise, if it's INSERT/UPDATE/DELETE with RETURNING, it returns
-	 * that. Otherwise, the function return type must be VOID.
+	 * Otherwise, if it's INSERT/UPDATE/DELETE/MERGE with RETURNING, it
+	 * returns that. Otherwise, the function return type must be VOID.
 	 *
 	 * Note: eventually replace this test with QueryReturnsTuples?	We'd need
 	 * a more general method of determining the output type, though.  Also, it
@@ -1684,7 +1684,8 @@ check_sql_fn_retval(List *queryTreeLists
 	else if (parse &&
 			 (parse->commandType == CMD_INSERT ||
 			  parse->commandType == CMD_UPDATE ||
-			  parse->commandType == CMD_DELETE) &&
+			  parse->commandType == CMD_DELETE ||
+			  parse->commandType == CMD_MERGE) &&
 			 parse->returningList)
 	{
 		tlist = parse->returningList;
diff --git a/src/backend/executor/nodeModifyTable.c b/src/backend/executor/nodeModifyTable.c
new file mode 100644
index f419c47..87ae37f
--- a/src/backend/executor/nodeModifyTable.c
+++ b/src/backend/executor/nodeModifyTable.c
@@ -36,8 +36,7 @@
  *		RETURNING tuple after completing each row insert, update, or delete.
  *		It must be called again to continue the operation.  Without RETURNING,
  *		we just loop within the node until all the work is done, then
- *		return NULL.  This avoids useless call/return overhead.  (MERGE does
- *		not support RETURNING.)
+ *		return NULL.  This avoids useless call/return overhead.
  */
 
 #include "postgres.h"
@@ -97,9 +96,6 @@ typedef struct ModifyTableContext
 										  TupleTableSlot *oldSlot,
 										  MergeActionState *relaction);
 
-	/* MERGE specific */
-	MergeActionState *relaction;	/* MERGE action in progress */
-
 	/*
 	 * Information about the changes that were made concurrently to a tuple
 	 * being updated or deleted
@@ -172,13 +168,14 @@ static TupleTableSlot *ExecMerge(ModifyT
 								 ItemPointer tupleid,
 								 bool canSetTag);
 static void ExecInitMerge(ModifyTableState *mtstate, EState *estate);
-static bool ExecMergeMatched(ModifyTableContext *context,
-							 ResultRelInfo *resultRelInfo,
-							 ItemPointer tupleid,
-							 bool canSetTag);
-static void ExecMergeNotMatched(ModifyTableContext *context,
-								ResultRelInfo *resultRelInfo,
-								bool canSetTag);
+static TupleTableSlot *ExecMergeMatched(ModifyTableContext *context,
+										ResultRelInfo *resultRelInfo,
+										ItemPointer tupleid,
+										bool canSetTag,
+										bool *matched);
+static TupleTableSlot *ExecMergeNotMatched(ModifyTableContext *context,
+										   ResultRelInfo *resultRelInfo,
+										   bool canSetTag);
 static TupleTableSlot *mergeGetUpdateNewTuple(ResultRelInfo *relinfo,
 											  TupleTableSlot *planSlot,
 											  TupleTableSlot *oldSlot,
@@ -987,7 +984,7 @@ ExecInsert(ModifyTableContext *context,
 		if (mtstate->operation == CMD_UPDATE)
 			wco_kind = WCO_RLS_UPDATE_CHECK;
 		else if (mtstate->operation == CMD_MERGE)
-			wco_kind = (context->relaction->mas_action->commandType == CMD_UPDATE) ?
+			wco_kind = (mtstate->mt_merge_action->mas_action->commandType == CMD_UPDATE) ?
 				WCO_RLS_UPDATE_CHECK : WCO_RLS_INSERT_CHECK;
 		else
 			wco_kind = WCO_RLS_INSERT_CHECK;
@@ -1838,7 +1835,7 @@ ExecCrossPartitionUpdate(ModifyTableCont
 			/* and project the new tuple to retry the UPDATE with */
 			context->cpUpdateRetrySlot =
 				context->GetUpdateNewTuple(resultRelInfo, epqslot, oldSlot,
-										   context->relaction);
+										   mtstate->mt_merge_action);
 			return false;
 		}
 	}
@@ -2054,7 +2051,7 @@ lreplace:
 		 * No luck, a retry is needed.  If running MERGE, we do not do so
 		 * here; instead let it handle that on its own rules.
 		 */
-		if (context->relaction != NULL)
+		if (context->mtstate->mt_merge_action != NULL)
 			return TM_Updated;
 
 		/*
@@ -2692,6 +2689,7 @@ static TupleTableSlot *
 ExecMerge(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 		  ItemPointer tupleid, bool canSetTag)
 {
+	TupleTableSlot *rslot = NULL;
 	bool		matched;
 
 	/*-----
@@ -2739,7 +2737,8 @@ ExecMerge(ModifyTableContext *context, R
 	 */
 	matched = tupleid != NULL;
 	if (matched)
-		matched = ExecMergeMatched(context, resultRelInfo, tupleid, canSetTag);
+		rslot = ExecMergeMatched(context, resultRelInfo, tupleid, canSetTag,
+								 &matched);
 
 	/*
 	 * Either we were dealing with a NOT MATCHED tuple or ExecMergeMatched()
@@ -2747,10 +2746,9 @@ ExecMerge(ModifyTableContext *context, R
 	 * matches.
 	 */
 	if (!matched)
-		ExecMergeNotMatched(context, resultRelInfo, canSetTag);
+		rslot = ExecMergeNotMatched(context, resultRelInfo, canSetTag);
 
-	/* No RETURNING support yet */
-	return NULL;
+	return rslot;
 }
 
 /*
@@ -2760,8 +2758,8 @@ ExecMerge(ModifyTableContext *context, R
  * We start from the first WHEN MATCHED action and check if the WHEN quals
  * pass, if any. If the WHEN quals for the first action do not pass, we
  * check the second, then the third and so on. If we reach to the end, no
- * action is taken and we return true, indicating that no further action is
- * required for this tuple.
+ * action is taken and "matched" is set to true, indicating that no further
+ * action is required for this tuple.
  *
  * If we do find a qualifying action, then we attempt to execute the action.
  *
@@ -2770,15 +2768,16 @@ ExecMerge(ModifyTableContext *context, R
  * with individual actions are evaluated by this routine via ExecQual, while
  * EvalPlanQual checks for the join quals. If EvalPlanQual tells us that the
  * updated tuple still passes the join quals, then we restart from the first
- * action to look for a qualifying action. Otherwise, we return false --
- * meaning that a NOT MATCHED action must now be executed for the current
- * source tuple.
+ * action to look for a qualifying action. Otherwise, "matched" is set to
+ * false -- meaning that a NOT MATCHED action must now be executed for the
+ * current source tuple.
  */
-static bool
+static TupleTableSlot *
 ExecMergeMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
-				 ItemPointer tupleid, bool canSetTag)
+				 ItemPointer tupleid, bool canSetTag, bool *matched)
 {
 	ModifyTableState *mtstate = context->mtstate;
+	TupleTableSlot *rslot = NULL;
 	TupleTableSlot *newslot;
 	EState	   *estate = context->estate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
@@ -2790,7 +2789,10 @@ ExecMergeMatched(ModifyTableContext *con
 	 * If there are no WHEN MATCHED actions, we are done.
 	 */
 	if (resultRelInfo->ri_matchedMergeAction == NIL)
-		return true;
+	{
+		*matched = true;
+		return NULL;
+	}
 
 	/*
 	 * Make tuple and any needed join variables available to ExecQual and
@@ -2869,7 +2871,7 @@ lmerge_matched:
 				 */
 				newslot = ExecProject(relaction->mas_proj);
 
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				context->GetUpdateNewTuple = mergeGetUpdateNewTuple;
 				context->cpUpdateRetrySlot = NULL;
 
@@ -2892,7 +2894,7 @@ lmerge_matched:
 				break;
 
 			case CMD_DELETE:
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				if (!ExecDeletePrologue(context, resultRelInfo, tupleid,
 										NULL, NULL))
 				{
@@ -2952,7 +2954,8 @@ lmerge_matched:
 				 * If the tuple was already deleted, return to let caller
 				 * handle it under NOT MATCHED clauses.
 				 */
-				return false;
+				*matched = false;
+				return NULL;
 
 			case TM_Updated:
 				{
@@ -3020,13 +3023,19 @@ lmerge_matched:
 							 * NOT MATCHED actions.
 							 */
 							if (TupIsNull(epqslot))
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							(void) ExecGetJunkAttribute(epqslot,
 														resultRelInfo->ri_RowIdAttNo,
 														&isNull);
 							if (isNull)
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							/*
 							 * When a tuple was updated and migrated to
@@ -3061,7 +3070,8 @@ lmerge_matched:
 							 * tuple already deleted; tell caller to run NOT
 							 * MATCHED actions
 							 */
-							return false;
+							*matched = false;
+							return NULL;
 
 						case TM_SelfModified:
 
@@ -3081,13 +3091,14 @@ lmerge_matched:
 										(errcode(ERRCODE_TRIGGERED_DATA_CHANGE_VIOLATION),
 										 errmsg("tuple to be updated or deleted was already modified by an operation triggered by the current command"),
 										 errhint("Consider using an AFTER trigger instead of a BEFORE trigger to propagate changes to other rows.")));
-							return false;
+							*matched = false;
+							return NULL;
 
 						default:
 							/* see table_tuple_lock call in ExecDelete() */
 							elog(ERROR, "unexpected table_tuple_lock status: %u",
 								 result);
-							return false;
+							return NULL;
 					}
 				}
 
@@ -3099,6 +3110,30 @@ lmerge_matched:
 				break;
 		}
 
+		/* Process RETURNING if present */
+		if (resultRelInfo->ri_projectReturning)
+		{
+			switch (commandType)
+			{
+				case CMD_UPDATE:
+					rslot = ExecProcessReturning(resultRelInfo, newslot,
+												 context->planSlot);
+					break;
+
+				case CMD_DELETE:
+					rslot = ExecProcessReturning(resultRelInfo,
+												 resultRelInfo->ri_oldTupleSlot,
+												 context->planSlot);
+					break;
+
+				case CMD_NOTHING:
+					break;
+
+				default:
+					elog(ERROR, "unknown action in MERGE WHEN MATCHED clause");
+			}
+		}
+
 		/*
 		 * We've activated one of the WHEN clauses, so we don't search
 		 * further. This is required behaviour, not an optimization.
@@ -3109,19 +3144,22 @@ lmerge_matched:
 	/*
 	 * Successfully executed an action or no qualifying action was found.
 	 */
-	return true;
+	*matched = true;
+
+	return rslot;
 }
 
 /*
  * Execute the first qualifying NOT MATCHED action.
  */
-static void
+static TupleTableSlot *
 ExecMergeNotMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 					bool canSetTag)
 {
 	ModifyTableState *mtstate = context->mtstate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
 	List	   *actionStates = NIL;
+	TupleTableSlot *rslot = NULL;
 	ListCell   *l;
 
 	/*
@@ -3171,10 +3209,10 @@ ExecMergeNotMatched(ModifyTableContext *
 				 * so we don't need to map the tuple here.
 				 */
 				newslot = ExecProject(action->mas_proj);
-				context->relaction = action;
+				mtstate->mt_merge_action = action;
 
-				(void) ExecInsert(context, mtstate->rootResultRelInfo, newslot,
-								  canSetTag, NULL, NULL);
+				rslot = ExecInsert(context, mtstate->rootResultRelInfo,
+								   newslot, canSetTag, NULL, NULL);
 				mtstate->mt_merge_inserted += 1;
 				break;
 			case CMD_NOTHING:
@@ -3190,6 +3228,8 @@ ExecMergeNotMatched(ModifyTableContext *
 		 */
 		break;
 	}
+
+	return rslot;
 }
 
 /*
@@ -3227,6 +3267,7 @@ ExecInitMerge(ModifyTableState *mtstate,
 		List	   *mergeActionList = lfirst(lc);
 		TupleDesc	relationDesc;
 		ListCell   *l;
+		int			action_idx;
 
 		resultRelInfo = mtstate->resultRelInfo + i;
 		i++;
@@ -3236,6 +3277,7 @@ ExecInitMerge(ModifyTableState *mtstate,
 		if (unlikely(!resultRelInfo->ri_projectNewInfoValid))
 			ExecInitMergeTupleSlots(mtstate, resultRelInfo);
 
+		action_idx = 1;
 		foreach(l, mergeActionList)
 		{
 			MergeAction *action = (MergeAction *) lfirst(l);
@@ -3250,6 +3292,7 @@ ExecInitMerge(ModifyTableState *mtstate,
 			 */
 			action_state = makeNode(MergeActionState);
 			action_state->mas_action = action;
+			action_state->mas_action_idx = action_idx++;
 			action_state->mas_whenqual = ExecInitQual((List *) action->qual,
 													  &mtstate->ps);
 
@@ -3386,6 +3429,64 @@ mergeGetUpdateNewTuple(ResultRelInfo *re
 }
 
 /*
+ * pg_merge_action() -
+ *	  SQL merge support function to retrieve the currently executing merge
+ *	  action command string ("INSERT", "UPDATE", or "DELETE").
+ */
+Datum
+pg_merge_action(PG_FUNCTION_ARGS)
+{
+	ModifyTableState *mtstate = (ModifyTableState *) fcinfo->context;
+	MergeActionState *relaction;
+
+	if (!mtstate || mtstate->operation != CMD_MERGE)
+		elog(ERROR, "merge support function called in non-merge context");
+
+	relaction = mtstate->mt_merge_action;
+	if (relaction)
+	{
+		CmdType		commandType = relaction->mas_action->commandType;
+
+		switch (commandType)
+		{
+			case CMD_INSERT:
+				PG_RETURN_TEXT_P(cstring_to_text("INSERT"));
+			case CMD_UPDATE:
+				PG_RETURN_TEXT_P(cstring_to_text("UPDATE"));
+			case CMD_DELETE:
+				PG_RETURN_TEXT_P(cstring_to_text("DELETE"));
+			case CMD_NOTHING:
+				PG_RETURN_NULL();
+			default:
+				elog(ERROR, "unrecognized commandType: %d", (int) commandType);
+		}
+	}
+
+	PG_RETURN_NULL();
+}
+
+/*
+ * pg_merge_when_clause() -
+ *	  SQL merge support function to retrieve the 1-based index of the
+ *	  currently executing merge WHEN clause.
+ */
+Datum
+pg_merge_when_clause(PG_FUNCTION_ARGS)
+{
+	ModifyTableState *mtstate = (ModifyTableState *) fcinfo->context;
+	MergeActionState *relaction;
+
+	if (!mtstate || mtstate->operation != CMD_MERGE)
+		elog(ERROR, "merge support function called in non-merge context");
+
+	relaction = mtstate->mt_merge_action;
+	if (relaction)
+		PG_RETURN_INT32((int32) relaction->mas_action_idx);
+
+	PG_RETURN_NULL();
+}
+
+/*
  * Process BEFORE EACH STATEMENT triggers
  */
 static void
@@ -3671,8 +3772,17 @@ ExecModifyTable(PlanState *pstate)
 				{
 					EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-					ExecMerge(&context, node->resultRelInfo, NULL, node->canSetTag);
-					continue;	/* no RETURNING support yet */
+					slot = ExecMerge(&context, node->resultRelInfo, NULL,
+									 node->canSetTag);
+
+					/*
+					 * If we got a RETURNING result, return it to the caller.
+					 * We'll continue the work on next call.
+					 */
+					if (slot)
+						return slot;
+
+					continue;	/* continue with the next tuple */
 				}
 
 				elog(ERROR, "tableoid is NULL");
@@ -3749,8 +3859,17 @@ ExecModifyTable(PlanState *pstate)
 					{
 						EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-						ExecMerge(&context, node->resultRelInfo, NULL, node->canSetTag);
-						continue;	/* no RETURNING support yet */
+						slot = ExecMerge(&context, node->resultRelInfo, NULL,
+										 node->canSetTag);
+
+						/*
+						 * If we got a RETURNING result, return it to the
+						 * caller.  We'll continue the work on next call.
+						 */
+						if (slot)
+							return slot;
+
+						continue;	/* continue with the next tuple */
 					}
 
 					elog(ERROR, "ctid is NULL");
@@ -3843,7 +3962,7 @@ ExecModifyTable(PlanState *pstate)
 				slot = internalGetUpdateNewTuple(resultRelInfo, context.planSlot,
 												 oldSlot, NULL);
 				context.GetUpdateNewTuple = internalGetUpdateNewTuple;
-				context.relaction = NULL;
+				node->mt_merge_action = NULL;
 
 				/* Now apply the update. */
 				slot = ExecUpdate(&context, resultRelInfo, tupleid, oldtuple,
diff --git a/src/backend/parser/analyze.c b/src/backend/parser/analyze.c
new file mode 100644
index 4a817b7..c13c866
--- a/src/backend/parser/analyze.c
+++ b/src/backend/parser/analyze.c
@@ -73,7 +73,6 @@ static void determineRecursiveColTypes(P
 									   Node *larg, List *nrtargetlist);
 static Query *transformReturnStmt(ParseState *pstate, ReturnStmt *stmt);
 static Query *transformUpdateStmt(ParseState *pstate, UpdateStmt *stmt);
-static List *transformReturningList(ParseState *pstate, List *returningList);
 static Query *transformPLAssignStmt(ParseState *pstate,
 									PLAssignStmt *stmt);
 static Query *transformDeclareCursorStmt(ParseState *pstate,
@@ -514,7 +513,8 @@ transformDeleteStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -940,7 +940,8 @@ transformInsertStmt(ParseState *pstate,
 	/* Process RETURNING, if any. */
 	if (stmt->returningList)
 		qry->returningList = transformReturningList(pstate,
-													stmt->returningList);
+													stmt->returningList,
+													EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -2405,7 +2406,8 @@ transformUpdateStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/*
 	 * Now we are done with SELECT-like processing, and can get on with
@@ -2499,10 +2501,11 @@ transformUpdateTargetList(ParseState *ps
 
 /*
  * transformReturningList -
- *	handle a RETURNING clause in INSERT/UPDATE/DELETE
+ *	handle a RETURNING clause in INSERT/UPDATE/DELETE/MERGE
  */
-static List *
-transformReturningList(ParseState *pstate, List *returningList)
+List *
+transformReturningList(ParseState *pstate, List *returningList,
+					   ParseExprKind exprKind)
 {
 	List	   *rlist;
 	int			save_next_resno;
@@ -2519,7 +2522,7 @@ transformReturningList(ParseState *pstat
 	pstate->p_next_resno = 1;
 
 	/* transform RETURNING identically to a SELECT targetlist */
-	rlist = transformTargetList(pstate, returningList, EXPR_KIND_RETURNING);
+	rlist = transformTargetList(pstate, returningList, exprKind);
 
 	/*
 	 * Complain if the nonempty tlist expanded to nothing (which is possible
diff --git a/src/backend/parser/gram.y b/src/backend/parser/gram.y
new file mode 100644
index a013838..4406430
--- a/src/backend/parser/gram.y
+++ b/src/backend/parser/gram.y
@@ -12241,6 +12241,7 @@ MergeStmt:
 			USING table_ref
 			ON a_expr
 			merge_when_list
+			returning_clause
 				{
 					MergeStmt  *m = makeNode(MergeStmt);
 
@@ -12249,6 +12250,7 @@ MergeStmt:
 					m->sourceRelation = $6;
 					m->joinCondition = $8;
 					m->mergeWhenClauses = $9;
+					m->returningList = $10;
 
 					$$ = (Node *) m;
 				}
diff --git a/src/backend/parser/parse_agg.c b/src/backend/parser/parse_agg.c
new file mode 100644
index f7a1046..c66b7cf
--- a/src/backend/parser/parse_agg.c
+++ b/src/backend/parser/parse_agg.c
@@ -455,6 +455,7 @@ check_agglevels_and_constraints(ParseSta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
@@ -903,6 +904,7 @@ transformWindowFuncCall(ParseState *psta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_cte.c b/src/backend/parser/parse_cte.c
new file mode 100644
index c5b1a49..ad3c525
--- a/src/backend/parser/parse_cte.c
+++ b/src/backend/parser/parse_cte.c
@@ -126,13 +126,6 @@ transformWithClause(ParseState *pstate,
 		CommonTableExpr *cte = (CommonTableExpr *) lfirst(lc);
 		ListCell   *rest;
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(cte->ctequery, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in WITH query"),
-					parser_errposition(pstate, cte->location));
-
 		for_each_cell(rest, withClause->ctes, lnext(withClause->ctes, lc))
 		{
 			CommonTableExpr *cte2 = (CommonTableExpr *) lfirst(rest);
@@ -153,7 +146,8 @@ transformWithClause(ParseState *pstate,
 			/* must be a data-modifying statement */
 			Assert(IsA(cte->ctequery, InsertStmt) ||
 				   IsA(cte->ctequery, UpdateStmt) ||
-				   IsA(cte->ctequery, DeleteStmt));
+				   IsA(cte->ctequery, DeleteStmt) ||
+				   IsA(cte->ctequery, MergeStmt));
 
 			pstate->p_hasModifyingCTE = true;
 		}
diff --git a/src/backend/parser/parse_expr.c b/src/backend/parser/parse_expr.c
new file mode 100644
index 53e904c..2416cd6
--- a/src/backend/parser/parse_expr.c
+++ b/src/backend/parser/parse_expr.c
@@ -482,6 +482,7 @@ transformColumnRef(ParseState *pstate, C
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CHECK_CONSTRAINT:
@@ -1708,6 +1709,7 @@ transformSubLink(ParseState *pstate, Sub
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CYCLE_MARK:
@@ -2997,6 +2999,7 @@ ParseExprKindName(ParseExprKind exprKind
 		case EXPR_KIND_OFFSET:
 			return "OFFSET";
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			return "RETURNING";
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
diff --git a/src/backend/parser/parse_func.c b/src/backend/parser/parse_func.c
new file mode 100644
index ca14f06..914624e
--- a/src/backend/parser/parse_func.c
+++ b/src/backend/parser/parse_func.c
@@ -31,6 +31,7 @@
 #include "parser/parse_target.h"
 #include "parser/parse_type.h"
 #include "utils/builtins.h"
+#include "utils/fmgroids.h"
 #include "utils/lsyscache.h"
 #include "utils/syscache.h"
 
@@ -348,6 +349,15 @@ ParseFuncOrColumn(ParseState *pstate, Li
 					 parser_errposition(pstate, location)));
 	}
 
+	/* Merge support functions are only allowed in MERGE's RETURNING list */
+	if (IsMergeSupportFunction(funcid) &&
+		pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+		ereport(ERROR,
+				(errcode(ERRCODE_WRONG_OBJECT_TYPE),
+				 errmsg("merge support function %s can only be called from the RETURNING list of a MERGE command",
+						NameListToString(funcname)),
+				 parser_errposition(pstate, location)));
+
 	/*
 	 * So far so good, so do some fdresult-type-specific processing.
 	 */
@@ -2602,6 +2612,7 @@ check_srf_call_placement(ParseState *pst
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_merge.c b/src/backend/parser/parse_merge.c
new file mode 100644
index d886637..40afa7c
--- a/src/backend/parser/parse_merge.c
+++ b/src/backend/parser/parse_merge.c
@@ -233,6 +233,10 @@ transformMergeStmt(ParseState *pstate, M
 	 */
 	qry->jointree = makeFromExpr(pstate->p_joinlist, joinExpr);
 
+	/* Transform the RETURNING list, if any */
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_MERGE_RETURNING);
+
 	/*
 	 * We now have a good query shape, so now look at the WHEN conditions and
 	 * action targetlists.
@@ -390,9 +394,6 @@ transformMergeStmt(ParseState *pstate, M
 
 	qry->mergeActionList = mergeActionList;
 
-	/* RETURNING could potentially be added in the future, but not in SQL std */
-	qry->returningList = NULL;
-
 	qry->hasTargetSRFs = false;
 	qry->hasSubLinks = pstate->p_hasSubLinks;
 
diff --git a/src/backend/parser/parse_relation.c b/src/backend/parser/parse_relation.c
new file mode 100644
index b490541..9fe7bcb
--- a/src/backend/parser/parse_relation.c
+++ b/src/backend/parser/parse_relation.c
@@ -2313,9 +2313,10 @@ addRangeTableEntryForCTE(ParseState *pst
 		cte->cterefcount++;
 
 	/*
-	 * We throw error if the CTE is INSERT/UPDATE/DELETE without RETURNING.
-	 * This won't get checked in case of a self-reference, but that's OK
-	 * because data-modifying CTEs aren't allowed to be recursive anyhow.
+	 * We throw error if the CTE is INSERT/UPDATE/DELETE/MERGE without
+	 * RETURNING.  This won't get checked in case of a self-reference, but
+	 * that's OK because data-modifying CTEs aren't allowed to be recursive
+	 * anyhow.
 	 */
 	if (IsA(cte->ctequery, Query))
 	{
diff --git a/src/backend/rewrite/rewriteHandler.c b/src/backend/rewrite/rewriteHandler.c
new file mode 100644
index c74bac2..939d61e
--- a/src/backend/rewrite/rewriteHandler.c
+++ b/src/backend/rewrite/rewriteHandler.c
@@ -3626,7 +3626,8 @@ RewriteQuery(Query *parsetree, List *rew
 			if (!(ctequery->commandType == CMD_SELECT ||
 				  ctequery->commandType == CMD_UPDATE ||
 				  ctequery->commandType == CMD_INSERT ||
-				  ctequery->commandType == CMD_DELETE))
+				  ctequery->commandType == CMD_DELETE ||
+				  ctequery->commandType == CMD_MERGE))
 			{
 				/*
 				 * Currently it could only be NOTIFY; this error message will
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
new file mode 100644
index c7d9d96..d393f51
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -2131,11 +2131,10 @@ QueryReturnsTuples(Query *parsetree)
 		case CMD_SELECT:
 			/* returns tuples */
 			return true;
-		case CMD_MERGE:
-			return false;
 		case CMD_INSERT:
 		case CMD_UPDATE:
 		case CMD_DELETE:
+		case CMD_MERGE:
 			/* the forms with RETURNING return tuples */
 			if (parsetree->returningList)
 				return true;
diff --git a/src/bin/psql/common.c b/src/bin/psql/common.c
new file mode 100644
index f907f5d..85eb868
--- a/src/bin/psql/common.c
+++ b/src/bin/psql/common.c
@@ -962,13 +962,17 @@ PrintQueryResult(PGresult *result, bool
 			else
 				success = true;
 
-			/* if it's INSERT/UPDATE/DELETE RETURNING, also print status */
+			/*
+			 * If it's INSERT/UPDATE/DELETE/MERGE RETURNING, also print
+			 * status.
+			 */
 			if (last || pset.show_all_results)
 			{
 				cmdstatus = PQcmdStatus(result);
 				if (strncmp(cmdstatus, "INSERT", 6) == 0 ||
 					strncmp(cmdstatus, "UPDATE", 6) == 0 ||
-					strncmp(cmdstatus, "DELETE", 6) == 0)
+					strncmp(cmdstatus, "DELETE", 6) == 0 ||
+					strncmp(cmdstatus, "MERGE", 5) == 0)
 					PrintQueryStatus(result, printStatusFout);
 			}
 
diff --git a/src/include/catalog/pg_proc.dat b/src/include/catalog/pg_proc.dat
new file mode 100644
index 86eb8e8..2c1c3ee
--- a/src/include/catalog/pg_proc.dat
+++ b/src/include/catalog/pg_proc.dat
@@ -11891,4 +11891,14 @@
   prorettype => 'bytea', proargtypes => 'pg_brin_minmax_multi_summary',
   prosrc => 'brin_minmax_multi_summary_send' },
 
+# MERGE support functions
+{ oid => '9499', descr => 'command type of current MERGE action',
+  proname => 'pg_merge_action',  provolatile => 'v',
+  prorettype => 'text', proargtypes => '',
+  prosrc => 'pg_merge_action' },
+{ oid => '9500', descr => 'index of current MERGE WHEN clause',
+  proname => 'pg_merge_when_clause',  provolatile => 'v',
+  prorettype => 'int4', proargtypes => '',
+  prosrc => 'pg_merge_when_clause' },
+
 ]
diff --git a/src/include/catalog/pg_proc.h b/src/include/catalog/pg_proc.h
new file mode 100644
index e7abe0b..e597826
--- a/src/include/catalog/pg_proc.h
+++ b/src/include/catalog/pg_proc.h
@@ -182,6 +182,11 @@ DECLARE_UNIQUE_INDEX(pg_proc_proname_arg
 #define PROARGMODE_VARIADIC 'v'
 #define PROARGMODE_TABLE	't'
 
+/* Is this a merge support function?  (Requires fmgroids.h) */
+#define IsMergeSupportFunction(oid) \
+	((oid) == F_PG_MERGE_ACTION || \
+	 (oid) == F_PG_MERGE_WHEN_CLAUSE)
+
 #endif							/* EXPOSE_TO_CLIENT_CODE */
 
 
diff --git a/src/include/nodes/execnodes.h b/src/include/nodes/execnodes.h
new file mode 100644
index 20f4c8b..5a0126d
--- a/src/include/nodes/execnodes.h
+++ b/src/include/nodes/execnodes.h
@@ -415,6 +415,7 @@ typedef struct MergeActionState
 	NodeTag		type;
 
 	MergeAction *mas_action;	/* associated MergeAction node */
+	int			mas_action_idx;	/* 1-based index of MergeAction node */
 	ProjectionInfo *mas_proj;	/* projection of the action's targetlist for
 								 * this rel */
 	ExprState  *mas_whenqual;	/* WHEN [NOT] MATCHED AND conditions */
@@ -1302,6 +1303,9 @@ typedef struct ModifyTableState
 	/* Flags showing which subcommands are present INS/UPD/DEL/DO NOTHING */
 	int			mt_merge_subcommands;
 
+	/* For MERGE, the action currently being executed */
+	MergeActionState *mt_merge_action;
+
 	/* tuple counters for MERGE */
 	double		mt_merge_inserted;
 	double		mt_merge_updated;
diff --git a/src/include/nodes/parsenodes.h b/src/include/nodes/parsenodes.h
new file mode 100644
index 89335d9..b001d45
--- a/src/include/nodes/parsenodes.h
+++ b/src/include/nodes/parsenodes.h
@@ -1775,6 +1775,7 @@ typedef struct MergeStmt
 	Node	   *sourceRelation; /* source relation */
 	Node	   *joinCondition;	/* join condition between source and target */
 	List	   *mergeWhenClauses;	/* list of MergeWhenClause(es) */
+	List	   *returningList;	/* list of expressions to return */
 	WithClause *withClause;		/* WITH clause */
 } MergeStmt;
 
diff --git a/src/include/parser/analyze.h b/src/include/parser/analyze.h
new file mode 100644
index 1cef183..78eb55e
--- a/src/include/parser/analyze.h
+++ b/src/include/parser/analyze.h
@@ -44,6 +44,8 @@ extern List *transformInsertRow(ParseSta
 								bool strip_indirection);
 extern List *transformUpdateTargetList(ParseState *pstate,
 									   List *origTlist);
+extern List *transformReturningList(ParseState *pstate, List *returningList,
+									ParseExprKind exprKind);
 extern Query *transformTopLevelStmt(ParseState *pstate, RawStmt *parseTree);
 extern Query *transformStmt(ParseState *pstate, Node *parseTree);
 
diff --git a/src/include/parser/parse_node.h b/src/include/parser/parse_node.h
new file mode 100644
index 1a37922..810a707
--- a/src/include/parser/parse_node.h
+++ b/src/include/parser/parse_node.h
@@ -61,7 +61,8 @@ typedef enum ParseExprKind
 	EXPR_KIND_DISTINCT_ON,		/* DISTINCT ON */
 	EXPR_KIND_LIMIT,			/* LIMIT */
 	EXPR_KIND_OFFSET,			/* OFFSET */
-	EXPR_KIND_RETURNING,		/* RETURNING */
+	EXPR_KIND_RETURNING,		/* RETURNING in INSERT/UPDATE/DELETE */
+	EXPR_KIND_MERGE_RETURNING,	/* RETURNING in MERGE */
 	EXPR_KIND_VALUES,			/* VALUES */
 	EXPR_KIND_VALUES_SINGLE,	/* single-row VALUES (in INSERT only) */
 	EXPR_KIND_CHECK_CONSTRAINT, /* CHECK constraint for a table */
diff --git a/src/test/regress/expected/merge.out b/src/test/regress/expected/merge.out
new file mode 100644
index bc53b21..be7128e
--- a/src/test/regress/expected/merge.out
+++ b/src/test/regress/expected/merge.out
@@ -123,20 +123,20 @@ ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
 ERROR:  name "target" specified more than once
 DETAIL:  The name is used both as MERGE target table and data source.
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
-ERROR:  MERGE not supported in WITH query
-LINE 1: WITH foo AS (
-             ^
--- used in COPY
+ERROR:  WITH query "foo" does not have a RETURNING clause
+LINE 4: ) SELECT * FROM foo;
+                        ^
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) TO stdout;
-ERROR:  MERGE not supported in COPY
+ERROR:  COPY query must have a RETURNING clause
 -- unsupported relation types
 -- view
 CREATE VIEW tv AS SELECT * FROM target;
@@ -1289,21 +1289,40 @@ WHEN MATCHED AND tid < 2 THEN
 ROLLBACK;
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
 	DELETE
-RETURNING *;
-ERROR:  syntax error at or near "RETURNING"
-LINE 10: RETURNING *;
-         ^
+RETURNING pg_merge_when_clause() AS when_clause,
+          pg_merge_action() AS merge_action,
+          t.*,
+          CASE pg_merge_action()
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ when_clause | merge_action | tid | balance |     description     
+-------------+--------------+-----+---------+---------------------
+           3 | DELETE       |   1 |     100 | Removed (1,100)
+           1 | UPDATE       |   2 |     220 | Added 20 to balance
+           2 | INSERT       |   4 |      40 | Inserted (4,40)
+(3 rows)
+
 ROLLBACK;
+-- error when using MERGE support functions outside MERGE
+SELECT pg_merge_action() FROM sq_target;
+ERROR:  merge support function pg_merge_action can only be called from the RETURNING list of a MERGE command
+LINE 1: SELECT pg_merge_action() FROM sq_target;
+               ^
+UPDATE sq_target SET balance = balance + 1 RETURNING pg_merge_when_clause();
+ERROR:  merge support function pg_merge_when_clause can only be called from the RETURNING list of a MERGE command
+LINE 1: ...ATE sq_target SET balance = balance + 1 RETURNING pg_merge_w...
+                                                             ^
 -- EXPLAIN
 CREATE TABLE ex_mtarget (a int, b int)
   WITH (autovacuum_enabled=off);
diff --git a/src/test/regress/sql/merge.sql b/src/test/regress/sql/merge.sql
new file mode 100644
index fdbcd70..e434a2b
--- a/src/test/regress/sql/merge.sql
+++ b/src/test/regress/sql/merge.sql
@@ -86,12 +86,12 @@ MERGE INTO target
 USING target
 ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
--- used in COPY
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
@@ -844,19 +844,29 @@ ROLLBACK;
 
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
 	DELETE
-RETURNING *;
+RETURNING pg_merge_when_clause() AS when_clause,
+          pg_merge_action() AS merge_action,
+          t.*,
+          CASE pg_merge_action()
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
 ROLLBACK;
 
+-- error when using MERGE support functions outside MERGE
+SELECT pg_merge_action() FROM sq_target;
+UPDATE sq_target SET balance = balance + 1 RETURNING pg_merge_when_clause();
+
 -- EXPLAIN
 CREATE TABLE ex_mtarget (a int, b int)
   WITH (autovacuum_enabled=off);

diff --git a/doc/src/sgml/dml.sgml b/doc/src/sgml/dml.sgml
new file mode 100644
index cbbc5e2..ff2a827
--- a/doc/src/sgml/dml.sgml
+++ b/doc/src/sgml/dml.sgml
@@ -283,10 +283,15 @@ DELETE FROM products;
    <secondary>RETURNING</secondary>
   </indexterm>
 
+  <indexterm zone="dml-returning">
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
   <para>
    Sometimes it is useful to obtain data from modified rows while they are
    being manipulated.  The <command>INSERT</command>, <command>UPDATE</command>,
-   and <command>DELETE</command> commands all have an
+   <command>DELETE</command>, and <command>MERGE</command> commands all have an
    optional <literal>RETURNING</literal> clause that supports this.  Use
    of <literal>RETURNING</literal> avoids performing an extra database query to
    collect the data, and is especially valuable when it would otherwise be
@@ -339,6 +344,24 @@ DELETE FROM products
 </programlisting>
   </para>
 
+  <para>
+   In a <command>MERGE</command>, the data available to <literal>RETURNING</literal> is
+   the content of the source row plus the content of the inserted, updated, or
+   deleted target row.  In addition, the merge support functions
+   <xref linkend="pg-merge-action"/> and <xref linkend="pg-merge-when-clause"/>
+   may be used to return information about which merge action was executed for
+   each row.  Since it is quite common for the source and target to have many
+   of the same columns, specifying <literal>RETURNING *</literal> can lead to a
+   lot of duplicated columns, so it is often more useful to just return the
+   target row.  For example:
+<programlisting>
+MERGE INTO products p USING new_products n ON p.product_no = n.product_no
+  WHEN NOT MATCHED THEN INSERT VALUES (n.product_no, n.name, n.price)
+  WHEN MATCHED THEN UPDATE SET name = n.name, price = n.price
+  RETURNING pg_merge_action(), p.*;
+</programlisting>
+  </para>
+
   <para>
    If there are triggers (<xref linkend="triggers"/>) on the target table,
    the data available to <literal>RETURNING</literal> is the row as modified by
diff --git a/doc/src/sgml/func.sgml b/doc/src/sgml/func.sgml
new file mode 100644
index e09e289..6927228
--- a/doc/src/sgml/func.sgml
+++ b/doc/src/sgml/func.sgml
@@ -21202,6 +21202,99 @@ SELECT count(*) FROM sometable;
 
  </sect1>
 
+ <sect1 id="functions-merge-support">
+  <title>Merge Support Functions</title>
+
+  <indexterm>
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
+  <para>
+   The merge support functions shown in
+   <xref linkend="functions-merge-support-table"/> may be used in the
+   <literal>RETURNING</literal> list of a <xref linkend="sql-merge"/> command
+   to return additional information about the action taken for each row.
+  </para>
+
+  <table id="functions-merge-support-table">
+    <title>Merge Support Functions</title>
+
+    <tgroup cols="1">
+     <thead>
+      <row>
+       <entry role="func_table_entry"><para role="func_signature">
+        Function
+       </para>
+       <para>
+        Description
+       </para></entry>
+      </row>
+     </thead>
+
+     <tbody>
+      <row>
+       <entry role="func_table_entry"><para role="func_signature">
+        <indexterm id="pg-merge-action">
+         <primary>pg_merge_action</primary>
+        </indexterm>
+        <function>pg_merge_action</function> ( )
+        <returnvalue>text</returnvalue>
+       </para>
+       <para>
+        Returns the action performed on the current row (<literal>'INSERT'</literal>,
+        <literal>'UPDATE'</literal>, or <literal>'DELETE'</literal>).
+       </para></entry>
+      </row>
+
+      <row>
+       <entry role="func_table_entry"><para role="func_signature">
+        <indexterm id="pg-merge-when-clause">
+         <primary>pg_merge_when_clause</primary>
+        </indexterm>
+        <function>pg_merge_when_clause</function> ( )
+        <returnvalue>integer</returnvalue>
+       </para>
+       <para>
+        Returns the 1-based index of the <literal>WHEN</literal> clause executed
+        for the current row.
+       </para></entry>
+      </row>
+     </tbody>
+    </tgroup>
+  </table>
+
+  <para>
+   Example:
+<screen><![CDATA[
+MERGE INTO products p
+  USING stock s ON p.product_id = s.product_id
+  WHEN MATCHED AND s.quantity > 0 THEN
+    UPDATE SET in_stock = true, quantity = s.quantity
+  WHEN MATCHED THEN
+    UPDATE SET in_stock = false, quantity = 0
+  WHEN NOT MATCHED THEN
+    INSERT (product_id, in_stock, quantity)
+      VALUES (s.product_id, true, s.quantity)
+  RETURNING pg_merge_action() AS action,
+            pg_merge_when_clause() AS when_clause,
+            p.*;
+
+ action | when_clause | product_id | in_stock | quantity 
+--------+-------------+------------+----------+----------
+ UPDATE |           1 |       1001 | t        |       50
+ UPDATE |           2 |       1002 | f        |        0
+ INSERT |           3 |       1003 | t        |       10
+]]></screen>
+  </para>
+
+  <para>
+   Note that it is an error to use these functions anywhere other than the
+   <literal>RETURNING</literal> list of a <command>MERGE</command> command.
+  </para>
+
+ </sect1>
+
  <sect1 id="functions-subquery">
   <title>Subquery Expressions</title>
 
diff --git a/doc/src/sgml/glossary.sgml b/doc/src/sgml/glossary.sgml
new file mode 100644
index 7c01a54..4317cfc
--- a/doc/src/sgml/glossary.sgml
+++ b/doc/src/sgml/glossary.sgml
@@ -1323,9 +1323,9 @@
      to a <glossterm linkend="glossary-client">client</glossterm> upon the
      completion of an <acronym>SQL</acronym> command, usually a
      <command>SELECT</command> but it can be an
-     <command>INSERT</command>, <command>UPDATE</command>, or
-     <command>DELETE</command> command if the <literal>RETURNING</literal>
-     clause is specified.
+     <command>INSERT</command>, <command>UPDATE</command>,
+     <command>DELETE</command>, or <command>MERGE</command> command if the
+     <literal>RETURNING</literal> clause is specified.
     </para>
     <para>
      The fact that a result set is a relation means that a query can be used
diff --git a/doc/src/sgml/plpgsql.sgml b/doc/src/sgml/plpgsql.sgml
new file mode 100644
index 8897a54..3249e9c
--- a/doc/src/sgml/plpgsql.sgml
+++ b/doc/src/sgml/plpgsql.sgml
@@ -1020,8 +1020,8 @@ INSERT INTO mytable VALUES (1,'one'), (2
     </para>
 
     <para>
-     If the command does return rows (for example <command>SELECT</command>,
-     or <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>
+     If the command does return rows (for example <command>SELECT</command>, or
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
      with <literal>RETURNING</literal>), there are two ways to proceed.
      When the command will return at most one row, or you only care about
      the first row of output, write the command as usual but add
@@ -1044,7 +1044,8 @@ INSERT INTO mytable VALUES (1,'one'), (2
      <application>PL/pgSQL</application> variable values can be
      automatically inserted into optimizable SQL commands, which
      are <command>SELECT</command>, <command>INSERT</command>,
-     <command>UPDATE</command>, <command>DELETE</command>, and certain
+     <command>UPDATE</command>, <command>DELETE</command>,
+     <command>MERGE</command> and certain
      utility commands that incorporate one of these, such
      as <command>EXPLAIN</command> and <command>CREATE TABLE ... AS
      SELECT</command>.  In these commands,
@@ -1148,6 +1149,7 @@ SELECT <replaceable>select_expressions</
 INSERT ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 UPDATE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 DELETE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
+MERGE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 </synopsis>
 
      where <replaceable>target</replaceable> can be a record variable, a row
@@ -1158,8 +1160,8 @@ DELETE ... RETURNING <replaceable>expres
      <literal>INTO</literal> clause) just as described above,
      and the plan is cached in the same way.
      This works for <command>SELECT</command>,
-     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
-     <literal>RETURNING</literal>, and certain utility commands
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
+     with <literal>RETURNING</literal>, and certain utility commands
      that return row sets, such as <command>EXPLAIN</command>.
      Except for the <literal>INTO</literal> clause, the SQL command is the same
      as it would be written outside <application>PL/pgSQL</application>.
@@ -1235,7 +1237,7 @@ END;
     </para>
 
     <para>
-     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
+     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command> with
      <literal>RETURNING</literal>, <application>PL/pgSQL</application> reports
      an error for more than one returned row, even when
      <literal>STRICT</literal> is not specified.  This is because there
diff --git a/doc/src/sgml/queries.sgml b/doc/src/sgml/queries.sgml
new file mode 100644
index 45741e7..0ca593b
--- a/doc/src/sgml/queries.sgml
+++ b/doc/src/sgml/queries.sgml
@@ -2062,10 +2062,11 @@ SELECT <replaceable>select_list</replace
    Table Expressions or <acronym>CTE</acronym>s, can be thought of as defining
    temporary tables that exist just for one query.  Each auxiliary statement
    in a <literal>WITH</literal> clause can be a <command>SELECT</command>,
-   <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>; and the
+   <command>INSERT</command>, <command>UPDATE</command>, <command>DELETE</command>,
+   or <command>MERGE</command>; and the
    <literal>WITH</literal> clause itself is attached to a primary statement that can
-   also be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>, or
-   <command>DELETE</command>.
+   also be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
+   <command>DELETE</command>, or <command>MERGE</command>.
   </para>
 
  <sect2 id="queries-with-select">
@@ -2587,7 +2588,8 @@ SELECT * FROM w AS w1 JOIN w AS w2 ON w1
   <para>
    The examples above only show <literal>WITH</literal> being used with
    <command>SELECT</command>, but it can be attached in the same way to
-   <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>.
+   <command>INSERT</command>, <command>UPDATE</command>,
+   <command>DELETE</command> or <command>MERGE</command>.
    In each case it effectively provides temporary table(s) that can
    be referred to in the main command.
   </para>
@@ -2598,7 +2600,8 @@ SELECT * FROM w AS w1 JOIN w AS w2 ON w1
 
    <para>
     You can use data-modifying statements (<command>INSERT</command>,
-    <command>UPDATE</command>, or <command>DELETE</command>) in <literal>WITH</literal>.  This
+    <command>UPDATE</command>, <command>DELETE</command>, or
+    <command>MERGE</command>) in <literal>WITH</literal>.  This
     allows you to perform several different operations in the same query.
     An example is:
 
diff --git a/doc/src/sgml/ref/copy.sgml b/doc/src/sgml/ref/copy.sgml
new file mode 100644
index c25b52d..d822470
--- a/doc/src/sgml/ref/copy.sgml
+++ b/doc/src/sgml/ref/copy.sgml
@@ -121,13 +121,15 @@ COPY { <replaceable class="parameter">ta
       A <link linkend="sql-select"><command>SELECT</command></link>,
       <link linkend="sql-values"><command>VALUES</command></link>,
       <link linkend="sql-insert"><command>INSERT</command></link>,
-      <link linkend="sql-update"><command>UPDATE</command></link>, or
-      <link linkend="sql-delete"><command>DELETE</command></link> command whose results are to be
+      <link linkend="sql-update"><command>UPDATE</command></link>,
+      <link linkend="sql-update"><command>DELETE</command></link>, or
+      <link linkend="sql-delete"><command>MERGE</command></link> command whose results are to be
       copied.  Note that parentheses are required around the query.
      </para>
      <para>
-      For <command>INSERT</command>, <command>UPDATE</command> and
-      <command>DELETE</command> queries a RETURNING clause must be provided,
+      For <command>INSERT</command>, <command>UPDATE</command>,
+      <command>DELETE</command>, and <command>MERGE</command> queries a
+      RETURNING clause must be provided,
       and the target relation must not have a conditional rule, nor
       an <literal>ALSO</literal> rule, nor an <literal>INSTEAD</literal> rule
       that expands to multiple statements.
diff --git a/doc/src/sgml/ref/merge.sgml b/doc/src/sgml/ref/merge.sgml
new file mode 100644
index 0995fe0..7eb641a
--- a/doc/src/sgml/ref/merge.sgml
+++ b/doc/src/sgml/ref/merge.sgml
@@ -25,6 +25,7 @@ PostgreSQL documentation
 MERGE INTO [ ONLY ] <replaceable class="parameter">target_table_name</replaceable> [ * ] [ [ AS ] <replaceable class="parameter">target_alias</replaceable> ]
 USING <replaceable class="parameter">data_source</replaceable> ON <replaceable class="parameter">join_condition</replaceable>
 <replaceable class="parameter">when_clause</replaceable> [...]
+[ RETURNING * | <replaceable class="parameter">output_expression</replaceable> [ [ AS ] <replaceable class="parameter">output_name</replaceable> ] [, ...] ]
 
 <phrase>where <replaceable class="parameter">data_source</replaceable> is:</phrase>
 
@@ -95,6 +96,20 @@ DELETE
   </para>
 
   <para>
+   The optional <literal>RETURNING</literal> clause causes <command>MERGE</command>
+   to compute and return value(s) based on each row inserted, updated, or
+   deleted.  Any expression using the source or target table's columns, or the
+   merge support functions <xref linkend="pg-merge-action"/> and
+   <xref linkend="pg-merge-when-clause"/> can be computed.  When an
+   <command>INSERT</command> or <command>UPDATE</command> action is performed,
+   the new values of the target table's columns are used.  When a
+   <command>DELETE</command> is performed, the old values of the target table's
+   columns are used.  The syntax of the <literal>RETURNING</literal> list is
+   identical to that of the output list of <command>SELECT</command>.
+
+  </para>
+
+  <para>
    There is no separate <literal>MERGE</literal> privilege.
    If you specify an update action, you must have the
    <literal>UPDATE</literal> privilege on the column(s)
@@ -125,6 +140,18 @@ DELETE
 
   <variablelist>
    <varlistentry>
+    <term><replaceable class="parameter">with_query</replaceable></term>
+    <listitem>
+     <para>
+      The <literal>WITH</literal> clause allows you to specify one or more
+      subqueries that can be referenced by name in the <command>MERGE</command>
+      query. See <xref linkend="queries-with"/> and <xref linkend="sql-select"/>
+      for details.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
     <term><replaceable class="parameter">target_table_name</replaceable></term>
     <listitem>
      <para>
@@ -399,13 +426,30 @@ DELETE
    </varlistentry>
 
    <varlistentry>
-    <term><replaceable class="parameter">with_query</replaceable></term>
+    <term><replaceable class="parameter">output_expression</replaceable></term>
     <listitem>
      <para>
-      The <literal>WITH</literal> clause allows you to specify one or more
-      subqueries that can be referenced by name in the <command>MERGE</command>
-      query. See <xref linkend="queries-with"/> and <xref linkend="sql-select"/>
-      for details.
+      An expression to be computed and returned by the <command>MERGE</command>
+      command after each row is merged.  The expression can use any columns of
+      the source or target tables, or the merge support functions
+      <xref linkend="pg-merge-action"/> and <xref linkend="pg-merge-when-clause"/>.
+     </para>
+     <para>
+      Writing <literal>*</literal> will return all columns from the source
+      table, followed by all columns from the target table.  Often this will
+      lead to a lot of duplication, since it is common for the source and
+      target tables to have a lot of the same columns.  This can be avoided by
+      qualifying the <literal>*</literal> with the name or alias of the source
+      or target table.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
+    <term><replaceable class="parameter">output_name</replaceable></term>
+    <listitem>
+     <para>
+      A name to use for a returned column.
      </para>
     </listitem>
    </varlistentry>
@@ -428,6 +472,13 @@ MERGE <replaceable class="parameter">tot
    were changed in any way.
   </para>
 
+  <para>
+   If the <command>MERGE</command> command contains a <literal>RETURNING</literal>
+   clause, the result will be similar to that of a <command>SELECT</command>
+   statement containing the columns and values defined in the
+   <literal>RETURNING</literal> list, computed over the row(s) inserted, updated,
+   or deleted by the command.
+  </para>
  </refsect1>
 
  <refsect1>
@@ -544,13 +595,6 @@ MERGE <replaceable class="parameter">tot
   </para>
 
   <para>
-   There is no <literal>RETURNING</literal> clause with
-   <command>MERGE</command>.  Actions of <command>INSERT</command>,
-   <command>UPDATE</command> and <command>DELETE</command> cannot contain
-   <literal>RETURNING</literal> or <literal>WITH</literal> clauses.
-  </para>
-
-  <para>
    When <command>MERGE</command> is run concurrently with other commands
    that modify the target table, the usual transaction isolation rules
    apply; see <xref linkend="transaction-iso"/> for an explanation
@@ -602,7 +646,8 @@ WHEN NOT MATCHED THEN
   <para>
    Attempt to insert a new stock item along with the quantity of stock. If
    the item already exists, instead update the stock count of the existing
-   item. Don't allow entries that have zero stock.
+   item. Don't allow entries that have zero stock. Return details of all
+   changes made.
 <programlisting>
 MERGE INTO wines w
 USING wine_stock_changes s
@@ -612,7 +657,8 @@ WHEN NOT MATCHED AND s.stock_delta > 0 T
 WHEN MATCHED AND w.stock + s.stock_delta > 0 THEN
   UPDATE SET stock = w.stock + s.stock_delta
 WHEN MATCHED THEN
-  DELETE;
+  DELETE
+RETURNING pg_merge_action(), w.*;
 </programlisting>
 
    The <literal>wine_stock_changes</literal> table might be, for example, a
@@ -627,7 +673,8 @@ WHEN MATCHED THEN
     This command conforms to the <acronym>SQL</acronym> standard.
   </para>
    <para>
-    The WITH clause and <literal>DO NOTHING</literal> action are extensions to
+    The WITH clause, <literal>DO NOTHING</literal> action, and
+    <literal>RETURNING</literal> clause are extensions to
     the <acronym>SQL</acronym> standard.
   </para>
  </refsect1>
diff --git a/doc/src/sgml/ref/select.sgml b/doc/src/sgml/ref/select.sgml
new file mode 100644
index 1f9538f..5840ddc
--- a/doc/src/sgml/ref/select.sgml
+++ b/doc/src/sgml/ref/select.sgml
@@ -74,7 +74,7 @@ SELECT [ ALL | DISTINCT [ ON ( <replacea
 
 <phrase>and <replaceable class="parameter">with_query</replaceable> is:</phrase>
 
-    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> )
+    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> | <replaceable class="parameter">merge</replaceable> )
         [ SEARCH { BREADTH | DEPTH } FIRST BY <replaceable>column_name</replaceable> [, ...] SET <replaceable>search_seq_col_name</replaceable> ]
         [ CYCLE <replaceable>column_name</replaceable> [, ...] SET <replaceable>cycle_mark_col_name</replaceable> [ TO <replaceable>cycle_mark_value</replaceable> DEFAULT <replaceable>cycle_mark_default</replaceable> ] USING <replaceable>cycle_path_col_name</replaceable> ]
 
@@ -227,10 +227,10 @@ TABLE [ ONLY ] <replaceable class="param
     The subqueries effectively act as temporary tables or views
     for the duration of the primary query.
     Each subquery can be a <command>SELECT</command>, <command>TABLE</command>, <command>VALUES</command>,
-    <command>INSERT</command>, <command>UPDATE</command> or
-    <command>DELETE</command> statement.
+    <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command> statement.
     When writing a data-modifying statement (<command>INSERT</command>,
-    <command>UPDATE</command> or <command>DELETE</command>) in
+    <command>UPDATE</command>, <command>DELETE</command>, or <command>MERGE</command>) in
     <literal>WITH</literal>, it is usual to include a <literal>RETURNING</literal> clause.
     It is the output of <literal>RETURNING</literal>, <emphasis>not</emphasis> the underlying
     table that the statement modifies, that forms the temporary table that is
@@ -2179,7 +2179,8 @@ SELECT 2+2;
 
    <para>
     <productname>PostgreSQL</productname> allows <command>INSERT</command>,
-    <command>UPDATE</command>, and <command>DELETE</command> to be used as <literal>WITH</literal>
+    <command>UPDATE</command>, <command>DELETE</command>, and
+    <command>MERGE</command> to be used as <literal>WITH</literal>
     queries.  This is not found in the SQL standard.
    </para>
   </refsect2>
diff --git a/doc/src/sgml/rowtypes.sgml b/doc/src/sgml/rowtypes.sgml
new file mode 100644
index 4d86f97..bbeac84
--- a/doc/src/sgml/rowtypes.sgml
+++ b/doc/src/sgml/rowtypes.sgml
@@ -348,7 +348,7 @@ SELECT m.* FROM some_table, LATERAL myfu
    column expansion of this kind when it appears at the top level of
    a <link linkend="queries-select-lists"><command>SELECT</command> output
    list</link>, a <link linkend="dml-returning"><literal>RETURNING</literal>
-   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>,
+   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>,
    a <link linkend="queries-values"><literal>VALUES</literal> clause</link>, or
    a <link linkend="sql-syntax-row-constructors">row constructor</link>.
    In all other contexts (including when nested inside one of those
diff --git a/doc/src/sgml/spi.sgml b/doc/src/sgml/spi.sgml
new file mode 100644
index 7581661..ad0a9b1
--- a/doc/src/sgml/spi.sgml
+++ b/doc/src/sgml/spi.sgml
@@ -301,8 +301,9 @@ SPI_execute("INSERT INTO foo SELECT * FR
    is returned in the global variable <varname>SPI_processed</varname>.
    If the return value of the function is <symbol>SPI_OK_SELECT</symbol>,
    <symbol>SPI_OK_INSERT_RETURNING</symbol>,
-   <symbol>SPI_OK_DELETE_RETURNING</symbol>, or
-   <symbol>SPI_OK_UPDATE_RETURNING</symbol>,
+   <symbol>SPI_OK_DELETE_RETURNING</symbol>,
+   <symbol>SPI_OK_UPDATE_RETURNING</symbol>, or
+   <symbol>SPI_OK_MERGE_RETURNING</symbol>
    then you can use the
    global pointer <literal>SPITupleTable *SPI_tuptable</literal> to
    access the result rows.  Some utility commands (such as
@@ -438,6 +439,15 @@ typedef struct SPITupleTable
     </varlistentry>
 
     <varlistentry>
+     <term><symbol>SPI_OK_MERGE</symbol></term>
+     <listitem>
+      <para>
+       if a <command>MERGE</command> was executed
+      </para>
+     </listitem>
+    </varlistentry>
+
+    <varlistentry>
      <term><symbol>SPI_OK_INSERT_RETURNING</symbol></term>
      <listitem>
       <para>
@@ -463,6 +473,15 @@ typedef struct SPITupleTable
       </para>
      </listitem>
     </varlistentry>
+
+    <varlistentry>
+     <term><symbol>SPI_OK_MERGE_RETURNING</symbol></term>
+     <listitem>
+      <para>
+       if a <command>MERGE RETURNING</command> was executed
+      </para>
+     </listitem>
+    </varlistentry>
 
     <varlistentry>
      <term><symbol>SPI_OK_UTILITY</symbol></term>
diff --git a/doc/src/sgml/xfunc.sgml b/doc/src/sgml/xfunc.sgml
new file mode 100644
index e2a5496..114c3d3
--- a/doc/src/sgml/xfunc.sgml
+++ b/doc/src/sgml/xfunc.sgml
@@ -177,7 +177,8 @@
     statements separated by semicolons.  A semicolon after the last
     statement is optional.  Unless the function is declared to return
     <type>void</type>, the last statement must be a <command>SELECT</command>,
-    or an <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>
+    or an <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command>
     that has a <literal>RETURNING</literal> clause.
    </para>
 
@@ -186,7 +187,8 @@
      language can be packaged together and defined as a function.
      Besides <command>SELECT</command> queries, the commands can include data
      modification queries (<command>INSERT</command>,
-     <command>UPDATE</command>, and <command>DELETE</command>), as well as
+     <command>UPDATE</command>, <command>DELETE</command> and
+     <command>MERGE</command>), as well as
      other SQL commands. (You cannot use transaction control commands, e.g.,
      <command>COMMIT</command>, <command>SAVEPOINT</command>, and some utility
      commands, e.g.,  <literal>VACUUM</literal>, in <acronym>SQL</acronym> functions.)
@@ -1235,8 +1237,9 @@ SELECT x, CASE WHEN x &gt; 0 THEN genera
 
     <note>
      <para>
-      If a function's last command is <command>INSERT</command>, <command>UPDATE</command>,
-      or <command>DELETE</command> with <literal>RETURNING</literal>, that command will
+      If a function's last command is <command>INSERT</command>,
+      <command>UPDATE</command>, <command>DELETE</command>, or
+      <command>MERGE</command> with <literal>RETURNING</literal>, that command will
       always be executed to completion, even if the function is not declared
       with <literal>SETOF</literal> or the calling query does not fetch all the
       result rows.  Any extra rows produced by the <literal>RETURNING</literal>
diff --git a/src/backend/commands/copy.c b/src/backend/commands/copy.c
new file mode 100644
index e34f583..aa3cca0
--- a/src/backend/commands/copy.c
+++ b/src/backend/commands/copy.c
@@ -274,12 +274,6 @@ DoCopy(ParseState *pstate, const CopyStm
 	{
 		Assert(stmt->query);
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(stmt->query, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in COPY"));
-
 		query = makeNode(RawStmt);
 		query->stmt = stmt->query;
 		query->stmt_location = stmt_location;
diff --git a/src/backend/commands/copyto.c b/src/backend/commands/copyto.c
new file mode 100644
index 8043b4e..e02d7d0
--- a/src/backend/commands/copyto.c
+++ b/src/backend/commands/copyto.c
@@ -510,7 +510,8 @@ BeginCopyTo(ParseState *pstate,
 		{
 			Assert(query->commandType == CMD_INSERT ||
 				   query->commandType == CMD_UPDATE ||
-				   query->commandType == CMD_DELETE);
+				   query->commandType == CMD_DELETE ||
+				   query->commandType == CMD_MERGE);
 
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
diff --git a/src/backend/executor/execExpr.c b/src/backend/executor/execExpr.c
new file mode 100644
index 812ead9..8572b01
--- a/src/backend/executor/execExpr.c
+++ b/src/backend/executor/execExpr.c
@@ -48,6 +48,7 @@
 #include "utils/array.h"
 #include "utils/builtins.h"
 #include "utils/datum.h"
+#include "utils/fmgroids.h"
 #include "utils/lsyscache.h"
 #include "utils/typcache.h"
 
@@ -2485,6 +2486,22 @@ ExecInitFunc(ExprEvalStep *scratch, Expr
 	InitFunctionCallInfoData(*fcinfo, flinfo,
 							 nargs, inputcollid, NULL, NULL);
 
+	/*
+	 * Merge support functions should only be called directly from a MERGE
+	 * command, and need access to the parent ModifyTableState.  The parser
+	 * should have checked that such functions only appear in the RETURNING
+	 * list of a MERGE, so this should never fail.
+	 */
+	if (IsMergeSupportFunction(funcid))
+	{
+		if (!state->parent ||
+			!IsA(state->parent, ModifyTableState) ||
+			((ModifyTableState *) state->parent)->operation != CMD_MERGE)
+			elog(ERROR, "merge support function called in non-merge context");
+
+		fcinfo->context = (Node *) state->parent;
+	}
+
 	/* Keep extra copies of this info to save an indirection at runtime */
 	scratch->d.func.fn_addr = flinfo->fn_addr;
 	scratch->d.func.nargs = nargs;
diff --git a/src/backend/executor/execPartition.c b/src/backend/executor/execPartition.c
new file mode 100644
index 651ad24..95bbf67
--- a/src/backend/executor/execPartition.c
+++ b/src/backend/executor/execPartition.c
@@ -545,8 +545,8 @@ ExecInitPartitionInfo(ModifyTableState *
 	 * Build WITH CHECK OPTION constraints for the partition.  Note that we
 	 * didn't build the withCheckOptionList for partitions within the planner,
 	 * but simple translation of varattnos will suffice.  This only occurs for
-	 * the INSERT case or in the case of UPDATE tuple routing where we didn't
-	 * find a result rel to reuse.
+	 * the INSERT case or in the case of UPDATE/MERGE tuple routing where we
+	 * didn't find a result rel to reuse.
 	 */
 	if (node && node->withCheckOptionLists != NIL)
 	{
@@ -557,13 +557,16 @@ ExecInitPartitionInfo(ModifyTableState *
 		/*
 		 * In the case of INSERT on a partitioned table, there is only one
 		 * plan.  Likewise, there is only one WCO list, not one per partition.
-		 * For UPDATE, there are as many WCO lists as there are plans.
+		 * For UPDATE/MERGE, there are as many WCO lists as there are plans.
 		 */
 		Assert((node->operation == CMD_INSERT &&
 				list_length(node->withCheckOptionLists) == 1 &&
 				list_length(node->resultRelations) == 1) ||
 			   (node->operation == CMD_UPDATE &&
 				list_length(node->withCheckOptionLists) ==
+				list_length(node->resultRelations)) ||
+			   (node->operation == CMD_MERGE &&
+				list_length(node->withCheckOptionLists) ==
 				list_length(node->resultRelations)));
 
 		/*
@@ -609,8 +612,8 @@ ExecInitPartitionInfo(ModifyTableState *
 	 * Build the RETURNING projection for the partition.  Note that we didn't
 	 * build the returningList for partitions within the planner, but simple
 	 * translation of varattnos will suffice.  This only occurs for the INSERT
-	 * case or in the case of UPDATE tuple routing where we didn't find a
-	 * result rel to reuse.
+	 * case or in the case of UPDATE/MERGE tuple routing where we didn't find
+	 * a result rel to reuse.
 	 */
 	if (node && node->returningLists != NIL)
 	{
@@ -624,6 +627,9 @@ ExecInitPartitionInfo(ModifyTableState *
 				list_length(node->resultRelations) == 1) ||
 			   (node->operation == CMD_UPDATE &&
 				list_length(node->returningLists) ==
+				list_length(node->resultRelations)) ||
+			   (node->operation == CMD_MERGE &&
+				list_length(node->returningLists) ==
 				list_length(node->resultRelations)));
 
 		/*
diff --git a/src/backend/executor/functions.c b/src/backend/executor/functions.c
new file mode 100644
index 50e06ec..7e4717a
--- a/src/backend/executor/functions.c
+++ b/src/backend/executor/functions.c
@@ -1665,8 +1665,8 @@ check_sql_fn_retval(List *queryTreeLists
 
 	/*
 	 * If it's a plain SELECT, it returns whatever the targetlist says.
-	 * Otherwise, if it's INSERT/UPDATE/DELETE with RETURNING, it returns
-	 * that. Otherwise, the function return type must be VOID.
+	 * Otherwise, if it's INSERT/UPDATE/DELETE/MERGE with RETURNING, it
+	 * returns that. Otherwise, the function return type must be VOID.
 	 *
 	 * Note: eventually replace this test with QueryReturnsTuples?	We'd need
 	 * a more general method of determining the output type, though.  Also, it
@@ -1684,7 +1684,8 @@ check_sql_fn_retval(List *queryTreeLists
 	else if (parse &&
 			 (parse->commandType == CMD_INSERT ||
 			  parse->commandType == CMD_UPDATE ||
-			  parse->commandType == CMD_DELETE) &&
+			  parse->commandType == CMD_DELETE ||
+			  parse->commandType == CMD_MERGE) &&
 			 parse->returningList)
 	{
 		tlist = parse->returningList;
diff --git a/src/backend/executor/nodeModifyTable.c b/src/backend/executor/nodeModifyTable.c
new file mode 100644
index 1ac6517..0bf94dc
--- a/src/backend/executor/nodeModifyTable.c
+++ b/src/backend/executor/nodeModifyTable.c
@@ -36,8 +36,7 @@
  *		RETURNING tuple after completing each row insert, update, or delete.
  *		It must be called again to continue the operation.  Without RETURNING,
  *		we just loop within the node until all the work is done, then
- *		return NULL.  This avoids useless call/return overhead.  (MERGE does
- *		not support RETURNING.)
+ *		return NULL.  This avoids useless call/return overhead.
  */
 
 #include "postgres.h"
@@ -97,9 +96,6 @@ typedef struct ModifyTableContext
 										  TupleTableSlot *oldSlot,
 										  MergeActionState *relaction);
 
-	/* MERGE specific */
-	MergeActionState *relaction;	/* MERGE action in progress */
-
 	/*
 	 * Information about the changes that were made concurrently to a tuple
 	 * being updated or deleted
@@ -172,13 +168,14 @@ static TupleTableSlot *ExecMerge(ModifyT
 								 ItemPointer tupleid,
 								 bool canSetTag);
 static void ExecInitMerge(ModifyTableState *mtstate, EState *estate);
-static bool ExecMergeMatched(ModifyTableContext *context,
-							 ResultRelInfo *resultRelInfo,
-							 ItemPointer tupleid,
-							 bool canSetTag);
-static void ExecMergeNotMatched(ModifyTableContext *context,
-								ResultRelInfo *resultRelInfo,
-								bool canSetTag);
+static TupleTableSlot *ExecMergeMatched(ModifyTableContext *context,
+										ResultRelInfo *resultRelInfo,
+										ItemPointer tupleid,
+										bool canSetTag,
+										bool *matched);
+static TupleTableSlot *ExecMergeNotMatched(ModifyTableContext *context,
+										   ResultRelInfo *resultRelInfo,
+										   bool canSetTag);
 static TupleTableSlot *mergeGetUpdateNewTuple(ResultRelInfo *relinfo,
 											  TupleTableSlot *planSlot,
 											  TupleTableSlot *oldSlot,
@@ -987,7 +984,7 @@ ExecInsert(ModifyTableContext *context,
 		if (mtstate->operation == CMD_UPDATE)
 			wco_kind = WCO_RLS_UPDATE_CHECK;
 		else if (mtstate->operation == CMD_MERGE)
-			wco_kind = (context->relaction->mas_action->commandType == CMD_UPDATE) ?
+			wco_kind = (mtstate->mt_merge_action->mas_action->commandType == CMD_UPDATE) ?
 				WCO_RLS_UPDATE_CHECK : WCO_RLS_INSERT_CHECK;
 		else
 			wco_kind = WCO_RLS_INSERT_CHECK;
@@ -1838,7 +1835,7 @@ ExecCrossPartitionUpdate(ModifyTableCont
 			/* and project the new tuple to retry the UPDATE with */
 			context->cpUpdateRetrySlot =
 				context->GetUpdateNewTuple(resultRelInfo, epqslot, oldSlot,
-										   context->relaction);
+										   mtstate->mt_merge_action);
 			return false;
 		}
 	}
@@ -2054,7 +2051,7 @@ lreplace:
 		 * No luck, a retry is needed.  If running MERGE, we do not do so
 		 * here; instead let it handle that on its own rules.
 		 */
-		if (context->relaction != NULL)
+		if (context->mtstate->mt_merge_action != NULL)
 			return TM_Updated;
 
 		/*
@@ -2692,6 +2689,7 @@ static TupleTableSlot *
 ExecMerge(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 		  ItemPointer tupleid, bool canSetTag)
 {
+	TupleTableSlot *rslot = NULL;
 	bool		matched;
 
 	/*-----
@@ -2739,7 +2737,8 @@ ExecMerge(ModifyTableContext *context, R
 	 */
 	matched = tupleid != NULL;
 	if (matched)
-		matched = ExecMergeMatched(context, resultRelInfo, tupleid, canSetTag);
+		rslot = ExecMergeMatched(context, resultRelInfo, tupleid, canSetTag,
+								 &matched);
 
 	/*
 	 * Either we were dealing with a NOT MATCHED tuple or ExecMergeMatched()
@@ -2747,10 +2746,9 @@ ExecMerge(ModifyTableContext *context, R
 	 * matches.
 	 */
 	if (!matched)
-		ExecMergeNotMatched(context, resultRelInfo, canSetTag);
+		rslot = ExecMergeNotMatched(context, resultRelInfo, canSetTag);
 
-	/* No RETURNING support yet */
-	return NULL;
+	return rslot;
 }
 
 /*
@@ -2760,8 +2758,8 @@ ExecMerge(ModifyTableContext *context, R
  * We start from the first WHEN MATCHED action and check if the WHEN quals
  * pass, if any. If the WHEN quals for the first action do not pass, we
  * check the second, then the third and so on. If we reach to the end, no
- * action is taken and we return true, indicating that no further action is
- * required for this tuple.
+ * action is taken and "matched" is set to true, indicating that no further
+ * action is required for this tuple.
  *
  * If we do find a qualifying action, then we attempt to execute the action.
  *
@@ -2770,15 +2768,16 @@ ExecMerge(ModifyTableContext *context, R
  * with individual actions are evaluated by this routine via ExecQual, while
  * EvalPlanQual checks for the join quals. If EvalPlanQual tells us that the
  * updated tuple still passes the join quals, then we restart from the first
- * action to look for a qualifying action. Otherwise, we return false --
- * meaning that a NOT MATCHED action must now be executed for the current
- * source tuple.
+ * action to look for a qualifying action. Otherwise, "matched" is set to
+ * false -- meaning that a NOT MATCHED action must now be executed for the
+ * current source tuple.
  */
-static bool
+static TupleTableSlot *
 ExecMergeMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
-				 ItemPointer tupleid, bool canSetTag)
+				 ItemPointer tupleid, bool canSetTag, bool *matched)
 {
 	ModifyTableState *mtstate = context->mtstate;
+	TupleTableSlot *rslot = NULL;
 	TupleTableSlot *newslot;
 	EState	   *estate = context->estate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
@@ -2790,7 +2789,10 @@ ExecMergeMatched(ModifyTableContext *con
 	 * If there are no WHEN MATCHED actions, we are done.
 	 */
 	if (resultRelInfo->ri_matchedMergeAction == NIL)
-		return true;
+	{
+		*matched = true;
+		return NULL;
+	}
 
 	/*
 	 * Make tuple and any needed join variables available to ExecQual and
@@ -2869,7 +2871,7 @@ lmerge_matched:
 				 */
 				newslot = ExecProject(relaction->mas_proj);
 
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				context->GetUpdateNewTuple = mergeGetUpdateNewTuple;
 				context->cpUpdateRetrySlot = NULL;
 
@@ -2892,7 +2894,7 @@ lmerge_matched:
 				break;
 
 			case CMD_DELETE:
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				if (!ExecDeletePrologue(context, resultRelInfo, tupleid,
 										NULL, NULL))
 				{
@@ -2952,7 +2954,8 @@ lmerge_matched:
 				 * If the tuple was already deleted, return to let caller
 				 * handle it under NOT MATCHED clauses.
 				 */
-				return false;
+				*matched = false;
+				return NULL;
 
 			case TM_Updated:
 				{
@@ -3020,13 +3023,19 @@ lmerge_matched:
 							 * NOT MATCHED actions.
 							 */
 							if (TupIsNull(epqslot))
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							(void) ExecGetJunkAttribute(epqslot,
 														resultRelInfo->ri_RowIdAttNo,
 														&isNull);
 							if (isNull)
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							/*
 							 * When a tuple was updated and migrated to
@@ -3061,7 +3070,8 @@ lmerge_matched:
 							 * tuple already deleted; tell caller to run NOT
 							 * MATCHED actions
 							 */
-							return false;
+							*matched = false;
+							return NULL;
 
 						case TM_SelfModified:
 
@@ -3081,13 +3091,14 @@ lmerge_matched:
 										(errcode(ERRCODE_TRIGGERED_DATA_CHANGE_VIOLATION),
 										 errmsg("tuple to be updated or deleted was already modified by an operation triggered by the current command"),
 										 errhint("Consider using an AFTER trigger instead of a BEFORE trigger to propagate changes to other rows.")));
-							return false;
+							*matched = false;
+							return NULL;
 
 						default:
 							/* see table_tuple_lock call in ExecDelete() */
 							elog(ERROR, "unexpected table_tuple_lock status: %u",
 								 result);
-							return false;
+							return NULL;
 					}
 				}
 
@@ -3099,6 +3110,30 @@ lmerge_matched:
 				break;
 		}
 
+		/* Process RETURNING if present */
+		if (resultRelInfo->ri_projectReturning)
+		{
+			switch (commandType)
+			{
+				case CMD_UPDATE:
+					rslot = ExecProcessReturning(resultRelInfo, newslot,
+												 context->planSlot);
+					break;
+
+				case CMD_DELETE:
+					rslot = ExecProcessReturning(resultRelInfo,
+												 resultRelInfo->ri_oldTupleSlot,
+												 context->planSlot);
+					break;
+
+				case CMD_NOTHING:
+					break;
+
+				default:
+					elog(ERROR, "unknown action in MERGE WHEN MATCHED clause");
+			}
+		}
+
 		/*
 		 * We've activated one of the WHEN clauses, so we don't search
 		 * further. This is required behaviour, not an optimization.
@@ -3109,19 +3144,22 @@ lmerge_matched:
 	/*
 	 * Successfully executed an action or no qualifying action was found.
 	 */
-	return true;
+	*matched = true;
+
+	return rslot;
 }
 
 /*
  * Execute the first qualifying NOT MATCHED action.
  */
-static void
+static TupleTableSlot *
 ExecMergeNotMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 					bool canSetTag)
 {
 	ModifyTableState *mtstate = context->mtstate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
 	List	   *actionStates = NIL;
+	TupleTableSlot *rslot = NULL;
 	ListCell   *l;
 
 	/*
@@ -3171,10 +3209,10 @@ ExecMergeNotMatched(ModifyTableContext *
 				 * so we don't need to map the tuple here.
 				 */
 				newslot = ExecProject(action->mas_proj);
-				context->relaction = action;
+				mtstate->mt_merge_action = action;
 
-				(void) ExecInsert(context, mtstate->rootResultRelInfo, newslot,
-								  canSetTag, NULL, NULL);
+				rslot = ExecInsert(context, mtstate->rootResultRelInfo,
+								   newslot, canSetTag, NULL, NULL);
 				mtstate->mt_merge_inserted += 1;
 				break;
 			case CMD_NOTHING:
@@ -3190,6 +3228,8 @@ ExecMergeNotMatched(ModifyTableContext *
 		 */
 		break;
 	}
+
+	return rslot;
 }
 
 /*
@@ -3386,6 +3426,64 @@ mergeGetUpdateNewTuple(ResultRelInfo *re
 }
 
 /*
+ * pg_merge_action() -
+ *	  SQL merge support function to retrieve the currently executing merge
+ *	  action command string ("INSERT", "UPDATE", or "DELETE").
+ */
+Datum
+pg_merge_action(PG_FUNCTION_ARGS)
+{
+	ModifyTableState *mtstate = (ModifyTableState *) fcinfo->context;
+	MergeActionState *relaction;
+
+	if (!mtstate || mtstate->operation != CMD_MERGE)
+		elog(ERROR, "merge support function called in non-merge context");
+
+	relaction = mtstate->mt_merge_action;
+	if (relaction)
+	{
+		CmdType		commandType = relaction->mas_action->commandType;
+
+		switch (commandType)
+		{
+			case CMD_INSERT:
+				PG_RETURN_TEXT_P(cstring_to_text("INSERT"));
+			case CMD_UPDATE:
+				PG_RETURN_TEXT_P(cstring_to_text("UPDATE"));
+			case CMD_DELETE:
+				PG_RETURN_TEXT_P(cstring_to_text("DELETE"));
+			case CMD_NOTHING:
+				PG_RETURN_NULL();
+			default:
+				elog(ERROR, "unrecognized commandType: %d", (int) commandType);
+		}
+	}
+
+	PG_RETURN_NULL();
+}
+
+/*
+ * pg_merge_when_clause() -
+ *	  SQL merge support function to retrieve the 1-based index of the
+ *	  currently executing merge WHEN clause.
+ */
+Datum
+pg_merge_when_clause(PG_FUNCTION_ARGS)
+{
+	ModifyTableState *mtstate = (ModifyTableState *) fcinfo->context;
+	MergeActionState *relaction;
+
+	if (!mtstate || mtstate->operation != CMD_MERGE)
+		elog(ERROR, "merge support function called in non-merge context");
+
+	relaction = mtstate->mt_merge_action;
+	if (relaction)
+		PG_RETURN_INT32((int32) relaction->mas_action->index);
+
+	PG_RETURN_NULL();
+}
+
+/*
  * Process BEFORE EACH STATEMENT triggers
  */
 static void
@@ -3671,8 +3769,17 @@ ExecModifyTable(PlanState *pstate)
 				{
 					EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-					ExecMerge(&context, node->resultRelInfo, NULL, node->canSetTag);
-					continue;	/* no RETURNING support yet */
+					slot = ExecMerge(&context, node->resultRelInfo, NULL,
+									 node->canSetTag);
+
+					/*
+					 * If we got a RETURNING result, return it to the caller.
+					 * We'll continue the work on next call.
+					 */
+					if (slot)
+						return slot;
+
+					continue;	/* continue with the next tuple */
 				}
 
 				elog(ERROR, "tableoid is NULL");
@@ -3749,8 +3856,17 @@ ExecModifyTable(PlanState *pstate)
 					{
 						EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-						ExecMerge(&context, node->resultRelInfo, NULL, node->canSetTag);
-						continue;	/* no RETURNING support yet */
+						slot = ExecMerge(&context, node->resultRelInfo, NULL,
+										 node->canSetTag);
+
+						/*
+						 * If we got a RETURNING result, return it to the
+						 * caller.  We'll continue the work on next call.
+						 */
+						if (slot)
+							return slot;
+
+						continue;	/* continue with the next tuple */
 					}
 
 					elog(ERROR, "ctid is NULL");
@@ -3843,7 +3959,7 @@ ExecModifyTable(PlanState *pstate)
 				slot = internalGetUpdateNewTuple(resultRelInfo, context.planSlot,
 												 oldSlot, NULL);
 				context.GetUpdateNewTuple = internalGetUpdateNewTuple;
-				context.relaction = NULL;
+				node->mt_merge_action = NULL;
 
 				/* Now apply the update. */
 				slot = ExecUpdate(&context, resultRelInfo, tupleid, oldtuple,
diff --git a/src/backend/executor/spi.c b/src/backend/executor/spi.c
new file mode 100644
index 61f03e3..f7f40ae
--- a/src/backend/executor/spi.c
+++ b/src/backend/executor/spi.c
@@ -2015,6 +2015,8 @@ SPI_result_code_string(int code)
 			return "SPI_OK_DELETE";
 		case SPI_OK_UPDATE:
 			return "SPI_OK_UPDATE";
+		case SPI_OK_MERGE:
+			return "SPI_OK_MERGE";
 		case SPI_OK_CURSOR:
 			return "SPI_OK_CURSOR";
 		case SPI_OK_INSERT_RETURNING:
@@ -2023,6 +2025,8 @@ SPI_result_code_string(int code)
 			return "SPI_OK_DELETE_RETURNING";
 		case SPI_OK_UPDATE_RETURNING:
 			return "SPI_OK_UPDATE_RETURNING";
+		case SPI_OK_MERGE_RETURNING:
+			return "SPI_OK_MERGE_RETURNING";
 		case SPI_OK_REWRITTEN:
 			return "SPI_OK_REWRITTEN";
 		case SPI_OK_REL_REGISTER:
@@ -2882,7 +2886,10 @@ _SPI_pquery(QueryDesc *queryDesc, bool f
 				res = SPI_OK_UPDATE;
 			break;
 		case CMD_MERGE:
-			res = SPI_OK_MERGE;
+			if (queryDesc->plannedstmt->hasReturning)
+				res = SPI_OK_MERGE_RETURNING;
+			else
+				res = SPI_OK_MERGE;
 			break;
 		default:
 			return SPI_ERROR_OPUNKNOWN;
diff --git a/src/backend/parser/analyze.c b/src/backend/parser/analyze.c
new file mode 100644
index e892df9..fe0f5cb
--- a/src/backend/parser/analyze.c
+++ b/src/backend/parser/analyze.c
@@ -73,7 +73,6 @@ static void determineRecursiveColTypes(P
 									   Node *larg, List *nrtargetlist);
 static Query *transformReturnStmt(ParseState *pstate, ReturnStmt *stmt);
 static Query *transformUpdateStmt(ParseState *pstate, UpdateStmt *stmt);
-static List *transformReturningList(ParseState *pstate, List *returningList);
 static Query *transformPLAssignStmt(ParseState *pstate,
 									PLAssignStmt *stmt);
 static Query *transformDeclareCursorStmt(ParseState *pstate,
@@ -514,7 +513,8 @@ transformDeleteStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -941,7 +941,8 @@ transformInsertStmt(ParseState *pstate,
 	/* Process RETURNING, if any. */
 	if (stmt->returningList)
 		qry->returningList = transformReturningList(pstate,
-													stmt->returningList);
+													stmt->returningList,
+													EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -2406,7 +2407,8 @@ transformUpdateStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/*
 	 * Now we are done with SELECT-like processing, and can get on with
@@ -2500,10 +2502,11 @@ transformUpdateTargetList(ParseState *ps
 
 /*
  * transformReturningList -
- *	handle a RETURNING clause in INSERT/UPDATE/DELETE
+ *	handle a RETURNING clause in INSERT/UPDATE/DELETE/MERGE
  */
-static List *
-transformReturningList(ParseState *pstate, List *returningList)
+List *
+transformReturningList(ParseState *pstate, List *returningList,
+					   ParseExprKind exprKind)
 {
 	List	   *rlist;
 	int			save_next_resno;
@@ -2520,7 +2523,7 @@ transformReturningList(ParseState *pstat
 	pstate->p_next_resno = 1;
 
 	/* transform RETURNING identically to a SELECT targetlist */
-	rlist = transformTargetList(pstate, returningList, EXPR_KIND_RETURNING);
+	rlist = transformTargetList(pstate, returningList, exprKind);
 
 	/*
 	 * Complain if the nonempty tlist expanded to nothing (which is possible
diff --git a/src/backend/parser/gram.y b/src/backend/parser/gram.y
new file mode 100644
index a013838..4406430
--- a/src/backend/parser/gram.y
+++ b/src/backend/parser/gram.y
@@ -12241,6 +12241,7 @@ MergeStmt:
 			USING table_ref
 			ON a_expr
 			merge_when_list
+			returning_clause
 				{
 					MergeStmt  *m = makeNode(MergeStmt);
 
@@ -12249,6 +12250,7 @@ MergeStmt:
 					m->sourceRelation = $6;
 					m->joinCondition = $8;
 					m->mergeWhenClauses = $9;
+					m->returningList = $10;
 
 					$$ = (Node *) m;
 				}
diff --git a/src/backend/parser/parse_agg.c b/src/backend/parser/parse_agg.c
new file mode 100644
index 4fbf80c..b8d655a
--- a/src/backend/parser/parse_agg.c
+++ b/src/backend/parser/parse_agg.c
@@ -456,6 +456,7 @@ check_agglevels_and_constraints(ParseSta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
@@ -904,6 +905,7 @@ transformWindowFuncCall(ParseState *psta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_cte.c b/src/backend/parser/parse_cte.c
new file mode 100644
index c5b1a49..ad3c525
--- a/src/backend/parser/parse_cte.c
+++ b/src/backend/parser/parse_cte.c
@@ -126,13 +126,6 @@ transformWithClause(ParseState *pstate,
 		CommonTableExpr *cte = (CommonTableExpr *) lfirst(lc);
 		ListCell   *rest;
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(cte->ctequery, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in WITH query"),
-					parser_errposition(pstate, cte->location));
-
 		for_each_cell(rest, withClause->ctes, lnext(withClause->ctes, lc))
 		{
 			CommonTableExpr *cte2 = (CommonTableExpr *) lfirst(rest);
@@ -153,7 +146,8 @@ transformWithClause(ParseState *pstate,
 			/* must be a data-modifying statement */
 			Assert(IsA(cte->ctequery, InsertStmt) ||
 				   IsA(cte->ctequery, UpdateStmt) ||
-				   IsA(cte->ctequery, DeleteStmt));
+				   IsA(cte->ctequery, DeleteStmt) ||
+				   IsA(cte->ctequery, MergeStmt));
 
 			pstate->p_hasModifyingCTE = true;
 		}
diff --git a/src/backend/parser/parse_expr.c b/src/backend/parser/parse_expr.c
new file mode 100644
index 7ff41ac..92cd8c1
--- a/src/backend/parser/parse_expr.c
+++ b/src/backend/parser/parse_expr.c
@@ -482,6 +482,7 @@ transformColumnRef(ParseState *pstate, C
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CHECK_CONSTRAINT:
@@ -1708,6 +1709,7 @@ transformSubLink(ParseState *pstate, Sub
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CYCLE_MARK:
@@ -3002,6 +3004,7 @@ ParseExprKindName(ParseExprKind exprKind
 		case EXPR_KIND_OFFSET:
 			return "OFFSET";
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			return "RETURNING";
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
diff --git a/src/backend/parser/parse_func.c b/src/backend/parser/parse_func.c
new file mode 100644
index ca14f06..914624e
--- a/src/backend/parser/parse_func.c
+++ b/src/backend/parser/parse_func.c
@@ -31,6 +31,7 @@
 #include "parser/parse_target.h"
 #include "parser/parse_type.h"
 #include "utils/builtins.h"
+#include "utils/fmgroids.h"
 #include "utils/lsyscache.h"
 #include "utils/syscache.h"
 
@@ -348,6 +349,15 @@ ParseFuncOrColumn(ParseState *pstate, Li
 					 parser_errposition(pstate, location)));
 	}
 
+	/* Merge support functions are only allowed in MERGE's RETURNING list */
+	if (IsMergeSupportFunction(funcid) &&
+		pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+		ereport(ERROR,
+				(errcode(ERRCODE_WRONG_OBJECT_TYPE),
+				 errmsg("merge support function %s can only be called from the RETURNING list of a MERGE command",
+						NameListToString(funcname)),
+				 parser_errposition(pstate, location)));
+
 	/*
 	 * So far so good, so do some fdresult-type-specific processing.
 	 */
@@ -2602,6 +2612,7 @@ check_srf_call_placement(ParseState *pst
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_merge.c b/src/backend/parser/parse_merge.c
new file mode 100644
index d886637..2e50eb0
--- a/src/backend/parser/parse_merge.c
+++ b/src/backend/parser/parse_merge.c
@@ -233,6 +233,10 @@ transformMergeStmt(ParseState *pstate, M
 	 */
 	qry->jointree = makeFromExpr(pstate->p_joinlist, joinExpr);
 
+	/* Transform the RETURNING list, if any */
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_MERGE_RETURNING);
+
 	/*
 	 * We now have a good query shape, so now look at the WHEN conditions and
 	 * action targetlists.
@@ -254,6 +258,7 @@ transformMergeStmt(ParseState *pstate, M
 		MergeAction *action;
 
 		action = makeNode(MergeAction);
+		action->index = foreach_current_index(l) + 1;
 		action->commandType = mergeWhenClause->commandType;
 		action->matched = mergeWhenClause->matched;
 
@@ -390,9 +395,6 @@ transformMergeStmt(ParseState *pstate, M
 
 	qry->mergeActionList = mergeActionList;
 
-	/* RETURNING could potentially be added in the future, but not in SQL std */
-	qry->returningList = NULL;
-
 	qry->hasTargetSRFs = false;
 	qry->hasSubLinks = pstate->p_hasSubLinks;
 
diff --git a/src/backend/parser/parse_relation.c b/src/backend/parser/parse_relation.c
new file mode 100644
index 52b4a6e..20e0c05
--- a/src/backend/parser/parse_relation.c
+++ b/src/backend/parser/parse_relation.c
@@ -2345,9 +2345,10 @@ addRangeTableEntryForCTE(ParseState *pst
 		cte->cterefcount++;
 
 	/*
-	 * We throw error if the CTE is INSERT/UPDATE/DELETE without RETURNING.
-	 * This won't get checked in case of a self-reference, but that's OK
-	 * because data-modifying CTEs aren't allowed to be recursive anyhow.
+	 * We throw error if the CTE is INSERT/UPDATE/DELETE/MERGE without
+	 * RETURNING.  This won't get checked in case of a self-reference, but
+	 * that's OK because data-modifying CTEs aren't allowed to be recursive
+	 * anyhow.
 	 */
 	if (IsA(cte->ctequery, Query))
 	{
diff --git a/src/backend/rewrite/rewriteHandler.c b/src/backend/rewrite/rewriteHandler.c
new file mode 100644
index c74bac2..939d61e
--- a/src/backend/rewrite/rewriteHandler.c
+++ b/src/backend/rewrite/rewriteHandler.c
@@ -3626,7 +3626,8 @@ RewriteQuery(Query *parsetree, List *rew
 			if (!(ctequery->commandType == CMD_SELECT ||
 				  ctequery->commandType == CMD_UPDATE ||
 				  ctequery->commandType == CMD_INSERT ||
-				  ctequery->commandType == CMD_DELETE))
+				  ctequery->commandType == CMD_DELETE ||
+				  ctequery->commandType == CMD_MERGE))
 			{
 				/*
 				 * Currently it could only be NOTIFY; this error message will
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
new file mode 100644
index c7d9d96..d393f51
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -2131,11 +2131,10 @@ QueryReturnsTuples(Query *parsetree)
 		case CMD_SELECT:
 			/* returns tuples */
 			return true;
-		case CMD_MERGE:
-			return false;
 		case CMD_INSERT:
 		case CMD_UPDATE:
 		case CMD_DELETE:
+		case CMD_MERGE:
 			/* the forms with RETURNING return tuples */
 			if (parsetree->returningList)
 				return true;
diff --git a/src/bin/psql/common.c b/src/bin/psql/common.c
new file mode 100644
index f907f5d..85eb868
--- a/src/bin/psql/common.c
+++ b/src/bin/psql/common.c
@@ -962,13 +962,17 @@ PrintQueryResult(PGresult *result, bool
 			else
 				success = true;
 
-			/* if it's INSERT/UPDATE/DELETE RETURNING, also print status */
+			/*
+			 * If it's INSERT/UPDATE/DELETE/MERGE RETURNING, also print
+			 * status.
+			 */
 			if (last || pset.show_all_results)
 			{
 				cmdstatus = PQcmdStatus(result);
 				if (strncmp(cmdstatus, "INSERT", 6) == 0 ||
 					strncmp(cmdstatus, "UPDATE", 6) == 0 ||
-					strncmp(cmdstatus, "DELETE", 6) == 0)
+					strncmp(cmdstatus, "DELETE", 6) == 0 ||
+					strncmp(cmdstatus, "MERGE", 5) == 0)
 					PrintQueryStatus(result, printStatusFout);
 			}
 
diff --git a/src/include/catalog/pg_proc.dat b/src/include/catalog/pg_proc.dat
new file mode 100644
index c0f2a8a..792060d
--- a/src/include/catalog/pg_proc.dat
+++ b/src/include/catalog/pg_proc.dat
@@ -11918,4 +11918,14 @@
   prorettype => 'bytea', proargtypes => 'pg_brin_minmax_multi_summary',
   prosrc => 'brin_minmax_multi_summary_send' },
 
+# MERGE support functions
+{ oid => '9499', descr => 'command type of current MERGE action',
+  proname => 'pg_merge_action',  provolatile => 'v',
+  prorettype => 'text', proargtypes => '',
+  prosrc => 'pg_merge_action' },
+{ oid => '9500', descr => 'index of current MERGE WHEN clause',
+  proname => 'pg_merge_when_clause',  provolatile => 'v',
+  prorettype => 'int4', proargtypes => '',
+  prosrc => 'pg_merge_when_clause' },
+
 ]
diff --git a/src/include/catalog/pg_proc.h b/src/include/catalog/pg_proc.h
new file mode 100644
index e7abe0b..e597826
--- a/src/include/catalog/pg_proc.h
+++ b/src/include/catalog/pg_proc.h
@@ -182,6 +182,11 @@ DECLARE_UNIQUE_INDEX(pg_proc_proname_arg
 #define PROARGMODE_VARIADIC 'v'
 #define PROARGMODE_TABLE	't'
 
+/* Is this a merge support function?  (Requires fmgroids.h) */
+#define IsMergeSupportFunction(oid) \
+	((oid) == F_PG_MERGE_ACTION || \
+	 (oid) == F_PG_MERGE_WHEN_CLAUSE)
+
 #endif							/* EXPOSE_TO_CLIENT_CODE */
 
 
diff --git a/src/include/executor/spi.h b/src/include/executor/spi.h
new file mode 100644
index d1de139..ed8787d
--- a/src/include/executor/spi.h
+++ b/src/include/executor/spi.h
@@ -97,6 +97,7 @@ typedef struct _SPI_plan *SPIPlanPtr;
 #define SPI_OK_REL_UNREGISTER	16
 #define SPI_OK_TD_REGISTER		17
 #define SPI_OK_MERGE			18
+#define SPI_OK_MERGE_RETURNING	19
 
 #define SPI_OPT_NONATOMIC		(1 << 0)
 
diff --git a/src/include/nodes/execnodes.h b/src/include/nodes/execnodes.h
new file mode 100644
index 20f4c8b..3d62fb9
--- a/src/include/nodes/execnodes.h
+++ b/src/include/nodes/execnodes.h
@@ -1302,6 +1302,9 @@ typedef struct ModifyTableState
 	/* Flags showing which subcommands are present INS/UPD/DEL/DO NOTHING */
 	int			mt_merge_subcommands;
 
+	/* For MERGE, the action currently being executed */
+	MergeActionState *mt_merge_action;
+
 	/* tuple counters for MERGE */
 	double		mt_merge_inserted;
 	double		mt_merge_updated;
diff --git a/src/include/nodes/parsenodes.h b/src/include/nodes/parsenodes.h
new file mode 100644
index 855da99..a35ab54
--- a/src/include/nodes/parsenodes.h
+++ b/src/include/nodes/parsenodes.h
@@ -1687,6 +1687,7 @@ typedef struct MergeWhenClause
 typedef struct MergeAction
 {
 	NodeTag		type;
+	int			index;			/* 1-based index of the clause */
 	bool		matched;		/* true=MATCHED, false=NOT MATCHED */
 	CmdType		commandType;	/* INSERT/UPDATE/DELETE/DO NOTHING */
 	/* OVERRIDING clause */
@@ -1801,6 +1802,7 @@ typedef struct MergeStmt
 	Node	   *sourceRelation; /* source relation */
 	Node	   *joinCondition;	/* join condition between source and target */
 	List	   *mergeWhenClauses;	/* list of MergeWhenClause(es) */
+	List	   *returningList;	/* list of expressions to return */
 	WithClause *withClause;		/* WITH clause */
 } MergeStmt;
 
diff --git a/src/include/parser/analyze.h b/src/include/parser/analyze.h
new file mode 100644
index 1cef183..78eb55e
--- a/src/include/parser/analyze.h
+++ b/src/include/parser/analyze.h
@@ -44,6 +44,8 @@ extern List *transformInsertRow(ParseSta
 								bool strip_indirection);
 extern List *transformUpdateTargetList(ParseState *pstate,
 									   List *origTlist);
+extern List *transformReturningList(ParseState *pstate, List *returningList,
+									ParseExprKind exprKind);
 extern Query *transformTopLevelStmt(ParseState *pstate, RawStmt *parseTree);
 extern Query *transformStmt(ParseState *pstate, Node *parseTree);
 
diff --git a/src/include/parser/parse_node.h b/src/include/parser/parse_node.h
new file mode 100644
index f589112..836819d
--- a/src/include/parser/parse_node.h
+++ b/src/include/parser/parse_node.h
@@ -61,7 +61,8 @@ typedef enum ParseExprKind
 	EXPR_KIND_DISTINCT_ON,		/* DISTINCT ON */
 	EXPR_KIND_LIMIT,			/* LIMIT */
 	EXPR_KIND_OFFSET,			/* OFFSET */
-	EXPR_KIND_RETURNING,		/* RETURNING */
+	EXPR_KIND_RETURNING,		/* RETURNING in INSERT/UPDATE/DELETE */
+	EXPR_KIND_MERGE_RETURNING,	/* RETURNING in MERGE */
 	EXPR_KIND_VALUES,			/* VALUES */
 	EXPR_KIND_VALUES_SINGLE,	/* single-row VALUES (in INSERT only) */
 	EXPR_KIND_CHECK_CONSTRAINT, /* CHECK constraint for a table */
diff --git a/src/pl/plpgsql/src/pl_exec.c b/src/pl/plpgsql/src/pl_exec.c
new file mode 100644
index 70a002a..143fb29
--- a/src/pl/plpgsql/src/pl_exec.c
+++ b/src/pl/plpgsql/src/pl_exec.c
@@ -4246,9 +4246,9 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 	/*
 	 * If we have INTO, then we only need one row back ... but if we have INTO
 	 * STRICT or extra check too_many_rows, ask for two rows, so that we can
-	 * verify the statement returns only one.  INSERT/UPDATE/DELETE are always
-	 * treated strictly. Without INTO, just run the statement to completion
-	 * (tcount = 0).
+	 * verify the statement returns only one.  INSERT/UPDATE/DELETE/MERGE are
+	 * always treated strictly. Without INTO, just run the statement to
+	 * completion (tcount = 0).
 	 *
 	 * We could just ask for two rows always when using INTO, but there are
 	 * some cases where demanding the extra row costs significant time, eg by
@@ -4286,10 +4286,11 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 			Assert(stmt->mod_stmt);
 			exec_set_found(estate, (SPI_processed != 0));
 			break;
@@ -4468,10 +4469,11 @@ exec_stmt_dynexecute(PLpgSQL_execstate *
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 		case SPI_OK_UTILITY:
 		case SPI_OK_REWRITTEN:
 			break;
diff --git a/src/pl/tcl/pltcl.c b/src/pl/tcl/pltcl.c
new file mode 100644
index 185d5be..58c4f50
--- a/src/pl/tcl/pltcl.c
+++ b/src/pl/tcl/pltcl.c
@@ -2441,6 +2441,7 @@ pltcl_process_SPI_result(Tcl_Interp *int
 		case SPI_OK_INSERT:
 		case SPI_OK_DELETE:
 		case SPI_OK_UPDATE:
+		case SPI_OK_MERGE:
 			Tcl_SetObjResult(interp, Tcl_NewWideIntObj(ntuples));
 			break;
 
@@ -2458,6 +2459,7 @@ pltcl_process_SPI_result(Tcl_Interp *int
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
+		case SPI_OK_MERGE_RETURNING:
 
 			/*
 			 * Process the tuples we got
diff --git a/src/test/regress/expected/merge.out b/src/test/regress/expected/merge.out
new file mode 100644
index bc53b21..b150726
--- a/src/test/regress/expected/merge.out
+++ b/src/test/regress/expected/merge.out
@@ -123,20 +123,20 @@ ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
 ERROR:  name "target" specified more than once
 DETAIL:  The name is used both as MERGE target table and data source.
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
-ERROR:  MERGE not supported in WITH query
-LINE 1: WITH foo AS (
-             ^
--- used in COPY
+ERROR:  WITH query "foo" does not have a RETURNING clause
+LINE 4: ) SELECT * FROM foo;
+                        ^
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) TO stdout;
-ERROR:  MERGE not supported in COPY
+ERROR:  COPY query must have a RETURNING clause
 -- unsupported relation types
 -- view
 CREATE VIEW tv AS SELECT * FROM target;
@@ -1289,20 +1289,187 @@ WHEN MATCHED AND tid < 2 THEN
 ROLLBACK;
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
-ERROR:  syntax error at or near "RETURNING"
-LINE 10: RETURNING *;
-         ^
+    DELETE
+RETURNING pg_merge_when_clause() AS when_clause,
+          pg_merge_action() AS merge_action,
+          t.*,
+          CASE pg_merge_action()
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ when_clause | merge_action | tid | balance |     description     
+-------------+--------------+-----+---------+---------------------
+           3 | DELETE       |   1 |     100 | Removed (1,100)
+           1 | UPDATE       |   2 |     220 | Added 20 to balance
+           2 | INSERT       |   4 |      40 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- error when using MERGE support functions outside MERGE
+SELECT pg_merge_action() FROM sq_target;
+ERROR:  merge support function pg_merge_action can only be called from the RETURNING list of a MERGE command
+LINE 1: SELECT pg_merge_action() FROM sq_target;
+               ^
+UPDATE sq_target SET balance = balance + 1 RETURNING pg_merge_when_clause();
+ERROR:  merge support function pg_merge_when_clause can only be called from the RETURNING list of a MERGE command
+LINE 1: ...ATE sq_target SET balance = balance + 1 RETURNING pg_merge_w...
+                                                             ^
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_when_clause() AS when_clause,
+              pg_merge_action() AS merge_action,
+              t.*,
+              CASE pg_merge_action()
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING merge_action, pg_merge_action() AS merge_log_action, l.*
+)
+SELECT * FROM m2;
+ merge_action | merge_log_action | tid |     last_change     
+--------------+------------------+-----+---------------------
+ DELETE       | UPDATE           |   1 | Removed (1,100)
+ UPDATE       | INSERT           |   2 | Added 20 to balance
+ INSERT       | INSERT           |   4 | Inserted (4,40)
+(3 rows)
+
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ tid |     last_change     
+-----+---------------------
+   1 | Removed (1,100)
+   2 | Added 20 to balance
+   4 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING v
+    ON tid = sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*
+) TO stdout;
+DELETE	1	100
+ROLLBACK;
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ action | tid | new_balance 
+--------+-----+-------------
+ DELETE |   1 |         100
+ UPDATE |   3 |         320
+ INSERT |   4 |         110
+(3 rows)
+
+ROLLBACK;
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ action | tid | balance 
+--------+-----+---------
+ DELETE |   1 |     100
+ INSERT |   4 |      40
+(2 rows)
+
+ROLLBACK;
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ r_action | r_tid | r_balance 
+----------+-------+-----------
+ DELETE   |     1 |       100
+ UPDATE   |     3 |       320
+ INSERT   |     4 |       110
+(3 rows)
+
 ROLLBACK;
 -- EXPLAIN
 CREATE TABLE ex_mtarget (a int, b int)
@@ -1499,7 +1666,7 @@ SELECT * FROM sq_target WHERE tid = 1;
 (1 row)
 
 ROLLBACK;
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 NOTICE:  drop cascades to view v
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1588,7 +1755,26 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (sid, delta, 'inserted by merge');
+    INSERT VALUES (sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
+ pg_merge_action | pg_merge_when_clause | tid | balance |           val            
+-----------------+----------------------+-----+---------+--------------------------
+ UPDATE          |                    1 |   2 |     110 | initial updated by merge
+ INSERT          |                    2 |   2 |      20 | inserted by merge
+ UPDATE          |                    1 |   4 |     330 | initial updated by merge
+ INSERT          |                    2 |   4 |      40 | inserted by merge
+ UPDATE          |                    1 |   6 |     550 | initial updated by merge
+ INSERT          |                    2 |   6 |      60 | inserted by merge
+ UPDATE          |                    1 |   8 |     770 | initial updated by merge
+ INSERT          |                    2 |   8 |      80 | inserted by merge
+ UPDATE          |                    1 |  10 |     990 | initial updated by merge
+ INSERT          |                    2 |  10 |     100 | inserted by merge
+ UPDATE          |                    1 |  12 |    1210 | initial updated by merge
+ INSERT          |                    2 |  12 |     120 | inserted by merge
+ UPDATE          |                    1 |  14 |    1430 | initial updated by merge
+ INSERT          |                    2 |  14 |     140 | inserted by merge
+(14 rows)
+
 SELECT * FROM pa_target ORDER BY tid;
  tid | balance |           val            
 -----+---------+--------------------------
@@ -1609,6 +1795,32 @@ SELECT * FROM pa_target ORDER BY tid;
 (14 rows)
 
 ROLLBACK;
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
+ pg_merge_action | pg_merge_when_clause | tid | balance |           val            
+-----------------+----------------------+-----+---------+--------------------------
+ UPDATE          |                    1 |   2 |     110 | initial updated by merge
+(1 row)
+
+SELECT * FROM pa_target ORDER BY tid;
+ tid | balance |           val            
+-----+---------+--------------------------
+   2 |     110 | initial updated by merge
+   3 |     300 | initial
+   5 |     500 | initial
+   7 |     700 | initial
+   9 |     900 | initial
+  11 |    1100 | initial
+  13 |    1300 | initial
+(7 rows)
+
+ROLLBACK;
 DROP TABLE pa_target CASCADE;
 -- The target table is partitioned in the same way, but this time by attaching
 -- partitions which have columns in different order, dropped columns etc.
@@ -1700,7 +1912,26 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (sid, delta, 'inserted by merge');
+    INSERT VALUES (sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
+ pg_merge_action | pg_merge_when_clause | tid | balance |           val            
+-----------------+----------------------+-----+---------+--------------------------
+ UPDATE          |                    1 |   2 |     110 | initial updated by merge
+ INSERT          |                    2 |   2 |      20 | inserted by merge
+ UPDATE          |                    1 |   4 |     330 | initial updated by merge
+ INSERT          |                    2 |   4 |      40 | inserted by merge
+ UPDATE          |                    1 |   6 |     550 | initial updated by merge
+ INSERT          |                    2 |   6 |      60 | inserted by merge
+ UPDATE          |                    1 |   8 |     770 | initial updated by merge
+ INSERT          |                    2 |   8 |      80 | inserted by merge
+ UPDATE          |                    1 |  10 |     990 | initial updated by merge
+ INSERT          |                    2 |  10 |     100 | inserted by merge
+ UPDATE          |                    1 |  12 |    1210 | initial updated by merge
+ INSERT          |                    2 |  12 |     120 | inserted by merge
+ UPDATE          |                    1 |  14 |    1430 | initial updated by merge
+ INSERT          |                    2 |  14 |     140 | inserted by merge
+(14 rows)
+
 SELECT * FROM pa_target ORDER BY tid;
  tid | balance |           val            
 -----+---------+--------------------------
@@ -1755,7 +1986,21 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
+ pg_merge_action | pg_merge_when_clause |          logts           | tid | balance |           val            
+-----------------+----------------------+--------------------------+-----+---------+--------------------------
+ UPDATE          |                    1 | Tue Jan 31 00:00:00 2017 |   1 |     110 | initial updated by merge
+ UPDATE          |                    1 | Tue Feb 28 00:00:00 2017 |   2 |     220 | initial updated by merge
+ INSERT          |                    2 | Sun Jan 15 00:00:00 2017 |   3 |      30 | inserted by merge
+ UPDATE          |                    1 | Tue Jan 31 00:00:00 2017 |   4 |     440 | initial updated by merge
+ UPDATE          |                    1 | Tue Feb 28 00:00:00 2017 |   5 |     550 | initial updated by merge
+ INSERT          |                    2 | Sun Jan 15 00:00:00 2017 |   6 |      60 | inserted by merge
+ UPDATE          |                    1 | Tue Jan 31 00:00:00 2017 |   7 |     770 | initial updated by merge
+ UPDATE          |                    1 | Tue Feb 28 00:00:00 2017 |   8 |     880 | initial updated by merge
+ INSERT          |                    2 | Sun Jan 15 00:00:00 2017 |   9 |      90 | inserted by merge
+(9 rows)
+
 SELECT * FROM pa_target ORDER BY tid;
           logts           | tid | balance |           val            
 --------------------------+-----+---------+--------------------------
diff --git a/src/test/regress/sql/merge.sql b/src/test/regress/sql/merge.sql
new file mode 100644
index fdbcd70..3bffb29
--- a/src/test/regress/sql/merge.sql
+++ b/src/test/regress/sql/merge.sql
@@ -86,12 +86,12 @@ MERGE INTO target
 USING target
 ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
--- used in COPY
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
@@ -844,17 +844,145 @@ ROLLBACK;
 
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
+    DELETE
+RETURNING pg_merge_when_clause() AS when_clause,
+          pg_merge_action() AS merge_action,
+          t.*,
+          CASE pg_merge_action()
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ROLLBACK;
+
+-- error when using MERGE support functions outside MERGE
+SELECT pg_merge_action() FROM sq_target;
+UPDATE sq_target SET balance = balance + 1 RETURNING pg_merge_when_clause();
+
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_when_clause() AS when_clause,
+              pg_merge_action() AS merge_action,
+              t.*,
+              CASE pg_merge_action()
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING merge_action, pg_merge_action() AS merge_log_action, l.*
+)
+SELECT * FROM m2;
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ROLLBACK;
+
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING v
+    ON tid = sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*
+) TO stdout;
+ROLLBACK;
+
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ROLLBACK;
+
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ROLLBACK;
+
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
 ROLLBACK;
 
 -- EXPLAIN
@@ -953,7 +1081,7 @@ WHEN MATCHED THEN
 SELECT * FROM sq_target WHERE tid = 1;
 ROLLBACK;
 
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1005,7 +1133,19 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (sid, delta, 'inserted by merge');
+    INSERT VALUES (sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
+SELECT * FROM pa_target ORDER BY tid;
+ROLLBACK;
+
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
@@ -1067,7 +1207,8 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (sid, delta, 'inserted by merge');
+    INSERT VALUES (sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
@@ -1109,7 +1250,8 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 

diff --git a/doc/src/sgml/dml.sgml b/doc/src/sgml/dml.sgml
new file mode 100644
index cbbc5e2..ff2a827
--- a/doc/src/sgml/dml.sgml
+++ b/doc/src/sgml/dml.sgml
@@ -283,10 +283,15 @@ DELETE FROM products;
    <secondary>RETURNING</secondary>
   </indexterm>
 
+  <indexterm zone="dml-returning">
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
   <para>
    Sometimes it is useful to obtain data from modified rows while they are
    being manipulated.  The <command>INSERT</command>, <command>UPDATE</command>,
-   and <command>DELETE</command> commands all have an
+   <command>DELETE</command>, and <command>MERGE</command> commands all have an
    optional <literal>RETURNING</literal> clause that supports this.  Use
    of <literal>RETURNING</literal> avoids performing an extra database query to
    collect the data, and is especially valuable when it would otherwise be
@@ -339,6 +344,24 @@ DELETE FROM products
 </programlisting>
   </para>
 
+  <para>
+   In a <command>MERGE</command>, the data available to <literal>RETURNING</literal> is
+   the content of the source row plus the content of the inserted, updated, or
+   deleted target row.  In addition, the merge support functions
+   <xref linkend="pg-merge-action"/> and <xref linkend="pg-merge-when-clause"/>
+   may be used to return information about which merge action was executed for
+   each row.  Since it is quite common for the source and target to have many
+   of the same columns, specifying <literal>RETURNING *</literal> can lead to a
+   lot of duplicated columns, so it is often more useful to just return the
+   target row.  For example:
+<programlisting>
+MERGE INTO products p USING new_products n ON p.product_no = n.product_no
+  WHEN NOT MATCHED THEN INSERT VALUES (n.product_no, n.name, n.price)
+  WHEN MATCHED THEN UPDATE SET name = n.name, price = n.price
+  RETURNING pg_merge_action(), p.*;
+</programlisting>
+  </para>
+
   <para>
    If there are triggers (<xref linkend="triggers"/>) on the target table,
    the data available to <literal>RETURNING</literal> is the row as modified by
diff --git a/doc/src/sgml/func.sgml b/doc/src/sgml/func.sgml
new file mode 100644
index 0cbdf63..224e9b8
--- a/doc/src/sgml/func.sgml
+++ b/doc/src/sgml/func.sgml
@@ -21216,6 +21216,99 @@ SELECT count(*) FROM sometable;
 
  </sect1>
 
+ <sect1 id="functions-merge-support">
+  <title>Merge Support Functions</title>
+
+  <indexterm>
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
+  <para>
+   The merge support functions shown in
+   <xref linkend="functions-merge-support-table"/> may be used in the
+   <literal>RETURNING</literal> list of a <xref linkend="sql-merge"/> command
+   to return additional information about the action taken for each row.
+  </para>
+
+  <table id="functions-merge-support-table">
+    <title>Merge Support Functions</title>
+
+    <tgroup cols="1">
+     <thead>
+      <row>
+       <entry role="func_table_entry"><para role="func_signature">
+        Function
+       </para>
+       <para>
+        Description
+       </para></entry>
+      </row>
+     </thead>
+
+     <tbody>
+      <row>
+       <entry role="func_table_entry"><para role="func_signature">
+        <indexterm id="pg-merge-action">
+         <primary>pg_merge_action</primary>
+        </indexterm>
+        <function>pg_merge_action</function> ( )
+        <returnvalue>text</returnvalue>
+       </para>
+       <para>
+        Returns the action performed on the current row (<literal>'INSERT'</literal>,
+        <literal>'UPDATE'</literal>, or <literal>'DELETE'</literal>).
+       </para></entry>
+      </row>
+
+      <row>
+       <entry role="func_table_entry"><para role="func_signature">
+        <indexterm id="pg-merge-when-clause">
+         <primary>pg_merge_when_clause</primary>
+        </indexterm>
+        <function>pg_merge_when_clause</function> ( )
+        <returnvalue>integer</returnvalue>
+       </para>
+       <para>
+        Returns the 1-based index of the <literal>WHEN</literal> clause executed
+        for the current row.
+       </para></entry>
+      </row>
+     </tbody>
+    </tgroup>
+  </table>
+
+  <para>
+   Example:
+<screen><![CDATA[
+MERGE INTO products p
+  USING stock s ON p.product_id = s.product_id
+  WHEN MATCHED AND s.quantity > 0 THEN
+    UPDATE SET in_stock = true, quantity = s.quantity
+  WHEN MATCHED THEN
+    UPDATE SET in_stock = false, quantity = 0
+  WHEN NOT MATCHED THEN
+    INSERT (product_id, in_stock, quantity)
+      VALUES (s.product_id, true, s.quantity)
+  RETURNING pg_merge_action() AS action,
+            pg_merge_when_clause() AS when_clause,
+            p.*;
+
+ action | when_clause | product_id | in_stock | quantity 
+--------+-------------+------------+----------+----------
+ UPDATE |           1 |       1001 | t        |       50
+ UPDATE |           2 |       1002 | f        |        0
+ INSERT |           3 |       1003 | t        |       10
+]]></screen>
+  </para>
+
+  <para>
+   Note that it is an error to use these functions anywhere other than the
+   <literal>RETURNING</literal> list of a <command>MERGE</command> command.
+  </para>
+
+ </sect1>
+
  <sect1 id="functions-subquery">
   <title>Subquery Expressions</title>
 
diff --git a/doc/src/sgml/glossary.sgml b/doc/src/sgml/glossary.sgml
new file mode 100644
index 7c01a54..4317cfc
--- a/doc/src/sgml/glossary.sgml
+++ b/doc/src/sgml/glossary.sgml
@@ -1323,9 +1323,9 @@
      to a <glossterm linkend="glossary-client">client</glossterm> upon the
      completion of an <acronym>SQL</acronym> command, usually a
      <command>SELECT</command> but it can be an
-     <command>INSERT</command>, <command>UPDATE</command>, or
-     <command>DELETE</command> command if the <literal>RETURNING</literal>
-     clause is specified.
+     <command>INSERT</command>, <command>UPDATE</command>,
+     <command>DELETE</command>, or <command>MERGE</command> command if the
+     <literal>RETURNING</literal> clause is specified.
     </para>
     <para>
      The fact that a result set is a relation means that a query can be used
diff --git a/doc/src/sgml/plpgsql.sgml b/doc/src/sgml/plpgsql.sgml
new file mode 100644
index 8897a54..3249e9c
--- a/doc/src/sgml/plpgsql.sgml
+++ b/doc/src/sgml/plpgsql.sgml
@@ -1020,8 +1020,8 @@ INSERT INTO mytable VALUES (1,'one'), (2
     </para>
 
     <para>
-     If the command does return rows (for example <command>SELECT</command>,
-     or <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>
+     If the command does return rows (for example <command>SELECT</command>, or
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
      with <literal>RETURNING</literal>), there are two ways to proceed.
      When the command will return at most one row, or you only care about
      the first row of output, write the command as usual but add
@@ -1044,7 +1044,8 @@ INSERT INTO mytable VALUES (1,'one'), (2
      <application>PL/pgSQL</application> variable values can be
      automatically inserted into optimizable SQL commands, which
      are <command>SELECT</command>, <command>INSERT</command>,
-     <command>UPDATE</command>, <command>DELETE</command>, and certain
+     <command>UPDATE</command>, <command>DELETE</command>,
+     <command>MERGE</command> and certain
      utility commands that incorporate one of these, such
      as <command>EXPLAIN</command> and <command>CREATE TABLE ... AS
      SELECT</command>.  In these commands,
@@ -1148,6 +1149,7 @@ SELECT <replaceable>select_expressions</
 INSERT ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 UPDATE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 DELETE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
+MERGE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 </synopsis>
 
      where <replaceable>target</replaceable> can be a record variable, a row
@@ -1158,8 +1160,8 @@ DELETE ... RETURNING <replaceable>expres
      <literal>INTO</literal> clause) just as described above,
      and the plan is cached in the same way.
      This works for <command>SELECT</command>,
-     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
-     <literal>RETURNING</literal>, and certain utility commands
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
+     with <literal>RETURNING</literal>, and certain utility commands
      that return row sets, such as <command>EXPLAIN</command>.
      Except for the <literal>INTO</literal> clause, the SQL command is the same
      as it would be written outside <application>PL/pgSQL</application>.
@@ -1235,7 +1237,7 @@ END;
     </para>
 
     <para>
-     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
+     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command> with
      <literal>RETURNING</literal>, <application>PL/pgSQL</application> reports
      an error for more than one returned row, even when
      <literal>STRICT</literal> is not specified.  This is because there
diff --git a/doc/src/sgml/queries.sgml b/doc/src/sgml/queries.sgml
new file mode 100644
index 45741e7..0ca593b
--- a/doc/src/sgml/queries.sgml
+++ b/doc/src/sgml/queries.sgml
@@ -2062,10 +2062,11 @@ SELECT <replaceable>select_list</replace
    Table Expressions or <acronym>CTE</acronym>s, can be thought of as defining
    temporary tables that exist just for one query.  Each auxiliary statement
    in a <literal>WITH</literal> clause can be a <command>SELECT</command>,
-   <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>; and the
+   <command>INSERT</command>, <command>UPDATE</command>, <command>DELETE</command>,
+   or <command>MERGE</command>; and the
    <literal>WITH</literal> clause itself is attached to a primary statement that can
-   also be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>, or
-   <command>DELETE</command>.
+   also be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
+   <command>DELETE</command>, or <command>MERGE</command>.
   </para>
 
  <sect2 id="queries-with-select">
@@ -2587,7 +2588,8 @@ SELECT * FROM w AS w1 JOIN w AS w2 ON w1
   <para>
    The examples above only show <literal>WITH</literal> being used with
    <command>SELECT</command>, but it can be attached in the same way to
-   <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>.
+   <command>INSERT</command>, <command>UPDATE</command>,
+   <command>DELETE</command> or <command>MERGE</command>.
    In each case it effectively provides temporary table(s) that can
    be referred to in the main command.
   </para>
@@ -2598,7 +2600,8 @@ SELECT * FROM w AS w1 JOIN w AS w2 ON w1
 
    <para>
     You can use data-modifying statements (<command>INSERT</command>,
-    <command>UPDATE</command>, or <command>DELETE</command>) in <literal>WITH</literal>.  This
+    <command>UPDATE</command>, <command>DELETE</command>, or
+    <command>MERGE</command>) in <literal>WITH</literal>.  This
     allows you to perform several different operations in the same query.
     An example is:
 
diff --git a/doc/src/sgml/ref/copy.sgml b/doc/src/sgml/ref/copy.sgml
new file mode 100644
index c25b52d..d822470
--- a/doc/src/sgml/ref/copy.sgml
+++ b/doc/src/sgml/ref/copy.sgml
@@ -121,13 +121,15 @@ COPY { <replaceable class="parameter">ta
       A <link linkend="sql-select"><command>SELECT</command></link>,
       <link linkend="sql-values"><command>VALUES</command></link>,
       <link linkend="sql-insert"><command>INSERT</command></link>,
-      <link linkend="sql-update"><command>UPDATE</command></link>, or
-      <link linkend="sql-delete"><command>DELETE</command></link> command whose results are to be
+      <link linkend="sql-update"><command>UPDATE</command></link>,
+      <link linkend="sql-update"><command>DELETE</command></link>, or
+      <link linkend="sql-delete"><command>MERGE</command></link> command whose results are to be
       copied.  Note that parentheses are required around the query.
      </para>
      <para>
-      For <command>INSERT</command>, <command>UPDATE</command> and
-      <command>DELETE</command> queries a RETURNING clause must be provided,
+      For <command>INSERT</command>, <command>UPDATE</command>,
+      <command>DELETE</command>, and <command>MERGE</command> queries a
+      RETURNING clause must be provided,
       and the target relation must not have a conditional rule, nor
       an <literal>ALSO</literal> rule, nor an <literal>INSTEAD</literal> rule
       that expands to multiple statements.
diff --git a/doc/src/sgml/ref/merge.sgml b/doc/src/sgml/ref/merge.sgml
new file mode 100644
index 0995fe0..7eb641a
--- a/doc/src/sgml/ref/merge.sgml
+++ b/doc/src/sgml/ref/merge.sgml
@@ -25,6 +25,7 @@ PostgreSQL documentation
 MERGE INTO [ ONLY ] <replaceable class="parameter">target_table_name</replaceable> [ * ] [ [ AS ] <replaceable class="parameter">target_alias</replaceable> ]
 USING <replaceable class="parameter">data_source</replaceable> ON <replaceable class="parameter">join_condition</replaceable>
 <replaceable class="parameter">when_clause</replaceable> [...]
+[ RETURNING * | <replaceable class="parameter">output_expression</replaceable> [ [ AS ] <replaceable class="parameter">output_name</replaceable> ] [, ...] ]
 
 <phrase>where <replaceable class="parameter">data_source</replaceable> is:</phrase>
 
@@ -95,6 +96,20 @@ DELETE
   </para>
 
   <para>
+   The optional <literal>RETURNING</literal> clause causes <command>MERGE</command>
+   to compute and return value(s) based on each row inserted, updated, or
+   deleted.  Any expression using the source or target table's columns, or the
+   merge support functions <xref linkend="pg-merge-action"/> and
+   <xref linkend="pg-merge-when-clause"/> can be computed.  When an
+   <command>INSERT</command> or <command>UPDATE</command> action is performed,
+   the new values of the target table's columns are used.  When a
+   <command>DELETE</command> is performed, the old values of the target table's
+   columns are used.  The syntax of the <literal>RETURNING</literal> list is
+   identical to that of the output list of <command>SELECT</command>.
+
+  </para>
+
+  <para>
    There is no separate <literal>MERGE</literal> privilege.
    If you specify an update action, you must have the
    <literal>UPDATE</literal> privilege on the column(s)
@@ -125,6 +140,18 @@ DELETE
 
   <variablelist>
    <varlistentry>
+    <term><replaceable class="parameter">with_query</replaceable></term>
+    <listitem>
+     <para>
+      The <literal>WITH</literal> clause allows you to specify one or more
+      subqueries that can be referenced by name in the <command>MERGE</command>
+      query. See <xref linkend="queries-with"/> and <xref linkend="sql-select"/>
+      for details.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
     <term><replaceable class="parameter">target_table_name</replaceable></term>
     <listitem>
      <para>
@@ -399,13 +426,30 @@ DELETE
    </varlistentry>
 
    <varlistentry>
-    <term><replaceable class="parameter">with_query</replaceable></term>
+    <term><replaceable class="parameter">output_expression</replaceable></term>
     <listitem>
      <para>
-      The <literal>WITH</literal> clause allows you to specify one or more
-      subqueries that can be referenced by name in the <command>MERGE</command>
-      query. See <xref linkend="queries-with"/> and <xref linkend="sql-select"/>
-      for details.
+      An expression to be computed and returned by the <command>MERGE</command>
+      command after each row is merged.  The expression can use any columns of
+      the source or target tables, or the merge support functions
+      <xref linkend="pg-merge-action"/> and <xref linkend="pg-merge-when-clause"/>.
+     </para>
+     <para>
+      Writing <literal>*</literal> will return all columns from the source
+      table, followed by all columns from the target table.  Often this will
+      lead to a lot of duplication, since it is common for the source and
+      target tables to have a lot of the same columns.  This can be avoided by
+      qualifying the <literal>*</literal> with the name or alias of the source
+      or target table.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
+    <term><replaceable class="parameter">output_name</replaceable></term>
+    <listitem>
+     <para>
+      A name to use for a returned column.
      </para>
     </listitem>
    </varlistentry>
@@ -428,6 +472,13 @@ MERGE <replaceable class="parameter">tot
    were changed in any way.
   </para>
 
+  <para>
+   If the <command>MERGE</command> command contains a <literal>RETURNING</literal>
+   clause, the result will be similar to that of a <command>SELECT</command>
+   statement containing the columns and values defined in the
+   <literal>RETURNING</literal> list, computed over the row(s) inserted, updated,
+   or deleted by the command.
+  </para>
  </refsect1>
 
  <refsect1>
@@ -544,13 +595,6 @@ MERGE <replaceable class="parameter">tot
   </para>
 
   <para>
-   There is no <literal>RETURNING</literal> clause with
-   <command>MERGE</command>.  Actions of <command>INSERT</command>,
-   <command>UPDATE</command> and <command>DELETE</command> cannot contain
-   <literal>RETURNING</literal> or <literal>WITH</literal> clauses.
-  </para>
-
-  <para>
    When <command>MERGE</command> is run concurrently with other commands
    that modify the target table, the usual transaction isolation rules
    apply; see <xref linkend="transaction-iso"/> for an explanation
@@ -602,7 +646,8 @@ WHEN NOT MATCHED THEN
   <para>
    Attempt to insert a new stock item along with the quantity of stock. If
    the item already exists, instead update the stock count of the existing
-   item. Don't allow entries that have zero stock.
+   item. Don't allow entries that have zero stock. Return details of all
+   changes made.
 <programlisting>
 MERGE INTO wines w
 USING wine_stock_changes s
@@ -612,7 +657,8 @@ WHEN NOT MATCHED AND s.stock_delta > 0 T
 WHEN MATCHED AND w.stock + s.stock_delta > 0 THEN
   UPDATE SET stock = w.stock + s.stock_delta
 WHEN MATCHED THEN
-  DELETE;
+  DELETE
+RETURNING pg_merge_action(), w.*;
 </programlisting>
 
    The <literal>wine_stock_changes</literal> table might be, for example, a
@@ -627,7 +673,8 @@ WHEN MATCHED THEN
     This command conforms to the <acronym>SQL</acronym> standard.
   </para>
    <para>
-    The WITH clause and <literal>DO NOTHING</literal> action are extensions to
+    The WITH clause, <literal>DO NOTHING</literal> action, and
+    <literal>RETURNING</literal> clause are extensions to
     the <acronym>SQL</acronym> standard.
   </para>
  </refsect1>
diff --git a/doc/src/sgml/ref/select.sgml b/doc/src/sgml/ref/select.sgml
new file mode 100644
index 1f9538f..5840ddc
--- a/doc/src/sgml/ref/select.sgml
+++ b/doc/src/sgml/ref/select.sgml
@@ -74,7 +74,7 @@ SELECT [ ALL | DISTINCT [ ON ( <replacea
 
 <phrase>and <replaceable class="parameter">with_query</replaceable> is:</phrase>
 
-    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> )
+    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> | <replaceable class="parameter">merge</replaceable> )
         [ SEARCH { BREADTH | DEPTH } FIRST BY <replaceable>column_name</replaceable> [, ...] SET <replaceable>search_seq_col_name</replaceable> ]
         [ CYCLE <replaceable>column_name</replaceable> [, ...] SET <replaceable>cycle_mark_col_name</replaceable> [ TO <replaceable>cycle_mark_value</replaceable> DEFAULT <replaceable>cycle_mark_default</replaceable> ] USING <replaceable>cycle_path_col_name</replaceable> ]
 
@@ -227,10 +227,10 @@ TABLE [ ONLY ] <replaceable class="param
     The subqueries effectively act as temporary tables or views
     for the duration of the primary query.
     Each subquery can be a <command>SELECT</command>, <command>TABLE</command>, <command>VALUES</command>,
-    <command>INSERT</command>, <command>UPDATE</command> or
-    <command>DELETE</command> statement.
+    <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command> statement.
     When writing a data-modifying statement (<command>INSERT</command>,
-    <command>UPDATE</command> or <command>DELETE</command>) in
+    <command>UPDATE</command>, <command>DELETE</command>, or <command>MERGE</command>) in
     <literal>WITH</literal>, it is usual to include a <literal>RETURNING</literal> clause.
     It is the output of <literal>RETURNING</literal>, <emphasis>not</emphasis> the underlying
     table that the statement modifies, that forms the temporary table that is
@@ -2179,7 +2179,8 @@ SELECT 2+2;
 
    <para>
     <productname>PostgreSQL</productname> allows <command>INSERT</command>,
-    <command>UPDATE</command>, and <command>DELETE</command> to be used as <literal>WITH</literal>
+    <command>UPDATE</command>, <command>DELETE</command>, and
+    <command>MERGE</command> to be used as <literal>WITH</literal>
     queries.  This is not found in the SQL standard.
    </para>
   </refsect2>
diff --git a/doc/src/sgml/rowtypes.sgml b/doc/src/sgml/rowtypes.sgml
new file mode 100644
index 4d86f97..bbeac84
--- a/doc/src/sgml/rowtypes.sgml
+++ b/doc/src/sgml/rowtypes.sgml
@@ -348,7 +348,7 @@ SELECT m.* FROM some_table, LATERAL myfu
    column expansion of this kind when it appears at the top level of
    a <link linkend="queries-select-lists"><command>SELECT</command> output
    list</link>, a <link linkend="dml-returning"><literal>RETURNING</literal>
-   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>,
+   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>,
    a <link linkend="queries-values"><literal>VALUES</literal> clause</link>, or
    a <link linkend="sql-syntax-row-constructors">row constructor</link>.
    In all other contexts (including when nested inside one of those
diff --git a/doc/src/sgml/spi.sgml b/doc/src/sgml/spi.sgml
new file mode 100644
index 651930a..ad0a9b1
--- a/doc/src/sgml/spi.sgml
+++ b/doc/src/sgml/spi.sgml
@@ -301,8 +301,9 @@ SPI_execute("INSERT INTO foo SELECT * FR
    is returned in the global variable <varname>SPI_processed</varname>.
    If the return value of the function is <symbol>SPI_OK_SELECT</symbol>,
    <symbol>SPI_OK_INSERT_RETURNING</symbol>,
-   <symbol>SPI_OK_DELETE_RETURNING</symbol>, or
-   <symbol>SPI_OK_UPDATE_RETURNING</symbol>,
+   <symbol>SPI_OK_DELETE_RETURNING</symbol>,
+   <symbol>SPI_OK_UPDATE_RETURNING</symbol>, or
+   <symbol>SPI_OK_MERGE_RETURNING</symbol>
    then you can use the
    global pointer <literal>SPITupleTable *SPI_tuptable</literal> to
    access the result rows.  Some utility commands (such as
@@ -472,6 +473,15 @@ typedef struct SPITupleTable
       </para>
      </listitem>
     </varlistentry>
+
+    <varlistentry>
+     <term><symbol>SPI_OK_MERGE_RETURNING</symbol></term>
+     <listitem>
+      <para>
+       if a <command>MERGE RETURNING</command> was executed
+      </para>
+     </listitem>
+    </varlistentry>
 
     <varlistentry>
      <term><symbol>SPI_OK_UTILITY</symbol></term>
diff --git a/doc/src/sgml/xfunc.sgml b/doc/src/sgml/xfunc.sgml
new file mode 100644
index e2a5496..114c3d3
--- a/doc/src/sgml/xfunc.sgml
+++ b/doc/src/sgml/xfunc.sgml
@@ -177,7 +177,8 @@
     statements separated by semicolons.  A semicolon after the last
     statement is optional.  Unless the function is declared to return
     <type>void</type>, the last statement must be a <command>SELECT</command>,
-    or an <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>
+    or an <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command>
     that has a <literal>RETURNING</literal> clause.
    </para>
 
@@ -186,7 +187,8 @@
      language can be packaged together and defined as a function.
      Besides <command>SELECT</command> queries, the commands can include data
      modification queries (<command>INSERT</command>,
-     <command>UPDATE</command>, and <command>DELETE</command>), as well as
+     <command>UPDATE</command>, <command>DELETE</command> and
+     <command>MERGE</command>), as well as
      other SQL commands. (You cannot use transaction control commands, e.g.,
      <command>COMMIT</command>, <command>SAVEPOINT</command>, and some utility
      commands, e.g.,  <literal>VACUUM</literal>, in <acronym>SQL</acronym> functions.)
@@ -1235,8 +1237,9 @@ SELECT x, CASE WHEN x &gt; 0 THEN genera
 
     <note>
      <para>
-      If a function's last command is <command>INSERT</command>, <command>UPDATE</command>,
-      or <command>DELETE</command> with <literal>RETURNING</literal>, that command will
+      If a function's last command is <command>INSERT</command>,
+      <command>UPDATE</command>, <command>DELETE</command>, or
+      <command>MERGE</command> with <literal>RETURNING</literal>, that command will
       always be executed to completion, even if the function is not declared
       with <literal>SETOF</literal> or the calling query does not fetch all the
       result rows.  Any extra rows produced by the <literal>RETURNING</literal>
diff --git a/src/backend/commands/copy.c b/src/backend/commands/copy.c
new file mode 100644
index e34f583..aa3cca0
--- a/src/backend/commands/copy.c
+++ b/src/backend/commands/copy.c
@@ -274,12 +274,6 @@ DoCopy(ParseState *pstate, const CopyStm
 	{
 		Assert(stmt->query);
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(stmt->query, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in COPY"));
-
 		query = makeNode(RawStmt);
 		query->stmt = stmt->query;
 		query->stmt_location = stmt_location;
diff --git a/src/backend/commands/copyto.c b/src/backend/commands/copyto.c
new file mode 100644
index 8043b4e..e02d7d0
--- a/src/backend/commands/copyto.c
+++ b/src/backend/commands/copyto.c
@@ -510,7 +510,8 @@ BeginCopyTo(ParseState *pstate,
 		{
 			Assert(query->commandType == CMD_INSERT ||
 				   query->commandType == CMD_UPDATE ||
-				   query->commandType == CMD_DELETE);
+				   query->commandType == CMD_DELETE ||
+				   query->commandType == CMD_MERGE);
 
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
diff --git a/src/backend/executor/execExpr.c b/src/backend/executor/execExpr.c
new file mode 100644
index 812ead9..8572b01
--- a/src/backend/executor/execExpr.c
+++ b/src/backend/executor/execExpr.c
@@ -48,6 +48,7 @@
 #include "utils/array.h"
 #include "utils/builtins.h"
 #include "utils/datum.h"
+#include "utils/fmgroids.h"
 #include "utils/lsyscache.h"
 #include "utils/typcache.h"
 
@@ -2485,6 +2486,22 @@ ExecInitFunc(ExprEvalStep *scratch, Expr
 	InitFunctionCallInfoData(*fcinfo, flinfo,
 							 nargs, inputcollid, NULL, NULL);
 
+	/*
+	 * Merge support functions should only be called directly from a MERGE
+	 * command, and need access to the parent ModifyTableState.  The parser
+	 * should have checked that such functions only appear in the RETURNING
+	 * list of a MERGE, so this should never fail.
+	 */
+	if (IsMergeSupportFunction(funcid))
+	{
+		if (!state->parent ||
+			!IsA(state->parent, ModifyTableState) ||
+			((ModifyTableState *) state->parent)->operation != CMD_MERGE)
+			elog(ERROR, "merge support function called in non-merge context");
+
+		fcinfo->context = (Node *) state->parent;
+	}
+
 	/* Keep extra copies of this info to save an indirection at runtime */
 	scratch->d.func.fn_addr = flinfo->fn_addr;
 	scratch->d.func.nargs = nargs;
diff --git a/src/backend/executor/execPartition.c b/src/backend/executor/execPartition.c
new file mode 100644
index fd6ca8a..95bbf67
--- a/src/backend/executor/execPartition.c
+++ b/src/backend/executor/execPartition.c
@@ -612,8 +612,8 @@ ExecInitPartitionInfo(ModifyTableState *
 	 * Build the RETURNING projection for the partition.  Note that we didn't
 	 * build the returningList for partitions within the planner, but simple
 	 * translation of varattnos will suffice.  This only occurs for the INSERT
-	 * case or in the case of UPDATE tuple routing where we didn't find a
-	 * result rel to reuse.
+	 * case or in the case of UPDATE/MERGE tuple routing where we didn't find
+	 * a result rel to reuse.
 	 */
 	if (node && node->returningLists != NIL)
 	{
@@ -622,12 +622,14 @@ ExecInitPartitionInfo(ModifyTableState *
 		List	   *returningList;
 
 		/* See the comment above for WCO lists. */
-		/* (except no RETURNING support for MERGE yet) */
 		Assert((node->operation == CMD_INSERT &&
 				list_length(node->returningLists) == 1 &&
 				list_length(node->resultRelations) == 1) ||
 			   (node->operation == CMD_UPDATE &&
 				list_length(node->returningLists) ==
+				list_length(node->resultRelations)) ||
+			   (node->operation == CMD_MERGE &&
+				list_length(node->returningLists) ==
 				list_length(node->resultRelations)));
 
 		/*
diff --git a/src/backend/executor/functions.c b/src/backend/executor/functions.c
new file mode 100644
index 50e06ec..7e4717a
--- a/src/backend/executor/functions.c
+++ b/src/backend/executor/functions.c
@@ -1665,8 +1665,8 @@ check_sql_fn_retval(List *queryTreeLists
 
 	/*
 	 * If it's a plain SELECT, it returns whatever the targetlist says.
-	 * Otherwise, if it's INSERT/UPDATE/DELETE with RETURNING, it returns
-	 * that. Otherwise, the function return type must be VOID.
+	 * Otherwise, if it's INSERT/UPDATE/DELETE/MERGE with RETURNING, it
+	 * returns that. Otherwise, the function return type must be VOID.
 	 *
 	 * Note: eventually replace this test with QueryReturnsTuples?	We'd need
 	 * a more general method of determining the output type, though.  Also, it
@@ -1684,7 +1684,8 @@ check_sql_fn_retval(List *queryTreeLists
 	else if (parse &&
 			 (parse->commandType == CMD_INSERT ||
 			  parse->commandType == CMD_UPDATE ||
-			  parse->commandType == CMD_DELETE) &&
+			  parse->commandType == CMD_DELETE ||
+			  parse->commandType == CMD_MERGE) &&
 			 parse->returningList)
 	{
 		tlist = parse->returningList;
diff --git a/src/backend/executor/nodeModifyTable.c b/src/backend/executor/nodeModifyTable.c
new file mode 100644
index 6f0543a..7e98890
--- a/src/backend/executor/nodeModifyTable.c
+++ b/src/backend/executor/nodeModifyTable.c
@@ -36,8 +36,7 @@
  *		RETURNING tuple after completing each row insert, update, or delete.
  *		It must be called again to continue the operation.  Without RETURNING,
  *		we just loop within the node until all the work is done, then
- *		return NULL.  This avoids useless call/return overhead.  (MERGE does
- *		not support RETURNING.)
+ *		return NULL.  This avoids useless call/return overhead.
  */
 
 #include "postgres.h"
@@ -97,9 +96,6 @@ typedef struct ModifyTableContext
 										  TupleTableSlot *oldSlot,
 										  MergeActionState *relaction);
 
-	/* MERGE specific */
-	MergeActionState *relaction;	/* MERGE action in progress */
-
 	/*
 	 * Information about the changes that were made concurrently to a tuple
 	 * being updated or deleted
@@ -172,13 +168,14 @@ static TupleTableSlot *ExecMerge(ModifyT
 								 ItemPointer tupleid,
 								 bool canSetTag);
 static void ExecInitMerge(ModifyTableState *mtstate, EState *estate);
-static bool ExecMergeMatched(ModifyTableContext *context,
-							 ResultRelInfo *resultRelInfo,
-							 ItemPointer tupleid,
-							 bool canSetTag);
-static void ExecMergeNotMatched(ModifyTableContext *context,
-								ResultRelInfo *resultRelInfo,
-								bool canSetTag);
+static TupleTableSlot *ExecMergeMatched(ModifyTableContext *context,
+										ResultRelInfo *resultRelInfo,
+										ItemPointer tupleid,
+										bool canSetTag,
+										bool *matched);
+static TupleTableSlot *ExecMergeNotMatched(ModifyTableContext *context,
+										   ResultRelInfo *resultRelInfo,
+										   bool canSetTag);
 static TupleTableSlot *mergeGetUpdateNewTuple(ResultRelInfo *relinfo,
 											  TupleTableSlot *planSlot,
 											  TupleTableSlot *oldSlot,
@@ -987,7 +984,7 @@ ExecInsert(ModifyTableContext *context,
 		if (mtstate->operation == CMD_UPDATE)
 			wco_kind = WCO_RLS_UPDATE_CHECK;
 		else if (mtstate->operation == CMD_MERGE)
-			wco_kind = (context->relaction->mas_action->commandType == CMD_UPDATE) ?
+			wco_kind = (mtstate->mt_merge_action->mas_action->commandType == CMD_UPDATE) ?
 				WCO_RLS_UPDATE_CHECK : WCO_RLS_INSERT_CHECK;
 		else
 			wco_kind = WCO_RLS_INSERT_CHECK;
@@ -1838,7 +1835,7 @@ ExecCrossPartitionUpdate(ModifyTableCont
 			/* and project the new tuple to retry the UPDATE with */
 			context->cpUpdateRetrySlot =
 				context->GetUpdateNewTuple(resultRelInfo, epqslot, oldSlot,
-										   context->relaction);
+										   mtstate->mt_merge_action);
 			return false;
 		}
 	}
@@ -2051,7 +2048,7 @@ lreplace:
 		 * No luck, a retry is needed.  If running MERGE, we do not do so
 		 * here; instead let it handle that on its own rules.
 		 */
-		if (context->relaction != NULL)
+		if (context->mtstate->mt_merge_action != NULL)
 			return TM_Updated;
 
 		/*
@@ -2689,6 +2686,7 @@ static TupleTableSlot *
 ExecMerge(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 		  ItemPointer tupleid, bool canSetTag)
 {
+	TupleTableSlot *rslot = NULL;
 	bool		matched;
 
 	/*-----
@@ -2736,7 +2734,8 @@ ExecMerge(ModifyTableContext *context, R
 	 */
 	matched = tupleid != NULL;
 	if (matched)
-		matched = ExecMergeMatched(context, resultRelInfo, tupleid, canSetTag);
+		rslot = ExecMergeMatched(context, resultRelInfo, tupleid, canSetTag,
+								 &matched);
 
 	/*
 	 * Either we were dealing with a NOT MATCHED tuple or ExecMergeMatched()
@@ -2744,10 +2743,9 @@ ExecMerge(ModifyTableContext *context, R
 	 * matches.
 	 */
 	if (!matched)
-		ExecMergeNotMatched(context, resultRelInfo, canSetTag);
+		rslot = ExecMergeNotMatched(context, resultRelInfo, canSetTag);
 
-	/* No RETURNING support yet */
-	return NULL;
+	return rslot;
 }
 
 /*
@@ -2757,8 +2755,8 @@ ExecMerge(ModifyTableContext *context, R
  * We start from the first WHEN MATCHED action and check if the WHEN quals
  * pass, if any. If the WHEN quals for the first action do not pass, we
  * check the second, then the third and so on. If we reach to the end, no
- * action is taken and we return true, indicating that no further action is
- * required for this tuple.
+ * action is taken and "matched" is set to true, indicating that no further
+ * action is required for this tuple.
  *
  * If we do find a qualifying action, then we attempt to execute the action.
  *
@@ -2767,15 +2765,16 @@ ExecMerge(ModifyTableContext *context, R
  * with individual actions are evaluated by this routine via ExecQual, while
  * EvalPlanQual checks for the join quals. If EvalPlanQual tells us that the
  * updated tuple still passes the join quals, then we restart from the first
- * action to look for a qualifying action. Otherwise, we return false --
- * meaning that a NOT MATCHED action must now be executed for the current
- * source tuple.
+ * action to look for a qualifying action. Otherwise, "matched" is set to
+ * false -- meaning that a NOT MATCHED action must now be executed for the
+ * current source tuple.
  */
-static bool
+static TupleTableSlot *
 ExecMergeMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
-				 ItemPointer tupleid, bool canSetTag)
+				 ItemPointer tupleid, bool canSetTag, bool *matched)
 {
 	ModifyTableState *mtstate = context->mtstate;
+	TupleTableSlot *rslot = NULL;
 	TupleTableSlot *newslot;
 	EState	   *estate = context->estate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
@@ -2787,7 +2786,10 @@ ExecMergeMatched(ModifyTableContext *con
 	 * If there are no WHEN MATCHED actions, we are done.
 	 */
 	if (resultRelInfo->ri_matchedMergeAction == NIL)
-		return true;
+	{
+		*matched = true;
+		return NULL;
+	}
 
 	/*
 	 * Make tuple and any needed join variables available to ExecQual and
@@ -2866,7 +2868,7 @@ lmerge_matched:
 				 */
 				newslot = ExecProject(relaction->mas_proj);
 
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				context->GetUpdateNewTuple = mergeGetUpdateNewTuple;
 				context->cpUpdateRetrySlot = NULL;
 
@@ -2889,7 +2891,7 @@ lmerge_matched:
 				break;
 
 			case CMD_DELETE:
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				if (!ExecDeletePrologue(context, resultRelInfo, tupleid,
 										NULL, NULL))
 				{
@@ -2949,7 +2951,8 @@ lmerge_matched:
 				 * If the tuple was already deleted, return to let caller
 				 * handle it under NOT MATCHED clauses.
 				 */
-				return false;
+				*matched = false;
+				return NULL;
 
 			case TM_Updated:
 				{
@@ -3016,13 +3019,19 @@ lmerge_matched:
 							 * NOT MATCHED actions.
 							 */
 							if (TupIsNull(epqslot))
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							(void) ExecGetJunkAttribute(epqslot,
 														resultRelInfo->ri_RowIdAttNo,
 														&isNull);
 							if (isNull)
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							/*
 							 * When a tuple was updated and migrated to
@@ -3057,7 +3066,8 @@ lmerge_matched:
 							 * tuple already deleted; tell caller to run NOT
 							 * MATCHED actions
 							 */
-							return false;
+							*matched = false;
+							return NULL;
 
 						case TM_SelfModified:
 
@@ -3077,13 +3087,14 @@ lmerge_matched:
 										(errcode(ERRCODE_TRIGGERED_DATA_CHANGE_VIOLATION),
 										 errmsg("tuple to be updated or deleted was already modified by an operation triggered by the current command"),
 										 errhint("Consider using an AFTER trigger instead of a BEFORE trigger to propagate changes to other rows.")));
-							return false;
+							*matched = false;
+							return NULL;
 
 						default:
 							/* see table_tuple_lock call in ExecDelete() */
 							elog(ERROR, "unexpected table_tuple_lock status: %u",
 								 result);
-							return false;
+							return NULL;
 					}
 				}
 
@@ -3095,6 +3106,30 @@ lmerge_matched:
 				break;
 		}
 
+		/* Process RETURNING if present */
+		if (resultRelInfo->ri_projectReturning)
+		{
+			switch (commandType)
+			{
+				case CMD_UPDATE:
+					rslot = ExecProcessReturning(resultRelInfo, newslot,
+												 context->planSlot);
+					break;
+
+				case CMD_DELETE:
+					rslot = ExecProcessReturning(resultRelInfo,
+												 resultRelInfo->ri_oldTupleSlot,
+												 context->planSlot);
+					break;
+
+				case CMD_NOTHING:
+					break;
+
+				default:
+					elog(ERROR, "unknown action in MERGE WHEN MATCHED clause");
+			}
+		}
+
 		/*
 		 * We've activated one of the WHEN clauses, so we don't search
 		 * further. This is required behaviour, not an optimization.
@@ -3105,19 +3140,22 @@ lmerge_matched:
 	/*
 	 * Successfully executed an action or no qualifying action was found.
 	 */
-	return true;
+	*matched = true;
+
+	return rslot;
 }
 
 /*
  * Execute the first qualifying NOT MATCHED action.
  */
-static void
+static TupleTableSlot *
 ExecMergeNotMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 					bool canSetTag)
 {
 	ModifyTableState *mtstate = context->mtstate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
 	List	   *actionStates = NIL;
+	TupleTableSlot *rslot = NULL;
 	ListCell   *l;
 
 	/*
@@ -3167,10 +3205,10 @@ ExecMergeNotMatched(ModifyTableContext *
 				 * so we don't need to map the tuple here.
 				 */
 				newslot = ExecProject(action->mas_proj);
-				context->relaction = action;
+				mtstate->mt_merge_action = action;
 
-				(void) ExecInsert(context, mtstate->rootResultRelInfo, newslot,
-								  canSetTag, NULL, NULL);
+				rslot = ExecInsert(context, mtstate->rootResultRelInfo,
+								   newslot, canSetTag, NULL, NULL);
 				mtstate->mt_merge_inserted += 1;
 				break;
 			case CMD_NOTHING:
@@ -3186,6 +3224,8 @@ ExecMergeNotMatched(ModifyTableContext *
 		 */
 		break;
 	}
+
+	return rslot;
 }
 
 /*
@@ -3382,6 +3422,64 @@ mergeGetUpdateNewTuple(ResultRelInfo *re
 }
 
 /*
+ * pg_merge_action() -
+ *	  SQL merge support function to retrieve the currently executing merge
+ *	  action command string ("INSERT", "UPDATE", or "DELETE").
+ */
+Datum
+pg_merge_action(PG_FUNCTION_ARGS)
+{
+	ModifyTableState *mtstate = (ModifyTableState *) fcinfo->context;
+	MergeActionState *relaction;
+
+	if (!mtstate || mtstate->operation != CMD_MERGE)
+		elog(ERROR, "merge support function called in non-merge context");
+
+	relaction = mtstate->mt_merge_action;
+	if (relaction)
+	{
+		CmdType		commandType = relaction->mas_action->commandType;
+
+		switch (commandType)
+		{
+			case CMD_INSERT:
+				PG_RETURN_TEXT_P(cstring_to_text("INSERT"));
+			case CMD_UPDATE:
+				PG_RETURN_TEXT_P(cstring_to_text("UPDATE"));
+			case CMD_DELETE:
+				PG_RETURN_TEXT_P(cstring_to_text("DELETE"));
+			case CMD_NOTHING:
+				PG_RETURN_NULL();
+			default:
+				elog(ERROR, "unrecognized commandType: %d", (int) commandType);
+		}
+	}
+
+	PG_RETURN_NULL();
+}
+
+/*
+ * pg_merge_when_clause() -
+ *	  SQL merge support function to retrieve the 1-based index of the
+ *	  currently executing merge WHEN clause.
+ */
+Datum
+pg_merge_when_clause(PG_FUNCTION_ARGS)
+{
+	ModifyTableState *mtstate = (ModifyTableState *) fcinfo->context;
+	MergeActionState *relaction;
+
+	if (!mtstate || mtstate->operation != CMD_MERGE)
+		elog(ERROR, "merge support function called in non-merge context");
+
+	relaction = mtstate->mt_merge_action;
+	if (relaction)
+		PG_RETURN_INT32((int32) relaction->mas_action->index);
+
+	PG_RETURN_NULL();
+}
+
+/*
  * Process BEFORE EACH STATEMENT triggers
  */
 static void
@@ -3667,8 +3765,17 @@ ExecModifyTable(PlanState *pstate)
 				{
 					EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-					ExecMerge(&context, node->resultRelInfo, NULL, node->canSetTag);
-					continue;	/* no RETURNING support yet */
+					slot = ExecMerge(&context, node->resultRelInfo, NULL,
+									 node->canSetTag);
+
+					/*
+					 * If we got a RETURNING result, return it to the caller.
+					 * We'll continue the work on next call.
+					 */
+					if (slot)
+						return slot;
+
+					continue;	/* continue with the next tuple */
 				}
 
 				elog(ERROR, "tableoid is NULL");
@@ -3745,8 +3852,17 @@ ExecModifyTable(PlanState *pstate)
 					{
 						EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-						ExecMerge(&context, node->resultRelInfo, NULL, node->canSetTag);
-						continue;	/* no RETURNING support yet */
+						slot = ExecMerge(&context, node->resultRelInfo, NULL,
+										 node->canSetTag);
+
+						/*
+						 * If we got a RETURNING result, return it to the
+						 * caller.  We'll continue the work on next call.
+						 */
+						if (slot)
+							return slot;
+
+						continue;	/* continue with the next tuple */
 					}
 
 					elog(ERROR, "ctid is NULL");
@@ -3839,7 +3955,7 @@ ExecModifyTable(PlanState *pstate)
 				slot = internalGetUpdateNewTuple(resultRelInfo, context.planSlot,
 												 oldSlot, NULL);
 				context.GetUpdateNewTuple = internalGetUpdateNewTuple;
-				context.relaction = NULL;
+				node->mt_merge_action = NULL;
 
 				/* Now apply the update. */
 				slot = ExecUpdate(&context, resultRelInfo, tupleid, oldtuple,
diff --git a/src/backend/executor/spi.c b/src/backend/executor/spi.c
new file mode 100644
index e3a170c..be88056
--- a/src/backend/executor/spi.c
+++ b/src/backend/executor/spi.c
@@ -2033,6 +2033,8 @@ SPI_result_code_string(int code)
 			return "SPI_OK_TD_REGISTER";
 		case SPI_OK_MERGE:
 			return "SPI_OK_MERGE";
+		case SPI_OK_MERGE_RETURNING:
+			return "SPI_OK_MERGE_RETURNING";
 	}
 	/* Unrecognized code ... return something useful ... */
 	sprintf(buf, "Unrecognized SPI code %d", code);
@@ -2886,7 +2888,10 @@ _SPI_pquery(QueryDesc *queryDesc, bool f
 				res = SPI_OK_UPDATE;
 			break;
 		case CMD_MERGE:
-			res = SPI_OK_MERGE;
+			if (queryDesc->plannedstmt->hasReturning)
+				res = SPI_OK_MERGE_RETURNING;
+			else
+				res = SPI_OK_MERGE;
 			break;
 		default:
 			return SPI_ERROR_OPUNKNOWN;
diff --git a/src/backend/parser/analyze.c b/src/backend/parser/analyze.c
new file mode 100644
index e892df9..fe0f5cb
--- a/src/backend/parser/analyze.c
+++ b/src/backend/parser/analyze.c
@@ -73,7 +73,6 @@ static void determineRecursiveColTypes(P
 									   Node *larg, List *nrtargetlist);
 static Query *transformReturnStmt(ParseState *pstate, ReturnStmt *stmt);
 static Query *transformUpdateStmt(ParseState *pstate, UpdateStmt *stmt);
-static List *transformReturningList(ParseState *pstate, List *returningList);
 static Query *transformPLAssignStmt(ParseState *pstate,
 									PLAssignStmt *stmt);
 static Query *transformDeclareCursorStmt(ParseState *pstate,
@@ -514,7 +513,8 @@ transformDeleteStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -941,7 +941,8 @@ transformInsertStmt(ParseState *pstate,
 	/* Process RETURNING, if any. */
 	if (stmt->returningList)
 		qry->returningList = transformReturningList(pstate,
-													stmt->returningList);
+													stmt->returningList,
+													EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -2406,7 +2407,8 @@ transformUpdateStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/*
 	 * Now we are done with SELECT-like processing, and can get on with
@@ -2500,10 +2502,11 @@ transformUpdateTargetList(ParseState *ps
 
 /*
  * transformReturningList -
- *	handle a RETURNING clause in INSERT/UPDATE/DELETE
+ *	handle a RETURNING clause in INSERT/UPDATE/DELETE/MERGE
  */
-static List *
-transformReturningList(ParseState *pstate, List *returningList)
+List *
+transformReturningList(ParseState *pstate, List *returningList,
+					   ParseExprKind exprKind)
 {
 	List	   *rlist;
 	int			save_next_resno;
@@ -2520,7 +2523,7 @@ transformReturningList(ParseState *pstat
 	pstate->p_next_resno = 1;
 
 	/* transform RETURNING identically to a SELECT targetlist */
-	rlist = transformTargetList(pstate, returningList, EXPR_KIND_RETURNING);
+	rlist = transformTargetList(pstate, returningList, exprKind);
 
 	/*
 	 * Complain if the nonempty tlist expanded to nothing (which is possible
diff --git a/src/backend/parser/gram.y b/src/backend/parser/gram.y
new file mode 100644
index a013838..4406430
--- a/src/backend/parser/gram.y
+++ b/src/backend/parser/gram.y
@@ -12241,6 +12241,7 @@ MergeStmt:
 			USING table_ref
 			ON a_expr
 			merge_when_list
+			returning_clause
 				{
 					MergeStmt  *m = makeNode(MergeStmt);
 
@@ -12249,6 +12250,7 @@ MergeStmt:
 					m->sourceRelation = $6;
 					m->joinCondition = $8;
 					m->mergeWhenClauses = $9;
+					m->returningList = $10;
 
 					$$ = (Node *) m;
 				}
diff --git a/src/backend/parser/parse_agg.c b/src/backend/parser/parse_agg.c
new file mode 100644
index 4fbf80c..b8d655a
--- a/src/backend/parser/parse_agg.c
+++ b/src/backend/parser/parse_agg.c
@@ -456,6 +456,7 @@ check_agglevels_and_constraints(ParseSta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
@@ -904,6 +905,7 @@ transformWindowFuncCall(ParseState *psta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_cte.c b/src/backend/parser/parse_cte.c
new file mode 100644
index c5b1a49..ad3c525
--- a/src/backend/parser/parse_cte.c
+++ b/src/backend/parser/parse_cte.c
@@ -126,13 +126,6 @@ transformWithClause(ParseState *pstate,
 		CommonTableExpr *cte = (CommonTableExpr *) lfirst(lc);
 		ListCell   *rest;
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(cte->ctequery, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in WITH query"),
-					parser_errposition(pstate, cte->location));
-
 		for_each_cell(rest, withClause->ctes, lnext(withClause->ctes, lc))
 		{
 			CommonTableExpr *cte2 = (CommonTableExpr *) lfirst(rest);
@@ -153,7 +146,8 @@ transformWithClause(ParseState *pstate,
 			/* must be a data-modifying statement */
 			Assert(IsA(cte->ctequery, InsertStmt) ||
 				   IsA(cte->ctequery, UpdateStmt) ||
-				   IsA(cte->ctequery, DeleteStmt));
+				   IsA(cte->ctequery, DeleteStmt) ||
+				   IsA(cte->ctequery, MergeStmt));
 
 			pstate->p_hasModifyingCTE = true;
 		}
diff --git a/src/backend/parser/parse_expr.c b/src/backend/parser/parse_expr.c
new file mode 100644
index 7ff41ac..92cd8c1
--- a/src/backend/parser/parse_expr.c
+++ b/src/backend/parser/parse_expr.c
@@ -482,6 +482,7 @@ transformColumnRef(ParseState *pstate, C
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CHECK_CONSTRAINT:
@@ -1708,6 +1709,7 @@ transformSubLink(ParseState *pstate, Sub
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CYCLE_MARK:
@@ -3002,6 +3004,7 @@ ParseExprKindName(ParseExprKind exprKind
 		case EXPR_KIND_OFFSET:
 			return "OFFSET";
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			return "RETURNING";
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
diff --git a/src/backend/parser/parse_func.c b/src/backend/parser/parse_func.c
new file mode 100644
index ca14f06..914624e
--- a/src/backend/parser/parse_func.c
+++ b/src/backend/parser/parse_func.c
@@ -31,6 +31,7 @@
 #include "parser/parse_target.h"
 #include "parser/parse_type.h"
 #include "utils/builtins.h"
+#include "utils/fmgroids.h"
 #include "utils/lsyscache.h"
 #include "utils/syscache.h"
 
@@ -348,6 +349,15 @@ ParseFuncOrColumn(ParseState *pstate, Li
 					 parser_errposition(pstate, location)));
 	}
 
+	/* Merge support functions are only allowed in MERGE's RETURNING list */
+	if (IsMergeSupportFunction(funcid) &&
+		pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+		ereport(ERROR,
+				(errcode(ERRCODE_WRONG_OBJECT_TYPE),
+				 errmsg("merge support function %s can only be called from the RETURNING list of a MERGE command",
+						NameListToString(funcname)),
+				 parser_errposition(pstate, location)));
+
 	/*
 	 * So far so good, so do some fdresult-type-specific processing.
 	 */
@@ -2602,6 +2612,7 @@ check_srf_call_placement(ParseState *pst
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_merge.c b/src/backend/parser/parse_merge.c
new file mode 100644
index d886637..2e50eb0
--- a/src/backend/parser/parse_merge.c
+++ b/src/backend/parser/parse_merge.c
@@ -233,6 +233,10 @@ transformMergeStmt(ParseState *pstate, M
 	 */
 	qry->jointree = makeFromExpr(pstate->p_joinlist, joinExpr);
 
+	/* Transform the RETURNING list, if any */
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_MERGE_RETURNING);
+
 	/*
 	 * We now have a good query shape, so now look at the WHEN conditions and
 	 * action targetlists.
@@ -254,6 +258,7 @@ transformMergeStmt(ParseState *pstate, M
 		MergeAction *action;
 
 		action = makeNode(MergeAction);
+		action->index = foreach_current_index(l) + 1;
 		action->commandType = mergeWhenClause->commandType;
 		action->matched = mergeWhenClause->matched;
 
@@ -390,9 +395,6 @@ transformMergeStmt(ParseState *pstate, M
 
 	qry->mergeActionList = mergeActionList;
 
-	/* RETURNING could potentially be added in the future, but not in SQL std */
-	qry->returningList = NULL;
-
 	qry->hasTargetSRFs = false;
 	qry->hasSubLinks = pstate->p_hasSubLinks;
 
diff --git a/src/backend/parser/parse_relation.c b/src/backend/parser/parse_relation.c
new file mode 100644
index de355dd..97ca651
--- a/src/backend/parser/parse_relation.c
+++ b/src/backend/parser/parse_relation.c
@@ -2345,9 +2345,10 @@ addRangeTableEntryForCTE(ParseState *pst
 		cte->cterefcount++;
 
 	/*
-	 * We throw error if the CTE is INSERT/UPDATE/DELETE without RETURNING.
-	 * This won't get checked in case of a self-reference, but that's OK
-	 * because data-modifying CTEs aren't allowed to be recursive anyhow.
+	 * We throw error if the CTE is INSERT/UPDATE/DELETE/MERGE without
+	 * RETURNING.  This won't get checked in case of a self-reference, but
+	 * that's OK because data-modifying CTEs aren't allowed to be recursive
+	 * anyhow.
 	 */
 	if (IsA(cte->ctequery, Query))
 	{
diff --git a/src/backend/rewrite/rewriteHandler.c b/src/backend/rewrite/rewriteHandler.c
new file mode 100644
index b6452e9..e753620
--- a/src/backend/rewrite/rewriteHandler.c
+++ b/src/backend/rewrite/rewriteHandler.c
@@ -3625,7 +3625,8 @@ RewriteQuery(Query *parsetree, List *rew
 			if (!(ctequery->commandType == CMD_SELECT ||
 				  ctequery->commandType == CMD_UPDATE ||
 				  ctequery->commandType == CMD_INSERT ||
-				  ctequery->commandType == CMD_DELETE))
+				  ctequery->commandType == CMD_DELETE ||
+				  ctequery->commandType == CMD_MERGE))
 			{
 				/*
 				 * Currently it could only be NOTIFY; this error message will
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
new file mode 100644
index c7d9d96..d393f51
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -2131,11 +2131,10 @@ QueryReturnsTuples(Query *parsetree)
 		case CMD_SELECT:
 			/* returns tuples */
 			return true;
-		case CMD_MERGE:
-			return false;
 		case CMD_INSERT:
 		case CMD_UPDATE:
 		case CMD_DELETE:
+		case CMD_MERGE:
 			/* the forms with RETURNING return tuples */
 			if (parsetree->returningList)
 				return true;
diff --git a/src/bin/psql/common.c b/src/bin/psql/common.c
new file mode 100644
index f907f5d..85eb868
--- a/src/bin/psql/common.c
+++ b/src/bin/psql/common.c
@@ -962,13 +962,17 @@ PrintQueryResult(PGresult *result, bool
 			else
 				success = true;
 
-			/* if it's INSERT/UPDATE/DELETE RETURNING, also print status */
+			/*
+			 * If it's INSERT/UPDATE/DELETE/MERGE RETURNING, also print
+			 * status.
+			 */
 			if (last || pset.show_all_results)
 			{
 				cmdstatus = PQcmdStatus(result);
 				if (strncmp(cmdstatus, "INSERT", 6) == 0 ||
 					strncmp(cmdstatus, "UPDATE", 6) == 0 ||
-					strncmp(cmdstatus, "DELETE", 6) == 0)
+					strncmp(cmdstatus, "DELETE", 6) == 0 ||
+					strncmp(cmdstatus, "MERGE", 5) == 0)
 					PrintQueryStatus(result, printStatusFout);
 			}
 
diff --git a/src/include/catalog/pg_proc.dat b/src/include/catalog/pg_proc.dat
new file mode 100644
index e2a7642..50c9b59
--- a/src/include/catalog/pg_proc.dat
+++ b/src/include/catalog/pg_proc.dat
@@ -11935,4 +11935,14 @@
   proname => 'any_value_transfn', prorettype => 'anyelement',
   proargtypes => 'anyelement anyelement', prosrc => 'any_value_transfn' },
 
+# MERGE support functions
+{ oid => '9499', descr => 'command type of current MERGE action',
+  proname => 'pg_merge_action',  provolatile => 'v',
+  prorettype => 'text', proargtypes => '',
+  prosrc => 'pg_merge_action' },
+{ oid => '9500', descr => 'index of current MERGE WHEN clause',
+  proname => 'pg_merge_when_clause',  provolatile => 'v',
+  prorettype => 'int4', proargtypes => '',
+  prosrc => 'pg_merge_when_clause' },
+
 ]
diff --git a/src/include/catalog/pg_proc.h b/src/include/catalog/pg_proc.h
new file mode 100644
index e7abe0b..e597826
--- a/src/include/catalog/pg_proc.h
+++ b/src/include/catalog/pg_proc.h
@@ -182,6 +182,11 @@ DECLARE_UNIQUE_INDEX(pg_proc_proname_arg
 #define PROARGMODE_VARIADIC 'v'
 #define PROARGMODE_TABLE	't'
 
+/* Is this a merge support function?  (Requires fmgroids.h) */
+#define IsMergeSupportFunction(oid) \
+	((oid) == F_PG_MERGE_ACTION || \
+	 (oid) == F_PG_MERGE_WHEN_CLAUSE)
+
 #endif							/* EXPOSE_TO_CLIENT_CODE */
 
 
diff --git a/src/include/executor/spi.h b/src/include/executor/spi.h
new file mode 100644
index d1de139..ed8787d
--- a/src/include/executor/spi.h
+++ b/src/include/executor/spi.h
@@ -97,6 +97,7 @@ typedef struct _SPI_plan *SPIPlanPtr;
 #define SPI_OK_REL_UNREGISTER	16
 #define SPI_OK_TD_REGISTER		17
 #define SPI_OK_MERGE			18
+#define SPI_OK_MERGE_RETURNING	19
 
 #define SPI_OPT_NONATOMIC		(1 << 0)
 
diff --git a/src/include/nodes/execnodes.h b/src/include/nodes/execnodes.h
new file mode 100644
index 20f4c8b..3d62fb9
--- a/src/include/nodes/execnodes.h
+++ b/src/include/nodes/execnodes.h
@@ -1302,6 +1302,9 @@ typedef struct ModifyTableState
 	/* Flags showing which subcommands are present INS/UPD/DEL/DO NOTHING */
 	int			mt_merge_subcommands;
 
+	/* For MERGE, the action currently being executed */
+	MergeActionState *mt_merge_action;
+
 	/* tuple counters for MERGE */
 	double		mt_merge_inserted;
 	double		mt_merge_updated;
diff --git a/src/include/nodes/parsenodes.h b/src/include/nodes/parsenodes.h
new file mode 100644
index f7d7f10..a959ac7
--- a/src/include/nodes/parsenodes.h
+++ b/src/include/nodes/parsenodes.h
@@ -1687,6 +1687,7 @@ typedef struct MergeWhenClause
 typedef struct MergeAction
 {
 	NodeTag		type;
+	int			index;			/* 1-based index of the clause */
 	bool		matched;		/* true=MATCHED, false=NOT MATCHED */
 	CmdType		commandType;	/* INSERT/UPDATE/DELETE/DO NOTHING */
 	/* OVERRIDING clause */
@@ -1806,6 +1807,7 @@ typedef struct MergeStmt
 	Node	   *sourceRelation; /* source relation */
 	Node	   *joinCondition;	/* join condition between source and target */
 	List	   *mergeWhenClauses;	/* list of MergeWhenClause(es) */
+	List	   *returningList;	/* list of expressions to return */
 	WithClause *withClause;		/* WITH clause */
 } MergeStmt;
 
diff --git a/src/include/parser/analyze.h b/src/include/parser/analyze.h
new file mode 100644
index 1cef183..78eb55e
--- a/src/include/parser/analyze.h
+++ b/src/include/parser/analyze.h
@@ -44,6 +44,8 @@ extern List *transformInsertRow(ParseSta
 								bool strip_indirection);
 extern List *transformUpdateTargetList(ParseState *pstate,
 									   List *origTlist);
+extern List *transformReturningList(ParseState *pstate, List *returningList,
+									ParseExprKind exprKind);
 extern Query *transformTopLevelStmt(ParseState *pstate, RawStmt *parseTree);
 extern Query *transformStmt(ParseState *pstate, Node *parseTree);
 
diff --git a/src/include/parser/parse_node.h b/src/include/parser/parse_node.h
new file mode 100644
index f589112..836819d
--- a/src/include/parser/parse_node.h
+++ b/src/include/parser/parse_node.h
@@ -61,7 +61,8 @@ typedef enum ParseExprKind
 	EXPR_KIND_DISTINCT_ON,		/* DISTINCT ON */
 	EXPR_KIND_LIMIT,			/* LIMIT */
 	EXPR_KIND_OFFSET,			/* OFFSET */
-	EXPR_KIND_RETURNING,		/* RETURNING */
+	EXPR_KIND_RETURNING,		/* RETURNING in INSERT/UPDATE/DELETE */
+	EXPR_KIND_MERGE_RETURNING,	/* RETURNING in MERGE */
 	EXPR_KIND_VALUES,			/* VALUES */
 	EXPR_KIND_VALUES_SINGLE,	/* single-row VALUES (in INSERT only) */
 	EXPR_KIND_CHECK_CONSTRAINT, /* CHECK constraint for a table */
diff --git a/src/pl/plpgsql/src/pl_exec.c b/src/pl/plpgsql/src/pl_exec.c
new file mode 100644
index ffd6d2e..502056d
--- a/src/pl/plpgsql/src/pl_exec.c
+++ b/src/pl/plpgsql/src/pl_exec.c
@@ -4246,9 +4246,9 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 	/*
 	 * If we have INTO, then we only need one row back ... but if we have INTO
 	 * STRICT or extra check too_many_rows, ask for two rows, so that we can
-	 * verify the statement returns only one.  INSERT/UPDATE/DELETE are always
-	 * treated strictly. Without INTO, just run the statement to completion
-	 * (tcount = 0).
+	 * verify the statement returns only one.  INSERT/UPDATE/DELETE/MERGE are
+	 * always treated strictly. Without INTO, just run the statement to
+	 * completion (tcount = 0).
 	 *
 	 * We could just ask for two rows always when using INTO, but there are
 	 * some cases where demanding the extra row costs significant time, eg by
@@ -4286,10 +4286,11 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 			Assert(stmt->mod_stmt);
 			exec_set_found(estate, (SPI_processed != 0));
 			break;
@@ -4468,10 +4469,11 @@ exec_stmt_dynexecute(PLpgSQL_execstate *
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 		case SPI_OK_UTILITY:
 		case SPI_OK_REWRITTEN:
 			break;
diff --git a/src/pl/tcl/pltcl.c b/src/pl/tcl/pltcl.c
new file mode 100644
index 499a9ea..58c4f50
--- a/src/pl/tcl/pltcl.c
+++ b/src/pl/tcl/pltcl.c
@@ -2459,6 +2459,7 @@ pltcl_process_SPI_result(Tcl_Interp *int
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
+		case SPI_OK_MERGE_RETURNING:
 
 			/*
 			 * Process the tuples we got
diff --git a/src/test/regress/expected/merge.out b/src/test/regress/expected/merge.out
new file mode 100644
index e1d3639..ef6dd55
--- a/src/test/regress/expected/merge.out
+++ b/src/test/regress/expected/merge.out
@@ -123,20 +123,20 @@ ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
 ERROR:  name "target" specified more than once
 DETAIL:  The name is used both as MERGE target table and data source.
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
-ERROR:  MERGE not supported in WITH query
-LINE 1: WITH foo AS (
-             ^
--- used in COPY
+ERROR:  WITH query "foo" does not have a RETURNING clause
+LINE 4: ) SELECT * FROM foo;
+                        ^
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) TO stdout;
-ERROR:  MERGE not supported in COPY
+ERROR:  COPY query must have a RETURNING clause
 -- unsupported relation types
 -- view
 CREATE VIEW tv AS SELECT * FROM target;
@@ -1289,20 +1289,187 @@ WHEN MATCHED AND tid < 2 THEN
 ROLLBACK;
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
-ERROR:  syntax error at or near "RETURNING"
-LINE 10: RETURNING *;
-         ^
+    DELETE
+RETURNING pg_merge_when_clause() AS when_clause,
+          pg_merge_action() AS merge_action,
+          t.*,
+          CASE pg_merge_action()
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ when_clause | merge_action | tid | balance |     description     
+-------------+--------------+-----+---------+---------------------
+           3 | DELETE       |   1 |     100 | Removed (1,100)
+           1 | UPDATE       |   2 |     220 | Added 20 to balance
+           2 | INSERT       |   4 |      40 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- error when using MERGE support functions outside MERGE
+SELECT pg_merge_action() FROM sq_target;
+ERROR:  merge support function pg_merge_action can only be called from the RETURNING list of a MERGE command
+LINE 1: SELECT pg_merge_action() FROM sq_target;
+               ^
+UPDATE sq_target SET balance = balance + 1 RETURNING pg_merge_when_clause();
+ERROR:  merge support function pg_merge_when_clause can only be called from the RETURNING list of a MERGE command
+LINE 1: ...ATE sq_target SET balance = balance + 1 RETURNING pg_merge_w...
+                                                             ^
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_when_clause() AS when_clause,
+              pg_merge_action() AS merge_action,
+              t.*,
+              CASE pg_merge_action()
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING merge_action, pg_merge_action() AS merge_log_action, l.*
+)
+SELECT * FROM m2;
+ merge_action | merge_log_action | tid |     last_change     
+--------------+------------------+-----+---------------------
+ DELETE       | UPDATE           |   1 | Removed (1,100)
+ UPDATE       | INSERT           |   2 | Added 20 to balance
+ INSERT       | INSERT           |   4 | Inserted (4,40)
+(3 rows)
+
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ tid |     last_change     
+-----+---------------------
+   1 | Removed (1,100)
+   2 | Added 20 to balance
+   4 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING v
+    ON tid = sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*
+) TO stdout;
+DELETE	1	100
+ROLLBACK;
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ action | tid | new_balance 
+--------+-----+-------------
+ DELETE |   1 |         100
+ UPDATE |   3 |         320
+ INSERT |   4 |         110
+(3 rows)
+
+ROLLBACK;
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ action | tid | balance 
+--------+-----+---------
+ DELETE |   1 |     100
+ INSERT |   4 |      40
+(2 rows)
+
+ROLLBACK;
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ r_action | r_tid | r_balance 
+----------+-------+-----------
+ DELETE   |     1 |       100
+ UPDATE   |     3 |       320
+ INSERT   |     4 |       110
+(3 rows)
+
 ROLLBACK;
 -- EXPLAIN
 CREATE TABLE ex_mtarget (a int, b int)
@@ -1499,7 +1666,7 @@ SELECT * FROM sq_target WHERE tid = 1;
 (1 row)
 
 ROLLBACK;
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 NOTICE:  drop cascades to view v
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1625,6 +1792,32 @@ SELECT * FROM pa_target ORDER BY tid;
 (14 rows)
 
 ROLLBACK;
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
+ pg_merge_action | pg_merge_when_clause | tid | balance |           val            
+-----------------+----------------------+-----+---------+--------------------------
+ UPDATE          |                    1 |   2 |     110 | initial updated by merge
+(1 row)
+
+SELECT * FROM pa_target ORDER BY tid;
+ tid | balance |           val            
+-----+---------+--------------------------
+   2 |     110 | initial updated by merge
+   3 |     300 | initial
+   5 |     500 | initial
+   7 |     700 | initial
+   9 |     900 | initial
+  11 |    1100 | initial
+  13 |    1300 | initial
+(7 rows)
+
+ROLLBACK;
 DROP TABLE pa_target CASCADE;
 -- The target table is partitioned in the same way, but this time by attaching
 -- partitions which have columns in different order, dropped columns etc.
@@ -1716,7 +1909,26 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (sid, delta, 'inserted by merge');
+    INSERT VALUES (sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
+ pg_merge_action | pg_merge_when_clause | tid | balance |           val            
+-----------------+----------------------+-----+---------+--------------------------
+ UPDATE          |                    1 |   2 |     110 | initial updated by merge
+ INSERT          |                    2 |   2 |      20 | inserted by merge
+ UPDATE          |                    1 |   4 |     330 | initial updated by merge
+ INSERT          |                    2 |   4 |      40 | inserted by merge
+ UPDATE          |                    1 |   6 |     550 | initial updated by merge
+ INSERT          |                    2 |   6 |      60 | inserted by merge
+ UPDATE          |                    1 |   8 |     770 | initial updated by merge
+ INSERT          |                    2 |   8 |      80 | inserted by merge
+ UPDATE          |                    1 |  10 |     990 | initial updated by merge
+ INSERT          |                    2 |  10 |     100 | inserted by merge
+ UPDATE          |                    1 |  12 |    1210 | initial updated by merge
+ INSERT          |                    2 |  12 |     120 | inserted by merge
+ UPDATE          |                    1 |  14 |    1430 | initial updated by merge
+ INSERT          |                    2 |  14 |     140 | inserted by merge
+(14 rows)
+
 SELECT * FROM pa_target ORDER BY tid;
  tid | balance |           val            
 -----+---------+--------------------------
@@ -1783,7 +1995,21 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
+ pg_merge_action | pg_merge_when_clause |          logts           | tid | balance |           val            
+-----------------+----------------------+--------------------------+-----+---------+--------------------------
+ UPDATE          |                    1 | Tue Jan 31 00:00:00 2017 |   1 |     110 | initial updated by merge
+ UPDATE          |                    1 | Tue Feb 28 00:00:00 2017 |   2 |     220 | initial updated by merge
+ INSERT          |                    2 | Sun Jan 15 00:00:00 2017 |   3 |      30 | inserted by merge
+ UPDATE          |                    1 | Tue Jan 31 00:00:00 2017 |   4 |     440 | initial updated by merge
+ UPDATE          |                    1 | Tue Feb 28 00:00:00 2017 |   5 |     550 | initial updated by merge
+ INSERT          |                    2 | Sun Jan 15 00:00:00 2017 |   6 |      60 | inserted by merge
+ UPDATE          |                    1 | Tue Jan 31 00:00:00 2017 |   7 |     770 | initial updated by merge
+ UPDATE          |                    1 | Tue Feb 28 00:00:00 2017 |   8 |     880 | initial updated by merge
+ INSERT          |                    2 | Sun Jan 15 00:00:00 2017 |   9 |      90 | inserted by merge
+(9 rows)
+
 SELECT * FROM pa_target ORDER BY tid;
           logts           | tid | balance |           val            
 --------------------------+-----+---------+--------------------------
diff --git a/src/test/regress/sql/merge.sql b/src/test/regress/sql/merge.sql
new file mode 100644
index f085416..dba6a19
--- a/src/test/regress/sql/merge.sql
+++ b/src/test/regress/sql/merge.sql
@@ -86,12 +86,12 @@ MERGE INTO target
 USING target
 ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
--- used in COPY
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
@@ -844,17 +844,145 @@ ROLLBACK;
 
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
+    DELETE
+RETURNING pg_merge_when_clause() AS when_clause,
+          pg_merge_action() AS merge_action,
+          t.*,
+          CASE pg_merge_action()
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ROLLBACK;
+
+-- error when using MERGE support functions outside MERGE
+SELECT pg_merge_action() FROM sq_target;
+UPDATE sq_target SET balance = balance + 1 RETURNING pg_merge_when_clause();
+
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_when_clause() AS when_clause,
+              pg_merge_action() AS merge_action,
+              t.*,
+              CASE pg_merge_action()
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING merge_action, pg_merge_action() AS merge_log_action, l.*
+)
+SELECT * FROM m2;
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ROLLBACK;
+
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING v
+    ON tid = sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*
+) TO stdout;
+ROLLBACK;
+
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ROLLBACK;
+
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ROLLBACK;
+
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
 ROLLBACK;
 
 -- EXPLAIN
@@ -953,7 +1081,7 @@ WHEN MATCHED THEN
 SELECT * FROM sq_target WHERE tid = 1;
 ROLLBACK;
 
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1020,6 +1148,17 @@ SELECT merge_func();
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
+SELECT * FROM pa_target ORDER BY tid;
+ROLLBACK;
+
 DROP TABLE pa_target CASCADE;
 
 -- The target table is partitioned in the same way, but this time by attaching
@@ -1078,7 +1217,8 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (sid, delta, 'inserted by merge');
+    INSERT VALUES (sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
@@ -1132,7 +1272,8 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 

diff --git a/doc/src/sgml/dml.sgml b/doc/src/sgml/dml.sgml
new file mode 100644
index cbbc5e2..ff2a827
--- a/doc/src/sgml/dml.sgml
+++ b/doc/src/sgml/dml.sgml
@@ -283,10 +283,15 @@ DELETE FROM products;
    <secondary>RETURNING</secondary>
   </indexterm>
 
+  <indexterm zone="dml-returning">
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
   <para>
    Sometimes it is useful to obtain data from modified rows while they are
    being manipulated.  The <command>INSERT</command>, <command>UPDATE</command>,
-   and <command>DELETE</command> commands all have an
+   <command>DELETE</command>, and <command>MERGE</command> commands all have an
    optional <literal>RETURNING</literal> clause that supports this.  Use
    of <literal>RETURNING</literal> avoids performing an extra database query to
    collect the data, and is especially valuable when it would otherwise be
@@ -339,6 +344,24 @@ DELETE FROM products
 </programlisting>
   </para>
 
+  <para>
+   In a <command>MERGE</command>, the data available to <literal>RETURNING</literal> is
+   the content of the source row plus the content of the inserted, updated, or
+   deleted target row.  In addition, the merge support functions
+   <xref linkend="pg-merge-action"/> and <xref linkend="pg-merge-when-clause"/>
+   may be used to return information about which merge action was executed for
+   each row.  Since it is quite common for the source and target to have many
+   of the same columns, specifying <literal>RETURNING *</literal> can lead to a
+   lot of duplicated columns, so it is often more useful to just return the
+   target row.  For example:
+<programlisting>
+MERGE INTO products p USING new_products n ON p.product_no = n.product_no
+  WHEN NOT MATCHED THEN INSERT VALUES (n.product_no, n.name, n.price)
+  WHEN MATCHED THEN UPDATE SET name = n.name, price = n.price
+  RETURNING pg_merge_action(), p.*;
+</programlisting>
+  </para>
+
   <para>
    If there are triggers (<xref linkend="triggers"/>) on the target table,
    the data available to <literal>RETURNING</literal> is the row as modified by
diff --git a/doc/src/sgml/func.sgml b/doc/src/sgml/func.sgml
new file mode 100644
index 0cbdf63..224e9b8
--- a/doc/src/sgml/func.sgml
+++ b/doc/src/sgml/func.sgml
@@ -21216,6 +21216,99 @@ SELECT count(*) FROM sometable;
 
  </sect1>
 
+ <sect1 id="functions-merge-support">
+  <title>Merge Support Functions</title>
+
+  <indexterm>
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
+  <para>
+   The merge support functions shown in
+   <xref linkend="functions-merge-support-table"/> may be used in the
+   <literal>RETURNING</literal> list of a <xref linkend="sql-merge"/> command
+   to return additional information about the action taken for each row.
+  </para>
+
+  <table id="functions-merge-support-table">
+    <title>Merge Support Functions</title>
+
+    <tgroup cols="1">
+     <thead>
+      <row>
+       <entry role="func_table_entry"><para role="func_signature">
+        Function
+       </para>
+       <para>
+        Description
+       </para></entry>
+      </row>
+     </thead>
+
+     <tbody>
+      <row>
+       <entry role="func_table_entry"><para role="func_signature">
+        <indexterm id="pg-merge-action">
+         <primary>pg_merge_action</primary>
+        </indexterm>
+        <function>pg_merge_action</function> ( )
+        <returnvalue>text</returnvalue>
+       </para>
+       <para>
+        Returns the action performed on the current row (<literal>'INSERT'</literal>,
+        <literal>'UPDATE'</literal>, or <literal>'DELETE'</literal>).
+       </para></entry>
+      </row>
+
+      <row>
+       <entry role="func_table_entry"><para role="func_signature">
+        <indexterm id="pg-merge-when-clause">
+         <primary>pg_merge_when_clause</primary>
+        </indexterm>
+        <function>pg_merge_when_clause</function> ( )
+        <returnvalue>integer</returnvalue>
+       </para>
+       <para>
+        Returns the 1-based index of the <literal>WHEN</literal> clause executed
+        for the current row.
+       </para></entry>
+      </row>
+     </tbody>
+    </tgroup>
+  </table>
+
+  <para>
+   Example:
+<screen><![CDATA[
+MERGE INTO products p
+  USING stock s ON p.product_id = s.product_id
+  WHEN MATCHED AND s.quantity > 0 THEN
+    UPDATE SET in_stock = true, quantity = s.quantity
+  WHEN MATCHED THEN
+    UPDATE SET in_stock = false, quantity = 0
+  WHEN NOT MATCHED THEN
+    INSERT (product_id, in_stock, quantity)
+      VALUES (s.product_id, true, s.quantity)
+  RETURNING pg_merge_action() AS action,
+            pg_merge_when_clause() AS when_clause,
+            p.*;
+
+ action | when_clause | product_id | in_stock | quantity 
+--------+-------------+------------+----------+----------
+ UPDATE |           1 |       1001 | t        |       50
+ UPDATE |           2 |       1002 | f        |        0
+ INSERT |           3 |       1003 | t        |       10
+]]></screen>
+  </para>
+
+  <para>
+   Note that it is an error to use these functions anywhere other than the
+   <literal>RETURNING</literal> list of a <command>MERGE</command> command.
+  </para>
+
+ </sect1>
+
  <sect1 id="functions-subquery">
   <title>Subquery Expressions</title>
 
diff --git a/doc/src/sgml/glossary.sgml b/doc/src/sgml/glossary.sgml
new file mode 100644
index 7c01a54..4317cfc
--- a/doc/src/sgml/glossary.sgml
+++ b/doc/src/sgml/glossary.sgml
@@ -1323,9 +1323,9 @@
      to a <glossterm linkend="glossary-client">client</glossterm> upon the
      completion of an <acronym>SQL</acronym> command, usually a
      <command>SELECT</command> but it can be an
-     <command>INSERT</command>, <command>UPDATE</command>, or
-     <command>DELETE</command> command if the <literal>RETURNING</literal>
-     clause is specified.
+     <command>INSERT</command>, <command>UPDATE</command>,
+     <command>DELETE</command>, or <command>MERGE</command> command if the
+     <literal>RETURNING</literal> clause is specified.
     </para>
     <para>
      The fact that a result set is a relation means that a query can be used
diff --git a/doc/src/sgml/plpgsql.sgml b/doc/src/sgml/plpgsql.sgml
new file mode 100644
index 7c8a49f..b839695
--- a/doc/src/sgml/plpgsql.sgml
+++ b/doc/src/sgml/plpgsql.sgml
@@ -1020,8 +1020,8 @@ INSERT INTO mytable VALUES (1,'one'), (2
     </para>
 
     <para>
-     If the command does return rows (for example <command>SELECT</command>,
-     or <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>
+     If the command does return rows (for example <command>SELECT</command>, or
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
      with <literal>RETURNING</literal>), there are two ways to proceed.
      When the command will return at most one row, or you only care about
      the first row of output, write the command as usual but add
@@ -1149,6 +1149,7 @@ SELECT <replaceable>select_expressions</
 INSERT ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 UPDATE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 DELETE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
+MERGE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 </synopsis>
 
      where <replaceable>target</replaceable> can be a record variable, a row
@@ -1159,8 +1160,8 @@ DELETE ... RETURNING <replaceable>expres
      <literal>INTO</literal> clause) just as described above,
      and the plan is cached in the same way.
      This works for <command>SELECT</command>,
-     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
-     <literal>RETURNING</literal>, and certain utility commands
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
+     with <literal>RETURNING</literal>, and certain utility commands
      that return row sets, such as <command>EXPLAIN</command>.
      Except for the <literal>INTO</literal> clause, the SQL command is the same
      as it would be written outside <application>PL/pgSQL</application>.
@@ -1236,7 +1237,7 @@ END;
     </para>
 
     <para>
-     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
+     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command> with
      <literal>RETURNING</literal>, <application>PL/pgSQL</application> reports
      an error for more than one returned row, even when
      <literal>STRICT</literal> is not specified.  This is because there
diff --git a/doc/src/sgml/queries.sgml b/doc/src/sgml/queries.sgml
new file mode 100644
index 6986ec5..9e9687b
--- a/doc/src/sgml/queries.sgml
+++ b/doc/src/sgml/queries.sgml
@@ -2062,9 +2062,10 @@ SELECT <replaceable>select_list</replace
    Table Expressions or <acronym>CTE</acronym>s, can be thought of as defining
    temporary tables that exist just for one query.  Each auxiliary statement
    in a <literal>WITH</literal> clause can be a <command>SELECT</command>,
-   <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>; and the
+   <command>INSERT</command>, <command>UPDATE</command>, <command>DELETE</command>,
+   or <command>MERGE</command>; and the
    <literal>WITH</literal> clause itself is attached to a primary statement that can
-   be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
+   also be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
    <command>DELETE</command>, or <command>MERGE</command>.
   </para>
 
@@ -2598,8 +2599,8 @@ SELECT * FROM w AS w1 JOIN w AS w2 ON w1
    <title>Data-Modifying Statements in <literal>WITH</literal></title>
 
    <para>
-    You can use most data-modifying statements (<command>INSERT</command>,
-    <command>UPDATE</command>, or <command>DELETE</command>, but not
+    You can use data-modifying statements (<command>INSERT</command>,
+    <command>UPDATE</command>, <command>DELETE</command>, or
     <command>MERGE</command>) in <literal>WITH</literal>.  This
     allows you to perform several different operations in the same query.
     An example is:
diff --git a/doc/src/sgml/ref/copy.sgml b/doc/src/sgml/ref/copy.sgml
new file mode 100644
index c25b52d..d822470
--- a/doc/src/sgml/ref/copy.sgml
+++ b/doc/src/sgml/ref/copy.sgml
@@ -121,13 +121,15 @@ COPY { <replaceable class="parameter">ta
       A <link linkend="sql-select"><command>SELECT</command></link>,
       <link linkend="sql-values"><command>VALUES</command></link>,
       <link linkend="sql-insert"><command>INSERT</command></link>,
-      <link linkend="sql-update"><command>UPDATE</command></link>, or
-      <link linkend="sql-delete"><command>DELETE</command></link> command whose results are to be
+      <link linkend="sql-update"><command>UPDATE</command></link>,
+      <link linkend="sql-update"><command>DELETE</command></link>, or
+      <link linkend="sql-delete"><command>MERGE</command></link> command whose results are to be
       copied.  Note that parentheses are required around the query.
      </para>
      <para>
-      For <command>INSERT</command>, <command>UPDATE</command> and
-      <command>DELETE</command> queries a RETURNING clause must be provided,
+      For <command>INSERT</command>, <command>UPDATE</command>,
+      <command>DELETE</command>, and <command>MERGE</command> queries a
+      RETURNING clause must be provided,
       and the target relation must not have a conditional rule, nor
       an <literal>ALSO</literal> rule, nor an <literal>INSTEAD</literal> rule
       that expands to multiple statements.
diff --git a/doc/src/sgml/ref/merge.sgml b/doc/src/sgml/ref/merge.sgml
new file mode 100644
index 0995fe0..7eb641a
--- a/doc/src/sgml/ref/merge.sgml
+++ b/doc/src/sgml/ref/merge.sgml
@@ -25,6 +25,7 @@ PostgreSQL documentation
 MERGE INTO [ ONLY ] <replaceable class="parameter">target_table_name</replaceable> [ * ] [ [ AS ] <replaceable class="parameter">target_alias</replaceable> ]
 USING <replaceable class="parameter">data_source</replaceable> ON <replaceable class="parameter">join_condition</replaceable>
 <replaceable class="parameter">when_clause</replaceable> [...]
+[ RETURNING * | <replaceable class="parameter">output_expression</replaceable> [ [ AS ] <replaceable class="parameter">output_name</replaceable> ] [, ...] ]
 
 <phrase>where <replaceable class="parameter">data_source</replaceable> is:</phrase>
 
@@ -95,6 +96,20 @@ DELETE
   </para>
 
   <para>
+   The optional <literal>RETURNING</literal> clause causes <command>MERGE</command>
+   to compute and return value(s) based on each row inserted, updated, or
+   deleted.  Any expression using the source or target table's columns, or the
+   merge support functions <xref linkend="pg-merge-action"/> and
+   <xref linkend="pg-merge-when-clause"/> can be computed.  When an
+   <command>INSERT</command> or <command>UPDATE</command> action is performed,
+   the new values of the target table's columns are used.  When a
+   <command>DELETE</command> is performed, the old values of the target table's
+   columns are used.  The syntax of the <literal>RETURNING</literal> list is
+   identical to that of the output list of <command>SELECT</command>.
+
+  </para>
+
+  <para>
    There is no separate <literal>MERGE</literal> privilege.
    If you specify an update action, you must have the
    <literal>UPDATE</literal> privilege on the column(s)
@@ -125,6 +140,18 @@ DELETE
 
   <variablelist>
    <varlistentry>
+    <term><replaceable class="parameter">with_query</replaceable></term>
+    <listitem>
+     <para>
+      The <literal>WITH</literal> clause allows you to specify one or more
+      subqueries that can be referenced by name in the <command>MERGE</command>
+      query. See <xref linkend="queries-with"/> and <xref linkend="sql-select"/>
+      for details.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
     <term><replaceable class="parameter">target_table_name</replaceable></term>
     <listitem>
      <para>
@@ -399,13 +426,30 @@ DELETE
    </varlistentry>
 
    <varlistentry>
-    <term><replaceable class="parameter">with_query</replaceable></term>
+    <term><replaceable class="parameter">output_expression</replaceable></term>
     <listitem>
      <para>
-      The <literal>WITH</literal> clause allows you to specify one or more
-      subqueries that can be referenced by name in the <command>MERGE</command>
-      query. See <xref linkend="queries-with"/> and <xref linkend="sql-select"/>
-      for details.
+      An expression to be computed and returned by the <command>MERGE</command>
+      command after each row is merged.  The expression can use any columns of
+      the source or target tables, or the merge support functions
+      <xref linkend="pg-merge-action"/> and <xref linkend="pg-merge-when-clause"/>.
+     </para>
+     <para>
+      Writing <literal>*</literal> will return all columns from the source
+      table, followed by all columns from the target table.  Often this will
+      lead to a lot of duplication, since it is common for the source and
+      target tables to have a lot of the same columns.  This can be avoided by
+      qualifying the <literal>*</literal> with the name or alias of the source
+      or target table.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
+    <term><replaceable class="parameter">output_name</replaceable></term>
+    <listitem>
+     <para>
+      A name to use for a returned column.
      </para>
     </listitem>
    </varlistentry>
@@ -428,6 +472,13 @@ MERGE <replaceable class="parameter">tot
    were changed in any way.
   </para>
 
+  <para>
+   If the <command>MERGE</command> command contains a <literal>RETURNING</literal>
+   clause, the result will be similar to that of a <command>SELECT</command>
+   statement containing the columns and values defined in the
+   <literal>RETURNING</literal> list, computed over the row(s) inserted, updated,
+   or deleted by the command.
+  </para>
  </refsect1>
 
  <refsect1>
@@ -544,13 +595,6 @@ MERGE <replaceable class="parameter">tot
   </para>
 
   <para>
-   There is no <literal>RETURNING</literal> clause with
-   <command>MERGE</command>.  Actions of <command>INSERT</command>,
-   <command>UPDATE</command> and <command>DELETE</command> cannot contain
-   <literal>RETURNING</literal> or <literal>WITH</literal> clauses.
-  </para>
-
-  <para>
    When <command>MERGE</command> is run concurrently with other commands
    that modify the target table, the usual transaction isolation rules
    apply; see <xref linkend="transaction-iso"/> for an explanation
@@ -602,7 +646,8 @@ WHEN NOT MATCHED THEN
   <para>
    Attempt to insert a new stock item along with the quantity of stock. If
    the item already exists, instead update the stock count of the existing
-   item. Don't allow entries that have zero stock.
+   item. Don't allow entries that have zero stock. Return details of all
+   changes made.
 <programlisting>
 MERGE INTO wines w
 USING wine_stock_changes s
@@ -612,7 +657,8 @@ WHEN NOT MATCHED AND s.stock_delta > 0 T
 WHEN MATCHED AND w.stock + s.stock_delta > 0 THEN
   UPDATE SET stock = w.stock + s.stock_delta
 WHEN MATCHED THEN
-  DELETE;
+  DELETE
+RETURNING pg_merge_action(), w.*;
 </programlisting>
 
    The <literal>wine_stock_changes</literal> table might be, for example, a
@@ -627,7 +673,8 @@ WHEN MATCHED THEN
     This command conforms to the <acronym>SQL</acronym> standard.
   </para>
    <para>
-    The WITH clause and <literal>DO NOTHING</literal> action are extensions to
+    The WITH clause, <literal>DO NOTHING</literal> action, and
+    <literal>RETURNING</literal> clause are extensions to
     the <acronym>SQL</acronym> standard.
   </para>
  </refsect1>
diff --git a/doc/src/sgml/ref/select.sgml b/doc/src/sgml/ref/select.sgml
new file mode 100644
index 1f9538f..5840ddc
--- a/doc/src/sgml/ref/select.sgml
+++ b/doc/src/sgml/ref/select.sgml
@@ -74,7 +74,7 @@ SELECT [ ALL | DISTINCT [ ON ( <replacea
 
 <phrase>and <replaceable class="parameter">with_query</replaceable> is:</phrase>
 
-    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> )
+    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> | <replaceable class="parameter">merge</replaceable> )
         [ SEARCH { BREADTH | DEPTH } FIRST BY <replaceable>column_name</replaceable> [, ...] SET <replaceable>search_seq_col_name</replaceable> ]
         [ CYCLE <replaceable>column_name</replaceable> [, ...] SET <replaceable>cycle_mark_col_name</replaceable> [ TO <replaceable>cycle_mark_value</replaceable> DEFAULT <replaceable>cycle_mark_default</replaceable> ] USING <replaceable>cycle_path_col_name</replaceable> ]
 
@@ -227,10 +227,10 @@ TABLE [ ONLY ] <replaceable class="param
     The subqueries effectively act as temporary tables or views
     for the duration of the primary query.
     Each subquery can be a <command>SELECT</command>, <command>TABLE</command>, <command>VALUES</command>,
-    <command>INSERT</command>, <command>UPDATE</command> or
-    <command>DELETE</command> statement.
+    <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command> statement.
     When writing a data-modifying statement (<command>INSERT</command>,
-    <command>UPDATE</command> or <command>DELETE</command>) in
+    <command>UPDATE</command>, <command>DELETE</command>, or <command>MERGE</command>) in
     <literal>WITH</literal>, it is usual to include a <literal>RETURNING</literal> clause.
     It is the output of <literal>RETURNING</literal>, <emphasis>not</emphasis> the underlying
     table that the statement modifies, that forms the temporary table that is
@@ -2179,7 +2179,8 @@ SELECT 2+2;
 
    <para>
     <productname>PostgreSQL</productname> allows <command>INSERT</command>,
-    <command>UPDATE</command>, and <command>DELETE</command> to be used as <literal>WITH</literal>
+    <command>UPDATE</command>, <command>DELETE</command>, and
+    <command>MERGE</command> to be used as <literal>WITH</literal>
     queries.  This is not found in the SQL standard.
    </para>
   </refsect2>
diff --git a/doc/src/sgml/rowtypes.sgml b/doc/src/sgml/rowtypes.sgml
new file mode 100644
index 4d86f97..bbeac84
--- a/doc/src/sgml/rowtypes.sgml
+++ b/doc/src/sgml/rowtypes.sgml
@@ -348,7 +348,7 @@ SELECT m.* FROM some_table, LATERAL myfu
    column expansion of this kind when it appears at the top level of
    a <link linkend="queries-select-lists"><command>SELECT</command> output
    list</link>, a <link linkend="dml-returning"><literal>RETURNING</literal>
-   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>,
+   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>,
    a <link linkend="queries-values"><literal>VALUES</literal> clause</link>, or
    a <link linkend="sql-syntax-row-constructors">row constructor</link>.
    In all other contexts (including when nested inside one of those
diff --git a/doc/src/sgml/spi.sgml b/doc/src/sgml/spi.sgml
new file mode 100644
index 651930a..ad0a9b1
--- a/doc/src/sgml/spi.sgml
+++ b/doc/src/sgml/spi.sgml
@@ -301,8 +301,9 @@ SPI_execute("INSERT INTO foo SELECT * FR
    is returned in the global variable <varname>SPI_processed</varname>.
    If the return value of the function is <symbol>SPI_OK_SELECT</symbol>,
    <symbol>SPI_OK_INSERT_RETURNING</symbol>,
-   <symbol>SPI_OK_DELETE_RETURNING</symbol>, or
-   <symbol>SPI_OK_UPDATE_RETURNING</symbol>,
+   <symbol>SPI_OK_DELETE_RETURNING</symbol>,
+   <symbol>SPI_OK_UPDATE_RETURNING</symbol>, or
+   <symbol>SPI_OK_MERGE_RETURNING</symbol>
    then you can use the
    global pointer <literal>SPITupleTable *SPI_tuptable</literal> to
    access the result rows.  Some utility commands (such as
@@ -472,6 +473,15 @@ typedef struct SPITupleTable
       </para>
      </listitem>
     </varlistentry>
+
+    <varlistentry>
+     <term><symbol>SPI_OK_MERGE_RETURNING</symbol></term>
+     <listitem>
+      <para>
+       if a <command>MERGE RETURNING</command> was executed
+      </para>
+     </listitem>
+    </varlistentry>
 
     <varlistentry>
      <term><symbol>SPI_OK_UTILITY</symbol></term>
diff --git a/doc/src/sgml/xfunc.sgml b/doc/src/sgml/xfunc.sgml
new file mode 100644
index 9620ea9..4f68efb
--- a/doc/src/sgml/xfunc.sgml
+++ b/doc/src/sgml/xfunc.sgml
@@ -177,7 +177,8 @@
     statements separated by semicolons.  A semicolon after the last
     statement is optional.  Unless the function is declared to return
     <type>void</type>, the last statement must be a <command>SELECT</command>,
-    or an <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>
+    or an <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command>
     that has a <literal>RETURNING</literal> clause.
    </para>
 
@@ -1236,8 +1237,9 @@ SELECT x, CASE WHEN x &gt; 0 THEN genera
 
     <note>
      <para>
-      If a function's last command is <command>INSERT</command>, <command>UPDATE</command>,
-      or <command>DELETE</command> with <literal>RETURNING</literal>, that command will
+      If a function's last command is <command>INSERT</command>,
+      <command>UPDATE</command>, <command>DELETE</command>, or
+      <command>MERGE</command> with <literal>RETURNING</literal>, that command will
       always be executed to completion, even if the function is not declared
       with <literal>SETOF</literal> or the calling query does not fetch all the
       result rows.  Any extra rows produced by the <literal>RETURNING</literal>
diff --git a/src/backend/commands/copy.c b/src/backend/commands/copy.c
new file mode 100644
index e34f583..aa3cca0
--- a/src/backend/commands/copy.c
+++ b/src/backend/commands/copy.c
@@ -274,12 +274,6 @@ DoCopy(ParseState *pstate, const CopyStm
 	{
 		Assert(stmt->query);
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(stmt->query, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in COPY"));
-
 		query = makeNode(RawStmt);
 		query->stmt = stmt->query;
 		query->stmt_location = stmt_location;
diff --git a/src/backend/commands/copyto.c b/src/backend/commands/copyto.c
new file mode 100644
index 8043b4e..e02d7d0
--- a/src/backend/commands/copyto.c
+++ b/src/backend/commands/copyto.c
@@ -510,7 +510,8 @@ BeginCopyTo(ParseState *pstate,
 		{
 			Assert(query->commandType == CMD_INSERT ||
 				   query->commandType == CMD_UPDATE ||
-				   query->commandType == CMD_DELETE);
+				   query->commandType == CMD_DELETE ||
+				   query->commandType == CMD_MERGE);
 
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
diff --git a/src/backend/executor/execExpr.c b/src/backend/executor/execExpr.c
new file mode 100644
index c61f23c..8ce6a4f
--- a/src/backend/executor/execExpr.c
+++ b/src/backend/executor/execExpr.c
@@ -48,6 +48,7 @@
 #include "utils/array.h"
 #include "utils/builtins.h"
 #include "utils/datum.h"
+#include "utils/fmgroids.h"
 #include "utils/lsyscache.h"
 #include "utils/typcache.h"
 
@@ -2488,6 +2489,22 @@ ExecInitFunc(ExprEvalStep *scratch, Expr
 	InitFunctionCallInfoData(*fcinfo, flinfo,
 							 nargs, inputcollid, NULL, NULL);
 
+	/*
+	 * Merge support functions should only be called directly from a MERGE
+	 * command, and need access to the parent ModifyTableState.  The parser
+	 * should have checked that such functions only appear in the RETURNING
+	 * list of a MERGE, so this should never fail.
+	 */
+	if (IsMergeSupportFunction(funcid))
+	{
+		if (!state->parent ||
+			!IsA(state->parent, ModifyTableState) ||
+			((ModifyTableState *) state->parent)->operation != CMD_MERGE)
+			elog(ERROR, "merge support function called in non-merge context");
+
+		fcinfo->context = (Node *) state->parent;
+	}
+
 	/* Keep extra copies of this info to save an indirection at runtime */
 	scratch->d.func.fn_addr = flinfo->fn_addr;
 	scratch->d.func.nargs = nargs;
diff --git a/src/backend/executor/execPartition.c b/src/backend/executor/execPartition.c
new file mode 100644
index fd6ca8a..95bbf67
--- a/src/backend/executor/execPartition.c
+++ b/src/backend/executor/execPartition.c
@@ -612,8 +612,8 @@ ExecInitPartitionInfo(ModifyTableState *
 	 * Build the RETURNING projection for the partition.  Note that we didn't
 	 * build the returningList for partitions within the planner, but simple
 	 * translation of varattnos will suffice.  This only occurs for the INSERT
-	 * case or in the case of UPDATE tuple routing where we didn't find a
-	 * result rel to reuse.
+	 * case or in the case of UPDATE/MERGE tuple routing where we didn't find
+	 * a result rel to reuse.
 	 */
 	if (node && node->returningLists != NIL)
 	{
@@ -622,12 +622,14 @@ ExecInitPartitionInfo(ModifyTableState *
 		List	   *returningList;
 
 		/* See the comment above for WCO lists. */
-		/* (except no RETURNING support for MERGE yet) */
 		Assert((node->operation == CMD_INSERT &&
 				list_length(node->returningLists) == 1 &&
 				list_length(node->resultRelations) == 1) ||
 			   (node->operation == CMD_UPDATE &&
 				list_length(node->returningLists) ==
+				list_length(node->resultRelations)) ||
+			   (node->operation == CMD_MERGE &&
+				list_length(node->returningLists) ==
 				list_length(node->resultRelations)));
 
 		/*
diff --git a/src/backend/executor/functions.c b/src/backend/executor/functions.c
new file mode 100644
index 50e06ec..7e4717a
--- a/src/backend/executor/functions.c
+++ b/src/backend/executor/functions.c
@@ -1665,8 +1665,8 @@ check_sql_fn_retval(List *queryTreeLists
 
 	/*
 	 * If it's a plain SELECT, it returns whatever the targetlist says.
-	 * Otherwise, if it's INSERT/UPDATE/DELETE with RETURNING, it returns
-	 * that. Otherwise, the function return type must be VOID.
+	 * Otherwise, if it's INSERT/UPDATE/DELETE/MERGE with RETURNING, it
+	 * returns that. Otherwise, the function return type must be VOID.
 	 *
 	 * Note: eventually replace this test with QueryReturnsTuples?	We'd need
 	 * a more general method of determining the output type, though.  Also, it
@@ -1684,7 +1684,8 @@ check_sql_fn_retval(List *queryTreeLists
 	else if (parse &&
 			 (parse->commandType == CMD_INSERT ||
 			  parse->commandType == CMD_UPDATE ||
-			  parse->commandType == CMD_DELETE) &&
+			  parse->commandType == CMD_DELETE ||
+			  parse->commandType == CMD_MERGE) &&
 			 parse->returningList)
 	{
 		tlist = parse->returningList;
diff --git a/src/backend/executor/nodeModifyTable.c b/src/backend/executor/nodeModifyTable.c
new file mode 100644
index 6f0543a..7e98890
--- a/src/backend/executor/nodeModifyTable.c
+++ b/src/backend/executor/nodeModifyTable.c
@@ -36,8 +36,7 @@
  *		RETURNING tuple after completing each row insert, update, or delete.
  *		It must be called again to continue the operation.  Without RETURNING,
  *		we just loop within the node until all the work is done, then
- *		return NULL.  This avoids useless call/return overhead.  (MERGE does
- *		not support RETURNING.)
+ *		return NULL.  This avoids useless call/return overhead.
  */
 
 #include "postgres.h"
@@ -97,9 +96,6 @@ typedef struct ModifyTableContext
 										  TupleTableSlot *oldSlot,
 										  MergeActionState *relaction);
 
-	/* MERGE specific */
-	MergeActionState *relaction;	/* MERGE action in progress */
-
 	/*
 	 * Information about the changes that were made concurrently to a tuple
 	 * being updated or deleted
@@ -172,13 +168,14 @@ static TupleTableSlot *ExecMerge(ModifyT
 								 ItemPointer tupleid,
 								 bool canSetTag);
 static void ExecInitMerge(ModifyTableState *mtstate, EState *estate);
-static bool ExecMergeMatched(ModifyTableContext *context,
-							 ResultRelInfo *resultRelInfo,
-							 ItemPointer tupleid,
-							 bool canSetTag);
-static void ExecMergeNotMatched(ModifyTableContext *context,
-								ResultRelInfo *resultRelInfo,
-								bool canSetTag);
+static TupleTableSlot *ExecMergeMatched(ModifyTableContext *context,
+										ResultRelInfo *resultRelInfo,
+										ItemPointer tupleid,
+										bool canSetTag,
+										bool *matched);
+static TupleTableSlot *ExecMergeNotMatched(ModifyTableContext *context,
+										   ResultRelInfo *resultRelInfo,
+										   bool canSetTag);
 static TupleTableSlot *mergeGetUpdateNewTuple(ResultRelInfo *relinfo,
 											  TupleTableSlot *planSlot,
 											  TupleTableSlot *oldSlot,
@@ -987,7 +984,7 @@ ExecInsert(ModifyTableContext *context,
 		if (mtstate->operation == CMD_UPDATE)
 			wco_kind = WCO_RLS_UPDATE_CHECK;
 		else if (mtstate->operation == CMD_MERGE)
-			wco_kind = (context->relaction->mas_action->commandType == CMD_UPDATE) ?
+			wco_kind = (mtstate->mt_merge_action->mas_action->commandType == CMD_UPDATE) ?
 				WCO_RLS_UPDATE_CHECK : WCO_RLS_INSERT_CHECK;
 		else
 			wco_kind = WCO_RLS_INSERT_CHECK;
@@ -1838,7 +1835,7 @@ ExecCrossPartitionUpdate(ModifyTableCont
 			/* and project the new tuple to retry the UPDATE with */
 			context->cpUpdateRetrySlot =
 				context->GetUpdateNewTuple(resultRelInfo, epqslot, oldSlot,
-										   context->relaction);
+										   mtstate->mt_merge_action);
 			return false;
 		}
 	}
@@ -2051,7 +2048,7 @@ lreplace:
 		 * No luck, a retry is needed.  If running MERGE, we do not do so
 		 * here; instead let it handle that on its own rules.
 		 */
-		if (context->relaction != NULL)
+		if (context->mtstate->mt_merge_action != NULL)
 			return TM_Updated;
 
 		/*
@@ -2689,6 +2686,7 @@ static TupleTableSlot *
 ExecMerge(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 		  ItemPointer tupleid, bool canSetTag)
 {
+	TupleTableSlot *rslot = NULL;
 	bool		matched;
 
 	/*-----
@@ -2736,7 +2734,8 @@ ExecMerge(ModifyTableContext *context, R
 	 */
 	matched = tupleid != NULL;
 	if (matched)
-		matched = ExecMergeMatched(context, resultRelInfo, tupleid, canSetTag);
+		rslot = ExecMergeMatched(context, resultRelInfo, tupleid, canSetTag,
+								 &matched);
 
 	/*
 	 * Either we were dealing with a NOT MATCHED tuple or ExecMergeMatched()
@@ -2744,10 +2743,9 @@ ExecMerge(ModifyTableContext *context, R
 	 * matches.
 	 */
 	if (!matched)
-		ExecMergeNotMatched(context, resultRelInfo, canSetTag);
+		rslot = ExecMergeNotMatched(context, resultRelInfo, canSetTag);
 
-	/* No RETURNING support yet */
-	return NULL;
+	return rslot;
 }
 
 /*
@@ -2757,8 +2755,8 @@ ExecMerge(ModifyTableContext *context, R
  * We start from the first WHEN MATCHED action and check if the WHEN quals
  * pass, if any. If the WHEN quals for the first action do not pass, we
  * check the second, then the third and so on. If we reach to the end, no
- * action is taken and we return true, indicating that no further action is
- * required for this tuple.
+ * action is taken and "matched" is set to true, indicating that no further
+ * action is required for this tuple.
  *
  * If we do find a qualifying action, then we attempt to execute the action.
  *
@@ -2767,15 +2765,16 @@ ExecMerge(ModifyTableContext *context, R
  * with individual actions are evaluated by this routine via ExecQual, while
  * EvalPlanQual checks for the join quals. If EvalPlanQual tells us that the
  * updated tuple still passes the join quals, then we restart from the first
- * action to look for a qualifying action. Otherwise, we return false --
- * meaning that a NOT MATCHED action must now be executed for the current
- * source tuple.
+ * action to look for a qualifying action. Otherwise, "matched" is set to
+ * false -- meaning that a NOT MATCHED action must now be executed for the
+ * current source tuple.
  */
-static bool
+static TupleTableSlot *
 ExecMergeMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
-				 ItemPointer tupleid, bool canSetTag)
+				 ItemPointer tupleid, bool canSetTag, bool *matched)
 {
 	ModifyTableState *mtstate = context->mtstate;
+	TupleTableSlot *rslot = NULL;
 	TupleTableSlot *newslot;
 	EState	   *estate = context->estate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
@@ -2787,7 +2786,10 @@ ExecMergeMatched(ModifyTableContext *con
 	 * If there are no WHEN MATCHED actions, we are done.
 	 */
 	if (resultRelInfo->ri_matchedMergeAction == NIL)
-		return true;
+	{
+		*matched = true;
+		return NULL;
+	}
 
 	/*
 	 * Make tuple and any needed join variables available to ExecQual and
@@ -2866,7 +2868,7 @@ lmerge_matched:
 				 */
 				newslot = ExecProject(relaction->mas_proj);
 
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				context->GetUpdateNewTuple = mergeGetUpdateNewTuple;
 				context->cpUpdateRetrySlot = NULL;
 
@@ -2889,7 +2891,7 @@ lmerge_matched:
 				break;
 
 			case CMD_DELETE:
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				if (!ExecDeletePrologue(context, resultRelInfo, tupleid,
 										NULL, NULL))
 				{
@@ -2949,7 +2951,8 @@ lmerge_matched:
 				 * If the tuple was already deleted, return to let caller
 				 * handle it under NOT MATCHED clauses.
 				 */
-				return false;
+				*matched = false;
+				return NULL;
 
 			case TM_Updated:
 				{
@@ -3016,13 +3019,19 @@ lmerge_matched:
 							 * NOT MATCHED actions.
 							 */
 							if (TupIsNull(epqslot))
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							(void) ExecGetJunkAttribute(epqslot,
 														resultRelInfo->ri_RowIdAttNo,
 														&isNull);
 							if (isNull)
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							/*
 							 * When a tuple was updated and migrated to
@@ -3057,7 +3066,8 @@ lmerge_matched:
 							 * tuple already deleted; tell caller to run NOT
 							 * MATCHED actions
 							 */
-							return false;
+							*matched = false;
+							return NULL;
 
 						case TM_SelfModified:
 
@@ -3077,13 +3087,14 @@ lmerge_matched:
 										(errcode(ERRCODE_TRIGGERED_DATA_CHANGE_VIOLATION),
 										 errmsg("tuple to be updated or deleted was already modified by an operation triggered by the current command"),
 										 errhint("Consider using an AFTER trigger instead of a BEFORE trigger to propagate changes to other rows.")));
-							return false;
+							*matched = false;
+							return NULL;
 
 						default:
 							/* see table_tuple_lock call in ExecDelete() */
 							elog(ERROR, "unexpected table_tuple_lock status: %u",
 								 result);
-							return false;
+							return NULL;
 					}
 				}
 
@@ -3095,6 +3106,30 @@ lmerge_matched:
 				break;
 		}
 
+		/* Process RETURNING if present */
+		if (resultRelInfo->ri_projectReturning)
+		{
+			switch (commandType)
+			{
+				case CMD_UPDATE:
+					rslot = ExecProcessReturning(resultRelInfo, newslot,
+												 context->planSlot);
+					break;
+
+				case CMD_DELETE:
+					rslot = ExecProcessReturning(resultRelInfo,
+												 resultRelInfo->ri_oldTupleSlot,
+												 context->planSlot);
+					break;
+
+				case CMD_NOTHING:
+					break;
+
+				default:
+					elog(ERROR, "unknown action in MERGE WHEN MATCHED clause");
+			}
+		}
+
 		/*
 		 * We've activated one of the WHEN clauses, so we don't search
 		 * further. This is required behaviour, not an optimization.
@@ -3105,19 +3140,22 @@ lmerge_matched:
 	/*
 	 * Successfully executed an action or no qualifying action was found.
 	 */
-	return true;
+	*matched = true;
+
+	return rslot;
 }
 
 /*
  * Execute the first qualifying NOT MATCHED action.
  */
-static void
+static TupleTableSlot *
 ExecMergeNotMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 					bool canSetTag)
 {
 	ModifyTableState *mtstate = context->mtstate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
 	List	   *actionStates = NIL;
+	TupleTableSlot *rslot = NULL;
 	ListCell   *l;
 
 	/*
@@ -3167,10 +3205,10 @@ ExecMergeNotMatched(ModifyTableContext *
 				 * so we don't need to map the tuple here.
 				 */
 				newslot = ExecProject(action->mas_proj);
-				context->relaction = action;
+				mtstate->mt_merge_action = action;
 
-				(void) ExecInsert(context, mtstate->rootResultRelInfo, newslot,
-								  canSetTag, NULL, NULL);
+				rslot = ExecInsert(context, mtstate->rootResultRelInfo,
+								   newslot, canSetTag, NULL, NULL);
 				mtstate->mt_merge_inserted += 1;
 				break;
 			case CMD_NOTHING:
@@ -3186,6 +3224,8 @@ ExecMergeNotMatched(ModifyTableContext *
 		 */
 		break;
 	}
+
+	return rslot;
 }
 
 /*
@@ -3382,6 +3422,64 @@ mergeGetUpdateNewTuple(ResultRelInfo *re
 }
 
 /*
+ * pg_merge_action() -
+ *	  SQL merge support function to retrieve the currently executing merge
+ *	  action command string ("INSERT", "UPDATE", or "DELETE").
+ */
+Datum
+pg_merge_action(PG_FUNCTION_ARGS)
+{
+	ModifyTableState *mtstate = (ModifyTableState *) fcinfo->context;
+	MergeActionState *relaction;
+
+	if (!mtstate || mtstate->operation != CMD_MERGE)
+		elog(ERROR, "merge support function called in non-merge context");
+
+	relaction = mtstate->mt_merge_action;
+	if (relaction)
+	{
+		CmdType		commandType = relaction->mas_action->commandType;
+
+		switch (commandType)
+		{
+			case CMD_INSERT:
+				PG_RETURN_TEXT_P(cstring_to_text("INSERT"));
+			case CMD_UPDATE:
+				PG_RETURN_TEXT_P(cstring_to_text("UPDATE"));
+			case CMD_DELETE:
+				PG_RETURN_TEXT_P(cstring_to_text("DELETE"));
+			case CMD_NOTHING:
+				PG_RETURN_NULL();
+			default:
+				elog(ERROR, "unrecognized commandType: %d", (int) commandType);
+		}
+	}
+
+	PG_RETURN_NULL();
+}
+
+/*
+ * pg_merge_when_clause() -
+ *	  SQL merge support function to retrieve the 1-based index of the
+ *	  currently executing merge WHEN clause.
+ */
+Datum
+pg_merge_when_clause(PG_FUNCTION_ARGS)
+{
+	ModifyTableState *mtstate = (ModifyTableState *) fcinfo->context;
+	MergeActionState *relaction;
+
+	if (!mtstate || mtstate->operation != CMD_MERGE)
+		elog(ERROR, "merge support function called in non-merge context");
+
+	relaction = mtstate->mt_merge_action;
+	if (relaction)
+		PG_RETURN_INT32((int32) relaction->mas_action->index);
+
+	PG_RETURN_NULL();
+}
+
+/*
  * Process BEFORE EACH STATEMENT triggers
  */
 static void
@@ -3667,8 +3765,17 @@ ExecModifyTable(PlanState *pstate)
 				{
 					EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-					ExecMerge(&context, node->resultRelInfo, NULL, node->canSetTag);
-					continue;	/* no RETURNING support yet */
+					slot = ExecMerge(&context, node->resultRelInfo, NULL,
+									 node->canSetTag);
+
+					/*
+					 * If we got a RETURNING result, return it to the caller.
+					 * We'll continue the work on next call.
+					 */
+					if (slot)
+						return slot;
+
+					continue;	/* continue with the next tuple */
 				}
 
 				elog(ERROR, "tableoid is NULL");
@@ -3745,8 +3852,17 @@ ExecModifyTable(PlanState *pstate)
 					{
 						EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-						ExecMerge(&context, node->resultRelInfo, NULL, node->canSetTag);
-						continue;	/* no RETURNING support yet */
+						slot = ExecMerge(&context, node->resultRelInfo, NULL,
+										 node->canSetTag);
+
+						/*
+						 * If we got a RETURNING result, return it to the
+						 * caller.  We'll continue the work on next call.
+						 */
+						if (slot)
+							return slot;
+
+						continue;	/* continue with the next tuple */
 					}
 
 					elog(ERROR, "ctid is NULL");
@@ -3839,7 +3955,7 @@ ExecModifyTable(PlanState *pstate)
 				slot = internalGetUpdateNewTuple(resultRelInfo, context.planSlot,
 												 oldSlot, NULL);
 				context.GetUpdateNewTuple = internalGetUpdateNewTuple;
-				context.relaction = NULL;
+				node->mt_merge_action = NULL;
 
 				/* Now apply the update. */
 				slot = ExecUpdate(&context, resultRelInfo, tupleid, oldtuple,
diff --git a/src/backend/executor/spi.c b/src/backend/executor/spi.c
new file mode 100644
index e3a170c..be88056
--- a/src/backend/executor/spi.c
+++ b/src/backend/executor/spi.c
@@ -2033,6 +2033,8 @@ SPI_result_code_string(int code)
 			return "SPI_OK_TD_REGISTER";
 		case SPI_OK_MERGE:
 			return "SPI_OK_MERGE";
+		case SPI_OK_MERGE_RETURNING:
+			return "SPI_OK_MERGE_RETURNING";
 	}
 	/* Unrecognized code ... return something useful ... */
 	sprintf(buf, "Unrecognized SPI code %d", code);
@@ -2886,7 +2888,10 @@ _SPI_pquery(QueryDesc *queryDesc, bool f
 				res = SPI_OK_UPDATE;
 			break;
 		case CMD_MERGE:
-			res = SPI_OK_MERGE;
+			if (queryDesc->plannedstmt->hasReturning)
+				res = SPI_OK_MERGE_RETURNING;
+			else
+				res = SPI_OK_MERGE;
 			break;
 		default:
 			return SPI_ERROR_OPUNKNOWN;
diff --git a/src/backend/parser/analyze.c b/src/backend/parser/analyze.c
new file mode 100644
index e892df9..fe0f5cb
--- a/src/backend/parser/analyze.c
+++ b/src/backend/parser/analyze.c
@@ -73,7 +73,6 @@ static void determineRecursiveColTypes(P
 									   Node *larg, List *nrtargetlist);
 static Query *transformReturnStmt(ParseState *pstate, ReturnStmt *stmt);
 static Query *transformUpdateStmt(ParseState *pstate, UpdateStmt *stmt);
-static List *transformReturningList(ParseState *pstate, List *returningList);
 static Query *transformPLAssignStmt(ParseState *pstate,
 									PLAssignStmt *stmt);
 static Query *transformDeclareCursorStmt(ParseState *pstate,
@@ -514,7 +513,8 @@ transformDeleteStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -941,7 +941,8 @@ transformInsertStmt(ParseState *pstate,
 	/* Process RETURNING, if any. */
 	if (stmt->returningList)
 		qry->returningList = transformReturningList(pstate,
-													stmt->returningList);
+													stmt->returningList,
+													EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -2406,7 +2407,8 @@ transformUpdateStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/*
 	 * Now we are done with SELECT-like processing, and can get on with
@@ -2500,10 +2502,11 @@ transformUpdateTargetList(ParseState *ps
 
 /*
  * transformReturningList -
- *	handle a RETURNING clause in INSERT/UPDATE/DELETE
+ *	handle a RETURNING clause in INSERT/UPDATE/DELETE/MERGE
  */
-static List *
-transformReturningList(ParseState *pstate, List *returningList)
+List *
+transformReturningList(ParseState *pstate, List *returningList,
+					   ParseExprKind exprKind)
 {
 	List	   *rlist;
 	int			save_next_resno;
@@ -2520,7 +2523,7 @@ transformReturningList(ParseState *pstat
 	pstate->p_next_resno = 1;
 
 	/* transform RETURNING identically to a SELECT targetlist */
-	rlist = transformTargetList(pstate, returningList, EXPR_KIND_RETURNING);
+	rlist = transformTargetList(pstate, returningList, exprKind);
 
 	/*
 	 * Complain if the nonempty tlist expanded to nothing (which is possible
diff --git a/src/backend/parser/gram.y b/src/backend/parser/gram.y
new file mode 100644
index a013838..4406430
--- a/src/backend/parser/gram.y
+++ b/src/backend/parser/gram.y
@@ -12241,6 +12241,7 @@ MergeStmt:
 			USING table_ref
 			ON a_expr
 			merge_when_list
+			returning_clause
 				{
 					MergeStmt  *m = makeNode(MergeStmt);
 
@@ -12249,6 +12250,7 @@ MergeStmt:
 					m->sourceRelation = $6;
 					m->joinCondition = $8;
 					m->mergeWhenClauses = $9;
+					m->returningList = $10;
 
 					$$ = (Node *) m;
 				}
diff --git a/src/backend/parser/parse_agg.c b/src/backend/parser/parse_agg.c
new file mode 100644
index 4fbf80c..b8d655a
--- a/src/backend/parser/parse_agg.c
+++ b/src/backend/parser/parse_agg.c
@@ -456,6 +456,7 @@ check_agglevels_and_constraints(ParseSta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
@@ -904,6 +905,7 @@ transformWindowFuncCall(ParseState *psta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_cte.c b/src/backend/parser/parse_cte.c
new file mode 100644
index c5b1a49..ad3c525
--- a/src/backend/parser/parse_cte.c
+++ b/src/backend/parser/parse_cte.c
@@ -126,13 +126,6 @@ transformWithClause(ParseState *pstate,
 		CommonTableExpr *cte = (CommonTableExpr *) lfirst(lc);
 		ListCell   *rest;
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(cte->ctequery, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in WITH query"),
-					parser_errposition(pstate, cte->location));
-
 		for_each_cell(rest, withClause->ctes, lnext(withClause->ctes, lc))
 		{
 			CommonTableExpr *cte2 = (CommonTableExpr *) lfirst(rest);
@@ -153,7 +146,8 @@ transformWithClause(ParseState *pstate,
 			/* must be a data-modifying statement */
 			Assert(IsA(cte->ctequery, InsertStmt) ||
 				   IsA(cte->ctequery, UpdateStmt) ||
-				   IsA(cte->ctequery, DeleteStmt));
+				   IsA(cte->ctequery, DeleteStmt) ||
+				   IsA(cte->ctequery, MergeStmt));
 
 			pstate->p_hasModifyingCTE = true;
 		}
diff --git a/src/backend/parser/parse_expr.c b/src/backend/parser/parse_expr.c
new file mode 100644
index 7ff41ac..92cd8c1
--- a/src/backend/parser/parse_expr.c
+++ b/src/backend/parser/parse_expr.c
@@ -482,6 +482,7 @@ transformColumnRef(ParseState *pstate, C
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CHECK_CONSTRAINT:
@@ -1708,6 +1709,7 @@ transformSubLink(ParseState *pstate, Sub
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CYCLE_MARK:
@@ -3002,6 +3004,7 @@ ParseExprKindName(ParseExprKind exprKind
 		case EXPR_KIND_OFFSET:
 			return "OFFSET";
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			return "RETURNING";
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
diff --git a/src/backend/parser/parse_func.c b/src/backend/parser/parse_func.c
new file mode 100644
index ca14f06..914624e
--- a/src/backend/parser/parse_func.c
+++ b/src/backend/parser/parse_func.c
@@ -31,6 +31,7 @@
 #include "parser/parse_target.h"
 #include "parser/parse_type.h"
 #include "utils/builtins.h"
+#include "utils/fmgroids.h"
 #include "utils/lsyscache.h"
 #include "utils/syscache.h"
 
@@ -348,6 +349,15 @@ ParseFuncOrColumn(ParseState *pstate, Li
 					 parser_errposition(pstate, location)));
 	}
 
+	/* Merge support functions are only allowed in MERGE's RETURNING list */
+	if (IsMergeSupportFunction(funcid) &&
+		pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+		ereport(ERROR,
+				(errcode(ERRCODE_WRONG_OBJECT_TYPE),
+				 errmsg("merge support function %s can only be called from the RETURNING list of a MERGE command",
+						NameListToString(funcname)),
+				 parser_errposition(pstate, location)));
+
 	/*
 	 * So far so good, so do some fdresult-type-specific processing.
 	 */
@@ -2602,6 +2612,7 @@ check_srf_call_placement(ParseState *pst
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_merge.c b/src/backend/parser/parse_merge.c
new file mode 100644
index d886637..2e50eb0
--- a/src/backend/parser/parse_merge.c
+++ b/src/backend/parser/parse_merge.c
@@ -233,6 +233,10 @@ transformMergeStmt(ParseState *pstate, M
 	 */
 	qry->jointree = makeFromExpr(pstate->p_joinlist, joinExpr);
 
+	/* Transform the RETURNING list, if any */
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_MERGE_RETURNING);
+
 	/*
 	 * We now have a good query shape, so now look at the WHEN conditions and
 	 * action targetlists.
@@ -254,6 +258,7 @@ transformMergeStmt(ParseState *pstate, M
 		MergeAction *action;
 
 		action = makeNode(MergeAction);
+		action->index = foreach_current_index(l) + 1;
 		action->commandType = mergeWhenClause->commandType;
 		action->matched = mergeWhenClause->matched;
 
@@ -390,9 +395,6 @@ transformMergeStmt(ParseState *pstate, M
 
 	qry->mergeActionList = mergeActionList;
 
-	/* RETURNING could potentially be added in the future, but not in SQL std */
-	qry->returningList = NULL;
-
 	qry->hasTargetSRFs = false;
 	qry->hasSubLinks = pstate->p_hasSubLinks;
 
diff --git a/src/backend/parser/parse_relation.c b/src/backend/parser/parse_relation.c
new file mode 100644
index de355dd..97ca651
--- a/src/backend/parser/parse_relation.c
+++ b/src/backend/parser/parse_relation.c
@@ -2345,9 +2345,10 @@ addRangeTableEntryForCTE(ParseState *pst
 		cte->cterefcount++;
 
 	/*
-	 * We throw error if the CTE is INSERT/UPDATE/DELETE without RETURNING.
-	 * This won't get checked in case of a self-reference, but that's OK
-	 * because data-modifying CTEs aren't allowed to be recursive anyhow.
+	 * We throw error if the CTE is INSERT/UPDATE/DELETE/MERGE without
+	 * RETURNING.  This won't get checked in case of a self-reference, but
+	 * that's OK because data-modifying CTEs aren't allowed to be recursive
+	 * anyhow.
 	 */
 	if (IsA(cte->ctequery, Query))
 	{
diff --git a/src/backend/rewrite/rewriteHandler.c b/src/backend/rewrite/rewriteHandler.c
new file mode 100644
index a614e3f..b337e15
--- a/src/backend/rewrite/rewriteHandler.c
+++ b/src/backend/rewrite/rewriteHandler.c
@@ -3639,7 +3639,8 @@ RewriteQuery(Query *parsetree, List *rew
 			if (!(ctequery->commandType == CMD_SELECT ||
 				  ctequery->commandType == CMD_UPDATE ||
 				  ctequery->commandType == CMD_INSERT ||
-				  ctequery->commandType == CMD_DELETE))
+				  ctequery->commandType == CMD_DELETE ||
+				  ctequery->commandType == CMD_MERGE))
 			{
 				/*
 				 * Currently it could only be NOTIFY; this error message will
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
new file mode 100644
index c7d9d96..d393f51
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -2131,11 +2131,10 @@ QueryReturnsTuples(Query *parsetree)
 		case CMD_SELECT:
 			/* returns tuples */
 			return true;
-		case CMD_MERGE:
-			return false;
 		case CMD_INSERT:
 		case CMD_UPDATE:
 		case CMD_DELETE:
+		case CMD_MERGE:
 			/* the forms with RETURNING return tuples */
 			if (parsetree->returningList)
 				return true;
diff --git a/src/bin/psql/common.c b/src/bin/psql/common.c
new file mode 100644
index f907f5d..85eb868
--- a/src/bin/psql/common.c
+++ b/src/bin/psql/common.c
@@ -962,13 +962,17 @@ PrintQueryResult(PGresult *result, bool
 			else
 				success = true;
 
-			/* if it's INSERT/UPDATE/DELETE RETURNING, also print status */
+			/*
+			 * If it's INSERT/UPDATE/DELETE/MERGE RETURNING, also print
+			 * status.
+			 */
 			if (last || pset.show_all_results)
 			{
 				cmdstatus = PQcmdStatus(result);
 				if (strncmp(cmdstatus, "INSERT", 6) == 0 ||
 					strncmp(cmdstatus, "UPDATE", 6) == 0 ||
-					strncmp(cmdstatus, "DELETE", 6) == 0)
+					strncmp(cmdstatus, "DELETE", 6) == 0 ||
+					strncmp(cmdstatus, "MERGE", 5) == 0)
 					PrintQueryStatus(result, printStatusFout);
 			}
 
diff --git a/src/include/catalog/pg_proc.dat b/src/include/catalog/pg_proc.dat
new file mode 100644
index e2a7642..50c9b59
--- a/src/include/catalog/pg_proc.dat
+++ b/src/include/catalog/pg_proc.dat
@@ -11935,4 +11935,14 @@
   proname => 'any_value_transfn', prorettype => 'anyelement',
   proargtypes => 'anyelement anyelement', prosrc => 'any_value_transfn' },
 
+# MERGE support functions
+{ oid => '9499', descr => 'command type of current MERGE action',
+  proname => 'pg_merge_action',  provolatile => 'v',
+  prorettype => 'text', proargtypes => '',
+  prosrc => 'pg_merge_action' },
+{ oid => '9500', descr => 'index of current MERGE WHEN clause',
+  proname => 'pg_merge_when_clause',  provolatile => 'v',
+  prorettype => 'int4', proargtypes => '',
+  prosrc => 'pg_merge_when_clause' },
+
 ]
diff --git a/src/include/catalog/pg_proc.h b/src/include/catalog/pg_proc.h
new file mode 100644
index e7abe0b..e597826
--- a/src/include/catalog/pg_proc.h
+++ b/src/include/catalog/pg_proc.h
@@ -182,6 +182,11 @@ DECLARE_UNIQUE_INDEX(pg_proc_proname_arg
 #define PROARGMODE_VARIADIC 'v'
 #define PROARGMODE_TABLE	't'
 
+/* Is this a merge support function?  (Requires fmgroids.h) */
+#define IsMergeSupportFunction(oid) \
+	((oid) == F_PG_MERGE_ACTION || \
+	 (oid) == F_PG_MERGE_WHEN_CLAUSE)
+
 #endif							/* EXPOSE_TO_CLIENT_CODE */
 
 
diff --git a/src/include/executor/spi.h b/src/include/executor/spi.h
new file mode 100644
index d1de139..ed8787d
--- a/src/include/executor/spi.h
+++ b/src/include/executor/spi.h
@@ -97,6 +97,7 @@ typedef struct _SPI_plan *SPIPlanPtr;
 #define SPI_OK_REL_UNREGISTER	16
 #define SPI_OK_TD_REGISTER		17
 #define SPI_OK_MERGE			18
+#define SPI_OK_MERGE_RETURNING	19
 
 #define SPI_OPT_NONATOMIC		(1 << 0)
 
diff --git a/src/include/nodes/execnodes.h b/src/include/nodes/execnodes.h
new file mode 100644
index 20f4c8b..3d62fb9
--- a/src/include/nodes/execnodes.h
+++ b/src/include/nodes/execnodes.h
@@ -1302,6 +1302,9 @@ typedef struct ModifyTableState
 	/* Flags showing which subcommands are present INS/UPD/DEL/DO NOTHING */
 	int			mt_merge_subcommands;
 
+	/* For MERGE, the action currently being executed */
+	MergeActionState *mt_merge_action;
+
 	/* tuple counters for MERGE */
 	double		mt_merge_inserted;
 	double		mt_merge_updated;
diff --git a/src/include/nodes/parsenodes.h b/src/include/nodes/parsenodes.h
new file mode 100644
index f7d7f10..a959ac7
--- a/src/include/nodes/parsenodes.h
+++ b/src/include/nodes/parsenodes.h
@@ -1687,6 +1687,7 @@ typedef struct MergeWhenClause
 typedef struct MergeAction
 {
 	NodeTag		type;
+	int			index;			/* 1-based index of the clause */
 	bool		matched;		/* true=MATCHED, false=NOT MATCHED */
 	CmdType		commandType;	/* INSERT/UPDATE/DELETE/DO NOTHING */
 	/* OVERRIDING clause */
@@ -1806,6 +1807,7 @@ typedef struct MergeStmt
 	Node	   *sourceRelation; /* source relation */
 	Node	   *joinCondition;	/* join condition between source and target */
 	List	   *mergeWhenClauses;	/* list of MergeWhenClause(es) */
+	List	   *returningList;	/* list of expressions to return */
 	WithClause *withClause;		/* WITH clause */
 } MergeStmt;
 
diff --git a/src/include/parser/analyze.h b/src/include/parser/analyze.h
new file mode 100644
index 1cef183..78eb55e
--- a/src/include/parser/analyze.h
+++ b/src/include/parser/analyze.h
@@ -44,6 +44,8 @@ extern List *transformInsertRow(ParseSta
 								bool strip_indirection);
 extern List *transformUpdateTargetList(ParseState *pstate,
 									   List *origTlist);
+extern List *transformReturningList(ParseState *pstate, List *returningList,
+									ParseExprKind exprKind);
 extern Query *transformTopLevelStmt(ParseState *pstate, RawStmt *parseTree);
 extern Query *transformStmt(ParseState *pstate, Node *parseTree);
 
diff --git a/src/include/parser/parse_node.h b/src/include/parser/parse_node.h
new file mode 100644
index f589112..836819d
--- a/src/include/parser/parse_node.h
+++ b/src/include/parser/parse_node.h
@@ -61,7 +61,8 @@ typedef enum ParseExprKind
 	EXPR_KIND_DISTINCT_ON,		/* DISTINCT ON */
 	EXPR_KIND_LIMIT,			/* LIMIT */
 	EXPR_KIND_OFFSET,			/* OFFSET */
-	EXPR_KIND_RETURNING,		/* RETURNING */
+	EXPR_KIND_RETURNING,		/* RETURNING in INSERT/UPDATE/DELETE */
+	EXPR_KIND_MERGE_RETURNING,	/* RETURNING in MERGE */
 	EXPR_KIND_VALUES,			/* VALUES */
 	EXPR_KIND_VALUES_SINGLE,	/* single-row VALUES (in INSERT only) */
 	EXPR_KIND_CHECK_CONSTRAINT, /* CHECK constraint for a table */
diff --git a/src/pl/plpgsql/src/pl_exec.c b/src/pl/plpgsql/src/pl_exec.c
new file mode 100644
index ffd6d2e..502056d
--- a/src/pl/plpgsql/src/pl_exec.c
+++ b/src/pl/plpgsql/src/pl_exec.c
@@ -4246,9 +4246,9 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 	/*
 	 * If we have INTO, then we only need one row back ... but if we have INTO
 	 * STRICT or extra check too_many_rows, ask for two rows, so that we can
-	 * verify the statement returns only one.  INSERT/UPDATE/DELETE are always
-	 * treated strictly. Without INTO, just run the statement to completion
-	 * (tcount = 0).
+	 * verify the statement returns only one.  INSERT/UPDATE/DELETE/MERGE are
+	 * always treated strictly. Without INTO, just run the statement to
+	 * completion (tcount = 0).
 	 *
 	 * We could just ask for two rows always when using INTO, but there are
 	 * some cases where demanding the extra row costs significant time, eg by
@@ -4286,10 +4286,11 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 			Assert(stmt->mod_stmt);
 			exec_set_found(estate, (SPI_processed != 0));
 			break;
@@ -4468,10 +4469,11 @@ exec_stmt_dynexecute(PLpgSQL_execstate *
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 		case SPI_OK_UTILITY:
 		case SPI_OK_REWRITTEN:
 			break;
diff --git a/src/pl/tcl/pltcl.c b/src/pl/tcl/pltcl.c
new file mode 100644
index 499a9ea..58c4f50
--- a/src/pl/tcl/pltcl.c
+++ b/src/pl/tcl/pltcl.c
@@ -2459,6 +2459,7 @@ pltcl_process_SPI_result(Tcl_Interp *int
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
+		case SPI_OK_MERGE_RETURNING:
 
 			/*
 			 * Process the tuples we got
diff --git a/src/test/regress/expected/merge.out b/src/test/regress/expected/merge.out
new file mode 100644
index e1d3639..ef6dd55
--- a/src/test/regress/expected/merge.out
+++ b/src/test/regress/expected/merge.out
@@ -123,20 +123,20 @@ ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
 ERROR:  name "target" specified more than once
 DETAIL:  The name is used both as MERGE target table and data source.
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
-ERROR:  MERGE not supported in WITH query
-LINE 1: WITH foo AS (
-             ^
--- used in COPY
+ERROR:  WITH query "foo" does not have a RETURNING clause
+LINE 4: ) SELECT * FROM foo;
+                        ^
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) TO stdout;
-ERROR:  MERGE not supported in COPY
+ERROR:  COPY query must have a RETURNING clause
 -- unsupported relation types
 -- view
 CREATE VIEW tv AS SELECT * FROM target;
@@ -1289,20 +1289,187 @@ WHEN MATCHED AND tid < 2 THEN
 ROLLBACK;
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
-ERROR:  syntax error at or near "RETURNING"
-LINE 10: RETURNING *;
-         ^
+    DELETE
+RETURNING pg_merge_when_clause() AS when_clause,
+          pg_merge_action() AS merge_action,
+          t.*,
+          CASE pg_merge_action()
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ when_clause | merge_action | tid | balance |     description     
+-------------+--------------+-----+---------+---------------------
+           3 | DELETE       |   1 |     100 | Removed (1,100)
+           1 | UPDATE       |   2 |     220 | Added 20 to balance
+           2 | INSERT       |   4 |      40 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- error when using MERGE support functions outside MERGE
+SELECT pg_merge_action() FROM sq_target;
+ERROR:  merge support function pg_merge_action can only be called from the RETURNING list of a MERGE command
+LINE 1: SELECT pg_merge_action() FROM sq_target;
+               ^
+UPDATE sq_target SET balance = balance + 1 RETURNING pg_merge_when_clause();
+ERROR:  merge support function pg_merge_when_clause can only be called from the RETURNING list of a MERGE command
+LINE 1: ...ATE sq_target SET balance = balance + 1 RETURNING pg_merge_w...
+                                                             ^
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_when_clause() AS when_clause,
+              pg_merge_action() AS merge_action,
+              t.*,
+              CASE pg_merge_action()
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING merge_action, pg_merge_action() AS merge_log_action, l.*
+)
+SELECT * FROM m2;
+ merge_action | merge_log_action | tid |     last_change     
+--------------+------------------+-----+---------------------
+ DELETE       | UPDATE           |   1 | Removed (1,100)
+ UPDATE       | INSERT           |   2 | Added 20 to balance
+ INSERT       | INSERT           |   4 | Inserted (4,40)
+(3 rows)
+
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ tid |     last_change     
+-----+---------------------
+   1 | Removed (1,100)
+   2 | Added 20 to balance
+   4 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING v
+    ON tid = sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*
+) TO stdout;
+DELETE	1	100
+ROLLBACK;
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ action | tid | new_balance 
+--------+-----+-------------
+ DELETE |   1 |         100
+ UPDATE |   3 |         320
+ INSERT |   4 |         110
+(3 rows)
+
+ROLLBACK;
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ action | tid | balance 
+--------+-----+---------
+ DELETE |   1 |     100
+ INSERT |   4 |      40
+(2 rows)
+
+ROLLBACK;
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ r_action | r_tid | r_balance 
+----------+-------+-----------
+ DELETE   |     1 |       100
+ UPDATE   |     3 |       320
+ INSERT   |     4 |       110
+(3 rows)
+
 ROLLBACK;
 -- EXPLAIN
 CREATE TABLE ex_mtarget (a int, b int)
@@ -1499,7 +1666,7 @@ SELECT * FROM sq_target WHERE tid = 1;
 (1 row)
 
 ROLLBACK;
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 NOTICE:  drop cascades to view v
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1625,6 +1792,32 @@ SELECT * FROM pa_target ORDER BY tid;
 (14 rows)
 
 ROLLBACK;
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
+ pg_merge_action | pg_merge_when_clause | tid | balance |           val            
+-----------------+----------------------+-----+---------+--------------------------
+ UPDATE          |                    1 |   2 |     110 | initial updated by merge
+(1 row)
+
+SELECT * FROM pa_target ORDER BY tid;
+ tid | balance |           val            
+-----+---------+--------------------------
+   2 |     110 | initial updated by merge
+   3 |     300 | initial
+   5 |     500 | initial
+   7 |     700 | initial
+   9 |     900 | initial
+  11 |    1100 | initial
+  13 |    1300 | initial
+(7 rows)
+
+ROLLBACK;
 DROP TABLE pa_target CASCADE;
 -- The target table is partitioned in the same way, but this time by attaching
 -- partitions which have columns in different order, dropped columns etc.
@@ -1716,7 +1909,26 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (sid, delta, 'inserted by merge');
+    INSERT VALUES (sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
+ pg_merge_action | pg_merge_when_clause | tid | balance |           val            
+-----------------+----------------------+-----+---------+--------------------------
+ UPDATE          |                    1 |   2 |     110 | initial updated by merge
+ INSERT          |                    2 |   2 |      20 | inserted by merge
+ UPDATE          |                    1 |   4 |     330 | initial updated by merge
+ INSERT          |                    2 |   4 |      40 | inserted by merge
+ UPDATE          |                    1 |   6 |     550 | initial updated by merge
+ INSERT          |                    2 |   6 |      60 | inserted by merge
+ UPDATE          |                    1 |   8 |     770 | initial updated by merge
+ INSERT          |                    2 |   8 |      80 | inserted by merge
+ UPDATE          |                    1 |  10 |     990 | initial updated by merge
+ INSERT          |                    2 |  10 |     100 | inserted by merge
+ UPDATE          |                    1 |  12 |    1210 | initial updated by merge
+ INSERT          |                    2 |  12 |     120 | inserted by merge
+ UPDATE          |                    1 |  14 |    1430 | initial updated by merge
+ INSERT          |                    2 |  14 |     140 | inserted by merge
+(14 rows)
+
 SELECT * FROM pa_target ORDER BY tid;
  tid | balance |           val            
 -----+---------+--------------------------
@@ -1783,7 +1995,21 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
+ pg_merge_action | pg_merge_when_clause |          logts           | tid | balance |           val            
+-----------------+----------------------+--------------------------+-----+---------+--------------------------
+ UPDATE          |                    1 | Tue Jan 31 00:00:00 2017 |   1 |     110 | initial updated by merge
+ UPDATE          |                    1 | Tue Feb 28 00:00:00 2017 |   2 |     220 | initial updated by merge
+ INSERT          |                    2 | Sun Jan 15 00:00:00 2017 |   3 |      30 | inserted by merge
+ UPDATE          |                    1 | Tue Jan 31 00:00:00 2017 |   4 |     440 | initial updated by merge
+ UPDATE          |                    1 | Tue Feb 28 00:00:00 2017 |   5 |     550 | initial updated by merge
+ INSERT          |                    2 | Sun Jan 15 00:00:00 2017 |   6 |      60 | inserted by merge
+ UPDATE          |                    1 | Tue Jan 31 00:00:00 2017 |   7 |     770 | initial updated by merge
+ UPDATE          |                    1 | Tue Feb 28 00:00:00 2017 |   8 |     880 | initial updated by merge
+ INSERT          |                    2 | Sun Jan 15 00:00:00 2017 |   9 |      90 | inserted by merge
+(9 rows)
+
 SELECT * FROM pa_target ORDER BY tid;
           logts           | tid | balance |           val            
 --------------------------+-----+---------+--------------------------
diff --git a/src/test/regress/sql/merge.sql b/src/test/regress/sql/merge.sql
new file mode 100644
index f085416..dba6a19
--- a/src/test/regress/sql/merge.sql
+++ b/src/test/regress/sql/merge.sql
@@ -86,12 +86,12 @@ MERGE INTO target
 USING target
 ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
--- used in COPY
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
@@ -844,17 +844,145 @@ ROLLBACK;
 
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
+    DELETE
+RETURNING pg_merge_when_clause() AS when_clause,
+          pg_merge_action() AS merge_action,
+          t.*,
+          CASE pg_merge_action()
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ROLLBACK;
+
+-- error when using MERGE support functions outside MERGE
+SELECT pg_merge_action() FROM sq_target;
+UPDATE sq_target SET balance = balance + 1 RETURNING pg_merge_when_clause();
+
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_when_clause() AS when_clause,
+              pg_merge_action() AS merge_action,
+              t.*,
+              CASE pg_merge_action()
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING merge_action, pg_merge_action() AS merge_log_action, l.*
+)
+SELECT * FROM m2;
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ROLLBACK;
+
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING v
+    ON tid = sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*
+) TO stdout;
+ROLLBACK;
+
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ROLLBACK;
+
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ROLLBACK;
+
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
 ROLLBACK;
 
 -- EXPLAIN
@@ -953,7 +1081,7 @@ WHEN MATCHED THEN
 SELECT * FROM sq_target WHERE tid = 1;
 ROLLBACK;
 
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1020,6 +1148,17 @@ SELECT merge_func();
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
+SELECT * FROM pa_target ORDER BY tid;
+ROLLBACK;
+
 DROP TABLE pa_target CASCADE;
 
 -- The target table is partitioned in the same way, but this time by attaching
@@ -1078,7 +1217,8 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (sid, delta, 'inserted by merge');
+    INSERT VALUES (sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
@@ -1132,7 +1272,8 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 

diff --git a/doc/src/sgml/dml.sgml b/doc/src/sgml/dml.sgml
new file mode 100644
index cbbc5e2..ff2a827
--- a/doc/src/sgml/dml.sgml
+++ b/doc/src/sgml/dml.sgml
@@ -283,10 +283,15 @@ DELETE FROM products;
    <secondary>RETURNING</secondary>
   </indexterm>
 
+  <indexterm zone="dml-returning">
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
   <para>
    Sometimes it is useful to obtain data from modified rows while they are
    being manipulated.  The <command>INSERT</command>, <command>UPDATE</command>,
-   and <command>DELETE</command> commands all have an
+   <command>DELETE</command>, and <command>MERGE</command> commands all have an
    optional <literal>RETURNING</literal> clause that supports this.  Use
    of <literal>RETURNING</literal> avoids performing an extra database query to
    collect the data, and is especially valuable when it would otherwise be
@@ -339,6 +344,24 @@ DELETE FROM products
 </programlisting>
   </para>
 
+  <para>
+   In a <command>MERGE</command>, the data available to <literal>RETURNING</literal> is
+   the content of the source row plus the content of the inserted, updated, or
+   deleted target row.  In addition, the merge support functions
+   <xref linkend="pg-merge-action"/> and <xref linkend="pg-merge-when-clause"/>
+   may be used to return information about which merge action was executed for
+   each row.  Since it is quite common for the source and target to have many
+   of the same columns, specifying <literal>RETURNING *</literal> can lead to a
+   lot of duplicated columns, so it is often more useful to just return the
+   target row.  For example:
+<programlisting>
+MERGE INTO products p USING new_products n ON p.product_no = n.product_no
+  WHEN NOT MATCHED THEN INSERT VALUES (n.product_no, n.name, n.price)
+  WHEN MATCHED THEN UPDATE SET name = n.name, price = n.price
+  RETURNING pg_merge_action(), p.*;
+</programlisting>
+  </para>
+
   <para>
    If there are triggers (<xref linkend="triggers"/>) on the target table,
    the data available to <literal>RETURNING</literal> is the row as modified by
diff --git a/doc/src/sgml/func.sgml b/doc/src/sgml/func.sgml
new file mode 100644
index 9c6107f..d07f90a
--- a/doc/src/sgml/func.sgml
+++ b/doc/src/sgml/func.sgml
@@ -21216,6 +21216,99 @@ SELECT count(*) FROM sometable;
 
  </sect1>
 
+ <sect1 id="functions-merge-support">
+  <title>Merge Support Functions</title>
+
+  <indexterm>
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
+  <para>
+   The merge support functions shown in
+   <xref linkend="functions-merge-support-table"/> may be used in the
+   <literal>RETURNING</literal> list of a <xref linkend="sql-merge"/> command
+   to return additional information about the action taken for each row.
+  </para>
+
+  <table id="functions-merge-support-table">
+    <title>Merge Support Functions</title>
+
+    <tgroup cols="1">
+     <thead>
+      <row>
+       <entry role="func_table_entry"><para role="func_signature">
+        Function
+       </para>
+       <para>
+        Description
+       </para></entry>
+      </row>
+     </thead>
+
+     <tbody>
+      <row>
+       <entry role="func_table_entry"><para role="func_signature">
+        <indexterm id="pg-merge-action">
+         <primary>pg_merge_action</primary>
+        </indexterm>
+        <function>pg_merge_action</function> ( )
+        <returnvalue>text</returnvalue>
+       </para>
+       <para>
+        Returns the action performed on the current row (<literal>'INSERT'</literal>,
+        <literal>'UPDATE'</literal>, or <literal>'DELETE'</literal>).
+       </para></entry>
+      </row>
+
+      <row>
+       <entry role="func_table_entry"><para role="func_signature">
+        <indexterm id="pg-merge-when-clause">
+         <primary>pg_merge_when_clause</primary>
+        </indexterm>
+        <function>pg_merge_when_clause</function> ( )
+        <returnvalue>integer</returnvalue>
+       </para>
+       <para>
+        Returns the 1-based index of the <literal>WHEN</literal> clause executed
+        for the current row.
+       </para></entry>
+      </row>
+     </tbody>
+    </tgroup>
+  </table>
+
+  <para>
+   Example:
+<screen><![CDATA[
+MERGE INTO products p
+  USING stock s ON p.product_id = s.product_id
+  WHEN MATCHED AND s.quantity > 0 THEN
+    UPDATE SET in_stock = true, quantity = s.quantity
+  WHEN MATCHED THEN
+    UPDATE SET in_stock = false, quantity = 0
+  WHEN NOT MATCHED THEN
+    INSERT (product_id, in_stock, quantity)
+      VALUES (s.product_id, true, s.quantity)
+  RETURNING pg_merge_action() AS action,
+            pg_merge_when_clause() AS when_clause,
+            p.*;
+
+ action | when_clause | product_id | in_stock | quantity 
+--------+-------------+------------+----------+----------
+ UPDATE |           1 |       1001 | t        |       50
+ UPDATE |           2 |       1002 | f        |        0
+ INSERT |           3 |       1003 | t        |       10
+]]></screen>
+  </para>
+
+  <para>
+   Note that it is an error to use these functions anywhere other than the
+   <literal>RETURNING</literal> list of a <command>MERGE</command> command.
+  </para>
+
+ </sect1>
+
  <sect1 id="functions-subquery">
   <title>Subquery Expressions</title>
 
diff --git a/doc/src/sgml/glossary.sgml b/doc/src/sgml/glossary.sgml
new file mode 100644
index 7c01a54..4317cfc
--- a/doc/src/sgml/glossary.sgml
+++ b/doc/src/sgml/glossary.sgml
@@ -1323,9 +1323,9 @@
      to a <glossterm linkend="glossary-client">client</glossterm> upon the
      completion of an <acronym>SQL</acronym> command, usually a
      <command>SELECT</command> but it can be an
-     <command>INSERT</command>, <command>UPDATE</command>, or
-     <command>DELETE</command> command if the <literal>RETURNING</literal>
-     clause is specified.
+     <command>INSERT</command>, <command>UPDATE</command>,
+     <command>DELETE</command>, or <command>MERGE</command> command if the
+     <literal>RETURNING</literal> clause is specified.
     </para>
     <para>
      The fact that a result set is a relation means that a query can be used
diff --git a/doc/src/sgml/plpgsql.sgml b/doc/src/sgml/plpgsql.sgml
new file mode 100644
index 7c8a49f..b839695
--- a/doc/src/sgml/plpgsql.sgml
+++ b/doc/src/sgml/plpgsql.sgml
@@ -1020,8 +1020,8 @@ INSERT INTO mytable VALUES (1,'one'), (2
     </para>
 
     <para>
-     If the command does return rows (for example <command>SELECT</command>,
-     or <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>
+     If the command does return rows (for example <command>SELECT</command>, or
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
      with <literal>RETURNING</literal>), there are two ways to proceed.
      When the command will return at most one row, or you only care about
      the first row of output, write the command as usual but add
@@ -1149,6 +1149,7 @@ SELECT <replaceable>select_expressions</
 INSERT ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 UPDATE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 DELETE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
+MERGE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 </synopsis>
 
      where <replaceable>target</replaceable> can be a record variable, a row
@@ -1159,8 +1160,8 @@ DELETE ... RETURNING <replaceable>expres
      <literal>INTO</literal> clause) just as described above,
      and the plan is cached in the same way.
      This works for <command>SELECT</command>,
-     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
-     <literal>RETURNING</literal>, and certain utility commands
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
+     with <literal>RETURNING</literal>, and certain utility commands
      that return row sets, such as <command>EXPLAIN</command>.
      Except for the <literal>INTO</literal> clause, the SQL command is the same
      as it would be written outside <application>PL/pgSQL</application>.
@@ -1236,7 +1237,7 @@ END;
     </para>
 
     <para>
-     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
+     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command> with
      <literal>RETURNING</literal>, <application>PL/pgSQL</application> reports
      an error for more than one returned row, even when
      <literal>STRICT</literal> is not specified.  This is because there
diff --git a/doc/src/sgml/queries.sgml b/doc/src/sgml/queries.sgml
new file mode 100644
index 6986ec5..9e9687b
--- a/doc/src/sgml/queries.sgml
+++ b/doc/src/sgml/queries.sgml
@@ -2062,9 +2062,10 @@ SELECT <replaceable>select_list</replace
    Table Expressions or <acronym>CTE</acronym>s, can be thought of as defining
    temporary tables that exist just for one query.  Each auxiliary statement
    in a <literal>WITH</literal> clause can be a <command>SELECT</command>,
-   <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>; and the
+   <command>INSERT</command>, <command>UPDATE</command>, <command>DELETE</command>,
+   or <command>MERGE</command>; and the
    <literal>WITH</literal> clause itself is attached to a primary statement that can
-   be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
+   also be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
    <command>DELETE</command>, or <command>MERGE</command>.
   </para>
 
@@ -2598,8 +2599,8 @@ SELECT * FROM w AS w1 JOIN w AS w2 ON w1
    <title>Data-Modifying Statements in <literal>WITH</literal></title>
 
    <para>
-    You can use most data-modifying statements (<command>INSERT</command>,
-    <command>UPDATE</command>, or <command>DELETE</command>, but not
+    You can use data-modifying statements (<command>INSERT</command>,
+    <command>UPDATE</command>, <command>DELETE</command>, or
     <command>MERGE</command>) in <literal>WITH</literal>.  This
     allows you to perform several different operations in the same query.
     An example is:
diff --git a/doc/src/sgml/ref/copy.sgml b/doc/src/sgml/ref/copy.sgml
new file mode 100644
index c25b52d..d822470
--- a/doc/src/sgml/ref/copy.sgml
+++ b/doc/src/sgml/ref/copy.sgml
@@ -121,13 +121,15 @@ COPY { <replaceable class="parameter">ta
       A <link linkend="sql-select"><command>SELECT</command></link>,
       <link linkend="sql-values"><command>VALUES</command></link>,
       <link linkend="sql-insert"><command>INSERT</command></link>,
-      <link linkend="sql-update"><command>UPDATE</command></link>, or
-      <link linkend="sql-delete"><command>DELETE</command></link> command whose results are to be
+      <link linkend="sql-update"><command>UPDATE</command></link>,
+      <link linkend="sql-update"><command>DELETE</command></link>, or
+      <link linkend="sql-delete"><command>MERGE</command></link> command whose results are to be
       copied.  Note that parentheses are required around the query.
      </para>
      <para>
-      For <command>INSERT</command>, <command>UPDATE</command> and
-      <command>DELETE</command> queries a RETURNING clause must be provided,
+      For <command>INSERT</command>, <command>UPDATE</command>,
+      <command>DELETE</command>, and <command>MERGE</command> queries a
+      RETURNING clause must be provided,
       and the target relation must not have a conditional rule, nor
       an <literal>ALSO</literal> rule, nor an <literal>INSTEAD</literal> rule
       that expands to multiple statements.
diff --git a/doc/src/sgml/ref/merge.sgml b/doc/src/sgml/ref/merge.sgml
new file mode 100644
index 0995fe0..7eb641a
--- a/doc/src/sgml/ref/merge.sgml
+++ b/doc/src/sgml/ref/merge.sgml
@@ -25,6 +25,7 @@ PostgreSQL documentation
 MERGE INTO [ ONLY ] <replaceable class="parameter">target_table_name</replaceable> [ * ] [ [ AS ] <replaceable class="parameter">target_alias</replaceable> ]
 USING <replaceable class="parameter">data_source</replaceable> ON <replaceable class="parameter">join_condition</replaceable>
 <replaceable class="parameter">when_clause</replaceable> [...]
+[ RETURNING * | <replaceable class="parameter">output_expression</replaceable> [ [ AS ] <replaceable class="parameter">output_name</replaceable> ] [, ...] ]
 
 <phrase>where <replaceable class="parameter">data_source</replaceable> is:</phrase>
 
@@ -95,6 +96,20 @@ DELETE
   </para>
 
   <para>
+   The optional <literal>RETURNING</literal> clause causes <command>MERGE</command>
+   to compute and return value(s) based on each row inserted, updated, or
+   deleted.  Any expression using the source or target table's columns, or the
+   merge support functions <xref linkend="pg-merge-action"/> and
+   <xref linkend="pg-merge-when-clause"/> can be computed.  When an
+   <command>INSERT</command> or <command>UPDATE</command> action is performed,
+   the new values of the target table's columns are used.  When a
+   <command>DELETE</command> is performed, the old values of the target table's
+   columns are used.  The syntax of the <literal>RETURNING</literal> list is
+   identical to that of the output list of <command>SELECT</command>.
+
+  </para>
+
+  <para>
    There is no separate <literal>MERGE</literal> privilege.
    If you specify an update action, you must have the
    <literal>UPDATE</literal> privilege on the column(s)
@@ -125,6 +140,18 @@ DELETE
 
   <variablelist>
    <varlistentry>
+    <term><replaceable class="parameter">with_query</replaceable></term>
+    <listitem>
+     <para>
+      The <literal>WITH</literal> clause allows you to specify one or more
+      subqueries that can be referenced by name in the <command>MERGE</command>
+      query. See <xref linkend="queries-with"/> and <xref linkend="sql-select"/>
+      for details.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
     <term><replaceable class="parameter">target_table_name</replaceable></term>
     <listitem>
      <para>
@@ -399,13 +426,30 @@ DELETE
    </varlistentry>
 
    <varlistentry>
-    <term><replaceable class="parameter">with_query</replaceable></term>
+    <term><replaceable class="parameter">output_expression</replaceable></term>
     <listitem>
      <para>
-      The <literal>WITH</literal> clause allows you to specify one or more
-      subqueries that can be referenced by name in the <command>MERGE</command>
-      query. See <xref linkend="queries-with"/> and <xref linkend="sql-select"/>
-      for details.
+      An expression to be computed and returned by the <command>MERGE</command>
+      command after each row is merged.  The expression can use any columns of
+      the source or target tables, or the merge support functions
+      <xref linkend="pg-merge-action"/> and <xref linkend="pg-merge-when-clause"/>.
+     </para>
+     <para>
+      Writing <literal>*</literal> will return all columns from the source
+      table, followed by all columns from the target table.  Often this will
+      lead to a lot of duplication, since it is common for the source and
+      target tables to have a lot of the same columns.  This can be avoided by
+      qualifying the <literal>*</literal> with the name or alias of the source
+      or target table.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
+    <term><replaceable class="parameter">output_name</replaceable></term>
+    <listitem>
+     <para>
+      A name to use for a returned column.
      </para>
     </listitem>
    </varlistentry>
@@ -428,6 +472,13 @@ MERGE <replaceable class="parameter">tot
    were changed in any way.
   </para>
 
+  <para>
+   If the <command>MERGE</command> command contains a <literal>RETURNING</literal>
+   clause, the result will be similar to that of a <command>SELECT</command>
+   statement containing the columns and values defined in the
+   <literal>RETURNING</literal> list, computed over the row(s) inserted, updated,
+   or deleted by the command.
+  </para>
  </refsect1>
 
  <refsect1>
@@ -544,13 +595,6 @@ MERGE <replaceable class="parameter">tot
   </para>
 
   <para>
-   There is no <literal>RETURNING</literal> clause with
-   <command>MERGE</command>.  Actions of <command>INSERT</command>,
-   <command>UPDATE</command> and <command>DELETE</command> cannot contain
-   <literal>RETURNING</literal> or <literal>WITH</literal> clauses.
-  </para>
-
-  <para>
    When <command>MERGE</command> is run concurrently with other commands
    that modify the target table, the usual transaction isolation rules
    apply; see <xref linkend="transaction-iso"/> for an explanation
@@ -602,7 +646,8 @@ WHEN NOT MATCHED THEN
   <para>
    Attempt to insert a new stock item along with the quantity of stock. If
    the item already exists, instead update the stock count of the existing
-   item. Don't allow entries that have zero stock.
+   item. Don't allow entries that have zero stock. Return details of all
+   changes made.
 <programlisting>
 MERGE INTO wines w
 USING wine_stock_changes s
@@ -612,7 +657,8 @@ WHEN NOT MATCHED AND s.stock_delta > 0 T
 WHEN MATCHED AND w.stock + s.stock_delta > 0 THEN
   UPDATE SET stock = w.stock + s.stock_delta
 WHEN MATCHED THEN
-  DELETE;
+  DELETE
+RETURNING pg_merge_action(), w.*;
 </programlisting>
 
    The <literal>wine_stock_changes</literal> table might be, for example, a
@@ -627,7 +673,8 @@ WHEN MATCHED THEN
     This command conforms to the <acronym>SQL</acronym> standard.
   </para>
    <para>
-    The WITH clause and <literal>DO NOTHING</literal> action are extensions to
+    The WITH clause, <literal>DO NOTHING</literal> action, and
+    <literal>RETURNING</literal> clause are extensions to
     the <acronym>SQL</acronym> standard.
   </para>
  </refsect1>
diff --git a/doc/src/sgml/ref/select.sgml b/doc/src/sgml/ref/select.sgml
new file mode 100644
index 1f9538f..5840ddc
--- a/doc/src/sgml/ref/select.sgml
+++ b/doc/src/sgml/ref/select.sgml
@@ -74,7 +74,7 @@ SELECT [ ALL | DISTINCT [ ON ( <replacea
 
 <phrase>and <replaceable class="parameter">with_query</replaceable> is:</phrase>
 
-    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> )
+    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> | <replaceable class="parameter">merge</replaceable> )
         [ SEARCH { BREADTH | DEPTH } FIRST BY <replaceable>column_name</replaceable> [, ...] SET <replaceable>search_seq_col_name</replaceable> ]
         [ CYCLE <replaceable>column_name</replaceable> [, ...] SET <replaceable>cycle_mark_col_name</replaceable> [ TO <replaceable>cycle_mark_value</replaceable> DEFAULT <replaceable>cycle_mark_default</replaceable> ] USING <replaceable>cycle_path_col_name</replaceable> ]
 
@@ -227,10 +227,10 @@ TABLE [ ONLY ] <replaceable class="param
     The subqueries effectively act as temporary tables or views
     for the duration of the primary query.
     Each subquery can be a <command>SELECT</command>, <command>TABLE</command>, <command>VALUES</command>,
-    <command>INSERT</command>, <command>UPDATE</command> or
-    <command>DELETE</command> statement.
+    <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command> statement.
     When writing a data-modifying statement (<command>INSERT</command>,
-    <command>UPDATE</command> or <command>DELETE</command>) in
+    <command>UPDATE</command>, <command>DELETE</command>, or <command>MERGE</command>) in
     <literal>WITH</literal>, it is usual to include a <literal>RETURNING</literal> clause.
     It is the output of <literal>RETURNING</literal>, <emphasis>not</emphasis> the underlying
     table that the statement modifies, that forms the temporary table that is
@@ -2179,7 +2179,8 @@ SELECT 2+2;
 
    <para>
     <productname>PostgreSQL</productname> allows <command>INSERT</command>,
-    <command>UPDATE</command>, and <command>DELETE</command> to be used as <literal>WITH</literal>
+    <command>UPDATE</command>, <command>DELETE</command>, and
+    <command>MERGE</command> to be used as <literal>WITH</literal>
     queries.  This is not found in the SQL standard.
    </para>
   </refsect2>
diff --git a/doc/src/sgml/rowtypes.sgml b/doc/src/sgml/rowtypes.sgml
new file mode 100644
index 4d86f97..bbeac84
--- a/doc/src/sgml/rowtypes.sgml
+++ b/doc/src/sgml/rowtypes.sgml
@@ -348,7 +348,7 @@ SELECT m.* FROM some_table, LATERAL myfu
    column expansion of this kind when it appears at the top level of
    a <link linkend="queries-select-lists"><command>SELECT</command> output
    list</link>, a <link linkend="dml-returning"><literal>RETURNING</literal>
-   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>,
+   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>,
    a <link linkend="queries-values"><literal>VALUES</literal> clause</link>, or
    a <link linkend="sql-syntax-row-constructors">row constructor</link>.
    In all other contexts (including when nested inside one of those
diff --git a/doc/src/sgml/spi.sgml b/doc/src/sgml/spi.sgml
new file mode 100644
index 651930a..ad0a9b1
--- a/doc/src/sgml/spi.sgml
+++ b/doc/src/sgml/spi.sgml
@@ -301,8 +301,9 @@ SPI_execute("INSERT INTO foo SELECT * FR
    is returned in the global variable <varname>SPI_processed</varname>.
    If the return value of the function is <symbol>SPI_OK_SELECT</symbol>,
    <symbol>SPI_OK_INSERT_RETURNING</symbol>,
-   <symbol>SPI_OK_DELETE_RETURNING</symbol>, or
-   <symbol>SPI_OK_UPDATE_RETURNING</symbol>,
+   <symbol>SPI_OK_DELETE_RETURNING</symbol>,
+   <symbol>SPI_OK_UPDATE_RETURNING</symbol>, or
+   <symbol>SPI_OK_MERGE_RETURNING</symbol>
    then you can use the
    global pointer <literal>SPITupleTable *SPI_tuptable</literal> to
    access the result rows.  Some utility commands (such as
@@ -472,6 +473,15 @@ typedef struct SPITupleTable
       </para>
      </listitem>
     </varlistentry>
+
+    <varlistentry>
+     <term><symbol>SPI_OK_MERGE_RETURNING</symbol></term>
+     <listitem>
+      <para>
+       if a <command>MERGE RETURNING</command> was executed
+      </para>
+     </listitem>
+    </varlistentry>
 
     <varlistentry>
      <term><symbol>SPI_OK_UTILITY</symbol></term>
diff --git a/doc/src/sgml/xfunc.sgml b/doc/src/sgml/xfunc.sgml
new file mode 100644
index 9620ea9..4f68efb
--- a/doc/src/sgml/xfunc.sgml
+++ b/doc/src/sgml/xfunc.sgml
@@ -177,7 +177,8 @@
     statements separated by semicolons.  A semicolon after the last
     statement is optional.  Unless the function is declared to return
     <type>void</type>, the last statement must be a <command>SELECT</command>,
-    or an <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>
+    or an <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command>
     that has a <literal>RETURNING</literal> clause.
    </para>
 
@@ -1236,8 +1237,9 @@ SELECT x, CASE WHEN x &gt; 0 THEN genera
 
     <note>
      <para>
-      If a function's last command is <command>INSERT</command>, <command>UPDATE</command>,
-      or <command>DELETE</command> with <literal>RETURNING</literal>, that command will
+      If a function's last command is <command>INSERT</command>,
+      <command>UPDATE</command>, <command>DELETE</command>, or
+      <command>MERGE</command> with <literal>RETURNING</literal>, that command will
       always be executed to completion, even if the function is not declared
       with <literal>SETOF</literal> or the calling query does not fetch all the
       result rows.  Any extra rows produced by the <literal>RETURNING</literal>
diff --git a/src/backend/commands/copy.c b/src/backend/commands/copy.c
new file mode 100644
index f3bbd91..3ca7d85
--- a/src/backend/commands/copy.c
+++ b/src/backend/commands/copy.c
@@ -277,12 +277,6 @@ DoCopy(ParseState *pstate, const CopyStm
 	{
 		Assert(stmt->query);
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(stmt->query, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in COPY"));
-
 		query = makeNode(RawStmt);
 		query->stmt = stmt->query;
 		query->stmt_location = stmt_location;
diff --git a/src/backend/commands/copyto.c b/src/backend/commands/copyto.c
new file mode 100644
index beea1ac..19b7f9f
--- a/src/backend/commands/copyto.c
+++ b/src/backend/commands/copyto.c
@@ -510,7 +510,8 @@ BeginCopyTo(ParseState *pstate,
 		{
 			Assert(query->commandType == CMD_INSERT ||
 				   query->commandType == CMD_UPDATE ||
-				   query->commandType == CMD_DELETE);
+				   query->commandType == CMD_DELETE ||
+				   query->commandType == CMD_MERGE);
 
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
diff --git a/src/backend/executor/execExpr.c b/src/backend/executor/execExpr.c
new file mode 100644
index c61f23c..8ce6a4f
--- a/src/backend/executor/execExpr.c
+++ b/src/backend/executor/execExpr.c
@@ -48,6 +48,7 @@
 #include "utils/array.h"
 #include "utils/builtins.h"
 #include "utils/datum.h"
+#include "utils/fmgroids.h"
 #include "utils/lsyscache.h"
 #include "utils/typcache.h"
 
@@ -2488,6 +2489,22 @@ ExecInitFunc(ExprEvalStep *scratch, Expr
 	InitFunctionCallInfoData(*fcinfo, flinfo,
 							 nargs, inputcollid, NULL, NULL);
 
+	/*
+	 * Merge support functions should only be called directly from a MERGE
+	 * command, and need access to the parent ModifyTableState.  The parser
+	 * should have checked that such functions only appear in the RETURNING
+	 * list of a MERGE, so this should never fail.
+	 */
+	if (IsMergeSupportFunction(funcid))
+	{
+		if (!state->parent ||
+			!IsA(state->parent, ModifyTableState) ||
+			((ModifyTableState *) state->parent)->operation != CMD_MERGE)
+			elog(ERROR, "merge support function called in non-merge context");
+
+		fcinfo->context = (Node *) state->parent;
+	}
+
 	/* Keep extra copies of this info to save an indirection at runtime */
 	scratch->d.func.fn_addr = flinfo->fn_addr;
 	scratch->d.func.nargs = nargs;
diff --git a/src/backend/executor/execPartition.c b/src/backend/executor/execPartition.c
new file mode 100644
index fd6ca8a..95bbf67
--- a/src/backend/executor/execPartition.c
+++ b/src/backend/executor/execPartition.c
@@ -612,8 +612,8 @@ ExecInitPartitionInfo(ModifyTableState *
 	 * Build the RETURNING projection for the partition.  Note that we didn't
 	 * build the returningList for partitions within the planner, but simple
 	 * translation of varattnos will suffice.  This only occurs for the INSERT
-	 * case or in the case of UPDATE tuple routing where we didn't find a
-	 * result rel to reuse.
+	 * case or in the case of UPDATE/MERGE tuple routing where we didn't find
+	 * a result rel to reuse.
 	 */
 	if (node && node->returningLists != NIL)
 	{
@@ -622,12 +622,14 @@ ExecInitPartitionInfo(ModifyTableState *
 		List	   *returningList;
 
 		/* See the comment above for WCO lists. */
-		/* (except no RETURNING support for MERGE yet) */
 		Assert((node->operation == CMD_INSERT &&
 				list_length(node->returningLists) == 1 &&
 				list_length(node->resultRelations) == 1) ||
 			   (node->operation == CMD_UPDATE &&
 				list_length(node->returningLists) ==
+				list_length(node->resultRelations)) ||
+			   (node->operation == CMD_MERGE &&
+				list_length(node->returningLists) ==
 				list_length(node->resultRelations)));
 
 		/*
diff --git a/src/backend/executor/functions.c b/src/backend/executor/functions.c
new file mode 100644
index 50e06ec..7e4717a
--- a/src/backend/executor/functions.c
+++ b/src/backend/executor/functions.c
@@ -1665,8 +1665,8 @@ check_sql_fn_retval(List *queryTreeLists
 
 	/*
 	 * If it's a plain SELECT, it returns whatever the targetlist says.
-	 * Otherwise, if it's INSERT/UPDATE/DELETE with RETURNING, it returns
-	 * that. Otherwise, the function return type must be VOID.
+	 * Otherwise, if it's INSERT/UPDATE/DELETE/MERGE with RETURNING, it
+	 * returns that. Otherwise, the function return type must be VOID.
 	 *
 	 * Note: eventually replace this test with QueryReturnsTuples?	We'd need
 	 * a more general method of determining the output type, though.  Also, it
@@ -1684,7 +1684,8 @@ check_sql_fn_retval(List *queryTreeLists
 	else if (parse &&
 			 (parse->commandType == CMD_INSERT ||
 			  parse->commandType == CMD_UPDATE ||
-			  parse->commandType == CMD_DELETE) &&
+			  parse->commandType == CMD_DELETE ||
+			  parse->commandType == CMD_MERGE) &&
 			 parse->returningList)
 	{
 		tlist = parse->returningList;
diff --git a/src/backend/executor/nodeModifyTable.c b/src/backend/executor/nodeModifyTable.c
new file mode 100644
index 3fa2b93..5b08d2e
--- a/src/backend/executor/nodeModifyTable.c
+++ b/src/backend/executor/nodeModifyTable.c
@@ -36,8 +36,7 @@
  *		RETURNING tuple after completing each row insert, update, or delete.
  *		It must be called again to continue the operation.  Without RETURNING,
  *		we just loop within the node until all the work is done, then
- *		return NULL.  This avoids useless call/return overhead.  (MERGE does
- *		not support RETURNING.)
+ *		return NULL.  This avoids useless call/return overhead.
  */
 
 #include "postgres.h"
@@ -88,9 +87,6 @@ typedef struct ModifyTableContext
 	 */
 	TupleTableSlot *planSlot;
 
-	/* MERGE specific */
-	MergeActionState *relaction;	/* MERGE action in progress */
-
 	/*
 	 * Information about the changes that were made concurrently to a tuple
 	 * being updated or deleted
@@ -153,13 +149,14 @@ static TupleTableSlot *ExecMerge(ModifyT
 								 ItemPointer tupleid,
 								 bool canSetTag);
 static void ExecInitMerge(ModifyTableState *mtstate, EState *estate);
-static bool ExecMergeMatched(ModifyTableContext *context,
-							 ResultRelInfo *resultRelInfo,
-							 ItemPointer tupleid,
-							 bool canSetTag);
-static void ExecMergeNotMatched(ModifyTableContext *context,
-								ResultRelInfo *resultRelInfo,
-								bool canSetTag);
+static TupleTableSlot *ExecMergeMatched(ModifyTableContext *context,
+										ResultRelInfo *resultRelInfo,
+										ItemPointer tupleid,
+										bool canSetTag,
+										bool *matched);
+static TupleTableSlot *ExecMergeNotMatched(ModifyTableContext *context,
+										   ResultRelInfo *resultRelInfo,
+										   bool canSetTag);
 
 
 /*
@@ -980,7 +977,7 @@ ExecInsert(ModifyTableContext *context,
 		if (mtstate->operation == CMD_UPDATE)
 			wco_kind = WCO_RLS_UPDATE_CHECK;
 		else if (mtstate->operation == CMD_MERGE)
-			wco_kind = (context->relaction->mas_action->commandType == CMD_UPDATE) ?
+			wco_kind = (mtstate->mt_merge_action->mas_action->commandType == CMD_UPDATE) ?
 				WCO_RLS_UPDATE_CHECK : WCO_RLS_INSERT_CHECK;
 		else
 			wco_kind = WCO_RLS_INSERT_CHECK;
@@ -1820,7 +1817,7 @@ ExecCrossPartitionUpdate(ModifyTableCont
 		 * additional rechecking, and might end up executing a different
 		 * action entirely).
 		 */
-		if (context->relaction != NULL)
+		if (mtstate->mt_merge_action != NULL)
 			return false;
 		else if (TupIsNull(epqslot))
 			return true;
@@ -2061,7 +2058,7 @@ lreplace:
 		 * No luck, a retry is needed.  If running MERGE, we do not do so
 		 * here; instead let it handle that on its own rules.
 		 */
-		if (context->relaction != NULL)
+		if (context->mtstate->mt_merge_action != NULL)
 			return TM_Updated;
 
 		/*
@@ -2703,6 +2700,7 @@ static TupleTableSlot *
 ExecMerge(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 		  ItemPointer tupleid, bool canSetTag)
 {
+	TupleTableSlot *rslot = NULL;
 	bool		matched;
 
 	/*-----
@@ -2750,7 +2748,8 @@ ExecMerge(ModifyTableContext *context, R
 	 */
 	matched = tupleid != NULL;
 	if (matched)
-		matched = ExecMergeMatched(context, resultRelInfo, tupleid, canSetTag);
+		rslot = ExecMergeMatched(context, resultRelInfo, tupleid, canSetTag,
+								 &matched);
 
 	/*
 	 * Either we were dealing with a NOT MATCHED tuple or ExecMergeMatched()
@@ -2758,10 +2757,9 @@ ExecMerge(ModifyTableContext *context, R
 	 * matches.
 	 */
 	if (!matched)
-		ExecMergeNotMatched(context, resultRelInfo, canSetTag);
+		rslot = ExecMergeNotMatched(context, resultRelInfo, canSetTag);
 
-	/* No RETURNING support yet */
-	return NULL;
+	return rslot;
 }
 
 /*
@@ -2771,8 +2769,8 @@ ExecMerge(ModifyTableContext *context, R
  * We start from the first WHEN MATCHED action and check if the WHEN quals
  * pass, if any. If the WHEN quals for the first action do not pass, we
  * check the second, then the third and so on. If we reach to the end, no
- * action is taken and we return true, indicating that no further action is
- * required for this tuple.
+ * action is taken and "matched" is set to true, indicating that no further
+ * action is required for this tuple.
  *
  * If we do find a qualifying action, then we attempt to execute the action.
  *
@@ -2781,15 +2779,16 @@ ExecMerge(ModifyTableContext *context, R
  * with individual actions are evaluated by this routine via ExecQual, while
  * EvalPlanQual checks for the join quals. If EvalPlanQual tells us that the
  * updated tuple still passes the join quals, then we restart from the first
- * action to look for a qualifying action. Otherwise, we return false --
- * meaning that a NOT MATCHED action must now be executed for the current
- * source tuple.
+ * action to look for a qualifying action. Otherwise, "matched" is set to
+ * false -- meaning that a NOT MATCHED action must now be executed for the
+ * current source tuple.
  */
-static bool
+static TupleTableSlot *
 ExecMergeMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
-				 ItemPointer tupleid, bool canSetTag)
+				 ItemPointer tupleid, bool canSetTag, bool *matched)
 {
 	ModifyTableState *mtstate = context->mtstate;
+	TupleTableSlot *rslot = NULL;
 	TupleTableSlot *newslot;
 	EState	   *estate = context->estate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
@@ -2801,7 +2800,10 @@ ExecMergeMatched(ModifyTableContext *con
 	 * If there are no WHEN MATCHED actions, we are done.
 	 */
 	if (resultRelInfo->ri_matchedMergeAction == NIL)
-		return true;
+	{
+		*matched = true;
+		return NULL;
+	}
 
 	/*
 	 * Make tuple and any needed join variables available to ExecQual and
@@ -2879,12 +2881,15 @@ lmerge_matched:
 				 */
 				newslot = ExecProject(relaction->mas_proj);
 
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				if (!ExecUpdatePrologue(context, resultRelInfo,
 										tupleid, NULL, newslot, &result))
 				{
 					if (result == TM_Ok)
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 					break;		/* concurrent update/delete */
 				}
 				result = ExecUpdateAct(context, resultRelInfo, tupleid, NULL,
@@ -2898,12 +2903,15 @@ lmerge_matched:
 				break;
 
 			case CMD_DELETE:
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				if (!ExecDeletePrologue(context, resultRelInfo, tupleid,
 										NULL, NULL, &result))
 				{
 					if (result == TM_Ok)
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 					break;		/* concurrent update/delete */
 				}
 				result = ExecDeleteAct(context, resultRelInfo, tupleid, false);
@@ -2959,7 +2967,8 @@ lmerge_matched:
 				 * If the tuple was already deleted, return to let caller
 				 * handle it under NOT MATCHED clauses.
 				 */
-				return false;
+				*matched = false;
+				return NULL;
 
 			case TM_Updated:
 				{
@@ -3005,13 +3014,19 @@ lmerge_matched:
 							 * NOT MATCHED actions.
 							 */
 							if (TupIsNull(epqslot))
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							(void) ExecGetJunkAttribute(epqslot,
 														resultRelInfo->ri_RowIdAttNo,
 														&isNull);
 							if (isNull)
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							/*
 							 * When a tuple was updated and migrated to
@@ -3046,7 +3061,8 @@ lmerge_matched:
 							 * tuple already deleted; tell caller to run NOT
 							 * MATCHED actions
 							 */
-							return false;
+							*matched = false;
+							return NULL;
 
 						case TM_SelfModified:
 
@@ -3066,13 +3082,14 @@ lmerge_matched:
 										(errcode(ERRCODE_TRIGGERED_DATA_CHANGE_VIOLATION),
 										 errmsg("tuple to be updated or deleted was already modified by an operation triggered by the current command"),
 										 errhint("Consider using an AFTER trigger instead of a BEFORE trigger to propagate changes to other rows.")));
-							return false;
+							*matched = false;
+							return NULL;
 
 						default:
 							/* see table_tuple_lock call in ExecDelete() */
 							elog(ERROR, "unexpected table_tuple_lock status: %u",
 								 result);
-							return false;
+							return NULL;
 					}
 				}
 
@@ -3084,6 +3101,30 @@ lmerge_matched:
 				break;
 		}
 
+		/* Process RETURNING if present */
+		if (resultRelInfo->ri_projectReturning)
+		{
+			switch (commandType)
+			{
+				case CMD_UPDATE:
+					rslot = ExecProcessReturning(resultRelInfo, newslot,
+												 context->planSlot);
+					break;
+
+				case CMD_DELETE:
+					rslot = ExecProcessReturning(resultRelInfo,
+												 resultRelInfo->ri_oldTupleSlot,
+												 context->planSlot);
+					break;
+
+				case CMD_NOTHING:
+					break;
+
+				default:
+					elog(ERROR, "unknown action in MERGE WHEN MATCHED clause");
+			}
+		}
+
 		/*
 		 * We've activated one of the WHEN clauses, so we don't search
 		 * further. This is required behaviour, not an optimization.
@@ -3094,19 +3135,22 @@ lmerge_matched:
 	/*
 	 * Successfully executed an action or no qualifying action was found.
 	 */
-	return true;
+	*matched = true;
+
+	return rslot;
 }
 
 /*
  * Execute the first qualifying NOT MATCHED action.
  */
-static void
+static TupleTableSlot *
 ExecMergeNotMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 					bool canSetTag)
 {
 	ModifyTableState *mtstate = context->mtstate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
 	List	   *actionStates = NIL;
+	TupleTableSlot *rslot = NULL;
 	ListCell   *l;
 
 	/*
@@ -3156,10 +3200,10 @@ ExecMergeNotMatched(ModifyTableContext *
 				 * so we don't need to map the tuple here.
 				 */
 				newslot = ExecProject(action->mas_proj);
-				context->relaction = action;
+				mtstate->mt_merge_action = action;
 
-				(void) ExecInsert(context, mtstate->rootResultRelInfo, newslot,
-								  canSetTag, NULL, NULL);
+				rslot = ExecInsert(context, mtstate->rootResultRelInfo,
+								   newslot, canSetTag, NULL, NULL);
 				mtstate->mt_merge_inserted += 1;
 				break;
 			case CMD_NOTHING:
@@ -3175,6 +3219,8 @@ ExecMergeNotMatched(ModifyTableContext *
 		 */
 		break;
 	}
+
+	return rslot;
 }
 
 /*
@@ -3352,6 +3398,64 @@ ExecInitMergeTupleSlots(ModifyTableState
 }
 
 /*
+ * pg_merge_action() -
+ *	  SQL merge support function to retrieve the currently executing merge
+ *	  action command string ("INSERT", "UPDATE", or "DELETE").
+ */
+Datum
+pg_merge_action(PG_FUNCTION_ARGS)
+{
+	ModifyTableState *mtstate = (ModifyTableState *) fcinfo->context;
+	MergeActionState *relaction;
+
+	if (!mtstate || mtstate->operation != CMD_MERGE)
+		elog(ERROR, "merge support function called in non-merge context");
+
+	relaction = mtstate->mt_merge_action;
+	if (relaction)
+	{
+		CmdType		commandType = relaction->mas_action->commandType;
+
+		switch (commandType)
+		{
+			case CMD_INSERT:
+				PG_RETURN_TEXT_P(cstring_to_text("INSERT"));
+			case CMD_UPDATE:
+				PG_RETURN_TEXT_P(cstring_to_text("UPDATE"));
+			case CMD_DELETE:
+				PG_RETURN_TEXT_P(cstring_to_text("DELETE"));
+			case CMD_NOTHING:
+				PG_RETURN_NULL();
+			default:
+				elog(ERROR, "unrecognized commandType: %d", (int) commandType);
+		}
+	}
+
+	PG_RETURN_NULL();
+}
+
+/*
+ * pg_merge_when_clause() -
+ *	  SQL merge support function to retrieve the 1-based index of the
+ *	  currently executing merge WHEN clause.
+ */
+Datum
+pg_merge_when_clause(PG_FUNCTION_ARGS)
+{
+	ModifyTableState *mtstate = (ModifyTableState *) fcinfo->context;
+	MergeActionState *relaction;
+
+	if (!mtstate || mtstate->operation != CMD_MERGE)
+		elog(ERROR, "merge support function called in non-merge context");
+
+	relaction = mtstate->mt_merge_action;
+	if (relaction)
+		PG_RETURN_INT32((int32) relaction->mas_action->index);
+
+	PG_RETURN_NULL();
+}
+
+/*
  * Process BEFORE EACH STATEMENT triggers
  */
 static void
@@ -3637,8 +3741,17 @@ ExecModifyTable(PlanState *pstate)
 				{
 					EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-					ExecMerge(&context, node->resultRelInfo, NULL, node->canSetTag);
-					continue;	/* no RETURNING support yet */
+					slot = ExecMerge(&context, node->resultRelInfo, NULL,
+									 node->canSetTag);
+
+					/*
+					 * If we got a RETURNING result, return it to the caller.
+					 * We'll continue the work on next call.
+					 */
+					if (slot)
+						return slot;
+
+					continue;	/* continue with the next tuple */
 				}
 
 				elog(ERROR, "tableoid is NULL");
@@ -3715,8 +3828,17 @@ ExecModifyTable(PlanState *pstate)
 					{
 						EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-						ExecMerge(&context, node->resultRelInfo, NULL, node->canSetTag);
-						continue;	/* no RETURNING support yet */
+						slot = ExecMerge(&context, node->resultRelInfo, NULL,
+										 node->canSetTag);
+
+						/*
+						 * If we got a RETURNING result, return it to the
+						 * caller.  We'll continue the work on next call.
+						 */
+						if (slot)
+							return slot;
+
+						continue;	/* continue with the next tuple */
 					}
 
 					elog(ERROR, "ctid is NULL");
@@ -3808,7 +3930,7 @@ ExecModifyTable(PlanState *pstate)
 				}
 				slot = ExecGetUpdateNewTuple(resultRelInfo, context.planSlot,
 											 oldSlot);
-				context.relaction = NULL;
+				node->mt_merge_action = NULL;
 
 				/* Now apply the update. */
 				slot = ExecUpdate(&context, resultRelInfo, tupleid, oldtuple,
diff --git a/src/backend/executor/spi.c b/src/backend/executor/spi.c
new file mode 100644
index e3a170c..be88056
--- a/src/backend/executor/spi.c
+++ b/src/backend/executor/spi.c
@@ -2033,6 +2033,8 @@ SPI_result_code_string(int code)
 			return "SPI_OK_TD_REGISTER";
 		case SPI_OK_MERGE:
 			return "SPI_OK_MERGE";
+		case SPI_OK_MERGE_RETURNING:
+			return "SPI_OK_MERGE_RETURNING";
 	}
 	/* Unrecognized code ... return something useful ... */
 	sprintf(buf, "Unrecognized SPI code %d", code);
@@ -2886,7 +2888,10 @@ _SPI_pquery(QueryDesc *queryDesc, bool f
 				res = SPI_OK_UPDATE;
 			break;
 		case CMD_MERGE:
-			res = SPI_OK_MERGE;
+			if (queryDesc->plannedstmt->hasReturning)
+				res = SPI_OK_MERGE_RETURNING;
+			else
+				res = SPI_OK_MERGE;
 			break;
 		default:
 			return SPI_ERROR_OPUNKNOWN;
diff --git a/src/backend/parser/analyze.c b/src/backend/parser/analyze.c
new file mode 100644
index e892df9..fe0f5cb
--- a/src/backend/parser/analyze.c
+++ b/src/backend/parser/analyze.c
@@ -73,7 +73,6 @@ static void determineRecursiveColTypes(P
 									   Node *larg, List *nrtargetlist);
 static Query *transformReturnStmt(ParseState *pstate, ReturnStmt *stmt);
 static Query *transformUpdateStmt(ParseState *pstate, UpdateStmt *stmt);
-static List *transformReturningList(ParseState *pstate, List *returningList);
 static Query *transformPLAssignStmt(ParseState *pstate,
 									PLAssignStmt *stmt);
 static Query *transformDeclareCursorStmt(ParseState *pstate,
@@ -514,7 +513,8 @@ transformDeleteStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -941,7 +941,8 @@ transformInsertStmt(ParseState *pstate,
 	/* Process RETURNING, if any. */
 	if (stmt->returningList)
 		qry->returningList = transformReturningList(pstate,
-													stmt->returningList);
+													stmt->returningList,
+													EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -2406,7 +2407,8 @@ transformUpdateStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/*
 	 * Now we are done with SELECT-like processing, and can get on with
@@ -2500,10 +2502,11 @@ transformUpdateTargetList(ParseState *ps
 
 /*
  * transformReturningList -
- *	handle a RETURNING clause in INSERT/UPDATE/DELETE
+ *	handle a RETURNING clause in INSERT/UPDATE/DELETE/MERGE
  */
-static List *
-transformReturningList(ParseState *pstate, List *returningList)
+List *
+transformReturningList(ParseState *pstate, List *returningList,
+					   ParseExprKind exprKind)
 {
 	List	   *rlist;
 	int			save_next_resno;
@@ -2520,7 +2523,7 @@ transformReturningList(ParseState *pstat
 	pstate->p_next_resno = 1;
 
 	/* transform RETURNING identically to a SELECT targetlist */
-	rlist = transformTargetList(pstate, returningList, EXPR_KIND_RETURNING);
+	rlist = transformTargetList(pstate, returningList, exprKind);
 
 	/*
 	 * Complain if the nonempty tlist expanded to nothing (which is possible
diff --git a/src/backend/parser/gram.y b/src/backend/parser/gram.y
new file mode 100644
index a013838..4406430
--- a/src/backend/parser/gram.y
+++ b/src/backend/parser/gram.y
@@ -12241,6 +12241,7 @@ MergeStmt:
 			USING table_ref
 			ON a_expr
 			merge_when_list
+			returning_clause
 				{
 					MergeStmt  *m = makeNode(MergeStmt);
 
@@ -12249,6 +12250,7 @@ MergeStmt:
 					m->sourceRelation = $6;
 					m->joinCondition = $8;
 					m->mergeWhenClauses = $9;
+					m->returningList = $10;
 
 					$$ = (Node *) m;
 				}
diff --git a/src/backend/parser/parse_agg.c b/src/backend/parser/parse_agg.c
new file mode 100644
index 4fbf80c..b8d655a
--- a/src/backend/parser/parse_agg.c
+++ b/src/backend/parser/parse_agg.c
@@ -456,6 +456,7 @@ check_agglevels_and_constraints(ParseSta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
@@ -904,6 +905,7 @@ transformWindowFuncCall(ParseState *psta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_cte.c b/src/backend/parser/parse_cte.c
new file mode 100644
index c5b1a49..ad3c525
--- a/src/backend/parser/parse_cte.c
+++ b/src/backend/parser/parse_cte.c
@@ -126,13 +126,6 @@ transformWithClause(ParseState *pstate,
 		CommonTableExpr *cte = (CommonTableExpr *) lfirst(lc);
 		ListCell   *rest;
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(cte->ctequery, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in WITH query"),
-					parser_errposition(pstate, cte->location));
-
 		for_each_cell(rest, withClause->ctes, lnext(withClause->ctes, lc))
 		{
 			CommonTableExpr *cte2 = (CommonTableExpr *) lfirst(rest);
@@ -153,7 +146,8 @@ transformWithClause(ParseState *pstate,
 			/* must be a data-modifying statement */
 			Assert(IsA(cte->ctequery, InsertStmt) ||
 				   IsA(cte->ctequery, UpdateStmt) ||
-				   IsA(cte->ctequery, DeleteStmt));
+				   IsA(cte->ctequery, DeleteStmt) ||
+				   IsA(cte->ctequery, MergeStmt));
 
 			pstate->p_hasModifyingCTE = true;
 		}
diff --git a/src/backend/parser/parse_expr.c b/src/backend/parser/parse_expr.c
new file mode 100644
index 78221d2..b4021f6
--- a/src/backend/parser/parse_expr.c
+++ b/src/backend/parser/parse_expr.c
@@ -482,6 +482,7 @@ transformColumnRef(ParseState *pstate, C
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CHECK_CONSTRAINT:
@@ -1708,6 +1709,7 @@ transformSubLink(ParseState *pstate, Sub
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CYCLE_MARK:
@@ -3002,6 +3004,7 @@ ParseExprKindName(ParseExprKind exprKind
 		case EXPR_KIND_OFFSET:
 			return "OFFSET";
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			return "RETURNING";
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
diff --git a/src/backend/parser/parse_func.c b/src/backend/parser/parse_func.c
new file mode 100644
index ca14f06..914624e
--- a/src/backend/parser/parse_func.c
+++ b/src/backend/parser/parse_func.c
@@ -31,6 +31,7 @@
 #include "parser/parse_target.h"
 #include "parser/parse_type.h"
 #include "utils/builtins.h"
+#include "utils/fmgroids.h"
 #include "utils/lsyscache.h"
 #include "utils/syscache.h"
 
@@ -348,6 +349,15 @@ ParseFuncOrColumn(ParseState *pstate, Li
 					 parser_errposition(pstate, location)));
 	}
 
+	/* Merge support functions are only allowed in MERGE's RETURNING list */
+	if (IsMergeSupportFunction(funcid) &&
+		pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+		ereport(ERROR,
+				(errcode(ERRCODE_WRONG_OBJECT_TYPE),
+				 errmsg("merge support function %s can only be called from the RETURNING list of a MERGE command",
+						NameListToString(funcname)),
+				 parser_errposition(pstate, location)));
+
 	/*
 	 * So far so good, so do some fdresult-type-specific processing.
 	 */
@@ -2602,6 +2612,7 @@ check_srf_call_placement(ParseState *pst
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_merge.c b/src/backend/parser/parse_merge.c
new file mode 100644
index d886637..2e50eb0
--- a/src/backend/parser/parse_merge.c
+++ b/src/backend/parser/parse_merge.c
@@ -233,6 +233,10 @@ transformMergeStmt(ParseState *pstate, M
 	 */
 	qry->jointree = makeFromExpr(pstate->p_joinlist, joinExpr);
 
+	/* Transform the RETURNING list, if any */
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_MERGE_RETURNING);
+
 	/*
 	 * We now have a good query shape, so now look at the WHEN conditions and
 	 * action targetlists.
@@ -254,6 +258,7 @@ transformMergeStmt(ParseState *pstate, M
 		MergeAction *action;
 
 		action = makeNode(MergeAction);
+		action->index = foreach_current_index(l) + 1;
 		action->commandType = mergeWhenClause->commandType;
 		action->matched = mergeWhenClause->matched;
 
@@ -390,9 +395,6 @@ transformMergeStmt(ParseState *pstate, M
 
 	qry->mergeActionList = mergeActionList;
 
-	/* RETURNING could potentially be added in the future, but not in SQL std */
-	qry->returningList = NULL;
-
 	qry->hasTargetSRFs = false;
 	qry->hasSubLinks = pstate->p_hasSubLinks;
 
diff --git a/src/backend/parser/parse_relation.c b/src/backend/parser/parse_relation.c
new file mode 100644
index 41d6049..794530b
--- a/src/backend/parser/parse_relation.c
+++ b/src/backend/parser/parse_relation.c
@@ -2345,9 +2345,10 @@ addRangeTableEntryForCTE(ParseState *pst
 		cte->cterefcount++;
 
 	/*
-	 * We throw error if the CTE is INSERT/UPDATE/DELETE without RETURNING.
-	 * This won't get checked in case of a self-reference, but that's OK
-	 * because data-modifying CTEs aren't allowed to be recursive anyhow.
+	 * We throw error if the CTE is INSERT/UPDATE/DELETE/MERGE without
+	 * RETURNING.  This won't get checked in case of a self-reference, but
+	 * that's OK because data-modifying CTEs aren't allowed to be recursive
+	 * anyhow.
 	 */
 	if (IsA(cte->ctequery, Query))
 	{
diff --git a/src/backend/rewrite/rewriteHandler.c b/src/backend/rewrite/rewriteHandler.c
new file mode 100644
index 980dc18..3689e54
--- a/src/backend/rewrite/rewriteHandler.c
+++ b/src/backend/rewrite/rewriteHandler.c
@@ -3655,7 +3655,8 @@ RewriteQuery(Query *parsetree, List *rew
 			if (!(ctequery->commandType == CMD_SELECT ||
 				  ctequery->commandType == CMD_UPDATE ||
 				  ctequery->commandType == CMD_INSERT ||
-				  ctequery->commandType == CMD_DELETE))
+				  ctequery->commandType == CMD_DELETE ||
+				  ctequery->commandType == CMD_MERGE))
 			{
 				/*
 				 * Currently it could only be NOTIFY; this error message will
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
new file mode 100644
index c7d9d96..d393f51
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -2131,11 +2131,10 @@ QueryReturnsTuples(Query *parsetree)
 		case CMD_SELECT:
 			/* returns tuples */
 			return true;
-		case CMD_MERGE:
-			return false;
 		case CMD_INSERT:
 		case CMD_UPDATE:
 		case CMD_DELETE:
+		case CMD_MERGE:
 			/* the forms with RETURNING return tuples */
 			if (parsetree->returningList)
 				return true;
diff --git a/src/bin/psql/common.c b/src/bin/psql/common.c
new file mode 100644
index f907f5d..85eb868
--- a/src/bin/psql/common.c
+++ b/src/bin/psql/common.c
@@ -962,13 +962,17 @@ PrintQueryResult(PGresult *result, bool
 			else
 				success = true;
 
-			/* if it's INSERT/UPDATE/DELETE RETURNING, also print status */
+			/*
+			 * If it's INSERT/UPDATE/DELETE/MERGE RETURNING, also print
+			 * status.
+			 */
 			if (last || pset.show_all_results)
 			{
 				cmdstatus = PQcmdStatus(result);
 				if (strncmp(cmdstatus, "INSERT", 6) == 0 ||
 					strncmp(cmdstatus, "UPDATE", 6) == 0 ||
-					strncmp(cmdstatus, "DELETE", 6) == 0)
+					strncmp(cmdstatus, "DELETE", 6) == 0 ||
+					strncmp(cmdstatus, "MERGE", 5) == 0)
 					PrintQueryStatus(result, printStatusFout);
 			}
 
diff --git a/src/include/catalog/pg_proc.dat b/src/include/catalog/pg_proc.dat
new file mode 100644
index 5055956..9e86bf9
--- a/src/include/catalog/pg_proc.dat
+++ b/src/include/catalog/pg_proc.dat
@@ -11939,4 +11939,14 @@
   proname => 'any_value_transfn', prorettype => 'anyelement',
   proargtypes => 'anyelement anyelement', prosrc => 'any_value_transfn' },
 
+# MERGE support functions
+{ oid => '9499', descr => 'command type of current MERGE action',
+  proname => 'pg_merge_action',  provolatile => 'v',
+  prorettype => 'text', proargtypes => '',
+  prosrc => 'pg_merge_action' },
+{ oid => '9500', descr => 'index of current MERGE WHEN clause',
+  proname => 'pg_merge_when_clause',  provolatile => 'v',
+  prorettype => 'int4', proargtypes => '',
+  prosrc => 'pg_merge_when_clause' },
+
 ]
diff --git a/src/include/catalog/pg_proc.h b/src/include/catalog/pg_proc.h
new file mode 100644
index e7abe0b..e597826
--- a/src/include/catalog/pg_proc.h
+++ b/src/include/catalog/pg_proc.h
@@ -182,6 +182,11 @@ DECLARE_UNIQUE_INDEX(pg_proc_proname_arg
 #define PROARGMODE_VARIADIC 'v'
 #define PROARGMODE_TABLE	't'
 
+/* Is this a merge support function?  (Requires fmgroids.h) */
+#define IsMergeSupportFunction(oid) \
+	((oid) == F_PG_MERGE_ACTION || \
+	 (oid) == F_PG_MERGE_WHEN_CLAUSE)
+
 #endif							/* EXPOSE_TO_CLIENT_CODE */
 
 
diff --git a/src/include/executor/spi.h b/src/include/executor/spi.h
new file mode 100644
index d1de139..ed8787d
--- a/src/include/executor/spi.h
+++ b/src/include/executor/spi.h
@@ -97,6 +97,7 @@ typedef struct _SPI_plan *SPIPlanPtr;
 #define SPI_OK_REL_UNREGISTER	16
 #define SPI_OK_TD_REGISTER		17
 #define SPI_OK_MERGE			18
+#define SPI_OK_MERGE_RETURNING	19
 
 #define SPI_OPT_NONATOMIC		(1 << 0)
 
diff --git a/src/include/nodes/execnodes.h b/src/include/nodes/execnodes.h
new file mode 100644
index bc67cb9..cc85e25
--- a/src/include/nodes/execnodes.h
+++ b/src/include/nodes/execnodes.h
@@ -1304,6 +1304,9 @@ typedef struct ModifyTableState
 	/* Flags showing which subcommands are present INS/UPD/DEL/DO NOTHING */
 	int			mt_merge_subcommands;
 
+	/* For MERGE, the action currently being executed */
+	MergeActionState *mt_merge_action;
+
 	/* tuple counters for MERGE */
 	double		mt_merge_inserted;
 	double		mt_merge_updated;
diff --git a/src/include/nodes/parsenodes.h b/src/include/nodes/parsenodes.h
new file mode 100644
index 371aa0f..3a0e076
--- a/src/include/nodes/parsenodes.h
+++ b/src/include/nodes/parsenodes.h
@@ -1686,6 +1686,7 @@ typedef struct MergeWhenClause
 typedef struct MergeAction
 {
 	NodeTag		type;
+	int			index;			/* 1-based index of the clause */
 	bool		matched;		/* true=MATCHED, false=NOT MATCHED */
 	CmdType		commandType;	/* INSERT/UPDATE/DELETE/DO NOTHING */
 	/* OVERRIDING clause */
@@ -1805,6 +1806,7 @@ typedef struct MergeStmt
 	Node	   *sourceRelation; /* source relation */
 	Node	   *joinCondition;	/* join condition between source and target */
 	List	   *mergeWhenClauses;	/* list of MergeWhenClause(es) */
+	List	   *returningList;	/* list of expressions to return */
 	WithClause *withClause;		/* WITH clause */
 } MergeStmt;
 
diff --git a/src/include/parser/analyze.h b/src/include/parser/analyze.h
new file mode 100644
index 1cef183..78eb55e
--- a/src/include/parser/analyze.h
+++ b/src/include/parser/analyze.h
@@ -44,6 +44,8 @@ extern List *transformInsertRow(ParseSta
 								bool strip_indirection);
 extern List *transformUpdateTargetList(ParseState *pstate,
 									   List *origTlist);
+extern List *transformReturningList(ParseState *pstate, List *returningList,
+									ParseExprKind exprKind);
 extern Query *transformTopLevelStmt(ParseState *pstate, RawStmt *parseTree);
 extern Query *transformStmt(ParseState *pstate, Node *parseTree);
 
diff --git a/src/include/parser/parse_node.h b/src/include/parser/parse_node.h
new file mode 100644
index f589112..836819d
--- a/src/include/parser/parse_node.h
+++ b/src/include/parser/parse_node.h
@@ -61,7 +61,8 @@ typedef enum ParseExprKind
 	EXPR_KIND_DISTINCT_ON,		/* DISTINCT ON */
 	EXPR_KIND_LIMIT,			/* LIMIT */
 	EXPR_KIND_OFFSET,			/* OFFSET */
-	EXPR_KIND_RETURNING,		/* RETURNING */
+	EXPR_KIND_RETURNING,		/* RETURNING in INSERT/UPDATE/DELETE */
+	EXPR_KIND_MERGE_RETURNING,	/* RETURNING in MERGE */
 	EXPR_KIND_VALUES,			/* VALUES */
 	EXPR_KIND_VALUES_SINGLE,	/* single-row VALUES (in INSERT only) */
 	EXPR_KIND_CHECK_CONSTRAINT, /* CHECK constraint for a table */
diff --git a/src/pl/plpgsql/src/pl_exec.c b/src/pl/plpgsql/src/pl_exec.c
new file mode 100644
index b0a2cac..b72e7b9
--- a/src/pl/plpgsql/src/pl_exec.c
+++ b/src/pl/plpgsql/src/pl_exec.c
@@ -4246,9 +4246,9 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 	/*
 	 * If we have INTO, then we only need one row back ... but if we have INTO
 	 * STRICT or extra check too_many_rows, ask for two rows, so that we can
-	 * verify the statement returns only one.  INSERT/UPDATE/DELETE are always
-	 * treated strictly. Without INTO, just run the statement to completion
-	 * (tcount = 0).
+	 * verify the statement returns only one.  INSERT/UPDATE/DELETE/MERGE are
+	 * always treated strictly. Without INTO, just run the statement to
+	 * completion (tcount = 0).
 	 *
 	 * We could just ask for two rows always when using INTO, but there are
 	 * some cases where demanding the extra row costs significant time, eg by
@@ -4286,10 +4286,11 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 			Assert(stmt->mod_stmt);
 			exec_set_found(estate, (SPI_processed != 0));
 			break;
@@ -4468,10 +4469,11 @@ exec_stmt_dynexecute(PLpgSQL_execstate *
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 		case SPI_OK_UTILITY:
 		case SPI_OK_REWRITTEN:
 			break;
diff --git a/src/pl/tcl/pltcl.c b/src/pl/tcl/pltcl.c
new file mode 100644
index 499a9ea..58c4f50
--- a/src/pl/tcl/pltcl.c
+++ b/src/pl/tcl/pltcl.c
@@ -2459,6 +2459,7 @@ pltcl_process_SPI_result(Tcl_Interp *int
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
+		case SPI_OK_MERGE_RETURNING:
 
 			/*
 			 * Process the tuples we got
diff --git a/src/test/regress/expected/merge.out b/src/test/regress/expected/merge.out
new file mode 100644
index e32afc3..5b60743
--- a/src/test/regress/expected/merge.out
+++ b/src/test/regress/expected/merge.out
@@ -123,20 +123,20 @@ ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
 ERROR:  name "target" specified more than once
 DETAIL:  The name is used both as MERGE target table and data source.
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
-ERROR:  MERGE not supported in WITH query
-LINE 1: WITH foo AS (
-             ^
--- used in COPY
+ERROR:  WITH query "foo" does not have a RETURNING clause
+LINE 4: ) SELECT * FROM foo;
+                        ^
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) TO stdout;
-ERROR:  MERGE not supported in COPY
+ERROR:  COPY query must have a RETURNING clause
 -- unsupported relation types
 -- view
 CREATE VIEW tv AS SELECT * FROM target;
@@ -1304,20 +1304,187 @@ WHEN MATCHED AND tid < 2 THEN
 ROLLBACK;
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
-ERROR:  syntax error at or near "RETURNING"
-LINE 10: RETURNING *;
-         ^
+    DELETE
+RETURNING pg_merge_when_clause() AS when_clause,
+          pg_merge_action() AS merge_action,
+          t.*,
+          CASE pg_merge_action()
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ when_clause | merge_action | tid | balance |     description     
+-------------+--------------+-----+---------+---------------------
+           3 | DELETE       |   1 |     100 | Removed (1,100)
+           1 | UPDATE       |   2 |     220 | Added 20 to balance
+           2 | INSERT       |   4 |      40 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- error when using MERGE support functions outside MERGE
+SELECT pg_merge_action() FROM sq_target;
+ERROR:  merge support function pg_merge_action can only be called from the RETURNING list of a MERGE command
+LINE 1: SELECT pg_merge_action() FROM sq_target;
+               ^
+UPDATE sq_target SET balance = balance + 1 RETURNING pg_merge_when_clause();
+ERROR:  merge support function pg_merge_when_clause can only be called from the RETURNING list of a MERGE command
+LINE 1: ...ATE sq_target SET balance = balance + 1 RETURNING pg_merge_w...
+                                                             ^
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_when_clause() AS when_clause,
+              pg_merge_action() AS merge_action,
+              t.*,
+              CASE pg_merge_action()
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING merge_action, pg_merge_action() AS merge_log_action, l.*
+)
+SELECT * FROM m2;
+ merge_action | merge_log_action | tid |     last_change     
+--------------+------------------+-----+---------------------
+ DELETE       | UPDATE           |   1 | Removed (1,100)
+ UPDATE       | INSERT           |   2 | Added 20 to balance
+ INSERT       | INSERT           |   4 | Inserted (4,40)
+(3 rows)
+
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ tid |     last_change     
+-----+---------------------
+   1 | Removed (1,100)
+   2 | Added 20 to balance
+   4 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING v
+    ON tid = sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*
+) TO stdout;
+DELETE	1	100
+ROLLBACK;
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ action | tid | new_balance 
+--------+-----+-------------
+ DELETE |   1 |         100
+ UPDATE |   3 |         320
+ INSERT |   4 |         110
+(3 rows)
+
+ROLLBACK;
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ action | tid | balance 
+--------+-----+---------
+ DELETE |   1 |     100
+ INSERT |   4 |      40
+(2 rows)
+
+ROLLBACK;
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ r_action | r_tid | r_balance 
+----------+-------+-----------
+ DELETE   |     1 |       100
+ UPDATE   |     3 |       320
+ INSERT   |     4 |       110
+(3 rows)
+
 ROLLBACK;
 -- EXPLAIN
 CREATE TABLE ex_mtarget (a int, b int)
@@ -1514,7 +1681,7 @@ SELECT * FROM sq_target WHERE tid = 1;
 (1 row)
 
 ROLLBACK;
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 NOTICE:  drop cascades to view v
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1640,6 +1807,32 @@ SELECT * FROM pa_target ORDER BY tid;
 (14 rows)
 
 ROLLBACK;
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
+ pg_merge_action | pg_merge_when_clause | tid | balance |           val            
+-----------------+----------------------+-----+---------+--------------------------
+ UPDATE          |                    1 |   2 |     110 | initial updated by merge
+(1 row)
+
+SELECT * FROM pa_target ORDER BY tid;
+ tid | balance |           val            
+-----+---------+--------------------------
+   2 |     110 | initial updated by merge
+   3 |     300 | initial
+   5 |     500 | initial
+   7 |     700 | initial
+   9 |     900 | initial
+  11 |    1100 | initial
+  13 |    1300 | initial
+(7 rows)
+
+ROLLBACK;
 DROP TABLE pa_target CASCADE;
 -- The target table is partitioned in the same way, but this time by attaching
 -- partitions which have columns in different order, dropped columns etc.
@@ -1731,7 +1924,26 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (sid, delta, 'inserted by merge');
+    INSERT VALUES (sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
+ pg_merge_action | pg_merge_when_clause | tid | balance |           val            
+-----------------+----------------------+-----+---------+--------------------------
+ UPDATE          |                    1 |   2 |     110 | initial updated by merge
+ INSERT          |                    2 |   2 |      20 | inserted by merge
+ UPDATE          |                    1 |   4 |     330 | initial updated by merge
+ INSERT          |                    2 |   4 |      40 | inserted by merge
+ UPDATE          |                    1 |   6 |     550 | initial updated by merge
+ INSERT          |                    2 |   6 |      60 | inserted by merge
+ UPDATE          |                    1 |   8 |     770 | initial updated by merge
+ INSERT          |                    2 |   8 |      80 | inserted by merge
+ UPDATE          |                    1 |  10 |     990 | initial updated by merge
+ INSERT          |                    2 |  10 |     100 | inserted by merge
+ UPDATE          |                    1 |  12 |    1210 | initial updated by merge
+ INSERT          |                    2 |  12 |     120 | inserted by merge
+ UPDATE          |                    1 |  14 |    1430 | initial updated by merge
+ INSERT          |                    2 |  14 |     140 | inserted by merge
+(14 rows)
+
 SELECT * FROM pa_target ORDER BY tid;
  tid | balance |           val            
 -----+---------+--------------------------
@@ -1798,7 +2010,21 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
+ pg_merge_action | pg_merge_when_clause |          logts           | tid | balance |           val            
+-----------------+----------------------+--------------------------+-----+---------+--------------------------
+ UPDATE          |                    1 | Tue Jan 31 00:00:00 2017 |   1 |     110 | initial updated by merge
+ UPDATE          |                    1 | Tue Feb 28 00:00:00 2017 |   2 |     220 | initial updated by merge
+ INSERT          |                    2 | Sun Jan 15 00:00:00 2017 |   3 |      30 | inserted by merge
+ UPDATE          |                    1 | Tue Jan 31 00:00:00 2017 |   4 |     440 | initial updated by merge
+ UPDATE          |                    1 | Tue Feb 28 00:00:00 2017 |   5 |     550 | initial updated by merge
+ INSERT          |                    2 | Sun Jan 15 00:00:00 2017 |   6 |      60 | inserted by merge
+ UPDATE          |                    1 | Tue Jan 31 00:00:00 2017 |   7 |     770 | initial updated by merge
+ UPDATE          |                    1 | Tue Feb 28 00:00:00 2017 |   8 |     880 | initial updated by merge
+ INSERT          |                    2 | Sun Jan 15 00:00:00 2017 |   9 |      90 | inserted by merge
+(9 rows)
+
 SELECT * FROM pa_target ORDER BY tid;
           logts           | tid | balance |           val            
 --------------------------+-----+---------+--------------------------
diff --git a/src/test/regress/sql/merge.sql b/src/test/regress/sql/merge.sql
new file mode 100644
index cae6902..e0f2442
--- a/src/test/regress/sql/merge.sql
+++ b/src/test/regress/sql/merge.sql
@@ -86,12 +86,12 @@ MERGE INTO target
 USING target
 ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
--- used in COPY
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
@@ -857,17 +857,145 @@ ROLLBACK;
 
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
+    DELETE
+RETURNING pg_merge_when_clause() AS when_clause,
+          pg_merge_action() AS merge_action,
+          t.*,
+          CASE pg_merge_action()
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ROLLBACK;
+
+-- error when using MERGE support functions outside MERGE
+SELECT pg_merge_action() FROM sq_target;
+UPDATE sq_target SET balance = balance + 1 RETURNING pg_merge_when_clause();
+
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_when_clause() AS when_clause,
+              pg_merge_action() AS merge_action,
+              t.*,
+              CASE pg_merge_action()
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING merge_action, pg_merge_action() AS merge_log_action, l.*
+)
+SELECT * FROM m2;
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ROLLBACK;
+
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING v
+    ON tid = sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*
+) TO stdout;
+ROLLBACK;
+
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ROLLBACK;
+
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ROLLBACK;
+
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
 ROLLBACK;
 
 -- EXPLAIN
@@ -966,7 +1094,7 @@ WHEN MATCHED THEN
 SELECT * FROM sq_target WHERE tid = 1;
 ROLLBACK;
 
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1033,6 +1161,17 @@ SELECT merge_func();
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
+SELECT * FROM pa_target ORDER BY tid;
+ROLLBACK;
+
 DROP TABLE pa_target CASCADE;
 
 -- The target table is partitioned in the same way, but this time by attaching
@@ -1091,7 +1230,8 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (sid, delta, 'inserted by merge');
+    INSERT VALUES (sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
@@ -1145,7 +1285,8 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 

diff --git a/doc/src/sgml/dml.sgml b/doc/src/sgml/dml.sgml
new file mode 100644
index cbbc5e2..ff2a827
--- a/doc/src/sgml/dml.sgml
+++ b/doc/src/sgml/dml.sgml
@@ -283,10 +283,15 @@ DELETE FROM products;
    <secondary>RETURNING</secondary>
   </indexterm>
 
+  <indexterm zone="dml-returning">
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
   <para>
    Sometimes it is useful to obtain data from modified rows while they are
    being manipulated.  The <command>INSERT</command>, <command>UPDATE</command>,
-   and <command>DELETE</command> commands all have an
+   <command>DELETE</command>, and <command>MERGE</command> commands all have an
    optional <literal>RETURNING</literal> clause that supports this.  Use
    of <literal>RETURNING</literal> avoids performing an extra database query to
    collect the data, and is especially valuable when it would otherwise be
@@ -339,6 +344,24 @@ DELETE FROM products
 </programlisting>
   </para>
 
+  <para>
+   In a <command>MERGE</command>, the data available to <literal>RETURNING</literal> is
+   the content of the source row plus the content of the inserted, updated, or
+   deleted target row.  In addition, the merge support functions
+   <xref linkend="pg-merge-action"/> and <xref linkend="pg-merge-when-clause"/>
+   may be used to return information about which merge action was executed for
+   each row.  Since it is quite common for the source and target to have many
+   of the same columns, specifying <literal>RETURNING *</literal> can lead to a
+   lot of duplicated columns, so it is often more useful to just return the
+   target row.  For example:
+<programlisting>
+MERGE INTO products p USING new_products n ON p.product_no = n.product_no
+  WHEN NOT MATCHED THEN INSERT VALUES (n.product_no, n.name, n.price)
+  WHEN MATCHED THEN UPDATE SET name = n.name, price = n.price
+  RETURNING pg_merge_action(), p.*;
+</programlisting>
+  </para>
+
   <para>
    If there are triggers (<xref linkend="triggers"/>) on the target table,
    the data available to <literal>RETURNING</literal> is the row as modified by
diff --git a/doc/src/sgml/func.sgml b/doc/src/sgml/func.sgml
new file mode 100644
index 5a47ce4..ca1b3cc
--- a/doc/src/sgml/func.sgml
+++ b/doc/src/sgml/func.sgml
@@ -21705,6 +21705,100 @@ SELECT count(*) FROM sometable;
 
  </sect1>
 
+ <sect1 id="functions-merge-support">
+  <title>Merge Support Functions</title>
+
+  <indexterm>
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
+  <para>
+   The merge support functions shown in
+   <xref linkend="functions-merge-support-table"/> may be used in the
+   <literal>RETURNING</literal> list of a <xref linkend="sql-merge"/> command
+   to return additional information about the action taken for each row.
+  </para>
+
+  <table id="functions-merge-support-table">
+    <title>Merge Support Functions</title>
+
+    <tgroup cols="1">
+     <thead>
+      <row>
+       <entry role="func_table_entry"><para role="func_signature">
+        Function
+       </para>
+       <para>
+        Description
+       </para></entry>
+      </row>
+     </thead>
+
+     <tbody>
+      <row>
+       <entry role="func_table_entry"><para role="func_signature">
+        <indexterm id="pg-merge-action">
+         <primary>pg_merge_action</primary>
+        </indexterm>
+        <function>pg_merge_action</function> ( )
+        <returnvalue>text</returnvalue>
+       </para>
+       <para>
+        Returns the action performed on the current row (<literal>'INSERT'</literal>,
+        <literal>'UPDATE'</literal>, or <literal>'DELETE'</literal>).
+       </para></entry>
+      </row>
+
+      <row>
+       <entry role="func_table_entry"><para role="func_signature">
+        <indexterm id="pg-merge-when-clause">
+         <primary>pg_merge_when_clause</primary>
+        </indexterm>
+        <function>pg_merge_when_clause</function> ( )
+        <returnvalue>integer</returnvalue>
+       </para>
+       <para>
+        Returns the 1-based index of the <literal>WHEN</literal> clause executed
+        for the current row.
+       </para></entry>
+      </row>
+     </tbody>
+    </tgroup>
+  </table>
+
+  <para>
+   Example:
+<screen><![CDATA[
+MERGE INTO products p
+  USING stock s ON p.product_id = s.product_id
+  WHEN MATCHED AND s.quantity > 0 THEN
+    UPDATE SET in_stock = true, quantity = s.quantity
+  WHEN MATCHED THEN
+    UPDATE SET in_stock = false, quantity = 0
+  WHEN NOT MATCHED THEN
+    INSERT (product_id, in_stock, quantity)
+      VALUES (s.product_id, true, s.quantity)
+  RETURNING pg_merge_action() AS action,
+            pg_merge_when_clause() AS when_clause,
+            p.*;
+
+ action | when_clause | product_id | in_stock | quantity 
+--------+-------------+------------+----------+----------
+ UPDATE |           1 |       1001 | t        |       50
+ UPDATE |           2 |       1002 | f        |        0
+ INSERT |           3 |       1003 | t        |       10
+]]></screen>
+  </para>
+
+  <para>
+   Note that these functions can only be used in the <literal>RETURNING</literal>
+   list of a <command>MERGE</command> command.  It is an error to use them in
+   any other part of a query, or in a subquery.
+  </para>
+
+ </sect1>
+
  <sect1 id="functions-subquery">
   <title>Subquery Expressions</title>
 
diff --git a/doc/src/sgml/glossary.sgml b/doc/src/sgml/glossary.sgml
new file mode 100644
index fe8def4..51a15ca
--- a/doc/src/sgml/glossary.sgml
+++ b/doc/src/sgml/glossary.sgml
@@ -1387,9 +1387,9 @@
      to a <glossterm linkend="glossary-client">client</glossterm> upon the
      completion of an <acronym>SQL</acronym> command, usually a
      <command>SELECT</command> but it can be an
-     <command>INSERT</command>, <command>UPDATE</command>, or
-     <command>DELETE</command> command if the <literal>RETURNING</literal>
-     clause is specified.
+     <command>INSERT</command>, <command>UPDATE</command>,
+     <command>DELETE</command>, or <command>MERGE</command> command if the
+     <literal>RETURNING</literal> clause is specified.
     </para>
     <para>
      The fact that a result set is a relation means that a query can be used
diff --git a/doc/src/sgml/plpgsql.sgml b/doc/src/sgml/plpgsql.sgml
new file mode 100644
index f55e901..6803240
--- a/doc/src/sgml/plpgsql.sgml
+++ b/doc/src/sgml/plpgsql.sgml
@@ -1020,8 +1020,8 @@ INSERT INTO mytable VALUES (1,'one'), (2
     </para>
 
     <para>
-     If the command does return rows (for example <command>SELECT</command>,
-     or <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>
+     If the command does return rows (for example <command>SELECT</command>, or
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
      with <literal>RETURNING</literal>), there are two ways to proceed.
      When the command will return at most one row, or you only care about
      the first row of output, write the command as usual but add
@@ -1149,6 +1149,7 @@ SELECT <replaceable>select_expressions</
 INSERT ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 UPDATE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 DELETE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
+MERGE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 </synopsis>
 
      where <replaceable>target</replaceable> can be a record variable, a row
@@ -1159,8 +1160,8 @@ DELETE ... RETURNING <replaceable>expres
      <literal>INTO</literal> clause) just as described above,
      and the plan is cached in the same way.
      This works for <command>SELECT</command>,
-     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
-     <literal>RETURNING</literal>, and certain utility commands
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
+     with <literal>RETURNING</literal>, and certain utility commands
      that return row sets, such as <command>EXPLAIN</command>.
      Except for the <literal>INTO</literal> clause, the SQL command is the same
      as it would be written outside <application>PL/pgSQL</application>.
@@ -1236,7 +1237,7 @@ END;
     </para>
 
     <para>
-     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
+     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command> with
      <literal>RETURNING</literal>, <application>PL/pgSQL</application> reports
      an error for more than one returned row, even when
      <literal>STRICT</literal> is not specified.  This is because there
diff --git a/doc/src/sgml/queries.sgml b/doc/src/sgml/queries.sgml
new file mode 100644
index 3f95849..9928373
--- a/doc/src/sgml/queries.sgml
+++ b/doc/src/sgml/queries.sgml
@@ -2062,9 +2062,10 @@ SELECT <replaceable>select_list</replace
    Table Expressions or <acronym>CTE</acronym>s, can be thought of as defining
    temporary tables that exist just for one query.  Each auxiliary statement
    in a <literal>WITH</literal> clause can be a <command>SELECT</command>,
-   <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>; and the
+   <command>INSERT</command>, <command>UPDATE</command>, <command>DELETE</command>,
+   or <command>MERGE</command>; and the
    <literal>WITH</literal> clause itself is attached to a primary statement that can
-   be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
+   also be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
    <command>DELETE</command>, or <command>MERGE</command>.
   </para>
 
@@ -2598,8 +2599,8 @@ SELECT * FROM w AS w1 JOIN w AS w2 ON w1
    <title>Data-Modifying Statements in <literal>WITH</literal></title>
 
    <para>
-    You can use most data-modifying statements (<command>INSERT</command>,
-    <command>UPDATE</command>, or <command>DELETE</command>, but not
+    You can use data-modifying statements (<command>INSERT</command>,
+    <command>UPDATE</command>, <command>DELETE</command>, or
     <command>MERGE</command>) in <literal>WITH</literal>.  This
     allows you to perform several different operations in the same query.
     An example is:
diff --git a/doc/src/sgml/ref/copy.sgml b/doc/src/sgml/ref/copy.sgml
new file mode 100644
index 5e591ed..dbc76df
--- a/doc/src/sgml/ref/copy.sgml
+++ b/doc/src/sgml/ref/copy.sgml
@@ -122,13 +122,16 @@ COPY { <replaceable class="parameter">ta
       A <link linkend="sql-select"><command>SELECT</command></link>,
       <link linkend="sql-values"><command>VALUES</command></link>,
       <link linkend="sql-insert"><command>INSERT</command></link>,
-      <link linkend="sql-update"><command>UPDATE</command></link>, or
-      <link linkend="sql-delete"><command>DELETE</command></link> command whose results are to be
-      copied.  Note that parentheses are required around the query.
+      <link linkend="sql-update"><command>UPDATE</command></link>,
+      <link linkend="sql-delete"><command>DELETE</command></link>, or
+      <link linkend="sql-merge"><command>MERGE</command></link> command
+      whose results are to be copied.  Note that parentheses are required
+      around the query.
      </para>
      <para>
-      For <command>INSERT</command>, <command>UPDATE</command> and
-      <command>DELETE</command> queries a RETURNING clause must be provided,
+      For <command>INSERT</command>, <command>UPDATE</command>,
+      <command>DELETE</command>, and <command>MERGE</command> queries a
+      RETURNING clause must be provided,
       and the target relation must not have a conditional rule, nor
       an <literal>ALSO</literal> rule, nor an <literal>INSTEAD</literal> rule
       that expands to multiple statements.
diff --git a/doc/src/sgml/ref/merge.sgml b/doc/src/sgml/ref/merge.sgml
new file mode 100644
index 0995fe0..7eb641a
--- a/doc/src/sgml/ref/merge.sgml
+++ b/doc/src/sgml/ref/merge.sgml
@@ -25,6 +25,7 @@ PostgreSQL documentation
 MERGE INTO [ ONLY ] <replaceable class="parameter">target_table_name</replaceable> [ * ] [ [ AS ] <replaceable class="parameter">target_alias</replaceable> ]
 USING <replaceable class="parameter">data_source</replaceable> ON <replaceable class="parameter">join_condition</replaceable>
 <replaceable class="parameter">when_clause</replaceable> [...]
+[ RETURNING * | <replaceable class="parameter">output_expression</replaceable> [ [ AS ] <replaceable class="parameter">output_name</replaceable> ] [, ...] ]
 
 <phrase>where <replaceable class="parameter">data_source</replaceable> is:</phrase>
 
@@ -95,6 +96,20 @@ DELETE
   </para>
 
   <para>
+   The optional <literal>RETURNING</literal> clause causes <command>MERGE</command>
+   to compute and return value(s) based on each row inserted, updated, or
+   deleted.  Any expression using the source or target table's columns, or the
+   merge support functions <xref linkend="pg-merge-action"/> and
+   <xref linkend="pg-merge-when-clause"/> can be computed.  When an
+   <command>INSERT</command> or <command>UPDATE</command> action is performed,
+   the new values of the target table's columns are used.  When a
+   <command>DELETE</command> is performed, the old values of the target table's
+   columns are used.  The syntax of the <literal>RETURNING</literal> list is
+   identical to that of the output list of <command>SELECT</command>.
+
+  </para>
+
+  <para>
    There is no separate <literal>MERGE</literal> privilege.
    If you specify an update action, you must have the
    <literal>UPDATE</literal> privilege on the column(s)
@@ -125,6 +140,18 @@ DELETE
 
   <variablelist>
    <varlistentry>
+    <term><replaceable class="parameter">with_query</replaceable></term>
+    <listitem>
+     <para>
+      The <literal>WITH</literal> clause allows you to specify one or more
+      subqueries that can be referenced by name in the <command>MERGE</command>
+      query. See <xref linkend="queries-with"/> and <xref linkend="sql-select"/>
+      for details.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
     <term><replaceable class="parameter">target_table_name</replaceable></term>
     <listitem>
      <para>
@@ -399,13 +426,30 @@ DELETE
    </varlistentry>
 
    <varlistentry>
-    <term><replaceable class="parameter">with_query</replaceable></term>
+    <term><replaceable class="parameter">output_expression</replaceable></term>
     <listitem>
      <para>
-      The <literal>WITH</literal> clause allows you to specify one or more
-      subqueries that can be referenced by name in the <command>MERGE</command>
-      query. See <xref linkend="queries-with"/> and <xref linkend="sql-select"/>
-      for details.
+      An expression to be computed and returned by the <command>MERGE</command>
+      command after each row is merged.  The expression can use any columns of
+      the source or target tables, or the merge support functions
+      <xref linkend="pg-merge-action"/> and <xref linkend="pg-merge-when-clause"/>.
+     </para>
+     <para>
+      Writing <literal>*</literal> will return all columns from the source
+      table, followed by all columns from the target table.  Often this will
+      lead to a lot of duplication, since it is common for the source and
+      target tables to have a lot of the same columns.  This can be avoided by
+      qualifying the <literal>*</literal> with the name or alias of the source
+      or target table.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
+    <term><replaceable class="parameter">output_name</replaceable></term>
+    <listitem>
+     <para>
+      A name to use for a returned column.
      </para>
     </listitem>
    </varlistentry>
@@ -428,6 +472,13 @@ MERGE <replaceable class="parameter">tot
    were changed in any way.
   </para>
 
+  <para>
+   If the <command>MERGE</command> command contains a <literal>RETURNING</literal>
+   clause, the result will be similar to that of a <command>SELECT</command>
+   statement containing the columns and values defined in the
+   <literal>RETURNING</literal> list, computed over the row(s) inserted, updated,
+   or deleted by the command.
+  </para>
  </refsect1>
 
  <refsect1>
@@ -544,13 +595,6 @@ MERGE <replaceable class="parameter">tot
   </para>
 
   <para>
-   There is no <literal>RETURNING</literal> clause with
-   <command>MERGE</command>.  Actions of <command>INSERT</command>,
-   <command>UPDATE</command> and <command>DELETE</command> cannot contain
-   <literal>RETURNING</literal> or <literal>WITH</literal> clauses.
-  </para>
-
-  <para>
    When <command>MERGE</command> is run concurrently with other commands
    that modify the target table, the usual transaction isolation rules
    apply; see <xref linkend="transaction-iso"/> for an explanation
@@ -602,7 +646,8 @@ WHEN NOT MATCHED THEN
   <para>
    Attempt to insert a new stock item along with the quantity of stock. If
    the item already exists, instead update the stock count of the existing
-   item. Don't allow entries that have zero stock.
+   item. Don't allow entries that have zero stock. Return details of all
+   changes made.
 <programlisting>
 MERGE INTO wines w
 USING wine_stock_changes s
@@ -612,7 +657,8 @@ WHEN NOT MATCHED AND s.stock_delta > 0 T
 WHEN MATCHED AND w.stock + s.stock_delta > 0 THEN
   UPDATE SET stock = w.stock + s.stock_delta
 WHEN MATCHED THEN
-  DELETE;
+  DELETE
+RETURNING pg_merge_action(), w.*;
 </programlisting>
 
    The <literal>wine_stock_changes</literal> table might be, for example, a
@@ -627,7 +673,8 @@ WHEN MATCHED THEN
     This command conforms to the <acronym>SQL</acronym> standard.
   </para>
    <para>
-    The WITH clause and <literal>DO NOTHING</literal> action are extensions to
+    The WITH clause, <literal>DO NOTHING</literal> action, and
+    <literal>RETURNING</literal> clause are extensions to
     the <acronym>SQL</acronym> standard.
   </para>
  </refsect1>
diff --git a/doc/src/sgml/ref/select.sgml b/doc/src/sgml/ref/select.sgml
new file mode 100644
index 0ee0cc7..f81bdea
--- a/doc/src/sgml/ref/select.sgml
+++ b/doc/src/sgml/ref/select.sgml
@@ -74,7 +74,7 @@ SELECT [ ALL | DISTINCT [ ON ( <replacea
 
 <phrase>and <replaceable class="parameter">with_query</replaceable> is:</phrase>
 
-    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> )
+    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> | <replaceable class="parameter">merge</replaceable> )
         [ SEARCH { BREADTH | DEPTH } FIRST BY <replaceable>column_name</replaceable> [, ...] SET <replaceable>search_seq_col_name</replaceable> ]
         [ CYCLE <replaceable>column_name</replaceable> [, ...] SET <replaceable>cycle_mark_col_name</replaceable> [ TO <replaceable>cycle_mark_value</replaceable> DEFAULT <replaceable>cycle_mark_default</replaceable> ] USING <replaceable>cycle_path_col_name</replaceable> ]
 
@@ -227,10 +227,10 @@ TABLE [ ONLY ] <replaceable class="param
     The subqueries effectively act as temporary tables or views
     for the duration of the primary query.
     Each subquery can be a <command>SELECT</command>, <command>TABLE</command>, <command>VALUES</command>,
-    <command>INSERT</command>, <command>UPDATE</command> or
-    <command>DELETE</command> statement.
+    <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command> statement.
     When writing a data-modifying statement (<command>INSERT</command>,
-    <command>UPDATE</command> or <command>DELETE</command>) in
+    <command>UPDATE</command>, <command>DELETE</command>, or <command>MERGE</command>) in
     <literal>WITH</literal>, it is usual to include a <literal>RETURNING</literal> clause.
     It is the output of <literal>RETURNING</literal>, <emphasis>not</emphasis> the underlying
     table that the statement modifies, that forms the temporary table that is
@@ -2179,7 +2179,8 @@ SELECT 2+2;
 
    <para>
     <productname>PostgreSQL</productname> allows <command>INSERT</command>,
-    <command>UPDATE</command>, and <command>DELETE</command> to be used as <literal>WITH</literal>
+    <command>UPDATE</command>, <command>DELETE</command>, and
+    <command>MERGE</command> to be used as <literal>WITH</literal>
     queries.  This is not found in the SQL standard.
    </para>
   </refsect2>
diff --git a/doc/src/sgml/rowtypes.sgml b/doc/src/sgml/rowtypes.sgml
new file mode 100644
index 4d86f97..bbeac84
--- a/doc/src/sgml/rowtypes.sgml
+++ b/doc/src/sgml/rowtypes.sgml
@@ -348,7 +348,7 @@ SELECT m.* FROM some_table, LATERAL myfu
    column expansion of this kind when it appears at the top level of
    a <link linkend="queries-select-lists"><command>SELECT</command> output
    list</link>, a <link linkend="dml-returning"><literal>RETURNING</literal>
-   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>,
+   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>,
    a <link linkend="queries-values"><literal>VALUES</literal> clause</link>, or
    a <link linkend="sql-syntax-row-constructors">row constructor</link>.
    In all other contexts (including when nested inside one of those
diff --git a/doc/src/sgml/spi.sgml b/doc/src/sgml/spi.sgml
new file mode 100644
index 651930a..ad0a9b1
--- a/doc/src/sgml/spi.sgml
+++ b/doc/src/sgml/spi.sgml
@@ -301,8 +301,9 @@ SPI_execute("INSERT INTO foo SELECT * FR
    is returned in the global variable <varname>SPI_processed</varname>.
    If the return value of the function is <symbol>SPI_OK_SELECT</symbol>,
    <symbol>SPI_OK_INSERT_RETURNING</symbol>,
-   <symbol>SPI_OK_DELETE_RETURNING</symbol>, or
-   <symbol>SPI_OK_UPDATE_RETURNING</symbol>,
+   <symbol>SPI_OK_DELETE_RETURNING</symbol>,
+   <symbol>SPI_OK_UPDATE_RETURNING</symbol>, or
+   <symbol>SPI_OK_MERGE_RETURNING</symbol>
    then you can use the
    global pointer <literal>SPITupleTable *SPI_tuptable</literal> to
    access the result rows.  Some utility commands (such as
@@ -472,6 +473,15 @@ typedef struct SPITupleTable
       </para>
      </listitem>
     </varlistentry>
+
+    <varlistentry>
+     <term><symbol>SPI_OK_MERGE_RETURNING</symbol></term>
+     <listitem>
+      <para>
+       if a <command>MERGE RETURNING</command> was executed
+      </para>
+     </listitem>
+    </varlistentry>
 
     <varlistentry>
      <term><symbol>SPI_OK_UTILITY</symbol></term>
diff --git a/doc/src/sgml/xfunc.sgml b/doc/src/sgml/xfunc.sgml
new file mode 100644
index 9620ea9..4f68efb
--- a/doc/src/sgml/xfunc.sgml
+++ b/doc/src/sgml/xfunc.sgml
@@ -177,7 +177,8 @@
     statements separated by semicolons.  A semicolon after the last
     statement is optional.  Unless the function is declared to return
     <type>void</type>, the last statement must be a <command>SELECT</command>,
-    or an <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>
+    or an <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command>
     that has a <literal>RETURNING</literal> clause.
    </para>
 
@@ -1236,8 +1237,9 @@ SELECT x, CASE WHEN x &gt; 0 THEN genera
 
     <note>
      <para>
-      If a function's last command is <command>INSERT</command>, <command>UPDATE</command>,
-      or <command>DELETE</command> with <literal>RETURNING</literal>, that command will
+      If a function's last command is <command>INSERT</command>,
+      <command>UPDATE</command>, <command>DELETE</command>, or
+      <command>MERGE</command> with <literal>RETURNING</literal>, that command will
       always be executed to completion, even if the function is not declared
       with <literal>SETOF</literal> or the calling query does not fetch all the
       result rows.  Any extra rows produced by the <literal>RETURNING</literal>
diff --git a/src/backend/commands/copy.c b/src/backend/commands/copy.c
new file mode 100644
index f14fae3..05f2e5b
--- a/src/backend/commands/copy.c
+++ b/src/backend/commands/copy.c
@@ -283,12 +283,6 @@ DoCopy(ParseState *pstate, const CopyStm
 	{
 		Assert(stmt->query);
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(stmt->query, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in COPY"));
-
 		query = makeNode(RawStmt);
 		query->stmt = stmt->query;
 		query->stmt_location = stmt_location;
diff --git a/src/backend/commands/copyto.c b/src/backend/commands/copyto.c
new file mode 100644
index 9e4b243..cf56f9f
--- a/src/backend/commands/copyto.c
+++ b/src/backend/commands/copyto.c
@@ -510,7 +510,8 @@ BeginCopyTo(ParseState *pstate,
 		{
 			Assert(query->commandType == CMD_INSERT ||
 				   query->commandType == CMD_UPDATE ||
-				   query->commandType == CMD_DELETE);
+				   query->commandType == CMD_DELETE ||
+				   query->commandType == CMD_MERGE);
 
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
diff --git a/src/backend/executor/execExpr.c b/src/backend/executor/execExpr.c
new file mode 100644
index e6e6168..5ea6c14
--- a/src/backend/executor/execExpr.c
+++ b/src/backend/executor/execExpr.c
@@ -48,6 +48,7 @@
 #include "utils/array.h"
 #include "utils/builtins.h"
 #include "utils/datum.h"
+#include "utils/fmgroids.h"
 #include "utils/lsyscache.h"
 #include "utils/typcache.h"
 
@@ -2603,6 +2604,22 @@ ExecInitFunc(ExprEvalStep *scratch, Expr
 	InitFunctionCallInfoData(*fcinfo, flinfo,
 							 nargs, inputcollid, NULL, NULL);
 
+	/*
+	 * Merge support functions should only be called directly from a MERGE
+	 * command, and need access to the parent ModifyTableState.  The parser
+	 * should have checked that such functions only appear in the RETURNING
+	 * list of a MERGE, so this should never fail.
+	 */
+	if (IsMergeSupportFunction(funcid))
+	{
+		if (!state->parent ||
+			!IsA(state->parent, ModifyTableState) ||
+			((ModifyTableState *) state->parent)->operation != CMD_MERGE)
+			elog(ERROR, "merge support function called in non-merge context");
+
+		fcinfo->context = (Node *) state->parent;
+	}
+
 	/* Keep extra copies of this info to save an indirection at runtime */
 	scratch->d.func.fn_addr = flinfo->fn_addr;
 	scratch->d.func.nargs = nargs;
diff --git a/src/backend/executor/execPartition.c b/src/backend/executor/execPartition.c
new file mode 100644
index eb8a87f..b9bd03c
--- a/src/backend/executor/execPartition.c
+++ b/src/backend/executor/execPartition.c
@@ -612,8 +612,8 @@ ExecInitPartitionInfo(ModifyTableState *
 	 * Build the RETURNING projection for the partition.  Note that we didn't
 	 * build the returningList for partitions within the planner, but simple
 	 * translation of varattnos will suffice.  This only occurs for the INSERT
-	 * case or in the case of UPDATE tuple routing where we didn't find a
-	 * result rel to reuse.
+	 * case or in the case of UPDATE/MERGE tuple routing where we didn't find
+	 * a result rel to reuse.
 	 */
 	if (node && node->returningLists != NIL)
 	{
@@ -622,12 +622,14 @@ ExecInitPartitionInfo(ModifyTableState *
 		List	   *returningList;
 
 		/* See the comment above for WCO lists. */
-		/* (except no RETURNING support for MERGE yet) */
 		Assert((node->operation == CMD_INSERT &&
 				list_length(node->returningLists) == 1 &&
 				list_length(node->resultRelations) == 1) ||
 			   (node->operation == CMD_UPDATE &&
 				list_length(node->returningLists) ==
+				list_length(node->resultRelations)) ||
+			   (node->operation == CMD_MERGE &&
+				list_length(node->returningLists) ==
 				list_length(node->resultRelations)));
 
 		/*
diff --git a/src/backend/executor/functions.c b/src/backend/executor/functions.c
new file mode 100644
index f55424e..f99f15f
--- a/src/backend/executor/functions.c
+++ b/src/backend/executor/functions.c
@@ -1660,8 +1660,8 @@ check_sql_fn_retval(List *queryTreeLists
 
 	/*
 	 * If it's a plain SELECT, it returns whatever the targetlist says.
-	 * Otherwise, if it's INSERT/UPDATE/DELETE with RETURNING, it returns
-	 * that. Otherwise, the function return type must be VOID.
+	 * Otherwise, if it's INSERT/UPDATE/DELETE/MERGE with RETURNING, it
+	 * returns that. Otherwise, the function return type must be VOID.
 	 *
 	 * Note: eventually replace this test with QueryReturnsTuples?	We'd need
 	 * a more general method of determining the output type, though.  Also, it
@@ -1679,7 +1679,8 @@ check_sql_fn_retval(List *queryTreeLists
 	else if (parse &&
 			 (parse->commandType == CMD_INSERT ||
 			  parse->commandType == CMD_UPDATE ||
-			  parse->commandType == CMD_DELETE) &&
+			  parse->commandType == CMD_DELETE ||
+			  parse->commandType == CMD_MERGE) &&
 			 parse->returningList)
 	{
 		tlist = parse->returningList;
diff --git a/src/backend/executor/nodeModifyTable.c b/src/backend/executor/nodeModifyTable.c
new file mode 100644
index 2a5fec8..b4c61ec
--- a/src/backend/executor/nodeModifyTable.c
+++ b/src/backend/executor/nodeModifyTable.c
@@ -36,8 +36,7 @@
  *		RETURNING tuple after completing each row insert, update, or delete.
  *		It must be called again to continue the operation.  Without RETURNING,
  *		we just loop within the node until all the work is done, then
- *		return NULL.  This avoids useless call/return overhead.  (MERGE does
- *		not support RETURNING.)
+ *		return NULL.  This avoids useless call/return overhead.
  */
 
 #include "postgres.h"
@@ -88,9 +87,6 @@ typedef struct ModifyTableContext
 	 */
 	TupleTableSlot *planSlot;
 
-	/* MERGE specific */
-	MergeActionState *relaction;	/* MERGE action in progress */
-
 	/*
 	 * Information about the changes that were made concurrently to a tuple
 	 * being updated or deleted
@@ -153,13 +149,14 @@ static TupleTableSlot *ExecMerge(ModifyT
 								 ItemPointer tupleid,
 								 bool canSetTag);
 static void ExecInitMerge(ModifyTableState *mtstate, EState *estate);
-static bool ExecMergeMatched(ModifyTableContext *context,
-							 ResultRelInfo *resultRelInfo,
-							 ItemPointer tupleid,
-							 bool canSetTag);
-static void ExecMergeNotMatched(ModifyTableContext *context,
-								ResultRelInfo *resultRelInfo,
-								bool canSetTag);
+static TupleTableSlot *ExecMergeMatched(ModifyTableContext *context,
+										ResultRelInfo *resultRelInfo,
+										ItemPointer tupleid,
+										bool canSetTag,
+										bool *matched);
+static TupleTableSlot *ExecMergeNotMatched(ModifyTableContext *context,
+										   ResultRelInfo *resultRelInfo,
+										   bool canSetTag);
 
 
 /*
@@ -979,7 +976,7 @@ ExecInsert(ModifyTableContext *context,
 		if (mtstate->operation == CMD_UPDATE)
 			wco_kind = WCO_RLS_UPDATE_CHECK;
 		else if (mtstate->operation == CMD_MERGE)
-			wco_kind = (context->relaction->mas_action->commandType == CMD_UPDATE) ?
+			wco_kind = (mtstate->mt_merge_action->mas_action->commandType == CMD_UPDATE) ?
 				WCO_RLS_UPDATE_CHECK : WCO_RLS_INSERT_CHECK;
 		else
 			wco_kind = WCO_RLS_INSERT_CHECK;
@@ -1828,7 +1825,7 @@ ExecCrossPartitionUpdate(ModifyTableCont
 		 * additional rechecking, and might end up executing a different
 		 * action entirely).
 		 */
-		if (context->relaction != NULL)
+		if (mtstate->mt_merge_action != NULL)
 			return false;
 		else if (TupIsNull(epqslot))
 			return true;
@@ -2069,7 +2066,7 @@ lreplace:
 		 * No luck, a retry is needed.  If running MERGE, we do not do so
 		 * here; instead let it handle that on its own rules.
 		 */
-		if (context->relaction != NULL)
+		if (context->mtstate->mt_merge_action != NULL)
 			return TM_Updated;
 
 		/*
@@ -2712,6 +2709,7 @@ static TupleTableSlot *
 ExecMerge(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 		  ItemPointer tupleid, bool canSetTag)
 {
+	TupleTableSlot *rslot = NULL;
 	bool		matched;
 
 	/*-----
@@ -2759,18 +2757,18 @@ ExecMerge(ModifyTableContext *context, R
 	 */
 	matched = tupleid != NULL;
 	if (matched)
-		matched = ExecMergeMatched(context, resultRelInfo, tupleid, canSetTag);
+		rslot = ExecMergeMatched(context, resultRelInfo, tupleid, canSetTag,
+								 &matched);
 
 	/*
-	 * Either we were dealing with a NOT MATCHED tuple or ExecMergeMatched()
-	 * returned "false", indicating the previously MATCHED tuple no longer
-	 * matches.
+	 * Deal with the NOT MATCHED case (either a NOT MATCHED tuple from the
+	 * join, or a previously MATCHED tuple for which ExecMergeMatched() set
+	 * "matched" to false, indicating that it no longer matches).
 	 */
 	if (!matched)
-		ExecMergeNotMatched(context, resultRelInfo, canSetTag);
+		rslot = ExecMergeNotMatched(context, resultRelInfo, canSetTag);
 
-	/* No RETURNING support yet */
-	return NULL;
+	return rslot;
 }
 
 /*
@@ -2780,8 +2778,8 @@ ExecMerge(ModifyTableContext *context, R
  * We start from the first WHEN MATCHED action and check if the WHEN quals
  * pass, if any. If the WHEN quals for the first action do not pass, we
  * check the second, then the third and so on. If we reach to the end, no
- * action is taken and we return true, indicating that no further action is
- * required for this tuple.
+ * action is taken and "matched" is set to true, indicating that no further
+ * action is required for this tuple.
  *
  * If we do find a qualifying action, then we attempt to execute the action.
  *
@@ -2790,15 +2788,16 @@ ExecMerge(ModifyTableContext *context, R
  * with individual actions are evaluated by this routine via ExecQual, while
  * EvalPlanQual checks for the join quals. If EvalPlanQual tells us that the
  * updated tuple still passes the join quals, then we restart from the first
- * action to look for a qualifying action. Otherwise, we return false --
- * meaning that a NOT MATCHED action must now be executed for the current
- * source tuple.
+ * action to look for a qualifying action. Otherwise, "matched" is set to
+ * false -- meaning that a NOT MATCHED action must now be executed for the
+ * current source tuple.
  */
-static bool
+static TupleTableSlot *
 ExecMergeMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
-				 ItemPointer tupleid, bool canSetTag)
+				 ItemPointer tupleid, bool canSetTag, bool *matched)
 {
 	ModifyTableState *mtstate = context->mtstate;
+	TupleTableSlot *rslot = NULL;
 	TupleTableSlot *newslot;
 	EState	   *estate = context->estate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
@@ -2810,7 +2809,10 @@ ExecMergeMatched(ModifyTableContext *con
 	 * If there are no WHEN MATCHED actions, we are done.
 	 */
 	if (resultRelInfo->ri_matchedMergeAction == NIL)
-		return true;
+	{
+		*matched = true;
+		return NULL;
+	}
 
 	/*
 	 * Make tuple and any needed join variables available to ExecQual and
@@ -2888,12 +2890,15 @@ lmerge_matched:
 				 */
 				newslot = ExecProject(relaction->mas_proj);
 
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				if (!ExecUpdatePrologue(context, resultRelInfo,
 										tupleid, NULL, newslot, &result))
 				{
 					if (result == TM_Ok)
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 					break;		/* concurrent update/delete */
 				}
 				result = ExecUpdateAct(context, resultRelInfo, tupleid, NULL,
@@ -2907,12 +2912,15 @@ lmerge_matched:
 				break;
 
 			case CMD_DELETE:
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				if (!ExecDeletePrologue(context, resultRelInfo, tupleid,
 										NULL, NULL, &result))
 				{
 					if (result == TM_Ok)
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 					break;		/* concurrent update/delete */
 				}
 				result = ExecDeleteAct(context, resultRelInfo, tupleid, false);
@@ -2968,7 +2976,8 @@ lmerge_matched:
 				 * If the tuple was already deleted, return to let caller
 				 * handle it under NOT MATCHED clauses.
 				 */
-				return false;
+				*matched = false;
+				return NULL;
 
 			case TM_Updated:
 				{
@@ -3014,13 +3023,19 @@ lmerge_matched:
 							 * NOT MATCHED actions.
 							 */
 							if (TupIsNull(epqslot))
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							(void) ExecGetJunkAttribute(epqslot,
 														resultRelInfo->ri_RowIdAttNo,
 														&isNull);
 							if (isNull)
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							/*
 							 * When a tuple was updated and migrated to
@@ -3055,7 +3070,8 @@ lmerge_matched:
 							 * tuple already deleted; tell caller to run NOT
 							 * MATCHED actions
 							 */
-							return false;
+							*matched = false;
+							return NULL;
 
 						case TM_SelfModified:
 
@@ -3075,13 +3091,14 @@ lmerge_matched:
 										(errcode(ERRCODE_TRIGGERED_DATA_CHANGE_VIOLATION),
 										 errmsg("tuple to be updated or deleted was already modified by an operation triggered by the current command"),
 										 errhint("Consider using an AFTER trigger instead of a BEFORE trigger to propagate changes to other rows.")));
-							return false;
+							*matched = false;
+							return NULL;
 
 						default:
 							/* see table_tuple_lock call in ExecDelete() */
 							elog(ERROR, "unexpected table_tuple_lock status: %u",
 								 result);
-							return false;
+							return NULL;
 					}
 				}
 
@@ -3093,6 +3110,30 @@ lmerge_matched:
 				break;
 		}
 
+		/* Process RETURNING if present */
+		if (resultRelInfo->ri_projectReturning)
+		{
+			switch (commandType)
+			{
+				case CMD_UPDATE:
+					rslot = ExecProcessReturning(resultRelInfo, newslot,
+												 context->planSlot);
+					break;
+
+				case CMD_DELETE:
+					rslot = ExecProcessReturning(resultRelInfo,
+												 resultRelInfo->ri_oldTupleSlot,
+												 context->planSlot);
+					break;
+
+				case CMD_NOTHING:
+					break;
+
+				default:
+					elog(ERROR, "unknown action in MERGE WHEN MATCHED clause");
+			}
+		}
+
 		/*
 		 * We've activated one of the WHEN clauses, so we don't search
 		 * further. This is required behaviour, not an optimization.
@@ -3103,19 +3144,22 @@ lmerge_matched:
 	/*
 	 * Successfully executed an action or no qualifying action was found.
 	 */
-	return true;
+	*matched = true;
+
+	return rslot;
 }
 
 /*
  * Execute the first qualifying NOT MATCHED action.
  */
-static void
+static TupleTableSlot *
 ExecMergeNotMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 					bool canSetTag)
 {
 	ModifyTableState *mtstate = context->mtstate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
 	List	   *actionStates = NIL;
+	TupleTableSlot *rslot = NULL;
 	ListCell   *l;
 
 	/*
@@ -3165,10 +3209,10 @@ ExecMergeNotMatched(ModifyTableContext *
 				 * so we don't need to map the tuple here.
 				 */
 				newslot = ExecProject(action->mas_proj);
-				context->relaction = action;
+				mtstate->mt_merge_action = action;
 
-				(void) ExecInsert(context, mtstate->rootResultRelInfo, newslot,
-								  canSetTag, NULL, NULL);
+				rslot = ExecInsert(context, mtstate->rootResultRelInfo,
+								   newslot, canSetTag, NULL, NULL);
 				mtstate->mt_merge_inserted += 1;
 				break;
 			case CMD_NOTHING:
@@ -3184,6 +3228,8 @@ ExecMergeNotMatched(ModifyTableContext *
 		 */
 		break;
 	}
+
+	return rslot;
 }
 
 /*
@@ -3361,6 +3407,64 @@ ExecInitMergeTupleSlots(ModifyTableState
 }
 
 /*
+ * pg_merge_action() -
+ *	  SQL merge support function to retrieve the currently executing merge
+ *	  action command string ("INSERT", "UPDATE", or "DELETE").
+ */
+Datum
+pg_merge_action(PG_FUNCTION_ARGS)
+{
+	ModifyTableState *mtstate = (ModifyTableState *) fcinfo->context;
+	MergeActionState *relaction;
+
+	if (!mtstate || mtstate->operation != CMD_MERGE)
+		elog(ERROR, "merge support function called in non-merge context");
+
+	relaction = mtstate->mt_merge_action;
+	if (relaction)
+	{
+		CmdType		commandType = relaction->mas_action->commandType;
+
+		switch (commandType)
+		{
+			case CMD_INSERT:
+				PG_RETURN_TEXT_P(cstring_to_text("INSERT"));
+			case CMD_UPDATE:
+				PG_RETURN_TEXT_P(cstring_to_text("UPDATE"));
+			case CMD_DELETE:
+				PG_RETURN_TEXT_P(cstring_to_text("DELETE"));
+			case CMD_NOTHING:
+				PG_RETURN_NULL();
+			default:
+				elog(ERROR, "unrecognized commandType: %d", (int) commandType);
+		}
+	}
+
+	PG_RETURN_NULL();
+}
+
+/*
+ * pg_merge_when_clause() -
+ *	  SQL merge support function to retrieve the 1-based index of the
+ *	  currently executing merge WHEN clause.
+ */
+Datum
+pg_merge_when_clause(PG_FUNCTION_ARGS)
+{
+	ModifyTableState *mtstate = (ModifyTableState *) fcinfo->context;
+	MergeActionState *relaction;
+
+	if (!mtstate || mtstate->operation != CMD_MERGE)
+		elog(ERROR, "merge support function called in non-merge context");
+
+	relaction = mtstate->mt_merge_action;
+	if (relaction)
+		PG_RETURN_INT32((int32) relaction->mas_action->index);
+
+	PG_RETURN_NULL();
+}
+
+/*
  * Process BEFORE EACH STATEMENT triggers
  */
 static void
@@ -3646,8 +3750,17 @@ ExecModifyTable(PlanState *pstate)
 				{
 					EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-					ExecMerge(&context, node->resultRelInfo, NULL, node->canSetTag);
-					continue;	/* no RETURNING support yet */
+					slot = ExecMerge(&context, node->resultRelInfo, NULL,
+									 node->canSetTag);
+
+					/*
+					 * If we got a RETURNING result, return it to the caller.
+					 * We'll continue the work on next call.
+					 */
+					if (slot)
+						return slot;
+
+					continue;	/* continue with the next tuple */
 				}
 
 				elog(ERROR, "tableoid is NULL");
@@ -3724,8 +3837,17 @@ ExecModifyTable(PlanState *pstate)
 					{
 						EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-						ExecMerge(&context, node->resultRelInfo, NULL, node->canSetTag);
-						continue;	/* no RETURNING support yet */
+						slot = ExecMerge(&context, node->resultRelInfo, NULL,
+										 node->canSetTag);
+
+						/*
+						 * If we got a RETURNING result, return it to the
+						 * caller.  We'll continue the work on next call.
+						 */
+						if (slot)
+							return slot;
+
+						continue;	/* continue with the next tuple */
 					}
 
 					elog(ERROR, "ctid is NULL");
@@ -3817,7 +3939,7 @@ ExecModifyTable(PlanState *pstate)
 				}
 				slot = ExecGetUpdateNewTuple(resultRelInfo, context.planSlot,
 											 oldSlot);
-				context.relaction = NULL;
+				node->mt_merge_action = NULL;
 
 				/* Now apply the update. */
 				slot = ExecUpdate(&context, resultRelInfo, tupleid, oldtuple,
diff --git a/src/backend/executor/spi.c b/src/backend/executor/spi.c
new file mode 100644
index 3397568..b9f443c
--- a/src/backend/executor/spi.c
+++ b/src/backend/executor/spi.c
@@ -2033,6 +2033,8 @@ SPI_result_code_string(int code)
 			return "SPI_OK_TD_REGISTER";
 		case SPI_OK_MERGE:
 			return "SPI_OK_MERGE";
+		case SPI_OK_MERGE_RETURNING:
+			return "SPI_OK_MERGE_RETURNING";
 	}
 	/* Unrecognized code ... return something useful ... */
 	sprintf(buf, "Unrecognized SPI code %d", code);
@@ -2886,7 +2888,10 @@ _SPI_pquery(QueryDesc *queryDesc, bool f
 				res = SPI_OK_UPDATE;
 			break;
 		case CMD_MERGE:
-			res = SPI_OK_MERGE;
+			if (queryDesc->plannedstmt->hasReturning)
+				res = SPI_OK_MERGE_RETURNING;
+			else
+				res = SPI_OK_MERGE;
 			break;
 		default:
 			return SPI_ERROR_OPUNKNOWN;
diff --git a/src/backend/parser/analyze.c b/src/backend/parser/analyze.c
new file mode 100644
index 4006632..6794511
--- a/src/backend/parser/analyze.c
+++ b/src/backend/parser/analyze.c
@@ -74,7 +74,6 @@ static void determineRecursiveColTypes(P
 									   Node *larg, List *nrtargetlist);
 static Query *transformReturnStmt(ParseState *pstate, ReturnStmt *stmt);
 static Query *transformUpdateStmt(ParseState *pstate, UpdateStmt *stmt);
-static List *transformReturningList(ParseState *pstate, List *returningList);
 static Query *transformPLAssignStmt(ParseState *pstate,
 									PLAssignStmt *stmt);
 static Query *transformDeclareCursorStmt(ParseState *pstate,
@@ -515,7 +514,8 @@ transformDeleteStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -942,7 +942,8 @@ transformInsertStmt(ParseState *pstate,
 	/* Process RETURNING, if any. */
 	if (stmt->returningList)
 		qry->returningList = transformReturningList(pstate,
-													stmt->returningList);
+													stmt->returningList,
+													EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -2407,7 +2408,8 @@ transformUpdateStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/*
 	 * Now we are done with SELECT-like processing, and can get on with
@@ -2501,10 +2503,11 @@ transformUpdateTargetList(ParseState *ps
 
 /*
  * transformReturningList -
- *	handle a RETURNING clause in INSERT/UPDATE/DELETE
+ *	handle a RETURNING clause in INSERT/UPDATE/DELETE/MERGE
  */
-static List *
-transformReturningList(ParseState *pstate, List *returningList)
+List *
+transformReturningList(ParseState *pstate, List *returningList,
+					   ParseExprKind exprKind)
 {
 	List	   *rlist;
 	int			save_next_resno;
@@ -2521,7 +2524,7 @@ transformReturningList(ParseState *pstat
 	pstate->p_next_resno = 1;
 
 	/* transform RETURNING identically to a SELECT targetlist */
-	rlist = transformTargetList(pstate, returningList, EXPR_KIND_RETURNING);
+	rlist = transformTargetList(pstate, returningList, exprKind);
 
 	/*
 	 * Complain if the nonempty tlist expanded to nothing (which is possible
diff --git a/src/backend/parser/gram.y b/src/backend/parser/gram.y
new file mode 100644
index 39ab7ea..5455607
--- a/src/backend/parser/gram.y
+++ b/src/backend/parser/gram.y
@@ -12249,6 +12249,7 @@ MergeStmt:
 			USING table_ref
 			ON a_expr
 			merge_when_list
+			returning_clause
 				{
 					MergeStmt  *m = makeNode(MergeStmt);
 
@@ -12257,6 +12258,7 @@ MergeStmt:
 					m->sourceRelation = $6;
 					m->joinCondition = $8;
 					m->mergeWhenClauses = $9;
+					m->returningList = $10;
 
 					$$ = (Node *) m;
 				}
diff --git a/src/backend/parser/parse_agg.c b/src/backend/parser/parse_agg.c
new file mode 100644
index 85cd47b..e72c37b
--- a/src/backend/parser/parse_agg.c
+++ b/src/backend/parser/parse_agg.c
@@ -456,6 +456,7 @@ check_agglevels_and_constraints(ParseSta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
@@ -903,6 +904,7 @@ transformWindowFuncCall(ParseState *psta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_cte.c b/src/backend/parser/parse_cte.c
new file mode 100644
index c5b1a49..ad3c525
--- a/src/backend/parser/parse_cte.c
+++ b/src/backend/parser/parse_cte.c
@@ -126,13 +126,6 @@ transformWithClause(ParseState *pstate,
 		CommonTableExpr *cte = (CommonTableExpr *) lfirst(lc);
 		ListCell   *rest;
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(cte->ctequery, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in WITH query"),
-					parser_errposition(pstate, cte->location));
-
 		for_each_cell(rest, withClause->ctes, lnext(withClause->ctes, lc))
 		{
 			CommonTableExpr *cte2 = (CommonTableExpr *) lfirst(rest);
@@ -153,7 +146,8 @@ transformWithClause(ParseState *pstate,
 			/* must be a data-modifying statement */
 			Assert(IsA(cte->ctequery, InsertStmt) ||
 				   IsA(cte->ctequery, UpdateStmt) ||
-				   IsA(cte->ctequery, DeleteStmt));
+				   IsA(cte->ctequery, DeleteStmt) ||
+				   IsA(cte->ctequery, MergeStmt));
 
 			pstate->p_hasModifyingCTE = true;
 		}
diff --git a/src/backend/parser/parse_expr.c b/src/backend/parser/parse_expr.c
new file mode 100644
index 346fd27..38fc8a9
--- a/src/backend/parser/parse_expr.c
+++ b/src/backend/parser/parse_expr.c
@@ -525,6 +525,7 @@ transformColumnRef(ParseState *pstate, C
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CHECK_CONSTRAINT:
@@ -1751,6 +1752,7 @@ transformSubLink(ParseState *pstate, Sub
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CYCLE_MARK:
@@ -3099,6 +3101,7 @@ ParseExprKindName(ParseExprKind exprKind
 		case EXPR_KIND_OFFSET:
 			return "OFFSET";
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			return "RETURNING";
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
diff --git a/src/backend/parser/parse_func.c b/src/backend/parser/parse_func.c
new file mode 100644
index b3f0b6a..5cc933e
--- a/src/backend/parser/parse_func.c
+++ b/src/backend/parser/parse_func.c
@@ -31,6 +31,7 @@
 #include "parser/parse_target.h"
 #include "parser/parse_type.h"
 #include "utils/builtins.h"
+#include "utils/fmgroids.h"
 #include "utils/lsyscache.h"
 #include "utils/syscache.h"
 
@@ -348,6 +349,15 @@ ParseFuncOrColumn(ParseState *pstate, Li
 					 parser_errposition(pstate, location)));
 	}
 
+	/* Merge support functions are only allowed in MERGE's RETURNING list */
+	if (IsMergeSupportFunction(funcid) &&
+		pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+		ereport(ERROR,
+				(errcode(ERRCODE_WRONG_OBJECT_TYPE),
+				 errmsg("merge support function %s can only be called from the RETURNING list of a MERGE command",
+						NameListToString(funcname)),
+				 parser_errposition(pstate, location)));
+
 	/*
 	 * So far so good, so do some fdresult-type-specific processing.
 	 */
@@ -2599,6 +2609,7 @@ check_srf_call_placement(ParseState *pst
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_merge.c b/src/backend/parser/parse_merge.c
new file mode 100644
index 91b1156..33852e9
--- a/src/backend/parser/parse_merge.c
+++ b/src/backend/parser/parse_merge.c
@@ -233,6 +233,10 @@ transformMergeStmt(ParseState *pstate, M
 	 */
 	qry->jointree = makeFromExpr(pstate->p_joinlist, joinExpr);
 
+	/* Transform the RETURNING list, if any */
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_MERGE_RETURNING);
+
 	/*
 	 * We now have a good query shape, so now look at the WHEN conditions and
 	 * action targetlists.
@@ -254,6 +258,7 @@ transformMergeStmt(ParseState *pstate, M
 		MergeAction *action;
 
 		action = makeNode(MergeAction);
+		action->index = foreach_current_index(l) + 1;
 		action->commandType = mergeWhenClause->commandType;
 		action->matched = mergeWhenClause->matched;
 
@@ -390,9 +395,6 @@ transformMergeStmt(ParseState *pstate, M
 
 	qry->mergeActionList = mergeActionList;
 
-	/* RETURNING could potentially be added in the future, but not in SQL std */
-	qry->returningList = NULL;
-
 	qry->hasTargetSRFs = false;
 	qry->hasSubLinks = pstate->p_hasSubLinks;
 
diff --git a/src/backend/parser/parse_relation.c b/src/backend/parser/parse_relation.c
new file mode 100644
index 41d6049..794530b
--- a/src/backend/parser/parse_relation.c
+++ b/src/backend/parser/parse_relation.c
@@ -2345,9 +2345,10 @@ addRangeTableEntryForCTE(ParseState *pst
 		cte->cterefcount++;
 
 	/*
-	 * We throw error if the CTE is INSERT/UPDATE/DELETE without RETURNING.
-	 * This won't get checked in case of a self-reference, but that's OK
-	 * because data-modifying CTEs aren't allowed to be recursive anyhow.
+	 * We throw error if the CTE is INSERT/UPDATE/DELETE/MERGE without
+	 * RETURNING.  This won't get checked in case of a self-reference, but
+	 * that's OK because data-modifying CTEs aren't allowed to be recursive
+	 * anyhow.
 	 */
 	if (IsA(cte->ctequery, Query))
 	{
diff --git a/src/backend/rewrite/rewriteHandler.c b/src/backend/rewrite/rewriteHandler.c
new file mode 100644
index b486ab5..1bd0ce2
--- a/src/backend/rewrite/rewriteHandler.c
+++ b/src/backend/rewrite/rewriteHandler.c
@@ -3629,9 +3629,9 @@ RewriteQuery(Query *parsetree, List *rew
 	ListCell   *lc1;
 
 	/*
-	 * First, recursively process any insert/update/delete statements in WITH
-	 * clauses.  (We have to do this first because the WITH clauses may get
-	 * copied into rule actions below.)
+	 * First, recursively process any insert/update/delete/merge statements in
+	 * WITH clauses.  (We have to do this first because the WITH clauses may
+	 * get copied into rule actions below.)
 	 */
 	foreach(lc1, parsetree->cteList)
 	{
@@ -3656,7 +3656,8 @@ RewriteQuery(Query *parsetree, List *rew
 			if (!(ctequery->commandType == CMD_SELECT ||
 				  ctequery->commandType == CMD_UPDATE ||
 				  ctequery->commandType == CMD_INSERT ||
-				  ctequery->commandType == CMD_DELETE))
+				  ctequery->commandType == CMD_DELETE ||
+				  ctequery->commandType == CMD_MERGE))
 			{
 				/*
 				 * Currently it could only be NOTIFY; this error message will
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
new file mode 100644
index 30b51bf..0ae3890
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -2141,11 +2141,10 @@ QueryReturnsTuples(Query *parsetree)
 		case CMD_SELECT:
 			/* returns tuples */
 			return true;
-		case CMD_MERGE:
-			return false;
 		case CMD_INSERT:
 		case CMD_UPDATE:
 		case CMD_DELETE:
+		case CMD_MERGE:
 			/* the forms with RETURNING return tuples */
 			if (parsetree->returningList)
 				return true;
diff --git a/src/backend/utils/adt/ruleutils.c b/src/backend/utils/adt/ruleutils.c
new file mode 100644
index d3a973d..f097d98
--- a/src/backend/utils/adt/ruleutils.c
+++ b/src/backend/utils/adt/ruleutils.c
@@ -7183,8 +7183,13 @@ get_merge_query_def(Query *query, depars
 			appendStringInfoString(buf, "DO NOTHING");
 	}
 
-	/* No RETURNING support in MERGE yet */
-	Assert(query->returningList == NIL);
+	/* Add RETURNING if present */
+	if (query->returningList)
+	{
+		appendContextKeyword(context, " RETURNING",
+							 -PRETTYINDENT_STD, PRETTYINDENT_STD, 1);
+		get_target_list(query->returningList, context, NULL, colNamesVisible);
+	}
 }
 
 
diff --git a/src/bin/psql/common.c b/src/bin/psql/common.c
new file mode 100644
index 5973df2..a1380f8
--- a/src/bin/psql/common.c
+++ b/src/bin/psql/common.c
@@ -982,13 +982,17 @@ PrintQueryResult(PGresult *result, bool
 			else
 				success = true;
 
-			/* if it's INSERT/UPDATE/DELETE RETURNING, also print status */
+			/*
+			 * If it's INSERT/UPDATE/DELETE/MERGE RETURNING, also print
+			 * status.
+			 */
 			if (last || pset.show_all_results)
 			{
 				cmdstatus = PQcmdStatus(result);
 				if (strncmp(cmdstatus, "INSERT", 6) == 0 ||
 					strncmp(cmdstatus, "UPDATE", 6) == 0 ||
-					strncmp(cmdstatus, "DELETE", 6) == 0)
+					strncmp(cmdstatus, "DELETE", 6) == 0 ||
+					strncmp(cmdstatus, "MERGE", 5) == 0)
 					PrintQueryStatus(result, printStatusFout);
 			}
 
diff --git a/src/include/catalog/pg_proc.dat b/src/include/catalog/pg_proc.dat
new file mode 100644
index 6996073..ac3e56b
--- a/src/include/catalog/pg_proc.dat
+++ b/src/include/catalog/pg_proc.dat
@@ -12043,4 +12043,14 @@
   proname => 'any_value_transfn', prorettype => 'anyelement',
   proargtypes => 'anyelement anyelement', prosrc => 'any_value_transfn' },
 
+# MERGE support functions
+{ oid => '9499', descr => 'command type of current MERGE action',
+  proname => 'pg_merge_action',  provolatile => 'v',
+  prorettype => 'text', proargtypes => '',
+  prosrc => 'pg_merge_action' },
+{ oid => '9500', descr => 'index of current MERGE WHEN clause',
+  proname => 'pg_merge_when_clause',  provolatile => 'v',
+  prorettype => 'int4', proargtypes => '',
+  prosrc => 'pg_merge_when_clause' },
+
 ]
diff --git a/src/include/catalog/pg_proc.h b/src/include/catalog/pg_proc.h
new file mode 100644
index e7abe0b..463add8
--- a/src/include/catalog/pg_proc.h
+++ b/src/include/catalog/pg_proc.h
@@ -182,6 +182,11 @@ DECLARE_UNIQUE_INDEX(pg_proc_proname_arg
 #define PROARGMODE_VARIADIC 'v'
 #define PROARGMODE_TABLE	't'
 
+/* Is this a merge support function?  (Requires fmgroids.h) */
+#define IsMergeSupportFunction(funcid) \
+	((funcid) == F_PG_MERGE_ACTION || \
+	 (funcid) == F_PG_MERGE_WHEN_CLAUSE)
+
 #endif							/* EXPOSE_TO_CLIENT_CODE */
 
 
diff --git a/src/include/executor/spi.h b/src/include/executor/spi.h
new file mode 100644
index d1de139..ed8787d
--- a/src/include/executor/spi.h
+++ b/src/include/executor/spi.h
@@ -97,6 +97,7 @@ typedef struct _SPI_plan *SPIPlanPtr;
 #define SPI_OK_REL_UNREGISTER	16
 #define SPI_OK_TD_REGISTER		17
 #define SPI_OK_MERGE			18
+#define SPI_OK_MERGE_RETURNING	19
 
 #define SPI_OPT_NONATOMIC		(1 << 0)
 
diff --git a/src/include/nodes/execnodes.h b/src/include/nodes/execnodes.h
new file mode 100644
index cb714f4..6708b6b
--- a/src/include/nodes/execnodes.h
+++ b/src/include/nodes/execnodes.h
@@ -1319,6 +1319,9 @@ typedef struct ModifyTableState
 	/* Flags showing which subcommands are present INS/UPD/DEL/DO NOTHING */
 	int			mt_merge_subcommands;
 
+	/* For MERGE, the action currently being executed */
+	MergeActionState *mt_merge_action;
+
 	/* tuple counters for MERGE */
 	double		mt_merge_inserted;
 	double		mt_merge_updated;
diff --git a/src/include/nodes/parsenodes.h b/src/include/nodes/parsenodes.h
new file mode 100644
index b3bec90..1e14f7b
--- a/src/include/nodes/parsenodes.h
+++ b/src/include/nodes/parsenodes.h
@@ -1687,6 +1687,7 @@ typedef struct MergeWhenClause
 typedef struct MergeAction
 {
 	NodeTag		type;
+	int			index;			/* 1-based index of the clause */
 	bool		matched;		/* true=MATCHED, false=NOT MATCHED */
 	CmdType		commandType;	/* INSERT/UPDATE/DELETE/DO NOTHING */
 	/* OVERRIDING clause */
@@ -1913,6 +1914,7 @@ typedef struct MergeStmt
 	Node	   *sourceRelation; /* source relation */
 	Node	   *joinCondition;	/* join condition between source and target */
 	List	   *mergeWhenClauses;	/* list of MergeWhenClause(es) */
+	List	   *returningList;	/* list of expressions to return */
 	WithClause *withClause;		/* WITH clause */
 } MergeStmt;
 
diff --git a/src/include/parser/analyze.h b/src/include/parser/analyze.h
new file mode 100644
index 1cef183..78eb55e
--- a/src/include/parser/analyze.h
+++ b/src/include/parser/analyze.h
@@ -44,6 +44,8 @@ extern List *transformInsertRow(ParseSta
 								bool strip_indirection);
 extern List *transformUpdateTargetList(ParseState *pstate,
 									   List *origTlist);
+extern List *transformReturningList(ParseState *pstate, List *returningList,
+									ParseExprKind exprKind);
 extern Query *transformTopLevelStmt(ParseState *pstate, RawStmt *parseTree);
 extern Query *transformStmt(ParseState *pstate, Node *parseTree);
 
diff --git a/src/include/parser/parse_node.h b/src/include/parser/parse_node.h
new file mode 100644
index f589112..836819d
--- a/src/include/parser/parse_node.h
+++ b/src/include/parser/parse_node.h
@@ -61,7 +61,8 @@ typedef enum ParseExprKind
 	EXPR_KIND_DISTINCT_ON,		/* DISTINCT ON */
 	EXPR_KIND_LIMIT,			/* LIMIT */
 	EXPR_KIND_OFFSET,			/* OFFSET */
-	EXPR_KIND_RETURNING,		/* RETURNING */
+	EXPR_KIND_RETURNING,		/* RETURNING in INSERT/UPDATE/DELETE */
+	EXPR_KIND_MERGE_RETURNING,	/* RETURNING in MERGE */
 	EXPR_KIND_VALUES,			/* VALUES */
 	EXPR_KIND_VALUES_SINGLE,	/* single-row VALUES (in INSERT only) */
 	EXPR_KIND_CHECK_CONSTRAINT, /* CHECK constraint for a table */
diff --git a/src/pl/plpgsql/src/pl_exec.c b/src/pl/plpgsql/src/pl_exec.c
new file mode 100644
index 4b76f76..fb6289c
--- a/src/pl/plpgsql/src/pl_exec.c
+++ b/src/pl/plpgsql/src/pl_exec.c
@@ -4267,9 +4267,9 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 	/*
 	 * If we have INTO, then we only need one row back ... but if we have INTO
 	 * STRICT or extra check too_many_rows, ask for two rows, so that we can
-	 * verify the statement returns only one.  INSERT/UPDATE/DELETE are always
-	 * treated strictly. Without INTO, just run the statement to completion
-	 * (tcount = 0).
+	 * verify the statement returns only one.  INSERT/UPDATE/DELETE/MERGE are
+	 * always treated strictly. Without INTO, just run the statement to
+	 * completion (tcount = 0).
 	 *
 	 * We could just ask for two rows always when using INTO, but there are
 	 * some cases where demanding the extra row costs significant time, eg by
@@ -4307,10 +4307,11 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 			Assert(stmt->mod_stmt);
 			exec_set_found(estate, (SPI_processed != 0));
 			break;
@@ -4489,10 +4490,11 @@ exec_stmt_dynexecute(PLpgSQL_execstate *
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 		case SPI_OK_UTILITY:
 		case SPI_OK_REWRITTEN:
 			break;
diff --git a/src/pl/tcl/pltcl.c b/src/pl/tcl/pltcl.c
new file mode 100644
index e8f9d7b..0c9f81e
--- a/src/pl/tcl/pltcl.c
+++ b/src/pl/tcl/pltcl.c
@@ -2456,6 +2456,7 @@ pltcl_process_SPI_result(Tcl_Interp *int
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
+		case SPI_OK_MERGE_RETURNING:
 
 			/*
 			 * Process the tuples we got
diff --git a/src/test/regress/expected/merge.out b/src/test/regress/expected/merge.out
new file mode 100644
index 133d421..da4dbc2
--- a/src/test/regress/expected/merge.out
+++ b/src/test/regress/expected/merge.out
@@ -123,20 +123,20 @@ ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
 ERROR:  name "target" specified more than once
 DETAIL:  The name is used both as MERGE target table and data source.
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
-ERROR:  MERGE not supported in WITH query
-LINE 1: WITH foo AS (
-             ^
--- used in COPY
+ERROR:  WITH query "foo" does not have a RETURNING clause
+LINE 4: ) SELECT * FROM foo;
+                        ^
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) TO stdout;
-ERROR:  MERGE not supported in COPY
+ERROR:  COPY query must have a RETURNING clause
 -- unsupported relation types
 -- view
 CREATE VIEW tv AS SELECT * FROM target;
@@ -1304,20 +1304,187 @@ WHEN MATCHED AND tid < 2 THEN
 ROLLBACK;
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
-ERROR:  syntax error at or near "RETURNING"
-LINE 10: RETURNING *;
-         ^
+    DELETE
+RETURNING pg_merge_when_clause() AS when_clause,
+          pg_merge_action() AS merge_action,
+          t.*,
+          CASE pg_merge_action()
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ when_clause | merge_action | tid | balance |     description     
+-------------+--------------+-----+---------+---------------------
+           3 | DELETE       |   1 |     100 | Removed (1,100)
+           1 | UPDATE       |   2 |     220 | Added 20 to balance
+           2 | INSERT       |   4 |      40 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- error when using MERGE support functions outside MERGE
+SELECT pg_merge_action() FROM sq_target;
+ERROR:  merge support function pg_merge_action can only be called from the RETURNING list of a MERGE command
+LINE 1: SELECT pg_merge_action() FROM sq_target;
+               ^
+UPDATE sq_target SET balance = balance + 1 RETURNING pg_merge_when_clause();
+ERROR:  merge support function pg_merge_when_clause can only be called from the RETURNING list of a MERGE command
+LINE 1: ...ATE sq_target SET balance = balance + 1 RETURNING pg_merge_w...
+                                                             ^
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_when_clause() AS when_clause,
+              pg_merge_action() AS merge_action,
+              t.*,
+              CASE pg_merge_action()
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING merge_action, pg_merge_action() AS merge_log_action, l.*
+)
+SELECT * FROM m2;
+ merge_action | merge_log_action | tid |     last_change     
+--------------+------------------+-----+---------------------
+ DELETE       | UPDATE           |   1 | Removed (1,100)
+ UPDATE       | INSERT           |   2 | Added 20 to balance
+ INSERT       | INSERT           |   4 | Inserted (4,40)
+(3 rows)
+
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ tid |     last_change     
+-----+---------------------
+   1 | Removed (1,100)
+   2 | Added 20 to balance
+   4 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING v
+    ON tid = sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*
+) TO stdout;
+DELETE	1	100
+ROLLBACK;
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ action | tid | new_balance 
+--------+-----+-------------
+ DELETE |   1 |         100
+ UPDATE |   3 |         320
+ INSERT |   4 |         110
+(3 rows)
+
+ROLLBACK;
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ action | tid | balance 
+--------+-----+---------
+ DELETE |   1 |     100
+ INSERT |   4 |      40
+(2 rows)
+
+ROLLBACK;
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ r_action | r_tid | r_balance 
+----------+-------+-----------
+ DELETE   |     1 |       100
+ UPDATE   |     3 |       320
+ INSERT   |     4 |       110
+(3 rows)
+
 ROLLBACK;
 -- EXPLAIN
 CREATE TABLE ex_mtarget (a int, b int)
@@ -1514,7 +1681,7 @@ SELECT * FROM sq_target WHERE tid = 1;
 (1 row)
 
 ROLLBACK;
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 NOTICE:  drop cascades to view v
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1640,6 +1807,32 @@ SELECT * FROM pa_target ORDER BY tid;
 (14 rows)
 
 ROLLBACK;
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
+ pg_merge_action | pg_merge_when_clause | tid | balance |           val            
+-----------------+----------------------+-----+---------+--------------------------
+ UPDATE          |                    1 |   2 |     110 | initial updated by merge
+(1 row)
+
+SELECT * FROM pa_target ORDER BY tid;
+ tid | balance |           val            
+-----+---------+--------------------------
+   2 |     110 | initial updated by merge
+   3 |     300 | initial
+   5 |     500 | initial
+   7 |     700 | initial
+   9 |     900 | initial
+  11 |    1100 | initial
+  13 |    1300 | initial
+(7 rows)
+
+ROLLBACK;
 DROP TABLE pa_target CASCADE;
 -- The target table is partitioned in the same way, but this time by attaching
 -- partitions which have columns in different order, dropped columns etc.
@@ -1731,7 +1924,26 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (sid, delta, 'inserted by merge');
+    INSERT VALUES (sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
+ pg_merge_action | pg_merge_when_clause | tid | balance |           val            
+-----------------+----------------------+-----+---------+--------------------------
+ UPDATE          |                    1 |   2 |     110 | initial updated by merge
+ INSERT          |                    2 |   2 |      20 | inserted by merge
+ UPDATE          |                    1 |   4 |     330 | initial updated by merge
+ INSERT          |                    2 |   4 |      40 | inserted by merge
+ UPDATE          |                    1 |   6 |     550 | initial updated by merge
+ INSERT          |                    2 |   6 |      60 | inserted by merge
+ UPDATE          |                    1 |   8 |     770 | initial updated by merge
+ INSERT          |                    2 |   8 |      80 | inserted by merge
+ UPDATE          |                    1 |  10 |     990 | initial updated by merge
+ INSERT          |                    2 |  10 |     100 | inserted by merge
+ UPDATE          |                    1 |  12 |    1210 | initial updated by merge
+ INSERT          |                    2 |  12 |     120 | inserted by merge
+ UPDATE          |                    1 |  14 |    1430 | initial updated by merge
+ INSERT          |                    2 |  14 |     140 | inserted by merge
+(14 rows)
+
 SELECT * FROM pa_target ORDER BY tid;
  tid | balance |           val            
 -----+---------+--------------------------
@@ -1798,7 +2010,21 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
+ pg_merge_action | pg_merge_when_clause |          logts           | tid | balance |           val            
+-----------------+----------------------+--------------------------+-----+---------+--------------------------
+ UPDATE          |                    1 | Tue Jan 31 00:00:00 2017 |   1 |     110 | initial updated by merge
+ UPDATE          |                    1 | Tue Feb 28 00:00:00 2017 |   2 |     220 | initial updated by merge
+ INSERT          |                    2 | Sun Jan 15 00:00:00 2017 |   3 |      30 | inserted by merge
+ UPDATE          |                    1 | Tue Jan 31 00:00:00 2017 |   4 |     440 | initial updated by merge
+ UPDATE          |                    1 | Tue Feb 28 00:00:00 2017 |   5 |     550 | initial updated by merge
+ INSERT          |                    2 | Sun Jan 15 00:00:00 2017 |   6 |      60 | inserted by merge
+ UPDATE          |                    1 | Tue Jan 31 00:00:00 2017 |   7 |     770 | initial updated by merge
+ UPDATE          |                    1 | Tue Feb 28 00:00:00 2017 |   8 |     880 | initial updated by merge
+ INSERT          |                    2 | Sun Jan 15 00:00:00 2017 |   9 |      90 | inserted by merge
+(9 rows)
+
 SELECT * FROM pa_target ORDER BY tid;
           logts           | tid | balance |           val            
 --------------------------+-----+---------+--------------------------
diff --git a/src/test/regress/expected/rules.out b/src/test/regress/expected/rules.out
new file mode 100644
index 7fd81e6..87a3686
--- a/src/test/regress/expected/rules.out
+++ b/src/test/regress/expected/rules.out
@@ -3590,7 +3590,8 @@ MERGE INTO rule_merge2 t USING (SELECT 1
 -- test deparsing
 CREATE TABLE sf_target(id int, data text, filling int[]);
 CREATE FUNCTION merge_sf_test()
- RETURNS void
+ RETURNS TABLE(pg_merge_action text, when_clause int,
+               a int, b text, id int, data text, filling int[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -3627,11 +3628,13 @@ WHEN NOT MATCHED
    VALUES (s.a, s.b, DEFAULT)
 WHEN NOT MATCHED
    THEN INSERT (filling[1], id)
-   VALUES (s.a, s.a);
+   VALUES (s.a, s.a)
+RETURNING
+   pg_merge_action(), pg_merge_when_clause() AS when_clause, *;
 END;
 \sf merge_sf_test
 CREATE OR REPLACE FUNCTION public.merge_sf_test()
- RETURNS void
+ RETURNS TABLE(pg_merge_action text, when_clause integer, a integer, b text, id integer, data text, filling integer[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -3668,7 +3671,14 @@ BEGIN ATOMIC
       VALUES (s.a, s.b, DEFAULT)
     WHEN NOT MATCHED
      THEN INSERT (filling[1], id)
-      VALUES (s.a, s.a);
+      VALUES (s.a, s.a)
+   RETURNING pg_merge_action() AS pg_merge_action,
+     pg_merge_when_clause() AS when_clause,
+     s.a,
+     s.b,
+     t.id,
+     t.data,
+     t.filling;
 END
 DROP FUNCTION merge_sf_test;
 DROP TABLE sf_target;
diff --git a/src/test/regress/sql/merge.sql b/src/test/regress/sql/merge.sql
new file mode 100644
index 4cf6db9..ae5d7c2
--- a/src/test/regress/sql/merge.sql
+++ b/src/test/regress/sql/merge.sql
@@ -86,12 +86,12 @@ MERGE INTO target
 USING target
 ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
--- used in COPY
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
@@ -857,17 +857,145 @@ ROLLBACK;
 
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
+    DELETE
+RETURNING pg_merge_when_clause() AS when_clause,
+          pg_merge_action() AS merge_action,
+          t.*,
+          CASE pg_merge_action()
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ROLLBACK;
+
+-- error when using MERGE support functions outside MERGE
+SELECT pg_merge_action() FROM sq_target;
+UPDATE sq_target SET balance = balance + 1 RETURNING pg_merge_when_clause();
+
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_when_clause() AS when_clause,
+              pg_merge_action() AS merge_action,
+              t.*,
+              CASE pg_merge_action()
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING merge_action, pg_merge_action() AS merge_log_action, l.*
+)
+SELECT * FROM m2;
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ROLLBACK;
+
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING v
+    ON tid = sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*
+) TO stdout;
+ROLLBACK;
+
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ROLLBACK;
+
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ROLLBACK;
+
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid > 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
 ROLLBACK;
 
 -- EXPLAIN
@@ -966,7 +1094,7 @@ WHEN MATCHED THEN
 SELECT * FROM sq_target WHERE tid = 1;
 ROLLBACK;
 
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1033,6 +1161,17 @@ SELECT merge_func();
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
+SELECT * FROM pa_target ORDER BY tid;
+ROLLBACK;
+
 DROP TABLE pa_target CASCADE;
 
 -- The target table is partitioned in the same way, but this time by attaching
@@ -1091,7 +1230,8 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (sid, delta, 'inserted by merge');
+    INSERT VALUES (sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
@@ -1145,7 +1285,8 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
diff --git a/src/test/regress/sql/rules.sql b/src/test/regress/sql/rules.sql
new file mode 100644
index 8b7e255..a33515b
--- a/src/test/regress/sql/rules.sql
+++ b/src/test/regress/sql/rules.sql
@@ -1281,7 +1281,8 @@ MERGE INTO rule_merge2 t USING (SELECT 1
 CREATE TABLE sf_target(id int, data text, filling int[]);
 
 CREATE FUNCTION merge_sf_test()
- RETURNS void
+ RETURNS TABLE(pg_merge_action text, when_clause int,
+               a int, b text, id int, data text, filling int[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -1318,7 +1319,9 @@ WHEN NOT MATCHED
    VALUES (s.a, s.b, DEFAULT)
 WHEN NOT MATCHED
    THEN INSERT (filling[1], id)
-   VALUES (s.a, s.a);
+   VALUES (s.a, s.a)
+RETURNING
+   pg_merge_action(), pg_merge_when_clause() AS when_clause, *;
 END;
 
 \sf merge_sf_test

diff --git a/doc/src/sgml/dml.sgml b/doc/src/sgml/dml.sgml
new file mode 100644
index cbbc5e2..ff2a827
--- a/doc/src/sgml/dml.sgml
+++ b/doc/src/sgml/dml.sgml
@@ -283,10 +283,15 @@ DELETE FROM products;
    <secondary>RETURNING</secondary>
   </indexterm>
 
+  <indexterm zone="dml-returning">
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
   <para>
    Sometimes it is useful to obtain data from modified rows while they are
    being manipulated.  The <command>INSERT</command>, <command>UPDATE</command>,
-   and <command>DELETE</command> commands all have an
+   <command>DELETE</command>, and <command>MERGE</command> commands all have an
    optional <literal>RETURNING</literal> clause that supports this.  Use
    of <literal>RETURNING</literal> avoids performing an extra database query to
    collect the data, and is especially valuable when it would otherwise be
@@ -339,6 +344,24 @@ DELETE FROM products
 </programlisting>
   </para>
 
+  <para>
+   In a <command>MERGE</command>, the data available to <literal>RETURNING</literal> is
+   the content of the source row plus the content of the inserted, updated, or
+   deleted target row.  In addition, the merge support functions
+   <xref linkend="pg-merge-action"/> and <xref linkend="pg-merge-when-clause"/>
+   may be used to return information about which merge action was executed for
+   each row.  Since it is quite common for the source and target to have many
+   of the same columns, specifying <literal>RETURNING *</literal> can lead to a
+   lot of duplicated columns, so it is often more useful to just return the
+   target row.  For example:
+<programlisting>
+MERGE INTO products p USING new_products n ON p.product_no = n.product_no
+  WHEN NOT MATCHED THEN INSERT VALUES (n.product_no, n.name, n.price)
+  WHEN MATCHED THEN UPDATE SET name = n.name, price = n.price
+  RETURNING pg_merge_action(), p.*;
+</programlisting>
+  </para>
+
   <para>
    If there are triggers (<xref linkend="triggers"/>) on the target table,
    the data available to <literal>RETURNING</literal> is the row as modified by
diff --git a/doc/src/sgml/func.sgml b/doc/src/sgml/func.sgml
new file mode 100644
index 5a47ce4..1406c18
--- a/doc/src/sgml/func.sgml
+++ b/doc/src/sgml/func.sgml
@@ -21705,6 +21705,94 @@ SELECT count(*) FROM sometable;
 
  </sect1>
 
+ <sect1 id="functions-merge-support">
+  <title>Merge Support Functions</title>
+
+  <indexterm>
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
+  <para>
+   The merge support functions shown in
+   <xref linkend="functions-merge-support-table"/> may be used in the
+   <literal>RETURNING</literal> list of a <xref linkend="sql-merge"/> command
+   to return additional information about the action taken for each row.
+  </para>
+
+  <table id="functions-merge-support-table">
+    <title>Merge Support Functions</title>
+
+    <tgroup cols="1">
+     <thead>
+      <row>
+       <entry role="func_table_entry"><para role="func_signature">
+        Function
+       </para>
+       <para>
+        Description
+       </para></entry>
+      </row>
+     </thead>
+
+     <tbody>
+      <row>
+       <entry role="func_table_entry"><para role="func_signature">
+        <indexterm id="pg-merge-action">
+         <primary>pg_merge_action</primary>
+        </indexterm>
+        <function>pg_merge_action</function> ( )
+        <returnvalue>text</returnvalue>
+       </para>
+       <para>
+        Returns the action performed on the current row (<literal>'INSERT'</literal>,
+        <literal>'UPDATE'</literal>, or <literal>'DELETE'</literal>).
+       </para></entry>
+      </row>
+
+      <row>
+       <entry role="func_table_entry"><para role="func_signature">
+        <indexterm id="pg-merge-when-clause">
+         <primary>pg_merge_when_clause</primary>
+        </indexterm>
+        <function>pg_merge_when_clause</function> ( )
+        <returnvalue>integer</returnvalue>
+       </para>
+       <para>
+        Returns the 1-based index of the <literal>WHEN</literal> clause executed
+        for the current row.
+       </para></entry>
+      </row>
+     </tbody>
+    </tgroup>
+  </table>
+
+  <para>
+   Example:
+<screen><![CDATA[
+MERGE INTO products p
+  USING stock s ON p.product_id = s.product_id
+  WHEN MATCHED AND s.quantity > 0 THEN
+    UPDATE SET in_stock = true, quantity = s.quantity
+  WHEN MATCHED THEN
+    UPDATE SET in_stock = false, quantity = 0
+  WHEN NOT MATCHED THEN
+    INSERT (product_id, in_stock, quantity)
+      VALUES (s.product_id, true, s.quantity)
+  RETURNING pg_merge_action() AS action,
+            pg_merge_when_clause() AS when_clause,
+            p.*;
+
+ action | when_clause | product_id | in_stock | quantity 
+--------+-------------+------------+----------+----------
+ UPDATE |           1 |       1001 | t        |       50
+ UPDATE |           2 |       1002 | f        |        0
+ INSERT |           3 |       1003 | t        |       10
+]]></screen>
+  </para>
+
+ </sect1>
+
  <sect1 id="functions-subquery">
   <title>Subquery Expressions</title>
 
diff --git a/doc/src/sgml/glossary.sgml b/doc/src/sgml/glossary.sgml
new file mode 100644
index fe8def4..51a15ca
--- a/doc/src/sgml/glossary.sgml
+++ b/doc/src/sgml/glossary.sgml
@@ -1387,9 +1387,9 @@
      to a <glossterm linkend="glossary-client">client</glossterm> upon the
      completion of an <acronym>SQL</acronym> command, usually a
      <command>SELECT</command> but it can be an
-     <command>INSERT</command>, <command>UPDATE</command>, or
-     <command>DELETE</command> command if the <literal>RETURNING</literal>
-     clause is specified.
+     <command>INSERT</command>, <command>UPDATE</command>,
+     <command>DELETE</command>, or <command>MERGE</command> command if the
+     <literal>RETURNING</literal> clause is specified.
     </para>
     <para>
      The fact that a result set is a relation means that a query can be used
diff --git a/doc/src/sgml/plpgsql.sgml b/doc/src/sgml/plpgsql.sgml
new file mode 100644
index f55e901..6803240
--- a/doc/src/sgml/plpgsql.sgml
+++ b/doc/src/sgml/plpgsql.sgml
@@ -1020,8 +1020,8 @@ INSERT INTO mytable VALUES (1,'one'), (2
     </para>
 
     <para>
-     If the command does return rows (for example <command>SELECT</command>,
-     or <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>
+     If the command does return rows (for example <command>SELECT</command>, or
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
      with <literal>RETURNING</literal>), there are two ways to proceed.
      When the command will return at most one row, or you only care about
      the first row of output, write the command as usual but add
@@ -1149,6 +1149,7 @@ SELECT <replaceable>select_expressions</
 INSERT ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 UPDATE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 DELETE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
+MERGE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 </synopsis>
 
      where <replaceable>target</replaceable> can be a record variable, a row
@@ -1159,8 +1160,8 @@ DELETE ... RETURNING <replaceable>expres
      <literal>INTO</literal> clause) just as described above,
      and the plan is cached in the same way.
      This works for <command>SELECT</command>,
-     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
-     <literal>RETURNING</literal>, and certain utility commands
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
+     with <literal>RETURNING</literal>, and certain utility commands
      that return row sets, such as <command>EXPLAIN</command>.
      Except for the <literal>INTO</literal> clause, the SQL command is the same
      as it would be written outside <application>PL/pgSQL</application>.
@@ -1236,7 +1237,7 @@ END;
     </para>
 
     <para>
-     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
+     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command> with
      <literal>RETURNING</literal>, <application>PL/pgSQL</application> reports
      an error for more than one returned row, even when
      <literal>STRICT</literal> is not specified.  This is because there
diff --git a/doc/src/sgml/queries.sgml b/doc/src/sgml/queries.sgml
new file mode 100644
index 3f95849..9928373
--- a/doc/src/sgml/queries.sgml
+++ b/doc/src/sgml/queries.sgml
@@ -2062,9 +2062,10 @@ SELECT <replaceable>select_list</replace
    Table Expressions or <acronym>CTE</acronym>s, can be thought of as defining
    temporary tables that exist just for one query.  Each auxiliary statement
    in a <literal>WITH</literal> clause can be a <command>SELECT</command>,
-   <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>; and the
+   <command>INSERT</command>, <command>UPDATE</command>, <command>DELETE</command>,
+   or <command>MERGE</command>; and the
    <literal>WITH</literal> clause itself is attached to a primary statement that can
-   be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
+   also be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
    <command>DELETE</command>, or <command>MERGE</command>.
   </para>
 
@@ -2598,8 +2599,8 @@ SELECT * FROM w AS w1 JOIN w AS w2 ON w1
    <title>Data-Modifying Statements in <literal>WITH</literal></title>
 
    <para>
-    You can use most data-modifying statements (<command>INSERT</command>,
-    <command>UPDATE</command>, or <command>DELETE</command>, but not
+    You can use data-modifying statements (<command>INSERT</command>,
+    <command>UPDATE</command>, <command>DELETE</command>, or
     <command>MERGE</command>) in <literal>WITH</literal>.  This
     allows you to perform several different operations in the same query.
     An example is:
diff --git a/doc/src/sgml/ref/copy.sgml b/doc/src/sgml/ref/copy.sgml
new file mode 100644
index 5e591ed..dbc76df
--- a/doc/src/sgml/ref/copy.sgml
+++ b/doc/src/sgml/ref/copy.sgml
@@ -122,13 +122,16 @@ COPY { <replaceable class="parameter">ta
       A <link linkend="sql-select"><command>SELECT</command></link>,
       <link linkend="sql-values"><command>VALUES</command></link>,
       <link linkend="sql-insert"><command>INSERT</command></link>,
-      <link linkend="sql-update"><command>UPDATE</command></link>, or
-      <link linkend="sql-delete"><command>DELETE</command></link> command whose results are to be
-      copied.  Note that parentheses are required around the query.
+      <link linkend="sql-update"><command>UPDATE</command></link>,
+      <link linkend="sql-delete"><command>DELETE</command></link>, or
+      <link linkend="sql-merge"><command>MERGE</command></link> command
+      whose results are to be copied.  Note that parentheses are required
+      around the query.
      </para>
      <para>
-      For <command>INSERT</command>, <command>UPDATE</command> and
-      <command>DELETE</command> queries a RETURNING clause must be provided,
+      For <command>INSERT</command>, <command>UPDATE</command>,
+      <command>DELETE</command>, and <command>MERGE</command> queries a
+      RETURNING clause must be provided,
       and the target relation must not have a conditional rule, nor
       an <literal>ALSO</literal> rule, nor an <literal>INSTEAD</literal> rule
       that expands to multiple statements.
diff --git a/doc/src/sgml/ref/merge.sgml b/doc/src/sgml/ref/merge.sgml
new file mode 100644
index 0995fe0..7eb641a
--- a/doc/src/sgml/ref/merge.sgml
+++ b/doc/src/sgml/ref/merge.sgml
@@ -25,6 +25,7 @@ PostgreSQL documentation
 MERGE INTO [ ONLY ] <replaceable class="parameter">target_table_name</replaceable> [ * ] [ [ AS ] <replaceable class="parameter">target_alias</replaceable> ]
 USING <replaceable class="parameter">data_source</replaceable> ON <replaceable class="parameter">join_condition</replaceable>
 <replaceable class="parameter">when_clause</replaceable> [...]
+[ RETURNING * | <replaceable class="parameter">output_expression</replaceable> [ [ AS ] <replaceable class="parameter">output_name</replaceable> ] [, ...] ]
 
 <phrase>where <replaceable class="parameter">data_source</replaceable> is:</phrase>
 
@@ -95,6 +96,20 @@ DELETE
   </para>
 
   <para>
+   The optional <literal>RETURNING</literal> clause causes <command>MERGE</command>
+   to compute and return value(s) based on each row inserted, updated, or
+   deleted.  Any expression using the source or target table's columns, or the
+   merge support functions <xref linkend="pg-merge-action"/> and
+   <xref linkend="pg-merge-when-clause"/> can be computed.  When an
+   <command>INSERT</command> or <command>UPDATE</command> action is performed,
+   the new values of the target table's columns are used.  When a
+   <command>DELETE</command> is performed, the old values of the target table's
+   columns are used.  The syntax of the <literal>RETURNING</literal> list is
+   identical to that of the output list of <command>SELECT</command>.
+
+  </para>
+
+  <para>
    There is no separate <literal>MERGE</literal> privilege.
    If you specify an update action, you must have the
    <literal>UPDATE</literal> privilege on the column(s)
@@ -125,6 +140,18 @@ DELETE
 
   <variablelist>
    <varlistentry>
+    <term><replaceable class="parameter">with_query</replaceable></term>
+    <listitem>
+     <para>
+      The <literal>WITH</literal> clause allows you to specify one or more
+      subqueries that can be referenced by name in the <command>MERGE</command>
+      query. See <xref linkend="queries-with"/> and <xref linkend="sql-select"/>
+      for details.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
     <term><replaceable class="parameter">target_table_name</replaceable></term>
     <listitem>
      <para>
@@ -399,13 +426,30 @@ DELETE
    </varlistentry>
 
    <varlistentry>
-    <term><replaceable class="parameter">with_query</replaceable></term>
+    <term><replaceable class="parameter">output_expression</replaceable></term>
     <listitem>
      <para>
-      The <literal>WITH</literal> clause allows you to specify one or more
-      subqueries that can be referenced by name in the <command>MERGE</command>
-      query. See <xref linkend="queries-with"/> and <xref linkend="sql-select"/>
-      for details.
+      An expression to be computed and returned by the <command>MERGE</command>
+      command after each row is merged.  The expression can use any columns of
+      the source or target tables, or the merge support functions
+      <xref linkend="pg-merge-action"/> and <xref linkend="pg-merge-when-clause"/>.
+     </para>
+     <para>
+      Writing <literal>*</literal> will return all columns from the source
+      table, followed by all columns from the target table.  Often this will
+      lead to a lot of duplication, since it is common for the source and
+      target tables to have a lot of the same columns.  This can be avoided by
+      qualifying the <literal>*</literal> with the name or alias of the source
+      or target table.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
+    <term><replaceable class="parameter">output_name</replaceable></term>
+    <listitem>
+     <para>
+      A name to use for a returned column.
      </para>
     </listitem>
    </varlistentry>
@@ -428,6 +472,13 @@ MERGE <replaceable class="parameter">tot
    were changed in any way.
   </para>
 
+  <para>
+   If the <command>MERGE</command> command contains a <literal>RETURNING</literal>
+   clause, the result will be similar to that of a <command>SELECT</command>
+   statement containing the columns and values defined in the
+   <literal>RETURNING</literal> list, computed over the row(s) inserted, updated,
+   or deleted by the command.
+  </para>
  </refsect1>
 
  <refsect1>
@@ -544,13 +595,6 @@ MERGE <replaceable class="parameter">tot
   </para>
 
   <para>
-   There is no <literal>RETURNING</literal> clause with
-   <command>MERGE</command>.  Actions of <command>INSERT</command>,
-   <command>UPDATE</command> and <command>DELETE</command> cannot contain
-   <literal>RETURNING</literal> or <literal>WITH</literal> clauses.
-  </para>
-
-  <para>
    When <command>MERGE</command> is run concurrently with other commands
    that modify the target table, the usual transaction isolation rules
    apply; see <xref linkend="transaction-iso"/> for an explanation
@@ -602,7 +646,8 @@ WHEN NOT MATCHED THEN
   <para>
    Attempt to insert a new stock item along with the quantity of stock. If
    the item already exists, instead update the stock count of the existing
-   item. Don't allow entries that have zero stock.
+   item. Don't allow entries that have zero stock. Return details of all
+   changes made.
 <programlisting>
 MERGE INTO wines w
 USING wine_stock_changes s
@@ -612,7 +657,8 @@ WHEN NOT MATCHED AND s.stock_delta > 0 T
 WHEN MATCHED AND w.stock + s.stock_delta > 0 THEN
   UPDATE SET stock = w.stock + s.stock_delta
 WHEN MATCHED THEN
-  DELETE;
+  DELETE
+RETURNING pg_merge_action(), w.*;
 </programlisting>
 
    The <literal>wine_stock_changes</literal> table might be, for example, a
@@ -627,7 +673,8 @@ WHEN MATCHED THEN
     This command conforms to the <acronym>SQL</acronym> standard.
   </para>
    <para>
-    The WITH clause and <literal>DO NOTHING</literal> action are extensions to
+    The WITH clause, <literal>DO NOTHING</literal> action, and
+    <literal>RETURNING</literal> clause are extensions to
     the <acronym>SQL</acronym> standard.
   </para>
  </refsect1>
diff --git a/doc/src/sgml/ref/select.sgml b/doc/src/sgml/ref/select.sgml
new file mode 100644
index 0ee0cc7..f81bdea
--- a/doc/src/sgml/ref/select.sgml
+++ b/doc/src/sgml/ref/select.sgml
@@ -74,7 +74,7 @@ SELECT [ ALL | DISTINCT [ ON ( <replacea
 
 <phrase>and <replaceable class="parameter">with_query</replaceable> is:</phrase>
 
-    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> )
+    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> | <replaceable class="parameter">merge</replaceable> )
         [ SEARCH { BREADTH | DEPTH } FIRST BY <replaceable>column_name</replaceable> [, ...] SET <replaceable>search_seq_col_name</replaceable> ]
         [ CYCLE <replaceable>column_name</replaceable> [, ...] SET <replaceable>cycle_mark_col_name</replaceable> [ TO <replaceable>cycle_mark_value</replaceable> DEFAULT <replaceable>cycle_mark_default</replaceable> ] USING <replaceable>cycle_path_col_name</replaceable> ]
 
@@ -227,10 +227,10 @@ TABLE [ ONLY ] <replaceable class="param
     The subqueries effectively act as temporary tables or views
     for the duration of the primary query.
     Each subquery can be a <command>SELECT</command>, <command>TABLE</command>, <command>VALUES</command>,
-    <command>INSERT</command>, <command>UPDATE</command> or
-    <command>DELETE</command> statement.
+    <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command> statement.
     When writing a data-modifying statement (<command>INSERT</command>,
-    <command>UPDATE</command> or <command>DELETE</command>) in
+    <command>UPDATE</command>, <command>DELETE</command>, or <command>MERGE</command>) in
     <literal>WITH</literal>, it is usual to include a <literal>RETURNING</literal> clause.
     It is the output of <literal>RETURNING</literal>, <emphasis>not</emphasis> the underlying
     table that the statement modifies, that forms the temporary table that is
@@ -2179,7 +2179,8 @@ SELECT 2+2;
 
    <para>
     <productname>PostgreSQL</productname> allows <command>INSERT</command>,
-    <command>UPDATE</command>, and <command>DELETE</command> to be used as <literal>WITH</literal>
+    <command>UPDATE</command>, <command>DELETE</command>, and
+    <command>MERGE</command> to be used as <literal>WITH</literal>
     queries.  This is not found in the SQL standard.
    </para>
   </refsect2>
diff --git a/doc/src/sgml/rowtypes.sgml b/doc/src/sgml/rowtypes.sgml
new file mode 100644
index 4d86f97..bbeac84
--- a/doc/src/sgml/rowtypes.sgml
+++ b/doc/src/sgml/rowtypes.sgml
@@ -348,7 +348,7 @@ SELECT m.* FROM some_table, LATERAL myfu
    column expansion of this kind when it appears at the top level of
    a <link linkend="queries-select-lists"><command>SELECT</command> output
    list</link>, a <link linkend="dml-returning"><literal>RETURNING</literal>
-   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>,
+   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>,
    a <link linkend="queries-values"><literal>VALUES</literal> clause</link>, or
    a <link linkend="sql-syntax-row-constructors">row constructor</link>.
    In all other contexts (including when nested inside one of those
diff --git a/doc/src/sgml/spi.sgml b/doc/src/sgml/spi.sgml
new file mode 100644
index 651930a..ad0a9b1
--- a/doc/src/sgml/spi.sgml
+++ b/doc/src/sgml/spi.sgml
@@ -301,8 +301,9 @@ SPI_execute("INSERT INTO foo SELECT * FR
    is returned in the global variable <varname>SPI_processed</varname>.
    If the return value of the function is <symbol>SPI_OK_SELECT</symbol>,
    <symbol>SPI_OK_INSERT_RETURNING</symbol>,
-   <symbol>SPI_OK_DELETE_RETURNING</symbol>, or
-   <symbol>SPI_OK_UPDATE_RETURNING</symbol>,
+   <symbol>SPI_OK_DELETE_RETURNING</symbol>,
+   <symbol>SPI_OK_UPDATE_RETURNING</symbol>, or
+   <symbol>SPI_OK_MERGE_RETURNING</symbol>
    then you can use the
    global pointer <literal>SPITupleTable *SPI_tuptable</literal> to
    access the result rows.  Some utility commands (such as
@@ -472,6 +473,15 @@ typedef struct SPITupleTable
       </para>
      </listitem>
     </varlistentry>
+
+    <varlistentry>
+     <term><symbol>SPI_OK_MERGE_RETURNING</symbol></term>
+     <listitem>
+      <para>
+       if a <command>MERGE RETURNING</command> was executed
+      </para>
+     </listitem>
+    </varlistentry>
 
     <varlistentry>
      <term><symbol>SPI_OK_UTILITY</symbol></term>
diff --git a/doc/src/sgml/xfunc.sgml b/doc/src/sgml/xfunc.sgml
new file mode 100644
index 9620ea9..4f68efb
--- a/doc/src/sgml/xfunc.sgml
+++ b/doc/src/sgml/xfunc.sgml
@@ -177,7 +177,8 @@
     statements separated by semicolons.  A semicolon after the last
     statement is optional.  Unless the function is declared to return
     <type>void</type>, the last statement must be a <command>SELECT</command>,
-    or an <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>
+    or an <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command>
     that has a <literal>RETURNING</literal> clause.
    </para>
 
@@ -1236,8 +1237,9 @@ SELECT x, CASE WHEN x &gt; 0 THEN genera
 
     <note>
      <para>
-      If a function's last command is <command>INSERT</command>, <command>UPDATE</command>,
-      or <command>DELETE</command> with <literal>RETURNING</literal>, that command will
+      If a function's last command is <command>INSERT</command>,
+      <command>UPDATE</command>, <command>DELETE</command>, or
+      <command>MERGE</command> with <literal>RETURNING</literal>, that command will
       always be executed to completion, even if the function is not declared
       with <literal>SETOF</literal> or the calling query does not fetch all the
       result rows.  Any extra rows produced by the <literal>RETURNING</literal>
diff --git a/src/backend/commands/copy.c b/src/backend/commands/copy.c
new file mode 100644
index f14fae3..05f2e5b
--- a/src/backend/commands/copy.c
+++ b/src/backend/commands/copy.c
@@ -283,12 +283,6 @@ DoCopy(ParseState *pstate, const CopyStm
 	{
 		Assert(stmt->query);
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(stmt->query, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in COPY"));
-
 		query = makeNode(RawStmt);
 		query->stmt = stmt->query;
 		query->stmt_location = stmt_location;
diff --git a/src/backend/commands/copyto.c b/src/backend/commands/copyto.c
new file mode 100644
index 9e4b243..cf56f9f
--- a/src/backend/commands/copyto.c
+++ b/src/backend/commands/copyto.c
@@ -510,7 +510,8 @@ BeginCopyTo(ParseState *pstate,
 		{
 			Assert(query->commandType == CMD_INSERT ||
 				   query->commandType == CMD_UPDATE ||
-				   query->commandType == CMD_DELETE);
+				   query->commandType == CMD_DELETE ||
+				   query->commandType == CMD_MERGE);
 
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
diff --git a/src/backend/executor/execPartition.c b/src/backend/executor/execPartition.c
new file mode 100644
index eb8a87f..b9bd03c
--- a/src/backend/executor/execPartition.c
+++ b/src/backend/executor/execPartition.c
@@ -612,8 +612,8 @@ ExecInitPartitionInfo(ModifyTableState *
 	 * Build the RETURNING projection for the partition.  Note that we didn't
 	 * build the returningList for partitions within the planner, but simple
 	 * translation of varattnos will suffice.  This only occurs for the INSERT
-	 * case or in the case of UPDATE tuple routing where we didn't find a
-	 * result rel to reuse.
+	 * case or in the case of UPDATE/MERGE tuple routing where we didn't find
+	 * a result rel to reuse.
 	 */
 	if (node && node->returningLists != NIL)
 	{
@@ -622,12 +622,14 @@ ExecInitPartitionInfo(ModifyTableState *
 		List	   *returningList;
 
 		/* See the comment above for WCO lists. */
-		/* (except no RETURNING support for MERGE yet) */
 		Assert((node->operation == CMD_INSERT &&
 				list_length(node->returningLists) == 1 &&
 				list_length(node->resultRelations) == 1) ||
 			   (node->operation == CMD_UPDATE &&
 				list_length(node->returningLists) ==
+				list_length(node->resultRelations)) ||
+			   (node->operation == CMD_MERGE &&
+				list_length(node->returningLists) ==
 				list_length(node->resultRelations)));
 
 		/*
diff --git a/src/backend/executor/functions.c b/src/backend/executor/functions.c
new file mode 100644
index f55424e..f99f15f
--- a/src/backend/executor/functions.c
+++ b/src/backend/executor/functions.c
@@ -1660,8 +1660,8 @@ check_sql_fn_retval(List *queryTreeLists
 
 	/*
 	 * If it's a plain SELECT, it returns whatever the targetlist says.
-	 * Otherwise, if it's INSERT/UPDATE/DELETE with RETURNING, it returns
-	 * that. Otherwise, the function return type must be VOID.
+	 * Otherwise, if it's INSERT/UPDATE/DELETE/MERGE with RETURNING, it
+	 * returns that. Otherwise, the function return type must be VOID.
 	 *
 	 * Note: eventually replace this test with QueryReturnsTuples?	We'd need
 	 * a more general method of determining the output type, though.  Also, it
@@ -1679,7 +1679,8 @@ check_sql_fn_retval(List *queryTreeLists
 	else if (parse &&
 			 (parse->commandType == CMD_INSERT ||
 			  parse->commandType == CMD_UPDATE ||
-			  parse->commandType == CMD_DELETE) &&
+			  parse->commandType == CMD_DELETE ||
+			  parse->commandType == CMD_MERGE) &&
 			 parse->returningList)
 	{
 		tlist = parse->returningList;
diff --git a/src/backend/executor/nodeModifyTable.c b/src/backend/executor/nodeModifyTable.c
new file mode 100644
index 2a5fec8..3f6f495
--- a/src/backend/executor/nodeModifyTable.c
+++ b/src/backend/executor/nodeModifyTable.c
@@ -36,8 +36,7 @@
  *		RETURNING tuple after completing each row insert, update, or delete.
  *		It must be called again to continue the operation.  Without RETURNING,
  *		we just loop within the node until all the work is done, then
- *		return NULL.  This avoids useless call/return overhead.  (MERGE does
- *		not support RETURNING.)
+ *		return NULL.  This avoids useless call/return overhead.
  */
 
 #include "postgres.h"
@@ -64,6 +63,10 @@
 #include "utils/rel.h"
 
 
+/* State available for merge support functions */
+static CmdType mergeActionCmdType = CMD_NOTHING;
+static int32 mergeActionIdx = 0;
+
 typedef struct MTTargetRelLookup
 {
 	Oid			relationOid;	/* hash key, must be first */
@@ -153,13 +156,14 @@ static TupleTableSlot *ExecMerge(ModifyT
 								 ItemPointer tupleid,
 								 bool canSetTag);
 static void ExecInitMerge(ModifyTableState *mtstate, EState *estate);
-static bool ExecMergeMatched(ModifyTableContext *context,
-							 ResultRelInfo *resultRelInfo,
-							 ItemPointer tupleid,
-							 bool canSetTag);
-static void ExecMergeNotMatched(ModifyTableContext *context,
-								ResultRelInfo *resultRelInfo,
-								bool canSetTag);
+static TupleTableSlot *ExecMergeMatched(ModifyTableContext *context,
+										ResultRelInfo *resultRelInfo,
+										ItemPointer tupleid,
+										bool canSetTag,
+										bool *matched);
+static TupleTableSlot *ExecMergeNotMatched(ModifyTableContext *context,
+										   ResultRelInfo *resultRelInfo,
+										   bool canSetTag);
 
 
 /*
@@ -237,6 +241,7 @@ ExecCheckPlanOutput(Relation resultRel,
 /*
  * ExecProcessReturning --- evaluate a RETURNING list
  *
+ * context: state of current ModifyTable operation
  * resultRelInfo: current result rel
  * tupleSlot: slot holding tuple actually inserted/updated/deleted
  * planSlot: slot holding tuple returned by top subplan node
@@ -247,12 +252,14 @@ ExecCheckPlanOutput(Relation resultRel,
  * Returns a slot holding the result tuple
  */
 static TupleTableSlot *
-ExecProcessReturning(ResultRelInfo *resultRelInfo,
+ExecProcessReturning(ModifyTableContext *context,
+					 ResultRelInfo *resultRelInfo,
 					 TupleTableSlot *tupleSlot,
 					 TupleTableSlot *planSlot)
 {
 	ProjectionInfo *projectReturning = resultRelInfo->ri_projectReturning;
 	ExprContext *econtext = projectReturning->pi_exprContext;
+	TupleTableSlot *rslot;
 
 	/* Make tuple and any needed join variables available to ExecProject */
 	if (tupleSlot)
@@ -266,8 +273,32 @@ ExecProcessReturning(ResultRelInfo *resu
 	econtext->ecxt_scantuple->tts_tableOid =
 		RelationGetRelid(resultRelInfo->ri_RelationDesc);
 
-	/* Compute the RETURNING expressions */
-	return ExecProject(projectReturning);
+	/*
+	 * Compute the RETURNING expressions. If this is a MERGE, additional state
+	 * is made available to any merge support functions in the RETURNING list.
+	 */
+	if (context->relaction)
+	{
+		CmdType		saved_mergeActionCmdType = mergeActionCmdType;
+		int32		saved_mergeActionIdx = mergeActionIdx;
+
+		mergeActionCmdType = context->relaction->mas_action->commandType;
+		mergeActionIdx = context->relaction->mas_action->index;
+		PG_TRY();
+		{
+			rslot = ExecProject(projectReturning);
+		}
+		PG_FINALLY();
+		{
+			mergeActionCmdType = saved_mergeActionCmdType;
+			mergeActionIdx = saved_mergeActionIdx;
+		}
+		PG_END_TRY();
+	}
+	else
+		rslot = ExecProject(projectReturning);
+
+	return rslot;
 }
 
 /*
@@ -1195,7 +1226,7 @@ ExecInsert(ModifyTableContext *context,
 
 	/* Process RETURNING if present */
 	if (resultRelInfo->ri_projectReturning)
-		result = ExecProcessReturning(resultRelInfo, slot, planSlot);
+		result = ExecProcessReturning(context, resultRelInfo, slot, planSlot);
 
 	if (inserted_tuple)
 		*inserted_tuple = slot;
@@ -1692,7 +1723,8 @@ ldelete:
 			}
 		}
 
-		rslot = ExecProcessReturning(resultRelInfo, slot, context->planSlot);
+		rslot = ExecProcessReturning(context, resultRelInfo, slot,
+									 context->planSlot);
 
 		/*
 		 * Before releasing the target tuple again, make sure rslot has a
@@ -2480,7 +2512,8 @@ redo_act:
 
 	/* Process RETURNING if present */
 	if (resultRelInfo->ri_projectReturning)
-		return ExecProcessReturning(resultRelInfo, slot, context->planSlot);
+		return ExecProcessReturning(context, resultRelInfo, slot,
+									context->planSlot);
 
 	return NULL;
 }
@@ -2712,6 +2745,7 @@ static TupleTableSlot *
 ExecMerge(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 		  ItemPointer tupleid, bool canSetTag)
 {
+	TupleTableSlot *rslot = NULL;
 	bool		matched;
 
 	/*-----
@@ -2759,18 +2793,18 @@ ExecMerge(ModifyTableContext *context, R
 	 */
 	matched = tupleid != NULL;
 	if (matched)
-		matched = ExecMergeMatched(context, resultRelInfo, tupleid, canSetTag);
+		rslot = ExecMergeMatched(context, resultRelInfo, tupleid, canSetTag,
+								 &matched);
 
 	/*
-	 * Either we were dealing with a NOT MATCHED tuple or ExecMergeMatched()
-	 * returned "false", indicating the previously MATCHED tuple no longer
-	 * matches.
+	 * Deal with the NOT MATCHED case (either a NOT MATCHED tuple from the
+	 * join, or a previously MATCHED tuple for which ExecMergeMatched() set
+	 * "matched" to false, indicating that it no longer matches).
 	 */
 	if (!matched)
-		ExecMergeNotMatched(context, resultRelInfo, canSetTag);
+		rslot = ExecMergeNotMatched(context, resultRelInfo, canSetTag);
 
-	/* No RETURNING support yet */
-	return NULL;
+	return rslot;
 }
 
 /*
@@ -2780,8 +2814,8 @@ ExecMerge(ModifyTableContext *context, R
  * We start from the first WHEN MATCHED action and check if the WHEN quals
  * pass, if any. If the WHEN quals for the first action do not pass, we
  * check the second, then the third and so on. If we reach to the end, no
- * action is taken and we return true, indicating that no further action is
- * required for this tuple.
+ * action is taken and "matched" is set to true, indicating that no further
+ * action is required for this tuple.
  *
  * If we do find a qualifying action, then we attempt to execute the action.
  *
@@ -2790,15 +2824,16 @@ ExecMerge(ModifyTableContext *context, R
  * with individual actions are evaluated by this routine via ExecQual, while
  * EvalPlanQual checks for the join quals. If EvalPlanQual tells us that the
  * updated tuple still passes the join quals, then we restart from the first
- * action to look for a qualifying action. Otherwise, we return false --
- * meaning that a NOT MATCHED action must now be executed for the current
- * source tuple.
+ * action to look for a qualifying action. Otherwise, "matched" is set to
+ * false -- meaning that a NOT MATCHED action must now be executed for the
+ * current source tuple.
  */
-static bool
+static TupleTableSlot *
 ExecMergeMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
-				 ItemPointer tupleid, bool canSetTag)
+				 ItemPointer tupleid, bool canSetTag, bool *matched)
 {
 	ModifyTableState *mtstate = context->mtstate;
+	TupleTableSlot *rslot = NULL;
 	TupleTableSlot *newslot;
 	EState	   *estate = context->estate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
@@ -2810,7 +2845,10 @@ ExecMergeMatched(ModifyTableContext *con
 	 * If there are no WHEN MATCHED actions, we are done.
 	 */
 	if (resultRelInfo->ri_matchedMergeAction == NIL)
-		return true;
+	{
+		*matched = true;
+		return NULL;
+	}
 
 	/*
 	 * Make tuple and any needed join variables available to ExecQual and
@@ -2893,7 +2931,10 @@ lmerge_matched:
 										tupleid, NULL, newslot, &result))
 				{
 					if (result == TM_Ok)
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 					break;		/* concurrent update/delete */
 				}
 				result = ExecUpdateAct(context, resultRelInfo, tupleid, NULL,
@@ -2912,7 +2953,10 @@ lmerge_matched:
 										NULL, NULL, &result))
 				{
 					if (result == TM_Ok)
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 					break;		/* concurrent update/delete */
 				}
 				result = ExecDeleteAct(context, resultRelInfo, tupleid, false);
@@ -2968,7 +3012,8 @@ lmerge_matched:
 				 * If the tuple was already deleted, return to let caller
 				 * handle it under NOT MATCHED clauses.
 				 */
-				return false;
+				*matched = false;
+				return NULL;
 
 			case TM_Updated:
 				{
@@ -3014,13 +3059,19 @@ lmerge_matched:
 							 * NOT MATCHED actions.
 							 */
 							if (TupIsNull(epqslot))
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							(void) ExecGetJunkAttribute(epqslot,
 														resultRelInfo->ri_RowIdAttNo,
 														&isNull);
 							if (isNull)
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							/*
 							 * When a tuple was updated and migrated to
@@ -3055,7 +3106,8 @@ lmerge_matched:
 							 * tuple already deleted; tell caller to run NOT
 							 * MATCHED actions
 							 */
-							return false;
+							*matched = false;
+							return NULL;
 
 						case TM_SelfModified:
 
@@ -3075,13 +3127,14 @@ lmerge_matched:
 										(errcode(ERRCODE_TRIGGERED_DATA_CHANGE_VIOLATION),
 										 errmsg("tuple to be updated or deleted was already modified by an operation triggered by the current command"),
 										 errhint("Consider using an AFTER trigger instead of a BEFORE trigger to propagate changes to other rows.")));
-							return false;
+							*matched = false;
+							return NULL;
 
 						default:
 							/* see table_tuple_lock call in ExecDelete() */
 							elog(ERROR, "unexpected table_tuple_lock status: %u",
 								 result);
-							return false;
+							return NULL;
 					}
 				}
 
@@ -3093,6 +3146,32 @@ lmerge_matched:
 				break;
 		}
 
+		/* Process RETURNING if present */
+		if (resultRelInfo->ri_projectReturning)
+		{
+			switch (commandType)
+			{
+				case CMD_UPDATE:
+					rslot = ExecProcessReturning(context,
+												 resultRelInfo, newslot,
+												 context->planSlot);
+					break;
+
+				case CMD_DELETE:
+					rslot = ExecProcessReturning(context,
+												 resultRelInfo,
+												 resultRelInfo->ri_oldTupleSlot,
+												 context->planSlot);
+					break;
+
+				case CMD_NOTHING:
+					break;
+
+				default:
+					elog(ERROR, "unknown action in MERGE WHEN MATCHED clause");
+			}
+		}
+
 		/*
 		 * We've activated one of the WHEN clauses, so we don't search
 		 * further. This is required behaviour, not an optimization.
@@ -3103,19 +3182,22 @@ lmerge_matched:
 	/*
 	 * Successfully executed an action or no qualifying action was found.
 	 */
-	return true;
+	*matched = true;
+
+	return rslot;
 }
 
 /*
  * Execute the first qualifying NOT MATCHED action.
  */
-static void
+static TupleTableSlot *
 ExecMergeNotMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 					bool canSetTag)
 {
 	ModifyTableState *mtstate = context->mtstate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
 	List	   *actionStates = NIL;
+	TupleTableSlot *rslot = NULL;
 	ListCell   *l;
 
 	/*
@@ -3167,8 +3249,8 @@ ExecMergeNotMatched(ModifyTableContext *
 				newslot = ExecProject(action->mas_proj);
 				context->relaction = action;
 
-				(void) ExecInsert(context, mtstate->rootResultRelInfo, newslot,
-								  canSetTag, NULL, NULL);
+				rslot = ExecInsert(context, mtstate->rootResultRelInfo,
+								   newslot, canSetTag, NULL, NULL);
 				mtstate->mt_merge_inserted += 1;
 				break;
 			case CMD_NOTHING:
@@ -3184,6 +3266,8 @@ ExecMergeNotMatched(ModifyTableContext *
 		 */
 		break;
 	}
+
+	return rslot;
 }
 
 /*
@@ -3361,6 +3445,44 @@ ExecInitMergeTupleSlots(ModifyTableState
 }
 
 /*
+ * pg_merge_action() -
+ *	  SQL merge support function to retrieve the currently executing merge
+ *	  action command string ("INSERT", "UPDATE", or "DELETE").
+ */
+Datum
+pg_merge_action(PG_FUNCTION_ARGS)
+{
+	switch (mergeActionCmdType)
+	{
+		case CMD_INSERT:
+			PG_RETURN_TEXT_P(cstring_to_text("INSERT"));
+		case CMD_UPDATE:
+			PG_RETURN_TEXT_P(cstring_to_text("UPDATE"));
+		case CMD_DELETE:
+			PG_RETURN_TEXT_P(cstring_to_text("DELETE"));
+		case CMD_NOTHING:
+			PG_RETURN_NULL();
+		default:
+			elog(ERROR, "unrecognized commandType: %d", (int) mergeActionCmdType);
+	}
+	PG_RETURN_NULL();
+}
+
+/*
+ * pg_merge_when_clause() -
+ *	  SQL merge support function to retrieve the 1-based index of the
+ *	  currently executing merge WHEN clause.
+ */
+Datum
+pg_merge_when_clause(PG_FUNCTION_ARGS)
+{
+	if (mergeActionIdx > 0)
+		PG_RETURN_INT32(mergeActionIdx);
+
+	PG_RETURN_NULL();
+}
+
+/*
  * Process BEFORE EACH STATEMENT triggers
  */
 static void
@@ -3615,6 +3737,9 @@ ExecModifyTable(PlanState *pstate)
 
 		context.planSlot = ExecProcNode(subplanstate);
 
+		/* Reset current MERGE action */
+		context.relaction = NULL;
+
 		/* No more tuples to process? */
 		if (TupIsNull(context.planSlot))
 			break;
@@ -3646,8 +3771,17 @@ ExecModifyTable(PlanState *pstate)
 				{
 					EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-					ExecMerge(&context, node->resultRelInfo, NULL, node->canSetTag);
-					continue;	/* no RETURNING support yet */
+					slot = ExecMerge(&context, node->resultRelInfo, NULL,
+									 node->canSetTag);
+
+					/*
+					 * If we got a RETURNING result, return it to the caller.
+					 * We'll continue the work on next call.
+					 */
+					if (slot)
+						return slot;
+
+					continue;	/* continue with the next tuple */
 				}
 
 				elog(ERROR, "tableoid is NULL");
@@ -3674,7 +3808,8 @@ ExecModifyTable(PlanState *pstate)
 			 * ExecProcessReturning by IterateDirectModify, so no need to
 			 * provide it here.
 			 */
-			slot = ExecProcessReturning(resultRelInfo, NULL, context.planSlot);
+			slot = ExecProcessReturning(&context, resultRelInfo, NULL,
+										context.planSlot);
 
 			return slot;
 		}
@@ -3724,8 +3859,17 @@ ExecModifyTable(PlanState *pstate)
 					{
 						EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-						ExecMerge(&context, node->resultRelInfo, NULL, node->canSetTag);
-						continue;	/* no RETURNING support yet */
+						slot = ExecMerge(&context, node->resultRelInfo, NULL,
+										 node->canSetTag);
+
+						/*
+						 * If we got a RETURNING result, return it to the
+						 * caller.  We'll continue the work on next call.
+						 */
+						if (slot)
+							return slot;
+
+						continue;	/* continue with the next tuple */
 					}
 
 					elog(ERROR, "ctid is NULL");
@@ -3817,7 +3961,6 @@ ExecModifyTable(PlanState *pstate)
 				}
 				slot = ExecGetUpdateNewTuple(resultRelInfo, context.planSlot,
 											 oldSlot);
-				context.relaction = NULL;
 
 				/* Now apply the update. */
 				slot = ExecUpdate(&context, resultRelInfo, tupleid, oldtuple,
diff --git a/src/backend/executor/spi.c b/src/backend/executor/spi.c
new file mode 100644
index 3397568..b9f443c
--- a/src/backend/executor/spi.c
+++ b/src/backend/executor/spi.c
@@ -2033,6 +2033,8 @@ SPI_result_code_string(int code)
 			return "SPI_OK_TD_REGISTER";
 		case SPI_OK_MERGE:
 			return "SPI_OK_MERGE";
+		case SPI_OK_MERGE_RETURNING:
+			return "SPI_OK_MERGE_RETURNING";
 	}
 	/* Unrecognized code ... return something useful ... */
 	sprintf(buf, "Unrecognized SPI code %d", code);
@@ -2886,7 +2888,10 @@ _SPI_pquery(QueryDesc *queryDesc, bool f
 				res = SPI_OK_UPDATE;
 			break;
 		case CMD_MERGE:
-			res = SPI_OK_MERGE;
+			if (queryDesc->plannedstmt->hasReturning)
+				res = SPI_OK_MERGE_RETURNING;
+			else
+				res = SPI_OK_MERGE;
 			break;
 		default:
 			return SPI_ERROR_OPUNKNOWN;
diff --git a/src/backend/optimizer/plan/subselect.c b/src/backend/optimizer/plan/subselect.c
new file mode 100644
index da2d8ab..2b898bf
--- a/src/backend/optimizer/plan/subselect.c
+++ b/src/backend/optimizer/plan/subselect.c
@@ -18,6 +18,7 @@
 
 #include "access/htup_details.h"
 #include "catalog/pg_operator.h"
+#include "catalog/pg_proc.h"
 #include "catalog/pg_type.h"
 #include "executor/executor.h"
 #include "miscadmin.h"
@@ -35,6 +36,7 @@
 #include "parser/parse_relation.h"
 #include "rewrite/rewriteManip.h"
 #include "utils/builtins.h"
+#include "utils/fmgroids.h"
 #include "utils/lsyscache.h"
 #include "utils/syscache.h"
 
@@ -1847,7 +1849,8 @@ convert_EXISTS_to_ANY(PlannerInfo *root,
 /*
  * Replace correlation vars (uplevel vars) with Params.
  *
- * Uplevel PlaceHolderVars and aggregates are replaced, too.
+ * Uplevel PlaceHolderVars, aggregates, GROUPING() expressions and merge
+ * support functions are replaced, too.
  *
  * Note: it is critical that this runs immediately after SS_process_sublinks.
  * Since we do not recurse into the arguments of uplevel PHVs and aggregates,
@@ -1901,6 +1904,22 @@ replace_correlation_vars_mutator(Node *n
 		if (((GroupingFunc *) node)->agglevelsup > 0)
 			return (Node *) replace_outer_grouping(root, (GroupingFunc *) node);
 	}
+	if (IsA(node, FuncExpr))
+	{
+		if (IsMergeSupportFunction(((FuncExpr *) node)->funcid))
+		{
+			Param	   *param;
+
+			/*
+			 * Replace with a Param, if it belongs to an upper-level MERGE
+			 * query. Otherwise, leave it be.
+			 */
+			param = replace_outer_merge_support_function(root,
+														 (FuncExpr *) node);
+			if (param)
+				return (Node *) param;
+		}
+	}
 	return expression_tree_mutator(node,
 								   replace_correlation_vars_mutator,
 								   (void *) root);
diff --git a/src/backend/optimizer/util/paramassign.c b/src/backend/optimizer/util/paramassign.c
new file mode 100644
index d6a923b..2b7b8d1
--- a/src/backend/optimizer/util/paramassign.c
+++ b/src/backend/optimizer/util/paramassign.c
@@ -307,6 +307,63 @@ replace_outer_grouping(PlannerInfo *root
 }
 
 /*
+ * Generate a Param node to replace the given FuncExpr expression, which is
+ * expected to be a merge support function, if it belongs to an upper-level
+ * MERGE query. Otherwise NULL will be returned.
+ *
+ * A NULL return value indicates either that the merge support function
+ * belongs to a MERGE query at this query level, or that there is no MERGE
+ * query at all. In both cases, the caller is expected to leave the FuncExpr
+ * node unaltered, so that it is executed at this query level (in the latter
+ * case, it is up to the merge support function to decide how to handle being
+ * called outside of a MERGE query).
+ */
+Param *
+replace_outer_merge_support_function(PlannerInfo *root, FuncExpr *func)
+{
+	Param	   *retval;
+	PlannerParamItem *pitem;
+	Index		levelsup;
+	Oid			ptype = exprType((Node *) func);
+
+	/* Find the upper-level MERGE query the function belongs to */
+	levelsup = 0;
+	while (root->parse->commandType != CMD_MERGE)
+	{
+		root = root->parent_root;
+		if (root == NULL)
+			return NULL;		/* Not in a MERGE query */
+		levelsup++;
+	}
+	if (levelsup == 0)
+		return NULL;			/* Local MERGE query at this level */
+
+	/*
+	 * It does not seem worthwhile to try to de-duplicate references to outer
+	 * merge support functions.  Just make a new slot every time.
+	 */
+	func = copyObject(func);
+
+	pitem = makeNode(PlannerParamItem);
+	pitem->item = (Node *) func;
+	pitem->paramId = list_length(root->glob->paramExecTypes);
+	root->glob->paramExecTypes = lappend_oid(root->glob->paramExecTypes,
+											 ptype);
+
+	root->plan_params = lappend(root->plan_params, pitem);
+
+	retval = makeNode(Param);
+	retval->paramkind = PARAM_EXEC;
+	retval->paramid = pitem->paramId;
+	retval->paramtype = ptype;
+	retval->paramtypmod = -1;
+	retval->paramcollid = InvalidOid;
+	retval->location = func->location;
+
+	return retval;
+}
+
+/*
  * Generate a Param node to replace the given Var,
  * which is expected to come from some upper NestLoop plan node.
  * Record the need for the Var in root->curOuterParams.
diff --git a/src/backend/parser/analyze.c b/src/backend/parser/analyze.c
new file mode 100644
index 4006632..d6eafab
--- a/src/backend/parser/analyze.c
+++ b/src/backend/parser/analyze.c
@@ -74,7 +74,6 @@ static void determineRecursiveColTypes(P
 									   Node *larg, List *nrtargetlist);
 static Query *transformReturnStmt(ParseState *pstate, ReturnStmt *stmt);
 static Query *transformUpdateStmt(ParseState *pstate, UpdateStmt *stmt);
-static List *transformReturningList(ParseState *pstate, List *returningList);
 static Query *transformPLAssignStmt(ParseState *pstate,
 									PLAssignStmt *stmt);
 static Query *transformDeclareCursorStmt(ParseState *pstate,
@@ -2501,9 +2500,9 @@ transformUpdateTargetList(ParseState *ps
 
 /*
  * transformReturningList -
- *	handle a RETURNING clause in INSERT/UPDATE/DELETE
+ *	handle a RETURNING clause in INSERT/UPDATE/DELETE/MERGE
  */
-static List *
+List *
 transformReturningList(ParseState *pstate, List *returningList)
 {
 	List	   *rlist;
diff --git a/src/backend/parser/gram.y b/src/backend/parser/gram.y
new file mode 100644
index 39ab7ea..5455607
--- a/src/backend/parser/gram.y
+++ b/src/backend/parser/gram.y
@@ -12249,6 +12249,7 @@ MergeStmt:
 			USING table_ref
 			ON a_expr
 			merge_when_list
+			returning_clause
 				{
 					MergeStmt  *m = makeNode(MergeStmt);
 
@@ -12257,6 +12258,7 @@ MergeStmt:
 					m->sourceRelation = $6;
 					m->joinCondition = $8;
 					m->mergeWhenClauses = $9;
+					m->returningList = $10;
 
 					$$ = (Node *) m;
 				}
diff --git a/src/backend/parser/parse_cte.c b/src/backend/parser/parse_cte.c
new file mode 100644
index c5b1a49..ad3c525
--- a/src/backend/parser/parse_cte.c
+++ b/src/backend/parser/parse_cte.c
@@ -126,13 +126,6 @@ transformWithClause(ParseState *pstate,
 		CommonTableExpr *cte = (CommonTableExpr *) lfirst(lc);
 		ListCell   *rest;
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(cte->ctequery, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in WITH query"),
-					parser_errposition(pstate, cte->location));
-
 		for_each_cell(rest, withClause->ctes, lnext(withClause->ctes, lc))
 		{
 			CommonTableExpr *cte2 = (CommonTableExpr *) lfirst(rest);
@@ -153,7 +146,8 @@ transformWithClause(ParseState *pstate,
 			/* must be a data-modifying statement */
 			Assert(IsA(cte->ctequery, InsertStmt) ||
 				   IsA(cte->ctequery, UpdateStmt) ||
-				   IsA(cte->ctequery, DeleteStmt));
+				   IsA(cte->ctequery, DeleteStmt) ||
+				   IsA(cte->ctequery, MergeStmt));
 
 			pstate->p_hasModifyingCTE = true;
 		}
diff --git a/src/backend/parser/parse_merge.c b/src/backend/parser/parse_merge.c
new file mode 100644
index 91b1156..4427f75
--- a/src/backend/parser/parse_merge.c
+++ b/src/backend/parser/parse_merge.c
@@ -233,6 +233,9 @@ transformMergeStmt(ParseState *pstate, M
 	 */
 	qry->jointree = makeFromExpr(pstate->p_joinlist, joinExpr);
 
+	/* Transform the RETURNING list, if any */
+	qry->returningList = transformReturningList(pstate, stmt->returningList);
+
 	/*
 	 * We now have a good query shape, so now look at the WHEN conditions and
 	 * action targetlists.
@@ -254,6 +257,7 @@ transformMergeStmt(ParseState *pstate, M
 		MergeAction *action;
 
 		action = makeNode(MergeAction);
+		action->index = foreach_current_index(l) + 1;
 		action->commandType = mergeWhenClause->commandType;
 		action->matched = mergeWhenClause->matched;
 
@@ -390,9 +394,6 @@ transformMergeStmt(ParseState *pstate, M
 
 	qry->mergeActionList = mergeActionList;
 
-	/* RETURNING could potentially be added in the future, but not in SQL std */
-	qry->returningList = NULL;
-
 	qry->hasTargetSRFs = false;
 	qry->hasSubLinks = pstate->p_hasSubLinks;
 
diff --git a/src/backend/parser/parse_relation.c b/src/backend/parser/parse_relation.c
new file mode 100644
index 41d6049..794530b
--- a/src/backend/parser/parse_relation.c
+++ b/src/backend/parser/parse_relation.c
@@ -2345,9 +2345,10 @@ addRangeTableEntryForCTE(ParseState *pst
 		cte->cterefcount++;
 
 	/*
-	 * We throw error if the CTE is INSERT/UPDATE/DELETE without RETURNING.
-	 * This won't get checked in case of a self-reference, but that's OK
-	 * because data-modifying CTEs aren't allowed to be recursive anyhow.
+	 * We throw error if the CTE is INSERT/UPDATE/DELETE/MERGE without
+	 * RETURNING.  This won't get checked in case of a self-reference, but
+	 * that's OK because data-modifying CTEs aren't allowed to be recursive
+	 * anyhow.
 	 */
 	if (IsA(cte->ctequery, Query))
 	{
diff --git a/src/backend/rewrite/rewriteHandler.c b/src/backend/rewrite/rewriteHandler.c
new file mode 100644
index b486ab5..1bd0ce2
--- a/src/backend/rewrite/rewriteHandler.c
+++ b/src/backend/rewrite/rewriteHandler.c
@@ -3629,9 +3629,9 @@ RewriteQuery(Query *parsetree, List *rew
 	ListCell   *lc1;
 
 	/*
-	 * First, recursively process any insert/update/delete statements in WITH
-	 * clauses.  (We have to do this first because the WITH clauses may get
-	 * copied into rule actions below.)
+	 * First, recursively process any insert/update/delete/merge statements in
+	 * WITH clauses.  (We have to do this first because the WITH clauses may
+	 * get copied into rule actions below.)
 	 */
 	foreach(lc1, parsetree->cteList)
 	{
@@ -3656,7 +3656,8 @@ RewriteQuery(Query *parsetree, List *rew
 			if (!(ctequery->commandType == CMD_SELECT ||
 				  ctequery->commandType == CMD_UPDATE ||
 				  ctequery->commandType == CMD_INSERT ||
-				  ctequery->commandType == CMD_DELETE))
+				  ctequery->commandType == CMD_DELETE ||
+				  ctequery->commandType == CMD_MERGE))
 			{
 				/*
 				 * Currently it could only be NOTIFY; this error message will
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
new file mode 100644
index 30b51bf..0ae3890
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -2141,11 +2141,10 @@ QueryReturnsTuples(Query *parsetree)
 		case CMD_SELECT:
 			/* returns tuples */
 			return true;
-		case CMD_MERGE:
-			return false;
 		case CMD_INSERT:
 		case CMD_UPDATE:
 		case CMD_DELETE:
+		case CMD_MERGE:
 			/* the forms with RETURNING return tuples */
 			if (parsetree->returningList)
 				return true;
diff --git a/src/backend/utils/adt/ruleutils.c b/src/backend/utils/adt/ruleutils.c
new file mode 100644
index d3a973d..f097d98
--- a/src/backend/utils/adt/ruleutils.c
+++ b/src/backend/utils/adt/ruleutils.c
@@ -7183,8 +7183,13 @@ get_merge_query_def(Query *query, depars
 			appendStringInfoString(buf, "DO NOTHING");
 	}
 
-	/* No RETURNING support in MERGE yet */
-	Assert(query->returningList == NIL);
+	/* Add RETURNING if present */
+	if (query->returningList)
+	{
+		appendContextKeyword(context, " RETURNING",
+							 -PRETTYINDENT_STD, PRETTYINDENT_STD, 1);
+		get_target_list(query->returningList, context, NULL, colNamesVisible);
+	}
 }
 
 
diff --git a/src/bin/psql/common.c b/src/bin/psql/common.c
new file mode 100644
index 5973df2..a1380f8
--- a/src/bin/psql/common.c
+++ b/src/bin/psql/common.c
@@ -982,13 +982,17 @@ PrintQueryResult(PGresult *result, bool
 			else
 				success = true;
 
-			/* if it's INSERT/UPDATE/DELETE RETURNING, also print status */
+			/*
+			 * If it's INSERT/UPDATE/DELETE/MERGE RETURNING, also print
+			 * status.
+			 */
 			if (last || pset.show_all_results)
 			{
 				cmdstatus = PQcmdStatus(result);
 				if (strncmp(cmdstatus, "INSERT", 6) == 0 ||
 					strncmp(cmdstatus, "UPDATE", 6) == 0 ||
-					strncmp(cmdstatus, "DELETE", 6) == 0)
+					strncmp(cmdstatus, "DELETE", 6) == 0 ||
+					strncmp(cmdstatus, "MERGE", 5) == 0)
 					PrintQueryStatus(result, printStatusFout);
 			}
 
diff --git a/src/include/catalog/pg_proc.dat b/src/include/catalog/pg_proc.dat
new file mode 100644
index 6996073..46c36af
--- a/src/include/catalog/pg_proc.dat
+++ b/src/include/catalog/pg_proc.dat
@@ -12043,4 +12043,14 @@
   proname => 'any_value_transfn', prorettype => 'anyelement',
   proargtypes => 'anyelement anyelement', prosrc => 'any_value_transfn' },
 
+# MERGE support functions
+{ oid => '9499', descr => 'command type of current MERGE action',
+  proname => 'pg_merge_action',  provolatile => 'v', proparallel => 'r',
+  prorettype => 'text', proargtypes => '',
+  prosrc => 'pg_merge_action' },
+{ oid => '9500', descr => 'index of current MERGE WHEN clause',
+  proname => 'pg_merge_when_clause',  provolatile => 'v', proparallel => 'r',
+  prorettype => 'int4', proargtypes => '',
+  prosrc => 'pg_merge_when_clause' },
+
 ]
diff --git a/src/include/catalog/pg_proc.h b/src/include/catalog/pg_proc.h
new file mode 100644
index e7abe0b..463add8
--- a/src/include/catalog/pg_proc.h
+++ b/src/include/catalog/pg_proc.h
@@ -182,6 +182,11 @@ DECLARE_UNIQUE_INDEX(pg_proc_proname_arg
 #define PROARGMODE_VARIADIC 'v'
 #define PROARGMODE_TABLE	't'
 
+/* Is this a merge support function?  (Requires fmgroids.h) */
+#define IsMergeSupportFunction(funcid) \
+	((funcid) == F_PG_MERGE_ACTION || \
+	 (funcid) == F_PG_MERGE_WHEN_CLAUSE)
+
 #endif							/* EXPOSE_TO_CLIENT_CODE */
 
 
diff --git a/src/include/executor/spi.h b/src/include/executor/spi.h
new file mode 100644
index d1de139..ed8787d
--- a/src/include/executor/spi.h
+++ b/src/include/executor/spi.h
@@ -97,6 +97,7 @@ typedef struct _SPI_plan *SPIPlanPtr;
 #define SPI_OK_REL_UNREGISTER	16
 #define SPI_OK_TD_REGISTER		17
 #define SPI_OK_MERGE			18
+#define SPI_OK_MERGE_RETURNING	19
 
 #define SPI_OPT_NONATOMIC		(1 << 0)
 
diff --git a/src/include/nodes/parsenodes.h b/src/include/nodes/parsenodes.h
new file mode 100644
index 88b03cc..79511e9
--- a/src/include/nodes/parsenodes.h
+++ b/src/include/nodes/parsenodes.h
@@ -1689,6 +1689,7 @@ typedef struct MergeWhenClause
 typedef struct MergeAction
 {
 	NodeTag		type;
+	int			index;			/* 1-based index of the clause */
 	bool		matched;		/* true=MATCHED, false=NOT MATCHED */
 	CmdType		commandType;	/* INSERT/UPDATE/DELETE/DO NOTHING */
 	/* OVERRIDING clause */
@@ -1915,6 +1916,7 @@ typedef struct MergeStmt
 	Node	   *sourceRelation; /* source relation */
 	Node	   *joinCondition;	/* join condition between source and target */
 	List	   *mergeWhenClauses;	/* list of MergeWhenClause(es) */
+	List	   *returningList;	/* list of expressions to return */
 	WithClause *withClause;		/* WITH clause */
 } MergeStmt;
 
diff --git a/src/include/optimizer/paramassign.h b/src/include/optimizer/paramassign.h
new file mode 100644
index 55c27a6..6f70c78
--- a/src/include/optimizer/paramassign.h
+++ b/src/include/optimizer/paramassign.h
@@ -20,6 +20,8 @@ extern Param *replace_outer_placeholderv
 										   PlaceHolderVar *phv);
 extern Param *replace_outer_agg(PlannerInfo *root, Aggref *agg);
 extern Param *replace_outer_grouping(PlannerInfo *root, GroupingFunc *grp);
+extern Param *replace_outer_merge_support_function(PlannerInfo *root,
+												   FuncExpr *func);
 extern Param *replace_nestloop_param_var(PlannerInfo *root, Var *var);
 extern Param *replace_nestloop_param_placeholdervar(PlannerInfo *root,
 													PlaceHolderVar *phv);
diff --git a/src/include/parser/analyze.h b/src/include/parser/analyze.h
new file mode 100644
index 1cef183..fd5848d
--- a/src/include/parser/analyze.h
+++ b/src/include/parser/analyze.h
@@ -44,6 +44,7 @@ extern List *transformInsertRow(ParseSta
 								bool strip_indirection);
 extern List *transformUpdateTargetList(ParseState *pstate,
 									   List *origTlist);
+extern List *transformReturningList(ParseState *pstate, List *returningList);
 extern Query *transformTopLevelStmt(ParseState *pstate, RawStmt *parseTree);
 extern Query *transformStmt(ParseState *pstate, Node *parseTree);
 
diff --git a/src/pl/plpgsql/src/pl_exec.c b/src/pl/plpgsql/src/pl_exec.c
new file mode 100644
index 4b76f76..fb6289c
--- a/src/pl/plpgsql/src/pl_exec.c
+++ b/src/pl/plpgsql/src/pl_exec.c
@@ -4267,9 +4267,9 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 	/*
 	 * If we have INTO, then we only need one row back ... but if we have INTO
 	 * STRICT or extra check too_many_rows, ask for two rows, so that we can
-	 * verify the statement returns only one.  INSERT/UPDATE/DELETE are always
-	 * treated strictly. Without INTO, just run the statement to completion
-	 * (tcount = 0).
+	 * verify the statement returns only one.  INSERT/UPDATE/DELETE/MERGE are
+	 * always treated strictly. Without INTO, just run the statement to
+	 * completion (tcount = 0).
 	 *
 	 * We could just ask for two rows always when using INTO, but there are
 	 * some cases where demanding the extra row costs significant time, eg by
@@ -4307,10 +4307,11 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 			Assert(stmt->mod_stmt);
 			exec_set_found(estate, (SPI_processed != 0));
 			break;
@@ -4489,10 +4490,11 @@ exec_stmt_dynexecute(PLpgSQL_execstate *
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 		case SPI_OK_UTILITY:
 		case SPI_OK_REWRITTEN:
 			break;
diff --git a/src/pl/tcl/pltcl.c b/src/pl/tcl/pltcl.c
new file mode 100644
index e8f9d7b..0c9f81e
--- a/src/pl/tcl/pltcl.c
+++ b/src/pl/tcl/pltcl.c
@@ -2456,6 +2456,7 @@ pltcl_process_SPI_result(Tcl_Interp *int
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
+		case SPI_OK_MERGE_RETURNING:
 
 			/*
 			 * Process the tuples we got
diff --git a/src/test/regress/expected/merge.out b/src/test/regress/expected/merge.out
new file mode 100644
index 133d421..e2cab70
--- a/src/test/regress/expected/merge.out
+++ b/src/test/regress/expected/merge.out
@@ -123,20 +123,20 @@ ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
 ERROR:  name "target" specified more than once
 DETAIL:  The name is used both as MERGE target table and data source.
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
-ERROR:  MERGE not supported in WITH query
-LINE 1: WITH foo AS (
-             ^
--- used in COPY
+ERROR:  WITH query "foo" does not have a RETURNING clause
+LINE 4: ) SELECT * FROM foo;
+                        ^
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) TO stdout;
-ERROR:  MERGE not supported in COPY
+ERROR:  COPY query must have a RETURNING clause
 -- unsupported relation types
 -- view
 CREATE VIEW tv AS SELECT * FROM target;
@@ -1304,20 +1304,184 @@ WHEN MATCHED AND tid < 2 THEN
 ROLLBACK;
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
+CREATE TABLE merge_actions(action text, abbrev text);
+INSERT INTO merge_actions VALUES ('INSERT', 'ins'), ('UPDATE', 'upd'), ('DELETE', 'del');
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
-ERROR:  syntax error at or near "RETURNING"
-LINE 10: RETURNING *;
-         ^
+    DELETE
+RETURNING pg_merge_when_clause() AS when_clause,
+          (SELECT abbrev FROM merge_actions
+            WHERE action = pg_merge_action()) AS merge_action,
+          t.*,
+          CASE pg_merge_action()
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ when_clause | merge_action | tid | balance |     description     
+-------------+--------------+-----+---------+---------------------
+           3 | del          |   1 |     100 | Removed (1,100)
+           1 | upd          |   2 |     220 | Added 20 to balance
+           2 | ins          |   4 |      40 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_when_clause() AS when_clause,
+              pg_merge_action() AS merge_action,
+              t.*,
+              CASE pg_merge_action()
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING merge_action, pg_merge_action() AS merge_log_action, l.*
+)
+SELECT * FROM m2;
+ merge_action | merge_log_action | tid |     last_change     
+--------------+------------------+-----+---------------------
+ DELETE       | UPDATE           |   1 | Removed (1,100)
+ UPDATE       | INSERT           |   2 | Added 20 to balance
+ INSERT       | INSERT           |   4 | Inserted (4,40)
+(3 rows)
+
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ tid |     last_change     
+-----+---------------------
+   1 | Removed (1,100)
+   2 | Added 20 to balance
+   4 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*
+) TO stdout;
+DELETE	1	100
+UPDATE	2	220
+INSERT	4	40
+ROLLBACK;
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ action | tid | new_balance 
+--------+-----+-------------
+ DELETE |   1 |         100
+ UPDATE |   3 |         320
+ INSERT |   4 |         110
+(3 rows)
+
+ROLLBACK;
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ action | tid | balance 
+--------+-----+---------
+ DELETE |   1 |     100
+ UPDATE |   2 |     220
+ INSERT |   4 |      40
+(3 rows)
+
+ROLLBACK;
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ r_action | r_tid | r_balance 
+----------+-------+-----------
+ DELETE   |     1 |       100
+ UPDATE   |     3 |       320
+ INSERT   |     4 |       110
+(3 rows)
+
 ROLLBACK;
 -- EXPLAIN
 CREATE TABLE ex_mtarget (a int, b int)
@@ -1514,7 +1678,7 @@ SELECT * FROM sq_target WHERE tid = 1;
 (1 row)
 
 ROLLBACK;
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 NOTICE:  drop cascades to view v
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1640,6 +1804,32 @@ SELECT * FROM pa_target ORDER BY tid;
 (14 rows)
 
 ROLLBACK;
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
+ pg_merge_action | pg_merge_when_clause | tid | balance |           val            
+-----------------+----------------------+-----+---------+--------------------------
+ UPDATE          |                    1 |   2 |     110 | initial updated by merge
+(1 row)
+
+SELECT * FROM pa_target ORDER BY tid;
+ tid | balance |           val            
+-----+---------+--------------------------
+   2 |     110 | initial updated by merge
+   3 |     300 | initial
+   5 |     500 | initial
+   7 |     700 | initial
+   9 |     900 | initial
+  11 |    1100 | initial
+  13 |    1300 | initial
+(7 rows)
+
+ROLLBACK;
 DROP TABLE pa_target CASCADE;
 -- The target table is partitioned in the same way, but this time by attaching
 -- partitions which have columns in different order, dropped columns etc.
@@ -1731,7 +1921,26 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (sid, delta, 'inserted by merge');
+    INSERT VALUES (sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
+ pg_merge_action | pg_merge_when_clause | tid | balance |           val            
+-----------------+----------------------+-----+---------+--------------------------
+ UPDATE          |                    1 |   2 |     110 | initial updated by merge
+ INSERT          |                    2 |   2 |      20 | inserted by merge
+ UPDATE          |                    1 |   4 |     330 | initial updated by merge
+ INSERT          |                    2 |   4 |      40 | inserted by merge
+ UPDATE          |                    1 |   6 |     550 | initial updated by merge
+ INSERT          |                    2 |   6 |      60 | inserted by merge
+ UPDATE          |                    1 |   8 |     770 | initial updated by merge
+ INSERT          |                    2 |   8 |      80 | inserted by merge
+ UPDATE          |                    1 |  10 |     990 | initial updated by merge
+ INSERT          |                    2 |  10 |     100 | inserted by merge
+ UPDATE          |                    1 |  12 |    1210 | initial updated by merge
+ INSERT          |                    2 |  12 |     120 | inserted by merge
+ UPDATE          |                    1 |  14 |    1430 | initial updated by merge
+ INSERT          |                    2 |  14 |     140 | inserted by merge
+(14 rows)
+
 SELECT * FROM pa_target ORDER BY tid;
  tid | balance |           val            
 -----+---------+--------------------------
@@ -1798,7 +2007,21 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
+ pg_merge_action | pg_merge_when_clause |          logts           | tid | balance |           val            
+-----------------+----------------------+--------------------------+-----+---------+--------------------------
+ UPDATE          |                    1 | Tue Jan 31 00:00:00 2017 |   1 |     110 | initial updated by merge
+ UPDATE          |                    1 | Tue Feb 28 00:00:00 2017 |   2 |     220 | initial updated by merge
+ INSERT          |                    2 | Sun Jan 15 00:00:00 2017 |   3 |      30 | inserted by merge
+ UPDATE          |                    1 | Tue Jan 31 00:00:00 2017 |   4 |     440 | initial updated by merge
+ UPDATE          |                    1 | Tue Feb 28 00:00:00 2017 |   5 |     550 | initial updated by merge
+ INSERT          |                    2 | Sun Jan 15 00:00:00 2017 |   6 |      60 | inserted by merge
+ UPDATE          |                    1 | Tue Jan 31 00:00:00 2017 |   7 |     770 | initial updated by merge
+ UPDATE          |                    1 | Tue Feb 28 00:00:00 2017 |   8 |     880 | initial updated by merge
+ INSERT          |                    2 | Sun Jan 15 00:00:00 2017 |   9 |      90 | inserted by merge
+(9 rows)
+
 SELECT * FROM pa_target ORDER BY tid;
           logts           | tid | balance |           val            
 --------------------------+-----+---------+--------------------------
diff --git a/src/test/regress/expected/rules.out b/src/test/regress/expected/rules.out
new file mode 100644
index 7fd81e6..87a3686
--- a/src/test/regress/expected/rules.out
+++ b/src/test/regress/expected/rules.out
@@ -3590,7 +3590,8 @@ MERGE INTO rule_merge2 t USING (SELECT 1
 -- test deparsing
 CREATE TABLE sf_target(id int, data text, filling int[]);
 CREATE FUNCTION merge_sf_test()
- RETURNS void
+ RETURNS TABLE(pg_merge_action text, when_clause int,
+               a int, b text, id int, data text, filling int[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -3627,11 +3628,13 @@ WHEN NOT MATCHED
    VALUES (s.a, s.b, DEFAULT)
 WHEN NOT MATCHED
    THEN INSERT (filling[1], id)
-   VALUES (s.a, s.a);
+   VALUES (s.a, s.a)
+RETURNING
+   pg_merge_action(), pg_merge_when_clause() AS when_clause, *;
 END;
 \sf merge_sf_test
 CREATE OR REPLACE FUNCTION public.merge_sf_test()
- RETURNS void
+ RETURNS TABLE(pg_merge_action text, when_clause integer, a integer, b text, id integer, data text, filling integer[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -3668,7 +3671,14 @@ BEGIN ATOMIC
       VALUES (s.a, s.b, DEFAULT)
     WHEN NOT MATCHED
      THEN INSERT (filling[1], id)
-      VALUES (s.a, s.a);
+      VALUES (s.a, s.a)
+   RETURNING pg_merge_action() AS pg_merge_action,
+     pg_merge_when_clause() AS when_clause,
+     s.a,
+     s.b,
+     t.id,
+     t.data,
+     t.filling;
 END
 DROP FUNCTION merge_sf_test;
 DROP TABLE sf_target;
diff --git a/src/test/regress/sql/merge.sql b/src/test/regress/sql/merge.sql
new file mode 100644
index 4cf6db9..6dfe4da
--- a/src/test/regress/sql/merge.sql
+++ b/src/test/regress/sql/merge.sql
@@ -86,12 +86,12 @@ MERGE INTO target
 USING target
 ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
--- used in COPY
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
@@ -857,17 +857,144 @@ ROLLBACK;
 
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
+CREATE TABLE merge_actions(action text, abbrev text);
+INSERT INTO merge_actions VALUES ('INSERT', 'ins'), ('UPDATE', 'upd'), ('DELETE', 'del');
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
+    DELETE
+RETURNING pg_merge_when_clause() AS when_clause,
+          (SELECT abbrev FROM merge_actions
+            WHERE action = pg_merge_action()) AS merge_action,
+          t.*,
+          CASE pg_merge_action()
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ROLLBACK;
+
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_when_clause() AS when_clause,
+              pg_merge_action() AS merge_action,
+              t.*,
+              CASE pg_merge_action()
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING merge_action, pg_merge_action() AS merge_log_action, l.*
+)
+SELECT * FROM m2;
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ROLLBACK;
+
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*
+) TO stdout;
+ROLLBACK;
+
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ROLLBACK;
+
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ROLLBACK;
+
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
 ROLLBACK;
 
 -- EXPLAIN
@@ -966,7 +1093,7 @@ WHEN MATCHED THEN
 SELECT * FROM sq_target WHERE tid = 1;
 ROLLBACK;
 
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1033,6 +1160,17 @@ SELECT merge_func();
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
+SELECT * FROM pa_target ORDER BY tid;
+ROLLBACK;
+
 DROP TABLE pa_target CASCADE;
 
 -- The target table is partitioned in the same way, but this time by attaching
@@ -1091,7 +1229,8 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (sid, delta, 'inserted by merge');
+    INSERT VALUES (sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
@@ -1145,7 +1284,8 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause(), t.*;
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
diff --git a/src/test/regress/sql/rules.sql b/src/test/regress/sql/rules.sql
new file mode 100644
index 8b7e255..a33515b
--- a/src/test/regress/sql/rules.sql
+++ b/src/test/regress/sql/rules.sql
@@ -1281,7 +1281,8 @@ MERGE INTO rule_merge2 t USING (SELECT 1
 CREATE TABLE sf_target(id int, data text, filling int[]);
 
 CREATE FUNCTION merge_sf_test()
- RETURNS void
+ RETURNS TABLE(pg_merge_action text, when_clause int,
+               a int, b text, id int, data text, filling int[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -1318,7 +1319,9 @@ WHEN NOT MATCHED
    VALUES (s.a, s.b, DEFAULT)
 WHEN NOT MATCHED
    THEN INSERT (filling[1], id)
-   VALUES (s.a, s.a);
+   VALUES (s.a, s.a)
+RETURNING
+   pg_merge_action(), pg_merge_when_clause() AS when_clause, *;
 END;
 
 \sf merge_sf_test

diff --git a/doc/src/sgml/dml.sgml b/doc/src/sgml/dml.sgml
new file mode 100644
index cbbc5e2..7f65f6e
--- a/doc/src/sgml/dml.sgml
+++ b/doc/src/sgml/dml.sgml
@@ -283,10 +283,15 @@ DELETE FROM products;
    <secondary>RETURNING</secondary>
   </indexterm>
 
+  <indexterm zone="dml-returning">
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
   <para>
    Sometimes it is useful to obtain data from modified rows while they are
    being manipulated.  The <command>INSERT</command>, <command>UPDATE</command>,
-   and <command>DELETE</command> commands all have an
+   <command>DELETE</command>, and <command>MERGE</command> commands all have an
    optional <literal>RETURNING</literal> clause that supports this.  Use
    of <literal>RETURNING</literal> avoids performing an extra database query to
    collect the data, and is especially valuable when it would otherwise be
@@ -339,6 +344,24 @@ DELETE FROM products
 </programlisting>
   </para>
 
+  <para>
+   In a <command>MERGE</command>, the data available to <literal>RETURNING</literal> is
+   the content of the source row plus the content of the inserted, updated, or
+   deleted target row.  In addition, the merge support functions
+   <xref linkend="pg-merge-action"/> and <xref linkend="pg-merge-when-clause-number"/>
+   may be used to return information about which merge action was executed for
+   each row.  Since it is quite common for the source and target to have many
+   of the same columns, specifying <literal>RETURNING *</literal> can lead to a
+   lot of duplicated columns, so it is often more useful to just return the
+   target row.  For example:
+<programlisting>
+MERGE INTO products p USING new_products n ON p.product_no = n.product_no
+  WHEN NOT MATCHED THEN INSERT VALUES (n.product_no, n.name, n.price)
+  WHEN MATCHED THEN UPDATE SET name = n.name, price = n.price
+  RETURNING pg_merge_action(), p.*;
+</programlisting>
+  </para>
+
   <para>
    If there are triggers (<xref linkend="triggers"/>) on the target table,
    the data available to <literal>RETURNING</literal> is the row as modified by
diff --git a/doc/src/sgml/func.sgml b/doc/src/sgml/func.sgml
new file mode 100644
index b948276..b84a95a
--- a/doc/src/sgml/func.sgml
+++ b/doc/src/sgml/func.sgml
@@ -21705,6 +21705,100 @@ SELECT count(*) FROM sometable;
 
  </sect1>
 
+ <sect1 id="functions-merge-support">
+  <title>Merge Support Functions</title>
+
+  <indexterm>
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
+  <para>
+   The merge support functions shown in
+   <xref linkend="functions-merge-support-table"/> may be used in the
+   <literal>RETURNING</literal> list of a <xref linkend="sql-merge"/> command
+   to return additional information about the action taken for each row.
+  </para>
+
+  <table id="functions-merge-support-table">
+    <title>Merge Support Functions</title>
+
+    <tgroup cols="1">
+     <thead>
+      <row>
+       <entry role="func_table_entry"><para role="func_signature">
+        Function
+       </para>
+       <para>
+        Description
+       </para></entry>
+      </row>
+     </thead>
+
+     <tbody>
+      <row>
+       <entry role="func_table_entry"><para role="func_signature">
+        <indexterm id="pg-merge-action">
+         <primary>pg_merge_action</primary>
+        </indexterm>
+        <function>pg_merge_action</function> ( )
+        <returnvalue>text</returnvalue>
+       </para>
+       <para>
+        Returns the action performed on the current row (<literal>'INSERT'</literal>,
+        <literal>'UPDATE'</literal>, or <literal>'DELETE'</literal>).
+       </para></entry>
+      </row>
+
+      <row>
+       <entry role="func_table_entry"><para role="func_signature">
+        <indexterm id="pg-merge-when-clause-number">
+         <primary>pg_merge_when_clause_number</primary>
+        </indexterm>
+        <function>pg_merge_when_clause_number</function> ( )
+        <returnvalue>integer</returnvalue>
+       </para>
+       <para>
+        Returns the 1-based index of the <literal>WHEN</literal> clause executed
+        for the current row.
+       </para></entry>
+      </row>
+     </tbody>
+    </tgroup>
+  </table>
+
+  <para>
+   Example:
+<screen><![CDATA[
+MERGE INTO products p
+  USING stock s ON p.product_id = s.product_id
+  WHEN MATCHED AND s.quantity > 0 THEN
+    UPDATE SET in_stock = true, quantity = s.quantity
+  WHEN MATCHED THEN
+    UPDATE SET in_stock = false, quantity = 0
+  WHEN NOT MATCHED THEN
+    INSERT (product_id, in_stock, quantity)
+      VALUES (s.product_id, true, s.quantity)
+  RETURNING pg_merge_action() AS action,
+            pg_merge_when_clause_number() AS clause_number,
+            p.*;
+
+ action | clause_number | product_id | in_stock | quantity 
+--------+---------------+------------+----------+----------
+ UPDATE |             1 |       1001 | t        |       50
+ UPDATE |             2 |       1002 | f        |        0
+ INSERT |             3 |       1003 | t        |       10
+]]></screen>
+  </para>
+
+  <para>
+   Note that these functions can only be used in the <literal>RETURNING</literal>
+   list of a <command>MERGE</command> command.  It is an error to use them in
+   any other part of a query.
+  </para>
+
+ </sect1>
+
  <sect1 id="functions-subquery">
   <title>Subquery Expressions</title>
 
diff --git a/doc/src/sgml/glossary.sgml b/doc/src/sgml/glossary.sgml
new file mode 100644
index fe8def4..51a15ca
--- a/doc/src/sgml/glossary.sgml
+++ b/doc/src/sgml/glossary.sgml
@@ -1387,9 +1387,9 @@
      to a <glossterm linkend="glossary-client">client</glossterm> upon the
      completion of an <acronym>SQL</acronym> command, usually a
      <command>SELECT</command> but it can be an
-     <command>INSERT</command>, <command>UPDATE</command>, or
-     <command>DELETE</command> command if the <literal>RETURNING</literal>
-     clause is specified.
+     <command>INSERT</command>, <command>UPDATE</command>,
+     <command>DELETE</command>, or <command>MERGE</command> command if the
+     <literal>RETURNING</literal> clause is specified.
     </para>
     <para>
      The fact that a result set is a relation means that a query can be used
diff --git a/doc/src/sgml/plpgsql.sgml b/doc/src/sgml/plpgsql.sgml
new file mode 100644
index f55e901..6803240
--- a/doc/src/sgml/plpgsql.sgml
+++ b/doc/src/sgml/plpgsql.sgml
@@ -1020,8 +1020,8 @@ INSERT INTO mytable VALUES (1,'one'), (2
     </para>
 
     <para>
-     If the command does return rows (for example <command>SELECT</command>,
-     or <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>
+     If the command does return rows (for example <command>SELECT</command>, or
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
      with <literal>RETURNING</literal>), there are two ways to proceed.
      When the command will return at most one row, or you only care about
      the first row of output, write the command as usual but add
@@ -1149,6 +1149,7 @@ SELECT <replaceable>select_expressions</
 INSERT ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 UPDATE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 DELETE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
+MERGE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 </synopsis>
 
      where <replaceable>target</replaceable> can be a record variable, a row
@@ -1159,8 +1160,8 @@ DELETE ... RETURNING <replaceable>expres
      <literal>INTO</literal> clause) just as described above,
      and the plan is cached in the same way.
      This works for <command>SELECT</command>,
-     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
-     <literal>RETURNING</literal>, and certain utility commands
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
+     with <literal>RETURNING</literal>, and certain utility commands
      that return row sets, such as <command>EXPLAIN</command>.
      Except for the <literal>INTO</literal> clause, the SQL command is the same
      as it would be written outside <application>PL/pgSQL</application>.
@@ -1236,7 +1237,7 @@ END;
     </para>
 
     <para>
-     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
+     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command> with
      <literal>RETURNING</literal>, <application>PL/pgSQL</application> reports
      an error for more than one returned row, even when
      <literal>STRICT</literal> is not specified.  This is because there
diff --git a/doc/src/sgml/queries.sgml b/doc/src/sgml/queries.sgml
new file mode 100644
index 3f95849..9928373
--- a/doc/src/sgml/queries.sgml
+++ b/doc/src/sgml/queries.sgml
@@ -2062,9 +2062,10 @@ SELECT <replaceable>select_list</replace
    Table Expressions or <acronym>CTE</acronym>s, can be thought of as defining
    temporary tables that exist just for one query.  Each auxiliary statement
    in a <literal>WITH</literal> clause can be a <command>SELECT</command>,
-   <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>; and the
+   <command>INSERT</command>, <command>UPDATE</command>, <command>DELETE</command>,
+   or <command>MERGE</command>; and the
    <literal>WITH</literal> clause itself is attached to a primary statement that can
-   be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
+   also be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
    <command>DELETE</command>, or <command>MERGE</command>.
   </para>
 
@@ -2598,8 +2599,8 @@ SELECT * FROM w AS w1 JOIN w AS w2 ON w1
    <title>Data-Modifying Statements in <literal>WITH</literal></title>
 
    <para>
-    You can use most data-modifying statements (<command>INSERT</command>,
-    <command>UPDATE</command>, or <command>DELETE</command>, but not
+    You can use data-modifying statements (<command>INSERT</command>,
+    <command>UPDATE</command>, <command>DELETE</command>, or
     <command>MERGE</command>) in <literal>WITH</literal>.  This
     allows you to perform several different operations in the same query.
     An example is:
diff --git a/doc/src/sgml/ref/copy.sgml b/doc/src/sgml/ref/copy.sgml
new file mode 100644
index 4d614a0..efaacb9
--- a/doc/src/sgml/ref/copy.sgml
+++ b/doc/src/sgml/ref/copy.sgml
@@ -122,13 +122,16 @@ COPY { <replaceable class="parameter">ta
       A <link linkend="sql-select"><command>SELECT</command></link>,
       <link linkend="sql-values"><command>VALUES</command></link>,
       <link linkend="sql-insert"><command>INSERT</command></link>,
-      <link linkend="sql-update"><command>UPDATE</command></link>, or
-      <link linkend="sql-delete"><command>DELETE</command></link> command whose results are to be
-      copied.  Note that parentheses are required around the query.
+      <link linkend="sql-update"><command>UPDATE</command></link>,
+      <link linkend="sql-delete"><command>DELETE</command></link>, or
+      <link linkend="sql-merge"><command>MERGE</command></link> command
+      whose results are to be copied.  Note that parentheses are required
+      around the query.
      </para>
      <para>
-      For <command>INSERT</command>, <command>UPDATE</command> and
-      <command>DELETE</command> queries a RETURNING clause must be provided,
+      For <command>INSERT</command>, <command>UPDATE</command>,
+      <command>DELETE</command>, and <command>MERGE</command> queries a
+      RETURNING clause must be provided,
       and the target relation must not have a conditional rule, nor
       an <literal>ALSO</literal> rule, nor an <literal>INSTEAD</literal> rule
       that expands to multiple statements.
diff --git a/doc/src/sgml/ref/merge.sgml b/doc/src/sgml/ref/merge.sgml
new file mode 100644
index 0995fe0..01dfc98
--- a/doc/src/sgml/ref/merge.sgml
+++ b/doc/src/sgml/ref/merge.sgml
@@ -25,6 +25,7 @@ PostgreSQL documentation
 MERGE INTO [ ONLY ] <replaceable class="parameter">target_table_name</replaceable> [ * ] [ [ AS ] <replaceable class="parameter">target_alias</replaceable> ]
 USING <replaceable class="parameter">data_source</replaceable> ON <replaceable class="parameter">join_condition</replaceable>
 <replaceable class="parameter">when_clause</replaceable> [...]
+[ RETURNING * | <replaceable class="parameter">output_expression</replaceable> [ [ AS ] <replaceable class="parameter">output_name</replaceable> ] [, ...] ]
 
 <phrase>where <replaceable class="parameter">data_source</replaceable> is:</phrase>
 
@@ -95,6 +96,20 @@ DELETE
   </para>
 
   <para>
+   The optional <literal>RETURNING</literal> clause causes <command>MERGE</command>
+   to compute and return value(s) based on each row inserted, updated, or
+   deleted.  Any expression using the source or target table's columns, or the
+   merge support functions <xref linkend="pg-merge-action"/> and
+   <xref linkend="pg-merge-when-clause-number"/> can be computed.  When an
+   <command>INSERT</command> or <command>UPDATE</command> action is performed,
+   the new values of the target table's columns are used.  When a
+   <command>DELETE</command> is performed, the old values of the target table's
+   columns are used.  The syntax of the <literal>RETURNING</literal> list is
+   identical to that of the output list of <command>SELECT</command>.
+
+  </para>
+
+  <para>
    There is no separate <literal>MERGE</literal> privilege.
    If you specify an update action, you must have the
    <literal>UPDATE</literal> privilege on the column(s)
@@ -125,6 +140,18 @@ DELETE
 
   <variablelist>
    <varlistentry>
+    <term><replaceable class="parameter">with_query</replaceable></term>
+    <listitem>
+     <para>
+      The <literal>WITH</literal> clause allows you to specify one or more
+      subqueries that can be referenced by name in the <command>MERGE</command>
+      query. See <xref linkend="queries-with"/> and <xref linkend="sql-select"/>
+      for details.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
     <term><replaceable class="parameter">target_table_name</replaceable></term>
     <listitem>
      <para>
@@ -399,13 +426,31 @@ DELETE
    </varlistentry>
 
    <varlistentry>
-    <term><replaceable class="parameter">with_query</replaceable></term>
+    <term><replaceable class="parameter">output_expression</replaceable></term>
     <listitem>
      <para>
-      The <literal>WITH</literal> clause allows you to specify one or more
-      subqueries that can be referenced by name in the <command>MERGE</command>
-      query. See <xref linkend="queries-with"/> and <xref linkend="sql-select"/>
-      for details.
+      An expression to be computed and returned by the <command>MERGE</command>
+      command after each row is merged.  The expression can use any columns of
+      the source or target tables, or the merge support functions
+      <xref linkend="pg-merge-action"/> and
+      <xref linkend="pg-merge-when-clause-number"/>.
+     </para>
+     <para>
+      Writing <literal>*</literal> will return all columns from the source
+      table, followed by all columns from the target table.  Often this will
+      lead to a lot of duplication, since it is common for the source and
+      target tables to have a lot of the same columns.  This can be avoided by
+      qualifying the <literal>*</literal> with the name or alias of the source
+      or target table.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
+    <term><replaceable class="parameter">output_name</replaceable></term>
+    <listitem>
+     <para>
+      A name to use for a returned column.
      </para>
     </listitem>
    </varlistentry>
@@ -428,6 +473,13 @@ MERGE <replaceable class="parameter">tot
    were changed in any way.
   </para>
 
+  <para>
+   If the <command>MERGE</command> command contains a <literal>RETURNING</literal>
+   clause, the result will be similar to that of a <command>SELECT</command>
+   statement containing the columns and values defined in the
+   <literal>RETURNING</literal> list, computed over the row(s) inserted, updated,
+   or deleted by the command.
+  </para>
  </refsect1>
 
  <refsect1>
@@ -544,13 +596,6 @@ MERGE <replaceable class="parameter">tot
   </para>
 
   <para>
-   There is no <literal>RETURNING</literal> clause with
-   <command>MERGE</command>.  Actions of <command>INSERT</command>,
-   <command>UPDATE</command> and <command>DELETE</command> cannot contain
-   <literal>RETURNING</literal> or <literal>WITH</literal> clauses.
-  </para>
-
-  <para>
    When <command>MERGE</command> is run concurrently with other commands
    that modify the target table, the usual transaction isolation rules
    apply; see <xref linkend="transaction-iso"/> for an explanation
@@ -602,7 +647,8 @@ WHEN NOT MATCHED THEN
   <para>
    Attempt to insert a new stock item along with the quantity of stock. If
    the item already exists, instead update the stock count of the existing
-   item. Don't allow entries that have zero stock.
+   item. Don't allow entries that have zero stock. Return details of all
+   changes made.
 <programlisting>
 MERGE INTO wines w
 USING wine_stock_changes s
@@ -612,7 +658,8 @@ WHEN NOT MATCHED AND s.stock_delta > 0 T
 WHEN MATCHED AND w.stock + s.stock_delta > 0 THEN
   UPDATE SET stock = w.stock + s.stock_delta
 WHEN MATCHED THEN
-  DELETE;
+  DELETE
+RETURNING pg_merge_action(), w.*;
 </programlisting>
 
    The <literal>wine_stock_changes</literal> table might be, for example, a
@@ -627,7 +674,8 @@ WHEN MATCHED THEN
     This command conforms to the <acronym>SQL</acronym> standard.
   </para>
    <para>
-    The WITH clause and <literal>DO NOTHING</literal> action are extensions to
+    The WITH clause, <literal>DO NOTHING</literal> action, and
+    <literal>RETURNING</literal> clause are extensions to
     the <acronym>SQL</acronym> standard.
   </para>
  </refsect1>
diff --git a/doc/src/sgml/ref/select.sgml b/doc/src/sgml/ref/select.sgml
new file mode 100644
index 0ee0cc7..f81bdea
--- a/doc/src/sgml/ref/select.sgml
+++ b/doc/src/sgml/ref/select.sgml
@@ -74,7 +74,7 @@ SELECT [ ALL | DISTINCT [ ON ( <replacea
 
 <phrase>and <replaceable class="parameter">with_query</replaceable> is:</phrase>
 
-    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> )
+    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> | <replaceable class="parameter">merge</replaceable> )
         [ SEARCH { BREADTH | DEPTH } FIRST BY <replaceable>column_name</replaceable> [, ...] SET <replaceable>search_seq_col_name</replaceable> ]
         [ CYCLE <replaceable>column_name</replaceable> [, ...] SET <replaceable>cycle_mark_col_name</replaceable> [ TO <replaceable>cycle_mark_value</replaceable> DEFAULT <replaceable>cycle_mark_default</replaceable> ] USING <replaceable>cycle_path_col_name</replaceable> ]
 
@@ -227,10 +227,10 @@ TABLE [ ONLY ] <replaceable class="param
     The subqueries effectively act as temporary tables or views
     for the duration of the primary query.
     Each subquery can be a <command>SELECT</command>, <command>TABLE</command>, <command>VALUES</command>,
-    <command>INSERT</command>, <command>UPDATE</command> or
-    <command>DELETE</command> statement.
+    <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command> statement.
     When writing a data-modifying statement (<command>INSERT</command>,
-    <command>UPDATE</command> or <command>DELETE</command>) in
+    <command>UPDATE</command>, <command>DELETE</command>, or <command>MERGE</command>) in
     <literal>WITH</literal>, it is usual to include a <literal>RETURNING</literal> clause.
     It is the output of <literal>RETURNING</literal>, <emphasis>not</emphasis> the underlying
     table that the statement modifies, that forms the temporary table that is
@@ -2179,7 +2179,8 @@ SELECT 2+2;
 
    <para>
     <productname>PostgreSQL</productname> allows <command>INSERT</command>,
-    <command>UPDATE</command>, and <command>DELETE</command> to be used as <literal>WITH</literal>
+    <command>UPDATE</command>, <command>DELETE</command>, and
+    <command>MERGE</command> to be used as <literal>WITH</literal>
     queries.  This is not found in the SQL standard.
    </para>
   </refsect2>
diff --git a/doc/src/sgml/rowtypes.sgml b/doc/src/sgml/rowtypes.sgml
new file mode 100644
index 4d86f97..bbeac84
--- a/doc/src/sgml/rowtypes.sgml
+++ b/doc/src/sgml/rowtypes.sgml
@@ -348,7 +348,7 @@ SELECT m.* FROM some_table, LATERAL myfu
    column expansion of this kind when it appears at the top level of
    a <link linkend="queries-select-lists"><command>SELECT</command> output
    list</link>, a <link linkend="dml-returning"><literal>RETURNING</literal>
-   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>,
+   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>,
    a <link linkend="queries-values"><literal>VALUES</literal> clause</link>, or
    a <link linkend="sql-syntax-row-constructors">row constructor</link>.
    In all other contexts (including when nested inside one of those
diff --git a/doc/src/sgml/spi.sgml b/doc/src/sgml/spi.sgml
new file mode 100644
index 16c3ab7..51d70c7
--- a/doc/src/sgml/spi.sgml
+++ b/doc/src/sgml/spi.sgml
@@ -301,8 +301,9 @@ SPI_execute("INSERT INTO foo SELECT * FR
    is returned in the global variable <varname>SPI_processed</varname>.
    If the return value of the function is <symbol>SPI_OK_SELECT</symbol>,
    <symbol>SPI_OK_INSERT_RETURNING</symbol>,
-   <symbol>SPI_OK_DELETE_RETURNING</symbol>, or
-   <symbol>SPI_OK_UPDATE_RETURNING</symbol>,
+   <symbol>SPI_OK_DELETE_RETURNING</symbol>,
+   <symbol>SPI_OK_UPDATE_RETURNING</symbol>, or
+   <symbol>SPI_OK_MERGE_RETURNING</symbol>
    then you can use the
    global pointer <literal>SPITupleTable *SPI_tuptable</literal> to
    access the result rows.  Some utility commands (such as
@@ -472,6 +473,15 @@ typedef struct SPITupleTable
       </para>
      </listitem>
     </varlistentry>
+
+    <varlistentry>
+     <term><symbol>SPI_OK_MERGE_RETURNING</symbol></term>
+     <listitem>
+      <para>
+       if a <command>MERGE RETURNING</command> was executed
+      </para>
+     </listitem>
+    </varlistentry>
 
     <varlistentry>
      <term><symbol>SPI_OK_UTILITY</symbol></term>
diff --git a/doc/src/sgml/xfunc.sgml b/doc/src/sgml/xfunc.sgml
new file mode 100644
index 9620ea9..4f68efb
--- a/doc/src/sgml/xfunc.sgml
+++ b/doc/src/sgml/xfunc.sgml
@@ -177,7 +177,8 @@
     statements separated by semicolons.  A semicolon after the last
     statement is optional.  Unless the function is declared to return
     <type>void</type>, the last statement must be a <command>SELECT</command>,
-    or an <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>
+    or an <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command>
     that has a <literal>RETURNING</literal> clause.
    </para>
 
@@ -1236,8 +1237,9 @@ SELECT x, CASE WHEN x &gt; 0 THEN genera
 
     <note>
      <para>
-      If a function's last command is <command>INSERT</command>, <command>UPDATE</command>,
-      or <command>DELETE</command> with <literal>RETURNING</literal>, that command will
+      If a function's last command is <command>INSERT</command>,
+      <command>UPDATE</command>, <command>DELETE</command>, or
+      <command>MERGE</command> with <literal>RETURNING</literal>, that command will
       always be executed to completion, even if the function is not declared
       with <literal>SETOF</literal> or the calling query does not fetch all the
       result rows.  Any extra rows produced by the <literal>RETURNING</literal>
diff --git a/src/backend/commands/copy.c b/src/backend/commands/copy.c
new file mode 100644
index f14fae3..05f2e5b
--- a/src/backend/commands/copy.c
+++ b/src/backend/commands/copy.c
@@ -283,12 +283,6 @@ DoCopy(ParseState *pstate, const CopyStm
 	{
 		Assert(stmt->query);
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(stmt->query, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in COPY"));
-
 		query = makeNode(RawStmt);
 		query->stmt = stmt->query;
 		query->stmt_location = stmt_location;
diff --git a/src/backend/commands/copyto.c b/src/backend/commands/copyto.c
new file mode 100644
index 9e4b243..cf56f9f
--- a/src/backend/commands/copyto.c
+++ b/src/backend/commands/copyto.c
@@ -510,7 +510,8 @@ BeginCopyTo(ParseState *pstate,
 		{
 			Assert(query->commandType == CMD_INSERT ||
 				   query->commandType == CMD_UPDATE ||
-				   query->commandType == CMD_DELETE);
+				   query->commandType == CMD_DELETE ||
+				   query->commandType == CMD_MERGE);
 
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
diff --git a/src/backend/executor/execExpr.c b/src/backend/executor/execExpr.c
new file mode 100644
index bf3a08c..7bdaac1
--- a/src/backend/executor/execExpr.c
+++ b/src/backend/executor/execExpr.c
@@ -48,6 +48,7 @@
 #include "utils/array.h"
 #include "utils/builtins.h"
 #include "utils/datum.h"
+#include "utils/fmgroids.h"
 #include "utils/lsyscache.h"
 #include "utils/typcache.h"
 
@@ -2590,6 +2591,22 @@ ExecInitFunc(ExprEvalStep *scratch, Expr
 	InitFunctionCallInfoData(*fcinfo, flinfo,
 							 nargs, inputcollid, NULL, NULL);
 
+	/*
+	 * Merge support functions should only be called from a MERGE command, and
+	 * need access to the parent ModifyTableState.  The parser should have
+	 * checked that such functions only appear in the RETURNING list of a
+	 * MERGE, so this should never fail.
+	 */
+	if (IsMergeSupportFunction(funcid))
+	{
+		if (!state->parent ||
+			!IsA(state->parent, ModifyTableState) ||
+			((ModifyTableState *) state->parent)->operation != CMD_MERGE)
+			elog(ERROR, "merge support function called in non-merge context");
+
+		fcinfo->context = (Node *) state->parent;
+	}
+
 	/* Keep extra copies of this info to save an indirection at runtime */
 	scratch->d.func.fn_addr = flinfo->fn_addr;
 	scratch->d.func.nargs = nargs;
diff --git a/src/backend/executor/execPartition.c b/src/backend/executor/execPartition.c
new file mode 100644
index eb8a87f..b9bd03c
--- a/src/backend/executor/execPartition.c
+++ b/src/backend/executor/execPartition.c
@@ -612,8 +612,8 @@ ExecInitPartitionInfo(ModifyTableState *
 	 * Build the RETURNING projection for the partition.  Note that we didn't
 	 * build the returningList for partitions within the planner, but simple
 	 * translation of varattnos will suffice.  This only occurs for the INSERT
-	 * case or in the case of UPDATE tuple routing where we didn't find a
-	 * result rel to reuse.
+	 * case or in the case of UPDATE/MERGE tuple routing where we didn't find
+	 * a result rel to reuse.
 	 */
 	if (node && node->returningLists != NIL)
 	{
@@ -622,12 +622,14 @@ ExecInitPartitionInfo(ModifyTableState *
 		List	   *returningList;
 
 		/* See the comment above for WCO lists. */
-		/* (except no RETURNING support for MERGE yet) */
 		Assert((node->operation == CMD_INSERT &&
 				list_length(node->returningLists) == 1 &&
 				list_length(node->resultRelations) == 1) ||
 			   (node->operation == CMD_UPDATE &&
 				list_length(node->returningLists) ==
+				list_length(node->resultRelations)) ||
+			   (node->operation == CMD_MERGE &&
+				list_length(node->returningLists) ==
 				list_length(node->resultRelations)));
 
 		/*
diff --git a/src/backend/executor/functions.c b/src/backend/executor/functions.c
new file mode 100644
index f55424e..f99f15f
--- a/src/backend/executor/functions.c
+++ b/src/backend/executor/functions.c
@@ -1660,8 +1660,8 @@ check_sql_fn_retval(List *queryTreeLists
 
 	/*
 	 * If it's a plain SELECT, it returns whatever the targetlist says.
-	 * Otherwise, if it's INSERT/UPDATE/DELETE with RETURNING, it returns
-	 * that. Otherwise, the function return type must be VOID.
+	 * Otherwise, if it's INSERT/UPDATE/DELETE/MERGE with RETURNING, it
+	 * returns that. Otherwise, the function return type must be VOID.
 	 *
 	 * Note: eventually replace this test with QueryReturnsTuples?	We'd need
 	 * a more general method of determining the output type, though.  Also, it
@@ -1679,7 +1679,8 @@ check_sql_fn_retval(List *queryTreeLists
 	else if (parse &&
 			 (parse->commandType == CMD_INSERT ||
 			  parse->commandType == CMD_UPDATE ||
-			  parse->commandType == CMD_DELETE) &&
+			  parse->commandType == CMD_DELETE ||
+			  parse->commandType == CMD_MERGE) &&
 			 parse->returningList)
 	{
 		tlist = parse->returningList;
diff --git a/src/backend/executor/nodeModifyTable.c b/src/backend/executor/nodeModifyTable.c
new file mode 100644
index 2a5fec8..460127e
--- a/src/backend/executor/nodeModifyTable.c
+++ b/src/backend/executor/nodeModifyTable.c
@@ -36,8 +36,7 @@
  *		RETURNING tuple after completing each row insert, update, or delete.
  *		It must be called again to continue the operation.  Without RETURNING,
  *		we just loop within the node until all the work is done, then
- *		return NULL.  This avoids useless call/return overhead.  (MERGE does
- *		not support RETURNING.)
+ *		return NULL.  This avoids useless call/return overhead.
  */
 
 #include "postgres.h"
@@ -88,9 +87,6 @@ typedef struct ModifyTableContext
 	 */
 	TupleTableSlot *planSlot;
 
-	/* MERGE specific */
-	MergeActionState *relaction;	/* MERGE action in progress */
-
 	/*
 	 * Information about the changes that were made concurrently to a tuple
 	 * being updated or deleted
@@ -153,13 +149,14 @@ static TupleTableSlot *ExecMerge(ModifyT
 								 ItemPointer tupleid,
 								 bool canSetTag);
 static void ExecInitMerge(ModifyTableState *mtstate, EState *estate);
-static bool ExecMergeMatched(ModifyTableContext *context,
-							 ResultRelInfo *resultRelInfo,
-							 ItemPointer tupleid,
-							 bool canSetTag);
-static void ExecMergeNotMatched(ModifyTableContext *context,
-								ResultRelInfo *resultRelInfo,
-								bool canSetTag);
+static TupleTableSlot *ExecMergeMatched(ModifyTableContext *context,
+										ResultRelInfo *resultRelInfo,
+										ItemPointer tupleid,
+										bool canSetTag,
+										bool *matched);
+static TupleTableSlot *ExecMergeNotMatched(ModifyTableContext *context,
+										   ResultRelInfo *resultRelInfo,
+										   bool canSetTag);
 
 
 /*
@@ -979,7 +976,7 @@ ExecInsert(ModifyTableContext *context,
 		if (mtstate->operation == CMD_UPDATE)
 			wco_kind = WCO_RLS_UPDATE_CHECK;
 		else if (mtstate->operation == CMD_MERGE)
-			wco_kind = (context->relaction->mas_action->commandType == CMD_UPDATE) ?
+			wco_kind = (mtstate->mt_merge_action->mas_action->commandType == CMD_UPDATE) ?
 				WCO_RLS_UPDATE_CHECK : WCO_RLS_INSERT_CHECK;
 		else
 			wco_kind = WCO_RLS_INSERT_CHECK;
@@ -1828,7 +1825,7 @@ ExecCrossPartitionUpdate(ModifyTableCont
 		 * additional rechecking, and might end up executing a different
 		 * action entirely).
 		 */
-		if (context->relaction != NULL)
+		if (mtstate->mt_merge_action != NULL)
 			return false;
 		else if (TupIsNull(epqslot))
 			return true;
@@ -2069,7 +2066,7 @@ lreplace:
 		 * No luck, a retry is needed.  If running MERGE, we do not do so
 		 * here; instead let it handle that on its own rules.
 		 */
-		if (context->relaction != NULL)
+		if (context->mtstate->mt_merge_action != NULL)
 			return TM_Updated;
 
 		/*
@@ -2712,6 +2709,7 @@ static TupleTableSlot *
 ExecMerge(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 		  ItemPointer tupleid, bool canSetTag)
 {
+	TupleTableSlot *rslot = NULL;
 	bool		matched;
 
 	/*-----
@@ -2759,18 +2757,18 @@ ExecMerge(ModifyTableContext *context, R
 	 */
 	matched = tupleid != NULL;
 	if (matched)
-		matched = ExecMergeMatched(context, resultRelInfo, tupleid, canSetTag);
+		rslot = ExecMergeMatched(context, resultRelInfo, tupleid, canSetTag,
+								 &matched);
 
 	/*
-	 * Either we were dealing with a NOT MATCHED tuple or ExecMergeMatched()
-	 * returned "false", indicating the previously MATCHED tuple no longer
-	 * matches.
+	 * Deal with the NOT MATCHED case (either a NOT MATCHED tuple from the
+	 * join, or a previously MATCHED tuple for which ExecMergeMatched() set
+	 * "matched" to false, indicating that it no longer matches).
 	 */
 	if (!matched)
-		ExecMergeNotMatched(context, resultRelInfo, canSetTag);
+		rslot = ExecMergeNotMatched(context, resultRelInfo, canSetTag);
 
-	/* No RETURNING support yet */
-	return NULL;
+	return rslot;
 }
 
 /*
@@ -2780,8 +2778,8 @@ ExecMerge(ModifyTableContext *context, R
  * We start from the first WHEN MATCHED action and check if the WHEN quals
  * pass, if any. If the WHEN quals for the first action do not pass, we
  * check the second, then the third and so on. If we reach to the end, no
- * action is taken and we return true, indicating that no further action is
- * required for this tuple.
+ * action is taken and "matched" is set to true, indicating that no further
+ * action is required for this tuple.
  *
  * If we do find a qualifying action, then we attempt to execute the action.
  *
@@ -2790,15 +2788,16 @@ ExecMerge(ModifyTableContext *context, R
  * with individual actions are evaluated by this routine via ExecQual, while
  * EvalPlanQual checks for the join quals. If EvalPlanQual tells us that the
  * updated tuple still passes the join quals, then we restart from the first
- * action to look for a qualifying action. Otherwise, we return false --
- * meaning that a NOT MATCHED action must now be executed for the current
- * source tuple.
+ * action to look for a qualifying action. Otherwise, "matched" is set to
+ * false -- meaning that a NOT MATCHED action must now be executed for the
+ * current source tuple.
  */
-static bool
+static TupleTableSlot *
 ExecMergeMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
-				 ItemPointer tupleid, bool canSetTag)
+				 ItemPointer tupleid, bool canSetTag, bool *matched)
 {
 	ModifyTableState *mtstate = context->mtstate;
+	TupleTableSlot *rslot = NULL;
 	TupleTableSlot *newslot;
 	EState	   *estate = context->estate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
@@ -2810,7 +2809,10 @@ ExecMergeMatched(ModifyTableContext *con
 	 * If there are no WHEN MATCHED actions, we are done.
 	 */
 	if (resultRelInfo->ri_matchedMergeAction == NIL)
-		return true;
+	{
+		*matched = true;
+		return NULL;
+	}
 
 	/*
 	 * Make tuple and any needed join variables available to ExecQual and
@@ -2888,12 +2890,15 @@ lmerge_matched:
 				 */
 				newslot = ExecProject(relaction->mas_proj);
 
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				if (!ExecUpdatePrologue(context, resultRelInfo,
 										tupleid, NULL, newslot, &result))
 				{
 					if (result == TM_Ok)
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 					break;		/* concurrent update/delete */
 				}
 				result = ExecUpdateAct(context, resultRelInfo, tupleid, NULL,
@@ -2907,12 +2912,15 @@ lmerge_matched:
 				break;
 
 			case CMD_DELETE:
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				if (!ExecDeletePrologue(context, resultRelInfo, tupleid,
 										NULL, NULL, &result))
 				{
 					if (result == TM_Ok)
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 					break;		/* concurrent update/delete */
 				}
 				result = ExecDeleteAct(context, resultRelInfo, tupleid, false);
@@ -2968,7 +2976,8 @@ lmerge_matched:
 				 * If the tuple was already deleted, return to let caller
 				 * handle it under NOT MATCHED clauses.
 				 */
-				return false;
+				*matched = false;
+				return NULL;
 
 			case TM_Updated:
 				{
@@ -3014,13 +3023,19 @@ lmerge_matched:
 							 * NOT MATCHED actions.
 							 */
 							if (TupIsNull(epqslot))
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							(void) ExecGetJunkAttribute(epqslot,
 														resultRelInfo->ri_RowIdAttNo,
 														&isNull);
 							if (isNull)
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							/*
 							 * When a tuple was updated and migrated to
@@ -3055,7 +3070,8 @@ lmerge_matched:
 							 * tuple already deleted; tell caller to run NOT
 							 * MATCHED actions
 							 */
-							return false;
+							*matched = false;
+							return NULL;
 
 						case TM_SelfModified:
 
@@ -3075,13 +3091,14 @@ lmerge_matched:
 										(errcode(ERRCODE_TRIGGERED_DATA_CHANGE_VIOLATION),
 										 errmsg("tuple to be updated or deleted was already modified by an operation triggered by the current command"),
 										 errhint("Consider using an AFTER trigger instead of a BEFORE trigger to propagate changes to other rows.")));
-							return false;
+							*matched = false;
+							return NULL;
 
 						default:
 							/* see table_tuple_lock call in ExecDelete() */
 							elog(ERROR, "unexpected table_tuple_lock status: %u",
 								 result);
-							return false;
+							return NULL;
 					}
 				}
 
@@ -3093,6 +3110,30 @@ lmerge_matched:
 				break;
 		}
 
+		/* Process RETURNING if present */
+		if (resultRelInfo->ri_projectReturning)
+		{
+			switch (commandType)
+			{
+				case CMD_UPDATE:
+					rslot = ExecProcessReturning(resultRelInfo, newslot,
+												 context->planSlot);
+					break;
+
+				case CMD_DELETE:
+					rslot = ExecProcessReturning(resultRelInfo,
+												 resultRelInfo->ri_oldTupleSlot,
+												 context->planSlot);
+					break;
+
+				case CMD_NOTHING:
+					break;
+
+				default:
+					elog(ERROR, "unknown action in MERGE WHEN MATCHED clause");
+			}
+		}
+
 		/*
 		 * We've activated one of the WHEN clauses, so we don't search
 		 * further. This is required behaviour, not an optimization.
@@ -3103,19 +3144,22 @@ lmerge_matched:
 	/*
 	 * Successfully executed an action or no qualifying action was found.
 	 */
-	return true;
+	*matched = true;
+
+	return rslot;
 }
 
 /*
  * Execute the first qualifying NOT MATCHED action.
  */
-static void
+static TupleTableSlot *
 ExecMergeNotMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 					bool canSetTag)
 {
 	ModifyTableState *mtstate = context->mtstate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
 	List	   *actionStates = NIL;
+	TupleTableSlot *rslot = NULL;
 	ListCell   *l;
 
 	/*
@@ -3165,10 +3209,10 @@ ExecMergeNotMatched(ModifyTableContext *
 				 * so we don't need to map the tuple here.
 				 */
 				newslot = ExecProject(action->mas_proj);
-				context->relaction = action;
+				mtstate->mt_merge_action = action;
 
-				(void) ExecInsert(context, mtstate->rootResultRelInfo, newslot,
-								  canSetTag, NULL, NULL);
+				rslot = ExecInsert(context, mtstate->rootResultRelInfo,
+								   newslot, canSetTag, NULL, NULL);
 				mtstate->mt_merge_inserted += 1;
 				break;
 			case CMD_NOTHING:
@@ -3184,6 +3228,8 @@ ExecMergeNotMatched(ModifyTableContext *
 		 */
 		break;
 	}
+
+	return rslot;
 }
 
 /*
@@ -3361,6 +3407,64 @@ ExecInitMergeTupleSlots(ModifyTableState
 }
 
 /*
+ * pg_merge_action() -
+ *	  SQL merge support function to retrieve the currently executing merge
+ *	  action command string ("INSERT", "UPDATE", or "DELETE").
+ */
+Datum
+pg_merge_action(PG_FUNCTION_ARGS)
+{
+	ModifyTableState *mtstate = (ModifyTableState *) fcinfo->context;
+	MergeActionState *relaction;
+
+	if (!mtstate || mtstate->operation != CMD_MERGE)
+		elog(ERROR, "merge support function called in non-merge context");
+
+	relaction = mtstate->mt_merge_action;
+	if (relaction)
+	{
+		CmdType		commandType = relaction->mas_action->commandType;
+
+		switch (commandType)
+		{
+			case CMD_INSERT:
+				PG_RETURN_TEXT_P(cstring_to_text("INSERT"));
+			case CMD_UPDATE:
+				PG_RETURN_TEXT_P(cstring_to_text("UPDATE"));
+			case CMD_DELETE:
+				PG_RETURN_TEXT_P(cstring_to_text("DELETE"));
+			case CMD_NOTHING:
+				PG_RETURN_NULL();
+			default:
+				elog(ERROR, "unrecognized commandType: %d", (int) commandType);
+		}
+	}
+
+	PG_RETURN_NULL();
+}
+
+/*
+ * pg_merge_when_clause_number() -
+ *	  SQL merge support function to retrieve the 1-based index of the
+ *	  currently executing merge WHEN clause.
+ */
+Datum
+pg_merge_when_clause_number(PG_FUNCTION_ARGS)
+{
+	ModifyTableState *mtstate = (ModifyTableState *) fcinfo->context;
+	MergeActionState *relaction;
+
+	if (!mtstate || mtstate->operation != CMD_MERGE)
+		elog(ERROR, "merge support function called in non-merge context");
+
+	relaction = mtstate->mt_merge_action;
+	if (relaction)
+		PG_RETURN_INT32((int32) relaction->mas_action->index);
+
+	PG_RETURN_NULL();
+}
+
+/*
  * Process BEFORE EACH STATEMENT triggers
  */
 static void
@@ -3646,8 +3750,17 @@ ExecModifyTable(PlanState *pstate)
 				{
 					EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-					ExecMerge(&context, node->resultRelInfo, NULL, node->canSetTag);
-					continue;	/* no RETURNING support yet */
+					slot = ExecMerge(&context, node->resultRelInfo, NULL,
+									 node->canSetTag);
+
+					/*
+					 * If we got a RETURNING result, return it to the caller.
+					 * We'll continue the work on next call.
+					 */
+					if (slot)
+						return slot;
+
+					continue;	/* continue with the next tuple */
 				}
 
 				elog(ERROR, "tableoid is NULL");
@@ -3724,8 +3837,17 @@ ExecModifyTable(PlanState *pstate)
 					{
 						EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-						ExecMerge(&context, node->resultRelInfo, NULL, node->canSetTag);
-						continue;	/* no RETURNING support yet */
+						slot = ExecMerge(&context, node->resultRelInfo, NULL,
+										 node->canSetTag);
+
+						/*
+						 * If we got a RETURNING result, return it to the
+						 * caller.  We'll continue the work on next call.
+						 */
+						if (slot)
+							return slot;
+
+						continue;	/* continue with the next tuple */
 					}
 
 					elog(ERROR, "ctid is NULL");
@@ -3817,7 +3939,7 @@ ExecModifyTable(PlanState *pstate)
 				}
 				slot = ExecGetUpdateNewTuple(resultRelInfo, context.planSlot,
 											 oldSlot);
-				context.relaction = NULL;
+				node->mt_merge_action = NULL;
 
 				/* Now apply the update. */
 				slot = ExecUpdate(&context, resultRelInfo, tupleid, oldtuple,
diff --git a/src/backend/executor/spi.c b/src/backend/executor/spi.c
new file mode 100644
index 3397568..b9f443c
--- a/src/backend/executor/spi.c
+++ b/src/backend/executor/spi.c
@@ -2033,6 +2033,8 @@ SPI_result_code_string(int code)
 			return "SPI_OK_TD_REGISTER";
 		case SPI_OK_MERGE:
 			return "SPI_OK_MERGE";
+		case SPI_OK_MERGE_RETURNING:
+			return "SPI_OK_MERGE_RETURNING";
 	}
 	/* Unrecognized code ... return something useful ... */
 	sprintf(buf, "Unrecognized SPI code %d", code);
@@ -2886,7 +2888,10 @@ _SPI_pquery(QueryDesc *queryDesc, bool f
 				res = SPI_OK_UPDATE;
 			break;
 		case CMD_MERGE:
-			res = SPI_OK_MERGE;
+			if (queryDesc->plannedstmt->hasReturning)
+				res = SPI_OK_MERGE_RETURNING;
+			else
+				res = SPI_OK_MERGE;
 			break;
 		default:
 			return SPI_ERROR_OPUNKNOWN;
diff --git a/src/backend/optimizer/plan/subselect.c b/src/backend/optimizer/plan/subselect.c
new file mode 100644
index 7a9fe88..7851761
--- a/src/backend/optimizer/plan/subselect.c
+++ b/src/backend/optimizer/plan/subselect.c
@@ -18,6 +18,7 @@
 
 #include "access/htup_details.h"
 #include "catalog/pg_operator.h"
+#include "catalog/pg_proc.h"
 #include "catalog/pg_type.h"
 #include "executor/executor.h"
 #include "miscadmin.h"
@@ -35,6 +36,7 @@
 #include "parser/parse_relation.h"
 #include "rewrite/rewriteManip.h"
 #include "utils/builtins.h"
+#include "utils/fmgroids.h"
 #include "utils/lsyscache.h"
 #include "utils/syscache.h"
 
@@ -1846,7 +1848,8 @@ convert_EXISTS_to_ANY(PlannerInfo *root,
 /*
  * Replace correlation vars (uplevel vars) with Params.
  *
- * Uplevel PlaceHolderVars and aggregates are replaced, too.
+ * Uplevel PlaceHolderVars, aggregates, GROUPING() expressions and merge
+ * support functions are replaced, too.
  *
  * Note: it is critical that this runs immediately after SS_process_sublinks.
  * Since we do not recurse into the arguments of uplevel PHVs and aggregates,
@@ -1900,6 +1903,22 @@ replace_correlation_vars_mutator(Node *n
 		if (((GroupingFunc *) node)->agglevelsup > 0)
 			return (Node *) replace_outer_grouping(root, (GroupingFunc *) node);
 	}
+	if (IsA(node, FuncExpr))
+	{
+		if (IsMergeSupportFunction(((FuncExpr *) node)->funcid))
+		{
+			Param	   *param;
+
+			/*
+			 * Replace with a Param, if it belongs to an upper-level MERGE
+			 * query. Otherwise, leave it be.
+			 */
+			param = replace_outer_merge_support_function(root,
+														 (FuncExpr *) node);
+			if (param)
+				return (Node *) param;
+		}
+	}
 	return expression_tree_mutator(node,
 								   replace_correlation_vars_mutator,
 								   (void *) root);
diff --git a/src/backend/optimizer/util/paramassign.c b/src/backend/optimizer/util/paramassign.c
new file mode 100644
index d6a923b..2b7b8d1
--- a/src/backend/optimizer/util/paramassign.c
+++ b/src/backend/optimizer/util/paramassign.c
@@ -307,6 +307,63 @@ replace_outer_grouping(PlannerInfo *root
 }
 
 /*
+ * Generate a Param node to replace the given FuncExpr expression, which is
+ * expected to be a merge support function, if it belongs to an upper-level
+ * MERGE query. Otherwise NULL will be returned.
+ *
+ * A NULL return value indicates either that the merge support function
+ * belongs to a MERGE query at this query level, or that there is no MERGE
+ * query at all. In both cases, the caller is expected to leave the FuncExpr
+ * node unaltered, so that it is executed at this query level (in the latter
+ * case, it is up to the merge support function to decide how to handle being
+ * called outside of a MERGE query).
+ */
+Param *
+replace_outer_merge_support_function(PlannerInfo *root, FuncExpr *func)
+{
+	Param	   *retval;
+	PlannerParamItem *pitem;
+	Index		levelsup;
+	Oid			ptype = exprType((Node *) func);
+
+	/* Find the upper-level MERGE query the function belongs to */
+	levelsup = 0;
+	while (root->parse->commandType != CMD_MERGE)
+	{
+		root = root->parent_root;
+		if (root == NULL)
+			return NULL;		/* Not in a MERGE query */
+		levelsup++;
+	}
+	if (levelsup == 0)
+		return NULL;			/* Local MERGE query at this level */
+
+	/*
+	 * It does not seem worthwhile to try to de-duplicate references to outer
+	 * merge support functions.  Just make a new slot every time.
+	 */
+	func = copyObject(func);
+
+	pitem = makeNode(PlannerParamItem);
+	pitem->item = (Node *) func;
+	pitem->paramId = list_length(root->glob->paramExecTypes);
+	root->glob->paramExecTypes = lappend_oid(root->glob->paramExecTypes,
+											 ptype);
+
+	root->plan_params = lappend(root->plan_params, pitem);
+
+	retval = makeNode(Param);
+	retval->paramkind = PARAM_EXEC;
+	retval->paramid = pitem->paramId;
+	retval->paramtype = ptype;
+	retval->paramtypmod = -1;
+	retval->paramcollid = InvalidOid;
+	retval->location = func->location;
+
+	return retval;
+}
+
+/*
  * Generate a Param node to replace the given Var,
  * which is expected to come from some upper NestLoop plan node.
  * Record the need for the Var in root->curOuterParams.
diff --git a/src/backend/parser/analyze.c b/src/backend/parser/analyze.c
new file mode 100644
index 4006632..6794511
--- a/src/backend/parser/analyze.c
+++ b/src/backend/parser/analyze.c
@@ -74,7 +74,6 @@ static void determineRecursiveColTypes(P
 									   Node *larg, List *nrtargetlist);
 static Query *transformReturnStmt(ParseState *pstate, ReturnStmt *stmt);
 static Query *transformUpdateStmt(ParseState *pstate, UpdateStmt *stmt);
-static List *transformReturningList(ParseState *pstate, List *returningList);
 static Query *transformPLAssignStmt(ParseState *pstate,
 									PLAssignStmt *stmt);
 static Query *transformDeclareCursorStmt(ParseState *pstate,
@@ -515,7 +514,8 @@ transformDeleteStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -942,7 +942,8 @@ transformInsertStmt(ParseState *pstate,
 	/* Process RETURNING, if any. */
 	if (stmt->returningList)
 		qry->returningList = transformReturningList(pstate,
-													stmt->returningList);
+													stmt->returningList,
+													EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -2407,7 +2408,8 @@ transformUpdateStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/*
 	 * Now we are done with SELECT-like processing, and can get on with
@@ -2501,10 +2503,11 @@ transformUpdateTargetList(ParseState *ps
 
 /*
  * transformReturningList -
- *	handle a RETURNING clause in INSERT/UPDATE/DELETE
+ *	handle a RETURNING clause in INSERT/UPDATE/DELETE/MERGE
  */
-static List *
-transformReturningList(ParseState *pstate, List *returningList)
+List *
+transformReturningList(ParseState *pstate, List *returningList,
+					   ParseExprKind exprKind)
 {
 	List	   *rlist;
 	int			save_next_resno;
@@ -2521,7 +2524,7 @@ transformReturningList(ParseState *pstat
 	pstate->p_next_resno = 1;
 
 	/* transform RETURNING identically to a SELECT targetlist */
-	rlist = transformTargetList(pstate, returningList, EXPR_KIND_RETURNING);
+	rlist = transformTargetList(pstate, returningList, exprKind);
 
 	/*
 	 * Complain if the nonempty tlist expanded to nothing (which is possible
diff --git a/src/backend/parser/gram.y b/src/backend/parser/gram.y
new file mode 100644
index 60080e8..3dd1063
--- a/src/backend/parser/gram.y
+++ b/src/backend/parser/gram.y
@@ -12255,6 +12255,7 @@ MergeStmt:
 			USING table_ref
 			ON a_expr
 			merge_when_list
+			returning_clause
 				{
 					MergeStmt  *m = makeNode(MergeStmt);
 
@@ -12263,6 +12264,7 @@ MergeStmt:
 					m->sourceRelation = $6;
 					m->joinCondition = $8;
 					m->mergeWhenClauses = $9;
+					m->returningList = $10;
 
 					$$ = (Node *) m;
 				}
diff --git a/src/backend/parser/parse_agg.c b/src/backend/parser/parse_agg.c
new file mode 100644
index 85cd47b..e72c37b
--- a/src/backend/parser/parse_agg.c
+++ b/src/backend/parser/parse_agg.c
@@ -456,6 +456,7 @@ check_agglevels_and_constraints(ParseSta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
@@ -903,6 +904,7 @@ transformWindowFuncCall(ParseState *psta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_cte.c b/src/backend/parser/parse_cte.c
new file mode 100644
index c5b1a49..ad3c525
--- a/src/backend/parser/parse_cte.c
+++ b/src/backend/parser/parse_cte.c
@@ -126,13 +126,6 @@ transformWithClause(ParseState *pstate,
 		CommonTableExpr *cte = (CommonTableExpr *) lfirst(lc);
 		ListCell   *rest;
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(cte->ctequery, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in WITH query"),
-					parser_errposition(pstate, cte->location));
-
 		for_each_cell(rest, withClause->ctes, lnext(withClause->ctes, lc))
 		{
 			CommonTableExpr *cte2 = (CommonTableExpr *) lfirst(rest);
@@ -153,7 +146,8 @@ transformWithClause(ParseState *pstate,
 			/* must be a data-modifying statement */
 			Assert(IsA(cte->ctequery, InsertStmt) ||
 				   IsA(cte->ctequery, UpdateStmt) ||
-				   IsA(cte->ctequery, DeleteStmt));
+				   IsA(cte->ctequery, DeleteStmt) ||
+				   IsA(cte->ctequery, MergeStmt));
 
 			pstate->p_hasModifyingCTE = true;
 		}
diff --git a/src/backend/parser/parse_expr.c b/src/backend/parser/parse_expr.c
new file mode 100644
index c08c063..3228624
--- a/src/backend/parser/parse_expr.c
+++ b/src/backend/parser/parse_expr.c
@@ -525,6 +525,7 @@ transformColumnRef(ParseState *pstate, C
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CHECK_CONSTRAINT:
@@ -1751,6 +1752,7 @@ transformSubLink(ParseState *pstate, Sub
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CYCLE_MARK:
@@ -3099,6 +3101,7 @@ ParseExprKindName(ParseExprKind exprKind
 		case EXPR_KIND_OFFSET:
 			return "OFFSET";
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			return "RETURNING";
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
diff --git a/src/backend/parser/parse_func.c b/src/backend/parser/parse_func.c
new file mode 100644
index b3f0b6a..cf7c0b9
--- a/src/backend/parser/parse_func.c
+++ b/src/backend/parser/parse_func.c
@@ -31,6 +31,7 @@
 #include "parser/parse_target.h"
 #include "parser/parse_type.h"
 #include "utils/builtins.h"
+#include "utils/fmgroids.h"
 #include "utils/lsyscache.h"
 #include "utils/syscache.h"
 
@@ -348,6 +349,26 @@ ParseFuncOrColumn(ParseState *pstate, Li
 					 parser_errposition(pstate, location)));
 	}
 
+	/* Merge support functions are only allowed in MERGE's RETURNING list */
+	if (IsMergeSupportFunction(funcid) &&
+		pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+	{
+		ParseState *parent_pstate = pstate->parentParseState;
+
+		/* May be in a subquery in RETURNING list */
+		while (parent_pstate &&
+			   parent_pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+			parent_pstate = parent_pstate->parentParseState;
+
+		if (!parent_pstate ||
+			parent_pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+			ereport(ERROR,
+					(errcode(ERRCODE_WRONG_OBJECT_TYPE),
+					 errmsg("merge support function %s can only be called from the RETURNING list of a MERGE command",
+							NameListToString(funcname)),
+					 parser_errposition(pstate, location)));
+	}
+
 	/*
 	 * So far so good, so do some fdresult-type-specific processing.
 	 */
@@ -2599,6 +2620,7 @@ check_srf_call_placement(ParseState *pst
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_merge.c b/src/backend/parser/parse_merge.c
new file mode 100644
index 91b1156..33852e9
--- a/src/backend/parser/parse_merge.c
+++ b/src/backend/parser/parse_merge.c
@@ -233,6 +233,10 @@ transformMergeStmt(ParseState *pstate, M
 	 */
 	qry->jointree = makeFromExpr(pstate->p_joinlist, joinExpr);
 
+	/* Transform the RETURNING list, if any */
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_MERGE_RETURNING);
+
 	/*
 	 * We now have a good query shape, so now look at the WHEN conditions and
 	 * action targetlists.
@@ -254,6 +258,7 @@ transformMergeStmt(ParseState *pstate, M
 		MergeAction *action;
 
 		action = makeNode(MergeAction);
+		action->index = foreach_current_index(l) + 1;
 		action->commandType = mergeWhenClause->commandType;
 		action->matched = mergeWhenClause->matched;
 
@@ -390,9 +395,6 @@ transformMergeStmt(ParseState *pstate, M
 
 	qry->mergeActionList = mergeActionList;
 
-	/* RETURNING could potentially be added in the future, but not in SQL std */
-	qry->returningList = NULL;
-
 	qry->hasTargetSRFs = false;
 	qry->hasSubLinks = pstate->p_hasSubLinks;
 
diff --git a/src/backend/parser/parse_relation.c b/src/backend/parser/parse_relation.c
new file mode 100644
index 864ea9b..cc19041
--- a/src/backend/parser/parse_relation.c
+++ b/src/backend/parser/parse_relation.c
@@ -2346,9 +2346,10 @@ addRangeTableEntryForCTE(ParseState *pst
 		cte->cterefcount++;
 
 	/*
-	 * We throw error if the CTE is INSERT/UPDATE/DELETE without RETURNING.
-	 * This won't get checked in case of a self-reference, but that's OK
-	 * because data-modifying CTEs aren't allowed to be recursive anyhow.
+	 * We throw error if the CTE is INSERT/UPDATE/DELETE/MERGE without
+	 * RETURNING.  This won't get checked in case of a self-reference, but
+	 * that's OK because data-modifying CTEs aren't allowed to be recursive
+	 * anyhow.
 	 */
 	if (IsA(cte->ctequery, Query))
 	{
diff --git a/src/backend/rewrite/rewriteHandler.c b/src/backend/rewrite/rewriteHandler.c
new file mode 100644
index b486ab5..1bd0ce2
--- a/src/backend/rewrite/rewriteHandler.c
+++ b/src/backend/rewrite/rewriteHandler.c
@@ -3629,9 +3629,9 @@ RewriteQuery(Query *parsetree, List *rew
 	ListCell   *lc1;
 
 	/*
-	 * First, recursively process any insert/update/delete statements in WITH
-	 * clauses.  (We have to do this first because the WITH clauses may get
-	 * copied into rule actions below.)
+	 * First, recursively process any insert/update/delete/merge statements in
+	 * WITH clauses.  (We have to do this first because the WITH clauses may
+	 * get copied into rule actions below.)
 	 */
 	foreach(lc1, parsetree->cteList)
 	{
@@ -3656,7 +3656,8 @@ RewriteQuery(Query *parsetree, List *rew
 			if (!(ctequery->commandType == CMD_SELECT ||
 				  ctequery->commandType == CMD_UPDATE ||
 				  ctequery->commandType == CMD_INSERT ||
-				  ctequery->commandType == CMD_DELETE))
+				  ctequery->commandType == CMD_DELETE ||
+				  ctequery->commandType == CMD_MERGE))
 			{
 				/*
 				 * Currently it could only be NOTIFY; this error message will
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
new file mode 100644
index e3ccf6c..6f3dc89
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -2142,11 +2142,10 @@ QueryReturnsTuples(Query *parsetree)
 		case CMD_SELECT:
 			/* returns tuples */
 			return true;
-		case CMD_MERGE:
-			return false;
 		case CMD_INSERT:
 		case CMD_UPDATE:
 		case CMD_DELETE:
+		case CMD_MERGE:
 			/* the forms with RETURNING return tuples */
 			if (parsetree->returningList)
 				return true;
diff --git a/src/backend/utils/adt/ruleutils.c b/src/backend/utils/adt/ruleutils.c
new file mode 100644
index fcb2f45..b47cfc7
--- a/src/backend/utils/adt/ruleutils.c
+++ b/src/backend/utils/adt/ruleutils.c
@@ -7183,8 +7183,13 @@ get_merge_query_def(Query *query, depars
 			appendStringInfoString(buf, "DO NOTHING");
 	}
 
-	/* No RETURNING support in MERGE yet */
-	Assert(query->returningList == NIL);
+	/* Add RETURNING if present */
+	if (query->returningList)
+	{
+		appendContextKeyword(context, " RETURNING",
+							 -PRETTYINDENT_STD, PRETTYINDENT_STD, 1);
+		get_target_list(query->returningList, context, NULL, colNamesVisible);
+	}
 }
 
 
diff --git a/src/bin/psql/common.c b/src/bin/psql/common.c
new file mode 100644
index 5973df2..a1380f8
--- a/src/bin/psql/common.c
+++ b/src/bin/psql/common.c
@@ -982,13 +982,17 @@ PrintQueryResult(PGresult *result, bool
 			else
 				success = true;
 
-			/* if it's INSERT/UPDATE/DELETE RETURNING, also print status */
+			/*
+			 * If it's INSERT/UPDATE/DELETE/MERGE RETURNING, also print
+			 * status.
+			 */
 			if (last || pset.show_all_results)
 			{
 				cmdstatus = PQcmdStatus(result);
 				if (strncmp(cmdstatus, "INSERT", 6) == 0 ||
 					strncmp(cmdstatus, "UPDATE", 6) == 0 ||
-					strncmp(cmdstatus, "DELETE", 6) == 0)
+					strncmp(cmdstatus, "DELETE", 6) == 0 ||
+					strncmp(cmdstatus, "MERGE", 5) == 0)
 					PrintQueryStatus(result, printStatusFout);
 			}
 
diff --git a/src/include/catalog/pg_proc.dat b/src/include/catalog/pg_proc.dat
new file mode 100644
index 6996073..2181eb8
--- a/src/include/catalog/pg_proc.dat
+++ b/src/include/catalog/pg_proc.dat
@@ -12043,4 +12043,14 @@
   proname => 'any_value_transfn', prorettype => 'anyelement',
   proargtypes => 'anyelement anyelement', prosrc => 'any_value_transfn' },
 
+# MERGE support functions
+{ oid => '9499', descr => 'command type of current MERGE action',
+  proname => 'pg_merge_action',  provolatile => 'v', proparallel => 'r',
+  prorettype => 'text', proargtypes => '',
+  prosrc => 'pg_merge_action' },
+{ oid => '9500', descr => 'index of current MERGE WHEN clause',
+  proname => 'pg_merge_when_clause_number',  provolatile => 'v',
+  proparallel => 'r', prorettype => 'int4', proargtypes => '',
+  prosrc => 'pg_merge_when_clause_number' },
+
 ]
diff --git a/src/include/catalog/pg_proc.h b/src/include/catalog/pg_proc.h
new file mode 100644
index e7abe0b..eb6cad9
--- a/src/include/catalog/pg_proc.h
+++ b/src/include/catalog/pg_proc.h
@@ -182,6 +182,11 @@ DECLARE_UNIQUE_INDEX(pg_proc_proname_arg
 #define PROARGMODE_VARIADIC 'v'
 #define PROARGMODE_TABLE	't'
 
+/* Is this a merge support function?  (Requires fmgroids.h) */
+#define IsMergeSupportFunction(funcid) \
+	((funcid) == F_PG_MERGE_ACTION || \
+	 (funcid) == F_PG_MERGE_WHEN_CLAUSE_NUMBER)
+
 #endif							/* EXPOSE_TO_CLIENT_CODE */
 
 
diff --git a/src/include/executor/spi.h b/src/include/executor/spi.h
new file mode 100644
index d1de139..ed8787d
--- a/src/include/executor/spi.h
+++ b/src/include/executor/spi.h
@@ -97,6 +97,7 @@ typedef struct _SPI_plan *SPIPlanPtr;
 #define SPI_OK_REL_UNREGISTER	16
 #define SPI_OK_TD_REGISTER		17
 #define SPI_OK_MERGE			18
+#define SPI_OK_MERGE_RETURNING	19
 
 #define SPI_OPT_NONATOMIC		(1 << 0)
 
diff --git a/src/include/nodes/execnodes.h b/src/include/nodes/execnodes.h
new file mode 100644
index cb714f4..6708b6b
--- a/src/include/nodes/execnodes.h
+++ b/src/include/nodes/execnodes.h
@@ -1319,6 +1319,9 @@ typedef struct ModifyTableState
 	/* Flags showing which subcommands are present INS/UPD/DEL/DO NOTHING */
 	int			mt_merge_subcommands;
 
+	/* For MERGE, the action currently being executed */
+	MergeActionState *mt_merge_action;
+
 	/* tuple counters for MERGE */
 	double		mt_merge_inserted;
 	double		mt_merge_updated;
diff --git a/src/include/nodes/parsenodes.h b/src/include/nodes/parsenodes.h
new file mode 100644
index efb5c3e..1d5d631
--- a/src/include/nodes/parsenodes.h
+++ b/src/include/nodes/parsenodes.h
@@ -1688,6 +1688,7 @@ typedef struct MergeWhenClause
 typedef struct MergeAction
 {
 	NodeTag		type;
+	int			index;			/* 1-based index of the clause */
 	bool		matched;		/* true=MATCHED, false=NOT MATCHED */
 	CmdType		commandType;	/* INSERT/UPDATE/DELETE/DO NOTHING */
 	/* OVERRIDING clause */
@@ -1914,6 +1915,7 @@ typedef struct MergeStmt
 	Node	   *sourceRelation; /* source relation */
 	Node	   *joinCondition;	/* join condition between source and target */
 	List	   *mergeWhenClauses;	/* list of MergeWhenClause(es) */
+	List	   *returningList;	/* list of expressions to return */
 	WithClause *withClause;		/* WITH clause */
 } MergeStmt;
 
diff --git a/src/include/optimizer/paramassign.h b/src/include/optimizer/paramassign.h
new file mode 100644
index 55c27a6..6f70c78
--- a/src/include/optimizer/paramassign.h
+++ b/src/include/optimizer/paramassign.h
@@ -20,6 +20,8 @@ extern Param *replace_outer_placeholderv
 										   PlaceHolderVar *phv);
 extern Param *replace_outer_agg(PlannerInfo *root, Aggref *agg);
 extern Param *replace_outer_grouping(PlannerInfo *root, GroupingFunc *grp);
+extern Param *replace_outer_merge_support_function(PlannerInfo *root,
+												   FuncExpr *func);
 extern Param *replace_nestloop_param_var(PlannerInfo *root, Var *var);
 extern Param *replace_nestloop_param_placeholdervar(PlannerInfo *root,
 													PlaceHolderVar *phv);
diff --git a/src/include/parser/analyze.h b/src/include/parser/analyze.h
new file mode 100644
index 1cef183..78eb55e
--- a/src/include/parser/analyze.h
+++ b/src/include/parser/analyze.h
@@ -44,6 +44,8 @@ extern List *transformInsertRow(ParseSta
 								bool strip_indirection);
 extern List *transformUpdateTargetList(ParseState *pstate,
 									   List *origTlist);
+extern List *transformReturningList(ParseState *pstate, List *returningList,
+									ParseExprKind exprKind);
 extern Query *transformTopLevelStmt(ParseState *pstate, RawStmt *parseTree);
 extern Query *transformStmt(ParseState *pstate, Node *parseTree);
 
diff --git a/src/include/parser/parse_node.h b/src/include/parser/parse_node.h
new file mode 100644
index f589112..836819d
--- a/src/include/parser/parse_node.h
+++ b/src/include/parser/parse_node.h
@@ -61,7 +61,8 @@ typedef enum ParseExprKind
 	EXPR_KIND_DISTINCT_ON,		/* DISTINCT ON */
 	EXPR_KIND_LIMIT,			/* LIMIT */
 	EXPR_KIND_OFFSET,			/* OFFSET */
-	EXPR_KIND_RETURNING,		/* RETURNING */
+	EXPR_KIND_RETURNING,		/* RETURNING in INSERT/UPDATE/DELETE */
+	EXPR_KIND_MERGE_RETURNING,	/* RETURNING in MERGE */
 	EXPR_KIND_VALUES,			/* VALUES */
 	EXPR_KIND_VALUES_SINGLE,	/* single-row VALUES (in INSERT only) */
 	EXPR_KIND_CHECK_CONSTRAINT, /* CHECK constraint for a table */
diff --git a/src/pl/plpgsql/src/pl_exec.c b/src/pl/plpgsql/src/pl_exec.c
new file mode 100644
index 4b76f76..fb6289c
--- a/src/pl/plpgsql/src/pl_exec.c
+++ b/src/pl/plpgsql/src/pl_exec.c
@@ -4267,9 +4267,9 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 	/*
 	 * If we have INTO, then we only need one row back ... but if we have INTO
 	 * STRICT or extra check too_many_rows, ask for two rows, so that we can
-	 * verify the statement returns only one.  INSERT/UPDATE/DELETE are always
-	 * treated strictly. Without INTO, just run the statement to completion
-	 * (tcount = 0).
+	 * verify the statement returns only one.  INSERT/UPDATE/DELETE/MERGE are
+	 * always treated strictly. Without INTO, just run the statement to
+	 * completion (tcount = 0).
 	 *
 	 * We could just ask for two rows always when using INTO, but there are
 	 * some cases where demanding the extra row costs significant time, eg by
@@ -4307,10 +4307,11 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 			Assert(stmt->mod_stmt);
 			exec_set_found(estate, (SPI_processed != 0));
 			break;
@@ -4489,10 +4490,11 @@ exec_stmt_dynexecute(PLpgSQL_execstate *
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 		case SPI_OK_UTILITY:
 		case SPI_OK_REWRITTEN:
 			break;
diff --git a/src/pl/tcl/pltcl.c b/src/pl/tcl/pltcl.c
new file mode 100644
index e8f9d7b..0c9f81e
--- a/src/pl/tcl/pltcl.c
+++ b/src/pl/tcl/pltcl.c
@@ -2456,6 +2456,7 @@ pltcl_process_SPI_result(Tcl_Interp *int
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
+		case SPI_OK_MERGE_RETURNING:
 
 			/*
 			 * Process the tuples we got
diff --git a/src/test/regress/expected/merge.out b/src/test/regress/expected/merge.out
new file mode 100644
index 133d421..5be3d2d
--- a/src/test/regress/expected/merge.out
+++ b/src/test/regress/expected/merge.out
@@ -123,20 +123,20 @@ ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
 ERROR:  name "target" specified more than once
 DETAIL:  The name is used both as MERGE target table and data source.
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
-ERROR:  MERGE not supported in WITH query
-LINE 1: WITH foo AS (
-             ^
--- used in COPY
+ERROR:  WITH query "foo" does not have a RETURNING clause
+LINE 4: ) SELECT * FROM foo;
+                        ^
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) TO stdout;
-ERROR:  MERGE not supported in COPY
+ERROR:  COPY query must have a RETURNING clause
 -- unsupported relation types
 -- view
 CREATE VIEW tv AS SELECT * FROM target;
@@ -1256,7 +1256,7 @@ BEGIN;
 MERGE INTO sq_target
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1271,7 +1271,7 @@ INSERT INTO sq_source (sid, balance, del
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1295,7 +1295,7 @@ WITH targq AS (
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1304,20 +1304,193 @@ WHEN MATCHED AND tid < 2 THEN
 ROLLBACK;
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
+CREATE TABLE merge_actions(action text, abbrev text);
+INSERT INTO merge_actions VALUES ('INSERT', 'ins'), ('UPDATE', 'upd'), ('DELETE', 'del');
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
-ERROR:  syntax error at or near "RETURNING"
-LINE 10: RETURNING *;
-         ^
+    DELETE
+RETURNING pg_merge_when_clause_number() AS clause_number,
+          (SELECT abbrev FROM merge_actions
+            WHERE action = pg_merge_action()) AS merge_action,
+          t.*,
+          CASE pg_merge_action()
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ clause_number | merge_action | tid | balance |     description     
+---------------+--------------+-----+---------+---------------------
+             3 | del          |   1 |     100 | Removed (1,100)
+             1 | upd          |   2 |     220 | Added 20 to balance
+             2 | ins          |   4 |      40 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- error when using MERGE support functions outside MERGE
+SELECT pg_merge_action() FROM sq_target;
+ERROR:  merge support function pg_merge_action can only be called from the RETURNING list of a MERGE command
+LINE 1: SELECT pg_merge_action() FROM sq_target;
+               ^
+UPDATE sq_target SET balance = balance + 1 RETURNING pg_merge_when_clause_number();
+ERROR:  merge support function pg_merge_when_clause_number can only be called from the RETURNING list of a MERGE command
+LINE 1: ...ATE sq_target SET balance = balance + 1 RETURNING pg_merge_w...
+                                                             ^
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_when_clause_number() AS clause_number,
+              pg_merge_action() AS merge_action,
+              t.*,
+              CASE pg_merge_action()
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING merge_action, pg_merge_action() AS merge_log_action, l.*
+)
+SELECT * FROM m2;
+ merge_action | merge_log_action | tid |     last_change     
+--------------+------------------+-----+---------------------
+ DELETE       | UPDATE           |   1 | Removed (1,100)
+ UPDATE       | INSERT           |   2 | Added 20 to balance
+ INSERT       | INSERT           |   4 | Inserted (4,40)
+(3 rows)
+
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ tid |     last_change     
+-----+---------------------
+   1 | Removed (1,100)
+   2 | Added 20 to balance
+   4 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*
+) TO stdout;
+DELETE	1	100
+UPDATE	2	220
+INSERT	4	40
+ROLLBACK;
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ action | tid | new_balance 
+--------+-----+-------------
+ DELETE |   1 |         100
+ UPDATE |   3 |         320
+ INSERT |   4 |         110
+(3 rows)
+
+ROLLBACK;
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ action | tid | balance 
+--------+-----+---------
+ DELETE |   1 |     100
+ UPDATE |   2 |     220
+ INSERT |   4 |      40
+(3 rows)
+
+ROLLBACK;
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ r_action | r_tid | r_balance 
+----------+-------+-----------
+ DELETE   |     1 |       100
+ UPDATE   |     3 |       320
+ INSERT   |     4 |       110
+(3 rows)
+
 ROLLBACK;
 -- EXPLAIN
 CREATE TABLE ex_mtarget (a int, b int)
@@ -1514,7 +1687,7 @@ SELECT * FROM sq_target WHERE tid = 1;
 (1 row)
 
 ROLLBACK;
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 NOTICE:  drop cascades to view v
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1640,6 +1813,32 @@ SELECT * FROM pa_target ORDER BY tid;
 (14 rows)
 
 ROLLBACK;
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING pg_merge_action(), pg_merge_when_clause_number(), t.*;
+ pg_merge_action | pg_merge_when_clause_number | tid | balance |           val            
+-----------------+-----------------------------+-----+---------+--------------------------
+ UPDATE          |                           1 |   2 |     110 | initial updated by merge
+(1 row)
+
+SELECT * FROM pa_target ORDER BY tid;
+ tid | balance |           val            
+-----+---------+--------------------------
+   2 |     110 | initial updated by merge
+   3 |     300 | initial
+   5 |     500 | initial
+   7 |     700 | initial
+   9 |     900 | initial
+  11 |    1100 | initial
+  13 |    1300 | initial
+(7 rows)
+
+ROLLBACK;
 DROP TABLE pa_target CASCADE;
 -- The target table is partitioned in the same way, but this time by attaching
 -- partitions which have columns in different order, dropped columns etc.
@@ -1731,7 +1930,26 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (sid, delta, 'inserted by merge');
+    INSERT VALUES (sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause_number(), t.*;
+ pg_merge_action | pg_merge_when_clause_number | tid | balance |           val            
+-----------------+-----------------------------+-----+---------+--------------------------
+ UPDATE          |                           1 |   2 |     110 | initial updated by merge
+ INSERT          |                           2 |   2 |      20 | inserted by merge
+ UPDATE          |                           1 |   4 |     330 | initial updated by merge
+ INSERT          |                           2 |   4 |      40 | inserted by merge
+ UPDATE          |                           1 |   6 |     550 | initial updated by merge
+ INSERT          |                           2 |   6 |      60 | inserted by merge
+ UPDATE          |                           1 |   8 |     770 | initial updated by merge
+ INSERT          |                           2 |   8 |      80 | inserted by merge
+ UPDATE          |                           1 |  10 |     990 | initial updated by merge
+ INSERT          |                           2 |  10 |     100 | inserted by merge
+ UPDATE          |                           1 |  12 |    1210 | initial updated by merge
+ INSERT          |                           2 |  12 |     120 | inserted by merge
+ UPDATE          |                           1 |  14 |    1430 | initial updated by merge
+ INSERT          |                           2 |  14 |     140 | inserted by merge
+(14 rows)
+
 SELECT * FROM pa_target ORDER BY tid;
  tid | balance |           val            
 -----+---------+--------------------------
@@ -1798,7 +2016,21 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause_number(), t.*;
+ pg_merge_action | pg_merge_when_clause_number |          logts           | tid | balance |           val            
+-----------------+-----------------------------+--------------------------+-----+---------+--------------------------
+ UPDATE          |                           1 | Tue Jan 31 00:00:00 2017 |   1 |     110 | initial updated by merge
+ UPDATE          |                           1 | Tue Feb 28 00:00:00 2017 |   2 |     220 | initial updated by merge
+ INSERT          |                           2 | Sun Jan 15 00:00:00 2017 |   3 |      30 | inserted by merge
+ UPDATE          |                           1 | Tue Jan 31 00:00:00 2017 |   4 |     440 | initial updated by merge
+ UPDATE          |                           1 | Tue Feb 28 00:00:00 2017 |   5 |     550 | initial updated by merge
+ INSERT          |                           2 | Sun Jan 15 00:00:00 2017 |   6 |      60 | inserted by merge
+ UPDATE          |                           1 | Tue Jan 31 00:00:00 2017 |   7 |     770 | initial updated by merge
+ UPDATE          |                           1 | Tue Feb 28 00:00:00 2017 |   8 |     880 | initial updated by merge
+ INSERT          |                           2 | Sun Jan 15 00:00:00 2017 |   9 |      90 | inserted by merge
+(9 rows)
+
 SELECT * FROM pa_target ORDER BY tid;
           logts           | tid | balance |           val            
 --------------------------+-----+---------+--------------------------
diff --git a/src/test/regress/expected/rules.out b/src/test/regress/expected/rules.out
new file mode 100644
index e07afcd..219f0ad
--- a/src/test/regress/expected/rules.out
+++ b/src/test/regress/expected/rules.out
@@ -3592,7 +3592,8 @@ MERGE INTO rule_merge2 t USING (SELECT 1
 -- test deparsing
 CREATE TABLE sf_target(id int, data text, filling int[]);
 CREATE FUNCTION merge_sf_test()
- RETURNS void
+ RETURNS TABLE(pg_merge_action text, when_clause int,
+               a int, b text, id int, data text, filling int[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -3629,11 +3630,13 @@ WHEN NOT MATCHED
    VALUES (s.a, s.b, DEFAULT)
 WHEN NOT MATCHED
    THEN INSERT (filling[1], id)
-   VALUES (s.a, s.a);
+   VALUES (s.a, s.a)
+RETURNING
+   pg_merge_action(), pg_merge_when_clause_number() AS clause_number, *;
 END;
 \sf merge_sf_test
 CREATE OR REPLACE FUNCTION public.merge_sf_test()
- RETURNS void
+ RETURNS TABLE(pg_merge_action text, when_clause integer, a integer, b text, id integer, data text, filling integer[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -3670,7 +3673,14 @@ BEGIN ATOMIC
       VALUES (s.a, s.b, DEFAULT)
     WHEN NOT MATCHED
      THEN INSERT (filling[1], id)
-      VALUES (s.a, s.a);
+      VALUES (s.a, s.a)
+   RETURNING pg_merge_action() AS pg_merge_action,
+     pg_merge_when_clause_number() AS clause_number,
+     s.a,
+     s.b,
+     t.id,
+     t.data,
+     t.filling;
 END
 DROP FUNCTION merge_sf_test;
 DROP TABLE sf_target;
diff --git a/src/test/regress/sql/merge.sql b/src/test/regress/sql/merge.sql
new file mode 100644
index 4cf6db9..8394d01
--- a/src/test/regress/sql/merge.sql
+++ b/src/test/regress/sql/merge.sql
@@ -86,12 +86,12 @@ MERGE INTO target
 USING target
 ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
--- used in COPY
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
@@ -816,7 +816,7 @@ BEGIN;
 MERGE INTO sq_target
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -829,7 +829,7 @@ INSERT INTO sq_source (sid, balance, del
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -847,7 +847,7 @@ WITH targq AS (
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -857,17 +857,148 @@ ROLLBACK;
 
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
+CREATE TABLE merge_actions(action text, abbrev text);
+INSERT INTO merge_actions VALUES ('INSERT', 'ins'), ('UPDATE', 'upd'), ('DELETE', 'del');
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
+    DELETE
+RETURNING pg_merge_when_clause_number() AS clause_number,
+          (SELECT abbrev FROM merge_actions
+            WHERE action = pg_merge_action()) AS merge_action,
+          t.*,
+          CASE pg_merge_action()
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ROLLBACK;
+
+-- error when using MERGE support functions outside MERGE
+SELECT pg_merge_action() FROM sq_target;
+UPDATE sq_target SET balance = balance + 1 RETURNING pg_merge_when_clause_number();
+
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_when_clause_number() AS clause_number,
+              pg_merge_action() AS merge_action,
+              t.*,
+              CASE pg_merge_action()
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING merge_action, pg_merge_action() AS merge_log_action, l.*
+)
+SELECT * FROM m2;
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ROLLBACK;
+
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*
+) TO stdout;
+ROLLBACK;
+
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ROLLBACK;
+
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ROLLBACK;
+
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
 ROLLBACK;
 
 -- EXPLAIN
@@ -966,7 +1097,7 @@ WHEN MATCHED THEN
 SELECT * FROM sq_target WHERE tid = 1;
 ROLLBACK;
 
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1033,6 +1164,17 @@ SELECT merge_func();
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING pg_merge_action(), pg_merge_when_clause_number(), t.*;
+SELECT * FROM pa_target ORDER BY tid;
+ROLLBACK;
+
 DROP TABLE pa_target CASCADE;
 
 -- The target table is partitioned in the same way, but this time by attaching
@@ -1091,7 +1233,8 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (sid, delta, 'inserted by merge');
+    INSERT VALUES (sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause_number(), t.*;
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
@@ -1145,7 +1288,8 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause_number(), t.*;
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
diff --git a/src/test/regress/sql/rules.sql b/src/test/regress/sql/rules.sql
new file mode 100644
index 8b7e255..5dbeb6b
--- a/src/test/regress/sql/rules.sql
+++ b/src/test/regress/sql/rules.sql
@@ -1281,7 +1281,8 @@ MERGE INTO rule_merge2 t USING (SELECT 1
 CREATE TABLE sf_target(id int, data text, filling int[]);
 
 CREATE FUNCTION merge_sf_test()
- RETURNS void
+ RETURNS TABLE(pg_merge_action text, when_clause int,
+               a int, b text, id int, data text, filling int[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -1318,7 +1319,9 @@ WHEN NOT MATCHED
    VALUES (s.a, s.b, DEFAULT)
 WHEN NOT MATCHED
    THEN INSERT (filling[1], id)
-   VALUES (s.a, s.a);
+   VALUES (s.a, s.a)
+RETURNING
+   pg_merge_action(), pg_merge_when_clause_number() AS clause_number, *;
 END;
 
 \sf merge_sf_test

diff --git a/doc/src/sgml/dml.sgml b/doc/src/sgml/dml.sgml
new file mode 100644
index cbbc5e2..7f65f6e
--- a/doc/src/sgml/dml.sgml
+++ b/doc/src/sgml/dml.sgml
@@ -283,10 +283,15 @@ DELETE FROM products;
    <secondary>RETURNING</secondary>
   </indexterm>
 
+  <indexterm zone="dml-returning">
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
   <para>
    Sometimes it is useful to obtain data from modified rows while they are
    being manipulated.  The <command>INSERT</command>, <command>UPDATE</command>,
-   and <command>DELETE</command> commands all have an
+   <command>DELETE</command>, and <command>MERGE</command> commands all have an
    optional <literal>RETURNING</literal> clause that supports this.  Use
    of <literal>RETURNING</literal> avoids performing an extra database query to
    collect the data, and is especially valuable when it would otherwise be
@@ -339,6 +344,24 @@ DELETE FROM products
 </programlisting>
   </para>
 
+  <para>
+   In a <command>MERGE</command>, the data available to <literal>RETURNING</literal> is
+   the content of the source row plus the content of the inserted, updated, or
+   deleted target row.  In addition, the merge support functions
+   <xref linkend="pg-merge-action"/> and <xref linkend="pg-merge-when-clause-number"/>
+   may be used to return information about which merge action was executed for
+   each row.  Since it is quite common for the source and target to have many
+   of the same columns, specifying <literal>RETURNING *</literal> can lead to a
+   lot of duplicated columns, so it is often more useful to just return the
+   target row.  For example:
+<programlisting>
+MERGE INTO products p USING new_products n ON p.product_no = n.product_no
+  WHEN NOT MATCHED THEN INSERT VALUES (n.product_no, n.name, n.price)
+  WHEN MATCHED THEN UPDATE SET name = n.name, price = n.price
+  RETURNING pg_merge_action(), p.*;
+</programlisting>
+  </para>
+
   <para>
    If there are triggers (<xref linkend="triggers"/>) on the target table,
    the data available to <literal>RETURNING</literal> is the row as modified by
diff --git a/doc/src/sgml/func.sgml b/doc/src/sgml/func.sgml
new file mode 100644
index be2f54c..e79e2ad
--- a/doc/src/sgml/func.sgml
+++ b/doc/src/sgml/func.sgml
@@ -21772,6 +21772,100 @@ SELECT count(*) FROM sometable;
 
  </sect1>
 
+ <sect1 id="functions-merge-support">
+  <title>Merge Support Functions</title>
+
+  <indexterm>
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
+  <para>
+   The merge support functions shown in
+   <xref linkend="functions-merge-support-table"/> may be used in the
+   <literal>RETURNING</literal> list of a <xref linkend="sql-merge"/> command
+   to return additional information about the action taken for each row.
+  </para>
+
+  <table id="functions-merge-support-table">
+    <title>Merge Support Functions</title>
+
+    <tgroup cols="1">
+     <thead>
+      <row>
+       <entry role="func_table_entry"><para role="func_signature">
+        Function
+       </para>
+       <para>
+        Description
+       </para></entry>
+      </row>
+     </thead>
+
+     <tbody>
+      <row>
+       <entry role="func_table_entry"><para role="func_signature">
+        <indexterm id="pg-merge-action">
+         <primary>pg_merge_action</primary>
+        </indexterm>
+        <function>pg_merge_action</function> ( )
+        <returnvalue>text</returnvalue>
+       </para>
+       <para>
+        Returns the action performed on the current row (<literal>'INSERT'</literal>,
+        <literal>'UPDATE'</literal>, or <literal>'DELETE'</literal>).
+       </para></entry>
+      </row>
+
+      <row>
+       <entry role="func_table_entry"><para role="func_signature">
+        <indexterm id="pg-merge-when-clause-number">
+         <primary>pg_merge_when_clause_number</primary>
+        </indexterm>
+        <function>pg_merge_when_clause_number</function> ( )
+        <returnvalue>integer</returnvalue>
+       </para>
+       <para>
+        Returns the 1-based index of the <literal>WHEN</literal> clause executed
+        for the current row.
+       </para></entry>
+      </row>
+     </tbody>
+    </tgroup>
+  </table>
+
+  <para>
+   Example:
+<screen><![CDATA[
+MERGE INTO products p
+  USING stock s ON p.product_id = s.product_id
+  WHEN MATCHED AND s.quantity > 0 THEN
+    UPDATE SET in_stock = true, quantity = s.quantity
+  WHEN MATCHED THEN
+    UPDATE SET in_stock = false, quantity = 0
+  WHEN NOT MATCHED THEN
+    INSERT (product_id, in_stock, quantity)
+      VALUES (s.product_id, true, s.quantity)
+  RETURNING pg_merge_action() AS action,
+            pg_merge_when_clause_number() AS clause_number,
+            p.*;
+
+ action | clause_number | product_id | in_stock | quantity 
+--------+---------------+------------+----------+----------
+ UPDATE |             1 |       1001 | t        |       50
+ UPDATE |             2 |       1002 | f        |        0
+ INSERT |             3 |       1003 | t        |       10
+]]></screen>
+  </para>
+
+  <para>
+   Note that these functions can only be used in the <literal>RETURNING</literal>
+   list of a <command>MERGE</command> command.  It is an error to use them in
+   any other part of a query.
+  </para>
+
+ </sect1>
+
  <sect1 id="functions-subquery">
   <title>Subquery Expressions</title>
 
diff --git a/doc/src/sgml/glossary.sgml b/doc/src/sgml/glossary.sgml
new file mode 100644
index fe8def4..51a15ca
--- a/doc/src/sgml/glossary.sgml
+++ b/doc/src/sgml/glossary.sgml
@@ -1387,9 +1387,9 @@
      to a <glossterm linkend="glossary-client">client</glossterm> upon the
      completion of an <acronym>SQL</acronym> command, usually a
      <command>SELECT</command> but it can be an
-     <command>INSERT</command>, <command>UPDATE</command>, or
-     <command>DELETE</command> command if the <literal>RETURNING</literal>
-     clause is specified.
+     <command>INSERT</command>, <command>UPDATE</command>,
+     <command>DELETE</command>, or <command>MERGE</command> command if the
+     <literal>RETURNING</literal> clause is specified.
     </para>
     <para>
      The fact that a result set is a relation means that a query can be used
diff --git a/doc/src/sgml/plpgsql.sgml b/doc/src/sgml/plpgsql.sgml
new file mode 100644
index f55e901..6803240
--- a/doc/src/sgml/plpgsql.sgml
+++ b/doc/src/sgml/plpgsql.sgml
@@ -1020,8 +1020,8 @@ INSERT INTO mytable VALUES (1,'one'), (2
     </para>
 
     <para>
-     If the command does return rows (for example <command>SELECT</command>,
-     or <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>
+     If the command does return rows (for example <command>SELECT</command>, or
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
      with <literal>RETURNING</literal>), there are two ways to proceed.
      When the command will return at most one row, or you only care about
      the first row of output, write the command as usual but add
@@ -1149,6 +1149,7 @@ SELECT <replaceable>select_expressions</
 INSERT ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 UPDATE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 DELETE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
+MERGE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 </synopsis>
 
      where <replaceable>target</replaceable> can be a record variable, a row
@@ -1159,8 +1160,8 @@ DELETE ... RETURNING <replaceable>expres
      <literal>INTO</literal> clause) just as described above,
      and the plan is cached in the same way.
      This works for <command>SELECT</command>,
-     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
-     <literal>RETURNING</literal>, and certain utility commands
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
+     with <literal>RETURNING</literal>, and certain utility commands
      that return row sets, such as <command>EXPLAIN</command>.
      Except for the <literal>INTO</literal> clause, the SQL command is the same
      as it would be written outside <application>PL/pgSQL</application>.
@@ -1236,7 +1237,7 @@ END;
     </para>
 
     <para>
-     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
+     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command> with
      <literal>RETURNING</literal>, <application>PL/pgSQL</application> reports
      an error for more than one returned row, even when
      <literal>STRICT</literal> is not specified.  This is because there
diff --git a/doc/src/sgml/queries.sgml b/doc/src/sgml/queries.sgml
new file mode 100644
index 3f95849..9928373
--- a/doc/src/sgml/queries.sgml
+++ b/doc/src/sgml/queries.sgml
@@ -2062,9 +2062,10 @@ SELECT <replaceable>select_list</replace
    Table Expressions or <acronym>CTE</acronym>s, can be thought of as defining
    temporary tables that exist just for one query.  Each auxiliary statement
    in a <literal>WITH</literal> clause can be a <command>SELECT</command>,
-   <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>; and the
+   <command>INSERT</command>, <command>UPDATE</command>, <command>DELETE</command>,
+   or <command>MERGE</command>; and the
    <literal>WITH</literal> clause itself is attached to a primary statement that can
-   be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
+   also be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
    <command>DELETE</command>, or <command>MERGE</command>.
   </para>
 
@@ -2598,8 +2599,8 @@ SELECT * FROM w AS w1 JOIN w AS w2 ON w1
    <title>Data-Modifying Statements in <literal>WITH</literal></title>
 
    <para>
-    You can use most data-modifying statements (<command>INSERT</command>,
-    <command>UPDATE</command>, or <command>DELETE</command>, but not
+    You can use data-modifying statements (<command>INSERT</command>,
+    <command>UPDATE</command>, <command>DELETE</command>, or
     <command>MERGE</command>) in <literal>WITH</literal>.  This
     allows you to perform several different operations in the same query.
     An example is:
diff --git a/doc/src/sgml/ref/copy.sgml b/doc/src/sgml/ref/copy.sgml
new file mode 100644
index 4d614a0..efaacb9
--- a/doc/src/sgml/ref/copy.sgml
+++ b/doc/src/sgml/ref/copy.sgml
@@ -122,13 +122,16 @@ COPY { <replaceable class="parameter">ta
       A <link linkend="sql-select"><command>SELECT</command></link>,
       <link linkend="sql-values"><command>VALUES</command></link>,
       <link linkend="sql-insert"><command>INSERT</command></link>,
-      <link linkend="sql-update"><command>UPDATE</command></link>, or
-      <link linkend="sql-delete"><command>DELETE</command></link> command whose results are to be
-      copied.  Note that parentheses are required around the query.
+      <link linkend="sql-update"><command>UPDATE</command></link>,
+      <link linkend="sql-delete"><command>DELETE</command></link>, or
+      <link linkend="sql-merge"><command>MERGE</command></link> command
+      whose results are to be copied.  Note that parentheses are required
+      around the query.
      </para>
      <para>
-      For <command>INSERT</command>, <command>UPDATE</command> and
-      <command>DELETE</command> queries a RETURNING clause must be provided,
+      For <command>INSERT</command>, <command>UPDATE</command>,
+      <command>DELETE</command>, and <command>MERGE</command> queries a
+      RETURNING clause must be provided,
       and the target relation must not have a conditional rule, nor
       an <literal>ALSO</literal> rule, nor an <literal>INSTEAD</literal> rule
       that expands to multiple statements.
diff --git a/doc/src/sgml/ref/merge.sgml b/doc/src/sgml/ref/merge.sgml
new file mode 100644
index 0995fe0..d1aa937
--- a/doc/src/sgml/ref/merge.sgml
+++ b/doc/src/sgml/ref/merge.sgml
@@ -25,6 +25,7 @@ PostgreSQL documentation
 MERGE INTO [ ONLY ] <replaceable class="parameter">target_table_name</replaceable> [ * ] [ [ AS ] <replaceable class="parameter">target_alias</replaceable> ]
 USING <replaceable class="parameter">data_source</replaceable> ON <replaceable class="parameter">join_condition</replaceable>
 <replaceable class="parameter">when_clause</replaceable> [...]
+[ RETURNING * | <replaceable class="parameter">output_expression</replaceable> [ [ AS ] <replaceable class="parameter">output_name</replaceable> ] [, ...] ]
 
 <phrase>where <replaceable class="parameter">data_source</replaceable> is:</phrase>
 
@@ -95,6 +96,19 @@ DELETE
   </para>
 
   <para>
+   The optional <literal>RETURNING</literal> clause causes <command>MERGE</command>
+   to compute and return value(s) based on each row inserted, updated, or
+   deleted.  Any expression using the source or target table's columns, or the
+   merge support functions <xref linkend="pg-merge-action"/> and
+   <xref linkend="pg-merge-when-clause-number"/> can be computed.  When an
+   <command>INSERT</command> or <command>UPDATE</command> action is performed,
+   the new values of the target table's columns are used.  When a
+   <command>DELETE</command> is performed, the old values of the target table's
+   columns are used.  The syntax of the <literal>RETURNING</literal> list is
+   identical to that of the output list of <command>SELECT</command>.
+  </para>
+
+  <para>
    There is no separate <literal>MERGE</literal> privilege.
    If you specify an update action, you must have the
    <literal>UPDATE</literal> privilege on the column(s)
@@ -125,6 +139,18 @@ DELETE
 
   <variablelist>
    <varlistentry>
+    <term><replaceable class="parameter">with_query</replaceable></term>
+    <listitem>
+     <para>
+      The <literal>WITH</literal> clause allows you to specify one or more
+      subqueries that can be referenced by name in the <command>MERGE</command>
+      query. See <xref linkend="queries-with"/> and <xref linkend="sql-select"/>
+      for details.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
     <term><replaceable class="parameter">target_table_name</replaceable></term>
     <listitem>
      <para>
@@ -399,13 +425,31 @@ DELETE
    </varlistentry>
 
    <varlistentry>
-    <term><replaceable class="parameter">with_query</replaceable></term>
+    <term><replaceable class="parameter">output_expression</replaceable></term>
     <listitem>
      <para>
-      The <literal>WITH</literal> clause allows you to specify one or more
-      subqueries that can be referenced by name in the <command>MERGE</command>
-      query. See <xref linkend="queries-with"/> and <xref linkend="sql-select"/>
-      for details.
+      An expression to be computed and returned by the <command>MERGE</command>
+      command after each row is merged.  The expression can use any columns of
+      the source or target tables, or the merge support functions
+      <xref linkend="pg-merge-action"/> and
+      <xref linkend="pg-merge-when-clause-number"/>.
+     </para>
+     <para>
+      Writing <literal>*</literal> will return all columns from the source
+      table, followed by all columns from the target table.  Often this will
+      lead to a lot of duplication, since it is common for the source and
+      target tables to have a lot of the same columns.  This can be avoided by
+      qualifying the <literal>*</literal> with the name or alias of the source
+      or target table.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
+    <term><replaceable class="parameter">output_name</replaceable></term>
+    <listitem>
+     <para>
+      A name to use for a returned column.
      </para>
     </listitem>
    </varlistentry>
@@ -428,6 +472,13 @@ MERGE <replaceable class="parameter">tot
    were changed in any way.
   </para>
 
+  <para>
+   If the <command>MERGE</command> command contains a <literal>RETURNING</literal>
+   clause, the result will be similar to that of a <command>SELECT</command>
+   statement containing the columns and values defined in the
+   <literal>RETURNING</literal> list, computed over the row(s) inserted, updated,
+   or deleted by the command.
+  </para>
  </refsect1>
 
  <refsect1>
@@ -544,13 +595,6 @@ MERGE <replaceable class="parameter">tot
   </para>
 
   <para>
-   There is no <literal>RETURNING</literal> clause with
-   <command>MERGE</command>.  Actions of <command>INSERT</command>,
-   <command>UPDATE</command> and <command>DELETE</command> cannot contain
-   <literal>RETURNING</literal> or <literal>WITH</literal> clauses.
-  </para>
-
-  <para>
    When <command>MERGE</command> is run concurrently with other commands
    that modify the target table, the usual transaction isolation rules
    apply; see <xref linkend="transaction-iso"/> for an explanation
@@ -602,7 +646,8 @@ WHEN NOT MATCHED THEN
   <para>
    Attempt to insert a new stock item along with the quantity of stock. If
    the item already exists, instead update the stock count of the existing
-   item. Don't allow entries that have zero stock.
+   item. Don't allow entries that have zero stock. Return details of all
+   changes made.
 <programlisting>
 MERGE INTO wines w
 USING wine_stock_changes s
@@ -612,7 +657,8 @@ WHEN NOT MATCHED AND s.stock_delta > 0 T
 WHEN MATCHED AND w.stock + s.stock_delta > 0 THEN
   UPDATE SET stock = w.stock + s.stock_delta
 WHEN MATCHED THEN
-  DELETE;
+  DELETE
+RETURNING pg_merge_action(), w.*;
 </programlisting>
 
    The <literal>wine_stock_changes</literal> table might be, for example, a
@@ -627,7 +673,8 @@ WHEN MATCHED THEN
     This command conforms to the <acronym>SQL</acronym> standard.
   </para>
    <para>
-    The WITH clause and <literal>DO NOTHING</literal> action are extensions to
+    The WITH clause, <literal>DO NOTHING</literal> action, and
+    <literal>RETURNING</literal> clause are extensions to
     the <acronym>SQL</acronym> standard.
   </para>
  </refsect1>
diff --git a/doc/src/sgml/ref/select.sgml b/doc/src/sgml/ref/select.sgml
new file mode 100644
index 0ee0cc7..f81bdea
--- a/doc/src/sgml/ref/select.sgml
+++ b/doc/src/sgml/ref/select.sgml
@@ -74,7 +74,7 @@ SELECT [ ALL | DISTINCT [ ON ( <replacea
 
 <phrase>and <replaceable class="parameter">with_query</replaceable> is:</phrase>
 
-    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> )
+    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> | <replaceable class="parameter">merge</replaceable> )
         [ SEARCH { BREADTH | DEPTH } FIRST BY <replaceable>column_name</replaceable> [, ...] SET <replaceable>search_seq_col_name</replaceable> ]
         [ CYCLE <replaceable>column_name</replaceable> [, ...] SET <replaceable>cycle_mark_col_name</replaceable> [ TO <replaceable>cycle_mark_value</replaceable> DEFAULT <replaceable>cycle_mark_default</replaceable> ] USING <replaceable>cycle_path_col_name</replaceable> ]
 
@@ -227,10 +227,10 @@ TABLE [ ONLY ] <replaceable class="param
     The subqueries effectively act as temporary tables or views
     for the duration of the primary query.
     Each subquery can be a <command>SELECT</command>, <command>TABLE</command>, <command>VALUES</command>,
-    <command>INSERT</command>, <command>UPDATE</command> or
-    <command>DELETE</command> statement.
+    <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command> statement.
     When writing a data-modifying statement (<command>INSERT</command>,
-    <command>UPDATE</command> or <command>DELETE</command>) in
+    <command>UPDATE</command>, <command>DELETE</command>, or <command>MERGE</command>) in
     <literal>WITH</literal>, it is usual to include a <literal>RETURNING</literal> clause.
     It is the output of <literal>RETURNING</literal>, <emphasis>not</emphasis> the underlying
     table that the statement modifies, that forms the temporary table that is
@@ -2179,7 +2179,8 @@ SELECT 2+2;
 
    <para>
     <productname>PostgreSQL</productname> allows <command>INSERT</command>,
-    <command>UPDATE</command>, and <command>DELETE</command> to be used as <literal>WITH</literal>
+    <command>UPDATE</command>, <command>DELETE</command>, and
+    <command>MERGE</command> to be used as <literal>WITH</literal>
     queries.  This is not found in the SQL standard.
    </para>
   </refsect2>
diff --git a/doc/src/sgml/rowtypes.sgml b/doc/src/sgml/rowtypes.sgml
new file mode 100644
index 4d86f97..bbeac84
--- a/doc/src/sgml/rowtypes.sgml
+++ b/doc/src/sgml/rowtypes.sgml
@@ -348,7 +348,7 @@ SELECT m.* FROM some_table, LATERAL myfu
    column expansion of this kind when it appears at the top level of
    a <link linkend="queries-select-lists"><command>SELECT</command> output
    list</link>, a <link linkend="dml-returning"><literal>RETURNING</literal>
-   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>,
+   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>,
    a <link linkend="queries-values"><literal>VALUES</literal> clause</link>, or
    a <link linkend="sql-syntax-row-constructors">row constructor</link>.
    In all other contexts (including when nested inside one of those
diff --git a/doc/src/sgml/spi.sgml b/doc/src/sgml/spi.sgml
new file mode 100644
index 16c3ab7..6410862
--- a/doc/src/sgml/spi.sgml
+++ b/doc/src/sgml/spi.sgml
@@ -301,8 +301,9 @@ SPI_execute("INSERT INTO foo SELECT * FR
    is returned in the global variable <varname>SPI_processed</varname>.
    If the return value of the function is <symbol>SPI_OK_SELECT</symbol>,
    <symbol>SPI_OK_INSERT_RETURNING</symbol>,
-   <symbol>SPI_OK_DELETE_RETURNING</symbol>, or
-   <symbol>SPI_OK_UPDATE_RETURNING</symbol>,
+   <symbol>SPI_OK_DELETE_RETURNING</symbol>,
+   <symbol>SPI_OK_UPDATE_RETURNING</symbol>, or
+   <symbol>SPI_OK_MERGE_RETURNING</symbol>,
    then you can use the
    global pointer <literal>SPITupleTable *SPI_tuptable</literal> to
    access the result rows.  Some utility commands (such as
@@ -472,6 +473,15 @@ typedef struct SPITupleTable
       </para>
      </listitem>
     </varlistentry>
+
+    <varlistentry>
+     <term><symbol>SPI_OK_MERGE_RETURNING</symbol></term>
+     <listitem>
+      <para>
+       if a <command>MERGE RETURNING</command> was executed
+      </para>
+     </listitem>
+    </varlistentry>
 
     <varlistentry>
      <term><symbol>SPI_OK_UTILITY</symbol></term>
diff --git a/doc/src/sgml/xfunc.sgml b/doc/src/sgml/xfunc.sgml
new file mode 100644
index 281c178..3aec197
--- a/doc/src/sgml/xfunc.sgml
+++ b/doc/src/sgml/xfunc.sgml
@@ -177,7 +177,8 @@
     statements separated by semicolons.  A semicolon after the last
     statement is optional.  Unless the function is declared to return
     <type>void</type>, the last statement must be a <command>SELECT</command>,
-    or an <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>
+    or an <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command>
     that has a <literal>RETURNING</literal> clause.
    </para>
 
@@ -1236,8 +1237,9 @@ SELECT x, CASE WHEN x &gt; 0 THEN genera
 
     <note>
      <para>
-      If a function's last command is <command>INSERT</command>, <command>UPDATE</command>,
-      or <command>DELETE</command> with <literal>RETURNING</literal>, that command will
+      If a function's last command is <command>INSERT</command>,
+      <command>UPDATE</command>, <command>DELETE</command>, or
+      <command>MERGE</command> with <literal>RETURNING</literal>, that command will
       always be executed to completion, even if the function is not declared
       with <literal>SETOF</literal> or the calling query does not fetch all the
       result rows.  Any extra rows produced by the <literal>RETURNING</literal>
diff --git a/src/backend/commands/copy.c b/src/backend/commands/copy.c
new file mode 100644
index f14fae3..05f2e5b
--- a/src/backend/commands/copy.c
+++ b/src/backend/commands/copy.c
@@ -283,12 +283,6 @@ DoCopy(ParseState *pstate, const CopyStm
 	{
 		Assert(stmt->query);
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(stmt->query, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in COPY"));
-
 		query = makeNode(RawStmt);
 		query->stmt = stmt->query;
 		query->stmt_location = stmt_location;
diff --git a/src/backend/commands/copyto.c b/src/backend/commands/copyto.c
new file mode 100644
index eaa3172..a50cd30
--- a/src/backend/commands/copyto.c
+++ b/src/backend/commands/copyto.c
@@ -510,7 +510,8 @@ BeginCopyTo(ParseState *pstate,
 		{
 			Assert(query->commandType == CMD_INSERT ||
 				   query->commandType == CMD_UPDATE ||
-				   query->commandType == CMD_DELETE);
+				   query->commandType == CMD_DELETE ||
+				   query->commandType == CMD_MERGE);
 
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
diff --git a/src/backend/executor/execExpr.c b/src/backend/executor/execExpr.c
new file mode 100644
index 2c62b0c..f09d512
--- a/src/backend/executor/execExpr.c
+++ b/src/backend/executor/execExpr.c
@@ -48,6 +48,7 @@
 #include "utils/array.h"
 #include "utils/builtins.h"
 #include "utils/datum.h"
+#include "utils/fmgroids.h"
 #include "utils/jsonfuncs.h"
 #include "utils/lsyscache.h"
 #include "utils/typcache.h"
@@ -2620,6 +2621,22 @@ ExecInitFunc(ExprEvalStep *scratch, Expr
 	InitFunctionCallInfoData(*fcinfo, flinfo,
 							 nargs, inputcollid, NULL, NULL);
 
+	/*
+	 * Merge support functions should only be called from a MERGE command, and
+	 * need access to the parent ModifyTableState.  The parser should have
+	 * checked that such functions only appear in the RETURNING list of a
+	 * MERGE, so this should never fail.
+	 */
+	if (IsMergeSupportFunction(funcid))
+	{
+		if (!state->parent ||
+			!IsA(state->parent, ModifyTableState) ||
+			((ModifyTableState *) state->parent)->operation != CMD_MERGE)
+			elog(ERROR, "merge support function called in non-merge context");
+
+		fcinfo->context = (Node *) state->parent;
+	}
+
 	/* Keep extra copies of this info to save an indirection at runtime */
 	scratch->d.func.fn_addr = flinfo->fn_addr;
 	scratch->d.func.nargs = nargs;
diff --git a/src/backend/executor/execPartition.c b/src/backend/executor/execPartition.c
new file mode 100644
index eb8a87f..b9bd03c
--- a/src/backend/executor/execPartition.c
+++ b/src/backend/executor/execPartition.c
@@ -612,8 +612,8 @@ ExecInitPartitionInfo(ModifyTableState *
 	 * Build the RETURNING projection for the partition.  Note that we didn't
 	 * build the returningList for partitions within the planner, but simple
 	 * translation of varattnos will suffice.  This only occurs for the INSERT
-	 * case or in the case of UPDATE tuple routing where we didn't find a
-	 * result rel to reuse.
+	 * case or in the case of UPDATE/MERGE tuple routing where we didn't find
+	 * a result rel to reuse.
 	 */
 	if (node && node->returningLists != NIL)
 	{
@@ -622,12 +622,14 @@ ExecInitPartitionInfo(ModifyTableState *
 		List	   *returningList;
 
 		/* See the comment above for WCO lists. */
-		/* (except no RETURNING support for MERGE yet) */
 		Assert((node->operation == CMD_INSERT &&
 				list_length(node->returningLists) == 1 &&
 				list_length(node->resultRelations) == 1) ||
 			   (node->operation == CMD_UPDATE &&
 				list_length(node->returningLists) ==
+				list_length(node->resultRelations)) ||
+			   (node->operation == CMD_MERGE &&
+				list_length(node->returningLists) ==
 				list_length(node->resultRelations)));
 
 		/*
diff --git a/src/backend/executor/functions.c b/src/backend/executor/functions.c
new file mode 100644
index f55424e..97874a4
--- a/src/backend/executor/functions.c
+++ b/src/backend/executor/functions.c
@@ -1660,8 +1660,8 @@ check_sql_fn_retval(List *queryTreeLists
 
 	/*
 	 * If it's a plain SELECT, it returns whatever the targetlist says.
-	 * Otherwise, if it's INSERT/UPDATE/DELETE with RETURNING, it returns
-	 * that. Otherwise, the function return type must be VOID.
+	 * Otherwise, if it's INSERT/UPDATE/DELETE/MERGE with RETURNING, it
+	 * returns that. Otherwise, the function return type must be VOID.
 	 *
 	 * Note: eventually replace this test with QueryReturnsTuples?	We'd need
 	 * a more general method of determining the output type, though.  Also, it
@@ -1679,7 +1679,8 @@ check_sql_fn_retval(List *queryTreeLists
 	else if (parse &&
 			 (parse->commandType == CMD_INSERT ||
 			  parse->commandType == CMD_UPDATE ||
-			  parse->commandType == CMD_DELETE) &&
+			  parse->commandType == CMD_DELETE ||
+			  parse->commandType == CMD_MERGE) &&
 			 parse->returningList)
 	{
 		tlist = parse->returningList;
@@ -1693,7 +1694,7 @@ check_sql_fn_retval(List *queryTreeLists
 				(errcode(ERRCODE_INVALID_FUNCTION_DEFINITION),
 				 errmsg("return type mismatch in function declared to return %s",
 						format_type_be(rettype)),
-				 errdetail("Function's final statement must be SELECT or INSERT/UPDATE/DELETE RETURNING.")));
+				 errdetail("Function's final statement must be SELECT or INSERT/UPDATE/DELETE/MERGE RETURNING.")));
 		return false;			/* keep compiler quiet */
 	}
 
diff --git a/src/backend/executor/nodeModifyTable.c b/src/backend/executor/nodeModifyTable.c
new file mode 100644
index 5005d8c..a6f6e32
--- a/src/backend/executor/nodeModifyTable.c
+++ b/src/backend/executor/nodeModifyTable.c
@@ -36,8 +36,7 @@
  *		RETURNING tuple after completing each row insert, update, or delete.
  *		It must be called again to continue the operation.  Without RETURNING,
  *		we just loop within the node until all the work is done, then
- *		return NULL.  This avoids useless call/return overhead.  (MERGE does
- *		not support RETURNING.)
+ *		return NULL.  This avoids useless call/return overhead.
  */
 
 #include "postgres.h"
@@ -88,9 +87,6 @@ typedef struct ModifyTableContext
 	 */
 	TupleTableSlot *planSlot;
 
-	/* MERGE specific */
-	MergeActionState *relaction;	/* MERGE action in progress */
-
 	/*
 	 * Information about the changes that were made concurrently to a tuple
 	 * being updated or deleted
@@ -153,13 +149,14 @@ static TupleTableSlot *ExecMerge(ModifyT
 								 ItemPointer tupleid,
 								 bool canSetTag);
 static void ExecInitMerge(ModifyTableState *mtstate, EState *estate);
-static bool ExecMergeMatched(ModifyTableContext *context,
-							 ResultRelInfo *resultRelInfo,
-							 ItemPointer tupleid,
-							 bool canSetTag);
-static void ExecMergeNotMatched(ModifyTableContext *context,
-								ResultRelInfo *resultRelInfo,
-								bool canSetTag);
+static TupleTableSlot *ExecMergeMatched(ModifyTableContext *context,
+										ResultRelInfo *resultRelInfo,
+										ItemPointer tupleid,
+										bool canSetTag,
+										bool *matched);
+static TupleTableSlot *ExecMergeNotMatched(ModifyTableContext *context,
+										   ResultRelInfo *resultRelInfo,
+										   bool canSetTag);
 
 
 /*
@@ -979,7 +976,7 @@ ExecInsert(ModifyTableContext *context,
 		if (mtstate->operation == CMD_UPDATE)
 			wco_kind = WCO_RLS_UPDATE_CHECK;
 		else if (mtstate->operation == CMD_MERGE)
-			wco_kind = (context->relaction->mas_action->commandType == CMD_UPDATE) ?
+			wco_kind = (mtstate->mt_merge_action->mas_action->commandType == CMD_UPDATE) ?
 				WCO_RLS_UPDATE_CHECK : WCO_RLS_INSERT_CHECK;
 		else
 			wco_kind = WCO_RLS_INSERT_CHECK;
@@ -1828,7 +1825,7 @@ ExecCrossPartitionUpdate(ModifyTableCont
 		 * additional rechecking, and might end up executing a different
 		 * action entirely).
 		 */
-		if (context->relaction != NULL)
+		if (mtstate->mt_merge_action != NULL)
 			return false;
 		else if (TupIsNull(epqslot))
 			return true;
@@ -2069,7 +2066,7 @@ lreplace:
 		 * No luck, a retry is needed.  If running MERGE, we do not do so
 		 * here; instead let it handle that on its own rules.
 		 */
-		if (context->relaction != NULL)
+		if (context->mtstate->mt_merge_action != NULL)
 			return TM_Updated;
 
 		/*
@@ -2712,6 +2709,7 @@ static TupleTableSlot *
 ExecMerge(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 		  ItemPointer tupleid, bool canSetTag)
 {
+	TupleTableSlot *rslot = NULL;
 	bool		matched;
 
 	/*-----
@@ -2759,18 +2757,18 @@ ExecMerge(ModifyTableContext *context, R
 	 */
 	matched = tupleid != NULL;
 	if (matched)
-		matched = ExecMergeMatched(context, resultRelInfo, tupleid, canSetTag);
+		rslot = ExecMergeMatched(context, resultRelInfo, tupleid, canSetTag,
+								 &matched);
 
 	/*
-	 * Either we were dealing with a NOT MATCHED tuple or ExecMergeMatched()
-	 * returned "false", indicating the previously MATCHED tuple no longer
-	 * matches.
+	 * Deal with the NOT MATCHED case (either a NOT MATCHED tuple from the
+	 * join, or a previously MATCHED tuple for which ExecMergeMatched() set
+	 * "matched" to false, indicating that it no longer matches).
 	 */
 	if (!matched)
-		ExecMergeNotMatched(context, resultRelInfo, canSetTag);
+		rslot = ExecMergeNotMatched(context, resultRelInfo, canSetTag);
 
-	/* No RETURNING support yet */
-	return NULL;
+	return rslot;
 }
 
 /*
@@ -2780,8 +2778,8 @@ ExecMerge(ModifyTableContext *context, R
  * We start from the first WHEN MATCHED action and check if the WHEN quals
  * pass, if any. If the WHEN quals for the first action do not pass, we
  * check the second, then the third and so on. If we reach to the end, no
- * action is taken and we return true, indicating that no further action is
- * required for this tuple.
+ * action is taken and "matched" is set to true, indicating that no further
+ * action is required for this tuple.
  *
  * If we do find a qualifying action, then we attempt to execute the action.
  *
@@ -2790,15 +2788,16 @@ ExecMerge(ModifyTableContext *context, R
  * with individual actions are evaluated by this routine via ExecQual, while
  * EvalPlanQual checks for the join quals. If EvalPlanQual tells us that the
  * updated tuple still passes the join quals, then we restart from the first
- * action to look for a qualifying action. Otherwise, we return false --
- * meaning that a NOT MATCHED action must now be executed for the current
- * source tuple.
+ * action to look for a qualifying action. Otherwise, "matched" is set to
+ * false -- meaning that a NOT MATCHED action must now be executed for the
+ * current source tuple.
  */
-static bool
+static TupleTableSlot *
 ExecMergeMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
-				 ItemPointer tupleid, bool canSetTag)
+				 ItemPointer tupleid, bool canSetTag, bool *matched)
 {
 	ModifyTableState *mtstate = context->mtstate;
+	TupleTableSlot *rslot = NULL;
 	TupleTableSlot *newslot;
 	EState	   *estate = context->estate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
@@ -2810,7 +2809,10 @@ ExecMergeMatched(ModifyTableContext *con
 	 * If there are no WHEN MATCHED actions, we are done.
 	 */
 	if (resultRelInfo->ri_matchedMergeAction == NIL)
-		return true;
+	{
+		*matched = true;
+		return NULL;
+	}
 
 	/*
 	 * Make tuple and any needed join variables available to ExecQual and
@@ -2889,12 +2891,15 @@ lmerge_matched:
 				 */
 				newslot = ExecProject(relaction->mas_proj);
 
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				if (!ExecUpdatePrologue(context, resultRelInfo,
 										tupleid, NULL, newslot, &result))
 				{
 					if (result == TM_Ok)
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 					break;		/* concurrent update/delete */
 				}
 				result = ExecUpdateAct(context, resultRelInfo, tupleid, NULL,
@@ -2908,12 +2913,15 @@ lmerge_matched:
 				break;
 
 			case CMD_DELETE:
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				if (!ExecDeletePrologue(context, resultRelInfo, tupleid,
 										NULL, NULL, &result))
 				{
 					if (result == TM_Ok)
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 					break;		/* concurrent update/delete */
 				}
 				result = ExecDeleteAct(context, resultRelInfo, tupleid, false);
@@ -2969,7 +2977,8 @@ lmerge_matched:
 				 * If the tuple was already deleted, return to let caller
 				 * handle it under NOT MATCHED clauses.
 				 */
-				return false;
+				*matched = false;
+				return NULL;
 
 			case TM_Updated:
 				{
@@ -3015,13 +3024,19 @@ lmerge_matched:
 							 * NOT MATCHED actions.
 							 */
 							if (TupIsNull(epqslot))
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							(void) ExecGetJunkAttribute(epqslot,
 														resultRelInfo->ri_RowIdAttNo,
 														&isNull);
 							if (isNull)
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							/*
 							 * When a tuple was updated and migrated to
@@ -3056,7 +3071,8 @@ lmerge_matched:
 							 * tuple already deleted; tell caller to run NOT
 							 * MATCHED actions
 							 */
-							return false;
+							*matched = false;
+							return NULL;
 
 						case TM_SelfModified:
 
@@ -3076,13 +3092,14 @@ lmerge_matched:
 										(errcode(ERRCODE_TRIGGERED_DATA_CHANGE_VIOLATION),
 										 errmsg("tuple to be updated or deleted was already modified by an operation triggered by the current command"),
 										 errhint("Consider using an AFTER trigger instead of a BEFORE trigger to propagate changes to other rows.")));
-							return false;
+							*matched = false;
+							return NULL;
 
 						default:
 							/* see table_tuple_lock call in ExecDelete() */
 							elog(ERROR, "unexpected table_tuple_lock status: %u",
 								 result);
-							return false;
+							return NULL;
 					}
 				}
 
@@ -3094,6 +3111,30 @@ lmerge_matched:
 				break;
 		}
 
+		/* Process RETURNING if present */
+		if (resultRelInfo->ri_projectReturning)
+		{
+			switch (commandType)
+			{
+				case CMD_UPDATE:
+					rslot = ExecProcessReturning(resultRelInfo, newslot,
+												 context->planSlot);
+					break;
+
+				case CMD_DELETE:
+					rslot = ExecProcessReturning(resultRelInfo,
+												 resultRelInfo->ri_oldTupleSlot,
+												 context->planSlot);
+					break;
+
+				case CMD_NOTHING:
+					break;
+
+				default:
+					elog(ERROR, "unknown action in MERGE WHEN MATCHED clause");
+			}
+		}
+
 		/*
 		 * We've activated one of the WHEN clauses, so we don't search
 		 * further. This is required behaviour, not an optimization.
@@ -3104,19 +3145,22 @@ lmerge_matched:
 	/*
 	 * Successfully executed an action or no qualifying action was found.
 	 */
-	return true;
+	*matched = true;
+
+	return rslot;
 }
 
 /*
  * Execute the first qualifying NOT MATCHED action.
  */
-static void
+static TupleTableSlot *
 ExecMergeNotMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 					bool canSetTag)
 {
 	ModifyTableState *mtstate = context->mtstate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
 	List	   *actionStates = NIL;
+	TupleTableSlot *rslot = NULL;
 	ListCell   *l;
 
 	/*
@@ -3166,10 +3210,10 @@ ExecMergeNotMatched(ModifyTableContext *
 				 * so we don't need to map the tuple here.
 				 */
 				newslot = ExecProject(action->mas_proj);
-				context->relaction = action;
+				mtstate->mt_merge_action = action;
 
-				(void) ExecInsert(context, mtstate->rootResultRelInfo, newslot,
-								  canSetTag, NULL, NULL);
+				rslot = ExecInsert(context, mtstate->rootResultRelInfo,
+								   newslot, canSetTag, NULL, NULL);
 				mtstate->mt_merge_inserted += 1;
 				break;
 			case CMD_NOTHING:
@@ -3185,6 +3229,8 @@ ExecMergeNotMatched(ModifyTableContext *
 		 */
 		break;
 	}
+
+	return rslot;
 }
 
 /*
@@ -3362,6 +3408,65 @@ ExecInitMergeTupleSlots(ModifyTableState
 }
 
 /*
+ * pg_merge_action() -
+ *	  SQL merge support function to retrieve the currently executing merge
+ *	  action command string ("INSERT", "UPDATE", or "DELETE").
+ */
+Datum
+pg_merge_action(PG_FUNCTION_ARGS)
+{
+	ModifyTableState *mtstate = (ModifyTableState *) fcinfo->context;
+	MergeActionState *relaction;
+	CmdType		commandType;
+
+	if (!mtstate || mtstate->operation != CMD_MERGE)
+		elog(ERROR, "merge support function called in non-merge context");
+
+	relaction = mtstate->mt_merge_action;
+	if (!relaction)
+		elog(ERROR, "no merge action in progress");
+
+	commandType = relaction->mas_action->commandType;
+	switch (commandType)
+	{
+		case CMD_INSERT:
+			PG_RETURN_TEXT_P(cstring_to_text("INSERT"));
+		case CMD_UPDATE:
+			PG_RETURN_TEXT_P(cstring_to_text("UPDATE"));
+		case CMD_DELETE:
+			PG_RETURN_TEXT_P(cstring_to_text("DELETE"));
+		case CMD_NOTHING:
+			elog(ERROR, "unexpected merge action: DO NOTHING");
+			break;
+		default:
+			elog(ERROR, "unrecognized commandType: %d", (int) commandType);
+	}
+
+	PG_RETURN_NULL();			/* keep compiler quiet */
+}
+
+/*
+ * pg_merge_when_clause_number() -
+ *	  SQL merge support function to retrieve the 1-based index of the
+ *	  currently executing merge WHEN clause.
+ */
+Datum
+pg_merge_when_clause_number(PG_FUNCTION_ARGS)
+{
+	ModifyTableState *mtstate = (ModifyTableState *) fcinfo->context;
+	MergeActionState *relaction;
+
+	if (!mtstate || mtstate->operation != CMD_MERGE)
+		elog(ERROR, "merge support function called in non-merge context");
+
+	relaction = mtstate->mt_merge_action;
+	if (!relaction)
+		elog(ERROR, "no merge action in progress");
+
+	PG_RETURN_INT32((int32) relaction->mas_action->index);
+}
+
+/*
  * Process BEFORE EACH STATEMENT triggers
  */
 static void
@@ -3647,8 +3752,17 @@ ExecModifyTable(PlanState *pstate)
 				{
 					EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-					ExecMerge(&context, node->resultRelInfo, NULL, node->canSetTag);
-					continue;	/* no RETURNING support yet */
+					slot = ExecMerge(&context, node->resultRelInfo, NULL,
+									 node->canSetTag);
+
+					/*
+					 * If we got a RETURNING result, return it to the caller.
+					 * We'll continue the work on next call.
+					 */
+					if (slot)
+						return slot;
+
+					continue;	/* continue with the next tuple */
 				}
 
 				elog(ERROR, "tableoid is NULL");
@@ -3725,8 +3839,17 @@ ExecModifyTable(PlanState *pstate)
 					{
 						EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-						ExecMerge(&context, node->resultRelInfo, NULL, node->canSetTag);
-						continue;	/* no RETURNING support yet */
+						slot = ExecMerge(&context, node->resultRelInfo, NULL,
+										 node->canSetTag);
+
+						/*
+						 * If we got a RETURNING result, return it to the
+						 * caller.  We'll continue the work on next call.
+						 */
+						if (slot)
+							return slot;
+
+						continue;	/* continue with the next tuple */
 					}
 
 					elog(ERROR, "ctid is NULL");
@@ -3818,7 +3941,7 @@ ExecModifyTable(PlanState *pstate)
 				}
 				slot = ExecGetUpdateNewTuple(resultRelInfo, context.planSlot,
 											 oldSlot);
-				context.relaction = NULL;
+				node->mt_merge_action = NULL;
 
 				/* Now apply the update. */
 				slot = ExecUpdate(&context, resultRelInfo, tupleid, oldtuple,
diff --git a/src/backend/executor/spi.c b/src/backend/executor/spi.c
new file mode 100644
index 3397568..b9f443c
--- a/src/backend/executor/spi.c
+++ b/src/backend/executor/spi.c
@@ -2033,6 +2033,8 @@ SPI_result_code_string(int code)
 			return "SPI_OK_TD_REGISTER";
 		case SPI_OK_MERGE:
 			return "SPI_OK_MERGE";
+		case SPI_OK_MERGE_RETURNING:
+			return "SPI_OK_MERGE_RETURNING";
 	}
 	/* Unrecognized code ... return something useful ... */
 	sprintf(buf, "Unrecognized SPI code %d", code);
@@ -2886,7 +2888,10 @@ _SPI_pquery(QueryDesc *queryDesc, bool f
 				res = SPI_OK_UPDATE;
 			break;
 		case CMD_MERGE:
-			res = SPI_OK_MERGE;
+			if (queryDesc->plannedstmt->hasReturning)
+				res = SPI_OK_MERGE_RETURNING;
+			else
+				res = SPI_OK_MERGE;
 			break;
 		default:
 			return SPI_ERROR_OPUNKNOWN;
diff --git a/src/backend/optimizer/plan/subselect.c b/src/backend/optimizer/plan/subselect.c
new file mode 100644
index 7a9fe88..d3ee1a2
--- a/src/backend/optimizer/plan/subselect.c
+++ b/src/backend/optimizer/plan/subselect.c
@@ -18,6 +18,7 @@
 
 #include "access/htup_details.h"
 #include "catalog/pg_operator.h"
+#include "catalog/pg_proc.h"
 #include "catalog/pg_type.h"
 #include "executor/executor.h"
 #include "miscadmin.h"
@@ -35,6 +36,7 @@
 #include "parser/parse_relation.h"
 #include "rewrite/rewriteManip.h"
 #include "utils/builtins.h"
+#include "utils/fmgroids.h"
 #include "utils/lsyscache.h"
 #include "utils/syscache.h"
 
@@ -1846,7 +1848,8 @@ convert_EXISTS_to_ANY(PlannerInfo *root,
 /*
  * Replace correlation vars (uplevel vars) with Params.
  *
- * Uplevel PlaceHolderVars and aggregates are replaced, too.
+ * Uplevel PlaceHolderVars, aggregates, GROUPING() expressions, and merge
+ * support functions are replaced, too.
  *
  * Note: it is critical that this runs immediately after SS_process_sublinks.
  * Since we do not recurse into the arguments of uplevel PHVs and aggregates,
@@ -1900,6 +1903,13 @@ replace_correlation_vars_mutator(Node *n
 		if (((GroupingFunc *) node)->agglevelsup > 0)
 			return (Node *) replace_outer_grouping(root, (GroupingFunc *) node);
 	}
+	if (IsA(node, FuncExpr) &&
+		IsMergeSupportFunction(((FuncExpr *) node)->funcid))
+	{
+		if (root->parse->commandType != CMD_MERGE)
+			return (Node *) replace_outer_merge_support_function(root,
+																 (FuncExpr *) node);
+	}
 	return expression_tree_mutator(node,
 								   replace_correlation_vars_mutator,
 								   (void *) root);
diff --git a/src/backend/optimizer/util/paramassign.c b/src/backend/optimizer/util/paramassign.c
new file mode 100644
index d6a923b..5ecf502
--- a/src/backend/optimizer/util/paramassign.c
+++ b/src/backend/optimizer/util/paramassign.c
@@ -307,6 +307,57 @@ replace_outer_grouping(PlannerInfo *root
 }
 
 /*
+ * Generate a Param node to replace the given FuncExpr node, which is expected
+ * to be a merge support function in the RETURNING list of an upper-level
+ * MERGE query.  Record the need for the FuncExpr in the proper upper-level
+ * root->plan_params.
+ */
+Param *
+replace_outer_merge_support_function(PlannerInfo *root, FuncExpr *func)
+{
+	Param	   *retval;
+	PlannerParamItem *pitem;
+	Oid			ptype = exprType((Node *) func);
+
+	Assert(root->parse->commandType != CMD_MERGE);
+
+	/*
+	 * The parser should have ensured that the merge support function is in
+	 * the RETURNING list of an upper-level MERGE query, so find that query.
+	 */
+	do
+	{
+		root = root->parent_root;
+		if (root == NULL)
+			elog(ERROR, "merge support function used outside MERGE");
+	} while (root->parse->commandType != CMD_MERGE);
+
+	/*
+	 * It does not seem worthwhile to try to de-duplicate references to outer
+	 * merge support functions.  Just make a new slot every time.
+	 */
+	func = copyObject(func);
+
+	pitem = makeNode(PlannerParamItem);
+	pitem->item = (Node *) func;
+	pitem->paramId = list_length(root->glob->paramExecTypes);
+	root->glob->paramExecTypes = lappend_oid(root->glob->paramExecTypes,
+											 ptype);
+
+	root->plan_params = lappend(root->plan_params, pitem);
+
+	retval = makeNode(Param);
+	retval->paramkind = PARAM_EXEC;
+	retval->paramid = pitem->paramId;
+	retval->paramtype = ptype;
+	retval->paramtypmod = -1;
+	retval->paramcollid = InvalidOid;
+	retval->location = func->location;
+
+	return retval;
+}
+
+/*
  * Generate a Param node to replace the given Var,
  * which is expected to come from some upper NestLoop plan node.
  * Record the need for the Var in root->curOuterParams.
diff --git a/src/backend/parser/analyze.c b/src/backend/parser/analyze.c
new file mode 100644
index 4006632..6794511
--- a/src/backend/parser/analyze.c
+++ b/src/backend/parser/analyze.c
@@ -74,7 +74,6 @@ static void determineRecursiveColTypes(P
 									   Node *larg, List *nrtargetlist);
 static Query *transformReturnStmt(ParseState *pstate, ReturnStmt *stmt);
 static Query *transformUpdateStmt(ParseState *pstate, UpdateStmt *stmt);
-static List *transformReturningList(ParseState *pstate, List *returningList);
 static Query *transformPLAssignStmt(ParseState *pstate,
 									PLAssignStmt *stmt);
 static Query *transformDeclareCursorStmt(ParseState *pstate,
@@ -515,7 +514,8 @@ transformDeleteStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -942,7 +942,8 @@ transformInsertStmt(ParseState *pstate,
 	/* Process RETURNING, if any. */
 	if (stmt->returningList)
 		qry->returningList = transformReturningList(pstate,
-													stmt->returningList);
+													stmt->returningList,
+													EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -2407,7 +2408,8 @@ transformUpdateStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/*
 	 * Now we are done with SELECT-like processing, and can get on with
@@ -2501,10 +2503,11 @@ transformUpdateTargetList(ParseState *ps
 
 /*
  * transformReturningList -
- *	handle a RETURNING clause in INSERT/UPDATE/DELETE
+ *	handle a RETURNING clause in INSERT/UPDATE/DELETE/MERGE
  */
-static List *
-transformReturningList(ParseState *pstate, List *returningList)
+List *
+transformReturningList(ParseState *pstate, List *returningList,
+					   ParseExprKind exprKind)
 {
 	List	   *rlist;
 	int			save_next_resno;
@@ -2521,7 +2524,7 @@ transformReturningList(ParseState *pstat
 	pstate->p_next_resno = 1;
 
 	/* transform RETURNING identically to a SELECT targetlist */
-	rlist = transformTargetList(pstate, returningList, EXPR_KIND_RETURNING);
+	rlist = transformTargetList(pstate, returningList, exprKind);
 
 	/*
 	 * Complain if the nonempty tlist expanded to nothing (which is possible
diff --git a/src/backend/parser/gram.y b/src/backend/parser/gram.y
new file mode 100644
index b3bdf94..dd2a74f
--- a/src/backend/parser/gram.y
+++ b/src/backend/parser/gram.y
@@ -12270,6 +12270,7 @@ MergeStmt:
 			USING table_ref
 			ON a_expr
 			merge_when_list
+			returning_clause
 				{
 					MergeStmt  *m = makeNode(MergeStmt);
 
@@ -12278,6 +12279,7 @@ MergeStmt:
 					m->sourceRelation = $6;
 					m->joinCondition = $8;
 					m->mergeWhenClauses = $9;
+					m->returningList = $10;
 
 					$$ = (Node *) m;
 				}
diff --git a/src/backend/parser/parse_agg.c b/src/backend/parser/parse_agg.c
new file mode 100644
index 85cd47b..e72c37b
--- a/src/backend/parser/parse_agg.c
+++ b/src/backend/parser/parse_agg.c
@@ -456,6 +456,7 @@ check_agglevels_and_constraints(ParseSta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
@@ -903,6 +904,7 @@ transformWindowFuncCall(ParseState *psta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_cte.c b/src/backend/parser/parse_cte.c
new file mode 100644
index c5b1a49..ad3c525
--- a/src/backend/parser/parse_cte.c
+++ b/src/backend/parser/parse_cte.c
@@ -126,13 +126,6 @@ transformWithClause(ParseState *pstate,
 		CommonTableExpr *cte = (CommonTableExpr *) lfirst(lc);
 		ListCell   *rest;
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(cte->ctequery, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in WITH query"),
-					parser_errposition(pstate, cte->location));
-
 		for_each_cell(rest, withClause->ctes, lnext(withClause->ctes, lc))
 		{
 			CommonTableExpr *cte2 = (CommonTableExpr *) lfirst(rest);
@@ -153,7 +146,8 @@ transformWithClause(ParseState *pstate,
 			/* must be a data-modifying statement */
 			Assert(IsA(cte->ctequery, InsertStmt) ||
 				   IsA(cte->ctequery, UpdateStmt) ||
-				   IsA(cte->ctequery, DeleteStmt));
+				   IsA(cte->ctequery, DeleteStmt) ||
+				   IsA(cte->ctequery, MergeStmt));
 
 			pstate->p_hasModifyingCTE = true;
 		}
diff --git a/src/backend/parser/parse_expr.c b/src/backend/parser/parse_expr.c
new file mode 100644
index aa72665..3957d7c
--- a/src/backend/parser/parse_expr.c
+++ b/src/backend/parser/parse_expr.c
@@ -541,6 +541,7 @@ transformColumnRef(ParseState *pstate, C
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CHECK_CONSTRAINT:
@@ -1767,6 +1768,7 @@ transformSubLink(ParseState *pstate, Sub
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CYCLE_MARK:
@@ -3115,6 +3117,7 @@ ParseExprKindName(ParseExprKind exprKind
 		case EXPR_KIND_OFFSET:
 			return "OFFSET";
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			return "RETURNING";
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
diff --git a/src/backend/parser/parse_func.c b/src/backend/parser/parse_func.c
new file mode 100644
index b3f0b6a..cf7c0b9
--- a/src/backend/parser/parse_func.c
+++ b/src/backend/parser/parse_func.c
@@ -31,6 +31,7 @@
 #include "parser/parse_target.h"
 #include "parser/parse_type.h"
 #include "utils/builtins.h"
+#include "utils/fmgroids.h"
 #include "utils/lsyscache.h"
 #include "utils/syscache.h"
 
@@ -348,6 +349,26 @@ ParseFuncOrColumn(ParseState *pstate, Li
 					 parser_errposition(pstate, location)));
 	}
 
+	/* Merge support functions are only allowed in MERGE's RETURNING list */
+	if (IsMergeSupportFunction(funcid) &&
+		pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+	{
+		ParseState *parent_pstate = pstate->parentParseState;
+
+		/* May be in a subquery in RETURNING list */
+		while (parent_pstate &&
+			   parent_pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+			parent_pstate = parent_pstate->parentParseState;
+
+		if (!parent_pstate ||
+			parent_pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+			ereport(ERROR,
+					(errcode(ERRCODE_WRONG_OBJECT_TYPE),
+					 errmsg("merge support function %s can only be called from the RETURNING list of a MERGE command",
+							NameListToString(funcname)),
+					 parser_errposition(pstate, location)));
+	}
+
 	/*
 	 * So far so good, so do some fdresult-type-specific processing.
 	 */
@@ -2599,6 +2620,7 @@ check_srf_call_placement(ParseState *pst
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_merge.c b/src/backend/parser/parse_merge.c
new file mode 100644
index 91b1156..33852e9
--- a/src/backend/parser/parse_merge.c
+++ b/src/backend/parser/parse_merge.c
@@ -233,6 +233,10 @@ transformMergeStmt(ParseState *pstate, M
 	 */
 	qry->jointree = makeFromExpr(pstate->p_joinlist, joinExpr);
 
+	/* Transform the RETURNING list, if any */
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_MERGE_RETURNING);
+
 	/*
 	 * We now have a good query shape, so now look at the WHEN conditions and
 	 * action targetlists.
@@ -254,6 +258,7 @@ transformMergeStmt(ParseState *pstate, M
 		MergeAction *action;
 
 		action = makeNode(MergeAction);
+		action->index = foreach_current_index(l) + 1;
 		action->commandType = mergeWhenClause->commandType;
 		action->matched = mergeWhenClause->matched;
 
@@ -390,9 +395,6 @@ transformMergeStmt(ParseState *pstate, M
 
 	qry->mergeActionList = mergeActionList;
 
-	/* RETURNING could potentially be added in the future, but not in SQL std */
-	qry->returningList = NULL;
-
 	qry->hasTargetSRFs = false;
 	qry->hasSubLinks = pstate->p_hasSubLinks;
 
diff --git a/src/backend/parser/parse_relation.c b/src/backend/parser/parse_relation.c
new file mode 100644
index 864ea9b..cc19041
--- a/src/backend/parser/parse_relation.c
+++ b/src/backend/parser/parse_relation.c
@@ -2346,9 +2346,10 @@ addRangeTableEntryForCTE(ParseState *pst
 		cte->cterefcount++;
 
 	/*
-	 * We throw error if the CTE is INSERT/UPDATE/DELETE without RETURNING.
-	 * This won't get checked in case of a self-reference, but that's OK
-	 * because data-modifying CTEs aren't allowed to be recursive anyhow.
+	 * We throw error if the CTE is INSERT/UPDATE/DELETE/MERGE without
+	 * RETURNING.  This won't get checked in case of a self-reference, but
+	 * that's OK because data-modifying CTEs aren't allowed to be recursive
+	 * anyhow.
 	 */
 	if (IsA(cte->ctequery, Query))
 	{
diff --git a/src/backend/rewrite/rewriteHandler.c b/src/backend/rewrite/rewriteHandler.c
new file mode 100644
index b486ab5..1bd0ce2
--- a/src/backend/rewrite/rewriteHandler.c
+++ b/src/backend/rewrite/rewriteHandler.c
@@ -3629,9 +3629,9 @@ RewriteQuery(Query *parsetree, List *rew
 	ListCell   *lc1;
 
 	/*
-	 * First, recursively process any insert/update/delete statements in WITH
-	 * clauses.  (We have to do this first because the WITH clauses may get
-	 * copied into rule actions below.)
+	 * First, recursively process any insert/update/delete/merge statements in
+	 * WITH clauses.  (We have to do this first because the WITH clauses may
+	 * get copied into rule actions below.)
 	 */
 	foreach(lc1, parsetree->cteList)
 	{
@@ -3656,7 +3656,8 @@ RewriteQuery(Query *parsetree, List *rew
 			if (!(ctequery->commandType == CMD_SELECT ||
 				  ctequery->commandType == CMD_UPDATE ||
 				  ctequery->commandType == CMD_INSERT ||
-				  ctequery->commandType == CMD_DELETE))
+				  ctequery->commandType == CMD_DELETE ||
+				  ctequery->commandType == CMD_MERGE))
 			{
 				/*
 				 * Currently it could only be NOTIFY; this error message will
diff --git a/src/backend/rewrite/rowsecurity.c b/src/backend/rewrite/rowsecurity.c
new file mode 100644
index b1620e4..966dcf0
--- a/src/backend/rewrite/rowsecurity.c
+++ b/src/backend/rewrite/rowsecurity.c
@@ -395,10 +395,10 @@ get_row_security_policies(Query *root, R
 	 * on the final action we take.
 	 *
 	 * We already fetched the SELECT policies above, to check existing rows,
-	 * but we must also check that new rows created by UPDATE actions are
-	 * visible, if SELECT rights are required for this relation. We don't do
-	 * this for INSERT actions, since an INSERT command would only do this
-	 * check if it had a RETURNING list, and MERGE does not support RETURNING.
+	 * but we must also check that new rows created by INSERT/UPDATE actions
+	 * are visible, if SELECT rights are required. For INSERT actions, we only
+	 * do this if RETURNING is specified, to be consistent with a plain INSERT
+	 * command, which can only require SELECT rights when RETURNING is used.
 	 *
 	 * We don't push the UPDATE/DELETE USING quals to the RTE because we don't
 	 * really want to apply them while scanning the relation since we don't
@@ -420,6 +420,8 @@ get_row_security_policies(Query *root, R
 		List	   *merge_delete_restrictive_policies;
 		List	   *merge_insert_permissive_policies;
 		List	   *merge_insert_restrictive_policies;
+		List	   *merge_select_permissive_policies = NIL;
+		List	   *merge_select_restrictive_policies = NIL;
 
 		/*
 		 * Fetch the UPDATE policies and set them up to execute on the
@@ -457,9 +459,6 @@ get_row_security_policies(Query *root, R
 		 */
 		if (perminfo->requiredPerms & ACL_SELECT)
 		{
-			List	   *merge_select_permissive_policies;
-			List	   *merge_select_restrictive_policies;
-
 			get_policies_for_relation(rel, CMD_SELECT, user_id,
 									  &merge_select_permissive_policies,
 									  &merge_select_restrictive_policies);
@@ -508,6 +507,21 @@ get_row_security_policies(Query *root, R
 							   withCheckOptions,
 							   hasSubLinks,
 							   false);
+
+		/*
+		 * Add ALL/SELECT policies as WCO_RLS_INSERT_CHECK WCOs, to ensure
+		 * that the inserted row is visible when executing an INSERT action,
+		 * if RETURNING is specified and SELECT rights are required for this
+		 * relation.
+		 */
+		if (perminfo->requiredPerms & ACL_SELECT && root->returningList)
+			add_with_check_options(rel, rt_index,
+								   WCO_RLS_INSERT_CHECK,
+								   merge_select_permissive_policies,
+								   merge_select_restrictive_policies,
+								   withCheckOptions,
+								   hasSubLinks,
+								   true);
 	}
 
 	table_close(rel, NoLock);
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
new file mode 100644
index e3ccf6c..6f3dc89
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -2142,11 +2142,10 @@ QueryReturnsTuples(Query *parsetree)
 		case CMD_SELECT:
 			/* returns tuples */
 			return true;
-		case CMD_MERGE:
-			return false;
 		case CMD_INSERT:
 		case CMD_UPDATE:
 		case CMD_DELETE:
+		case CMD_MERGE:
 			/* the forms with RETURNING return tuples */
 			if (parsetree->returningList)
 				return true;
diff --git a/src/backend/utils/adt/ruleutils.c b/src/backend/utils/adt/ruleutils.c
new file mode 100644
index 03f2835..25f49fa
--- a/src/backend/utils/adt/ruleutils.c
+++ b/src/backend/utils/adt/ruleutils.c
@@ -7183,8 +7183,13 @@ get_merge_query_def(Query *query, depars
 			appendStringInfoString(buf, "DO NOTHING");
 	}
 
-	/* No RETURNING support in MERGE yet */
-	Assert(query->returningList == NIL);
+	/* Add RETURNING if present */
+	if (query->returningList)
+	{
+		appendContextKeyword(context, " RETURNING",
+							 -PRETTYINDENT_STD, PRETTYINDENT_STD, 1);
+		get_target_list(query->returningList, context, NULL, colNamesVisible);
+	}
 }
 
 
diff --git a/src/bin/psql/common.c b/src/bin/psql/common.c
new file mode 100644
index 10ad1f2..3f42ca4
--- a/src/bin/psql/common.c
+++ b/src/bin/psql/common.c
@@ -982,13 +982,17 @@ PrintQueryResult(PGresult *result, bool
 			else
 				success = true;
 
-			/* if it's INSERT/UPDATE/DELETE RETURNING, also print status */
+			/*
+			 * If it's INSERT/UPDATE/DELETE/MERGE RETURNING, also print
+			 * status.
+			 */
 			if (last || pset.show_all_results)
 			{
 				cmdstatus = PQcmdStatus(result);
 				if (strncmp(cmdstatus, "INSERT", 6) == 0 ||
 					strncmp(cmdstatus, "UPDATE", 6) == 0 ||
-					strncmp(cmdstatus, "DELETE", 6) == 0)
+					strncmp(cmdstatus, "DELETE", 6) == 0 ||
+					strncmp(cmdstatus, "MERGE", 5) == 0)
 					PrintQueryStatus(result, printStatusFout);
 			}
 
diff --git a/src/include/catalog/pg_proc.dat b/src/include/catalog/pg_proc.dat
new file mode 100644
index 12fac15..88a3b29
--- a/src/include/catalog/pg_proc.dat
+++ b/src/include/catalog/pg_proc.dat
@@ -12049,4 +12049,14 @@
   proname => 'any_value_transfn', prorettype => 'anyelement',
   proargtypes => 'anyelement anyelement', prosrc => 'any_value_transfn' },
 
+# MERGE support functions
+{ oid => '9499', descr => 'command type of current MERGE action',
+  proname => 'pg_merge_action',  provolatile => 'v', proparallel => 'r',
+  prorettype => 'text', proargtypes => '',
+  prosrc => 'pg_merge_action' },
+{ oid => '9500', descr => 'index of current MERGE WHEN clause',
+  proname => 'pg_merge_when_clause_number',  provolatile => 'v',
+  proparallel => 'r', prorettype => 'int4', proargtypes => '',
+  prosrc => 'pg_merge_when_clause_number' },
+
 ]
diff --git a/src/include/catalog/pg_proc.h b/src/include/catalog/pg_proc.h
new file mode 100644
index e7abe0b..eb6cad9
--- a/src/include/catalog/pg_proc.h
+++ b/src/include/catalog/pg_proc.h
@@ -182,6 +182,11 @@ DECLARE_UNIQUE_INDEX(pg_proc_proname_arg
 #define PROARGMODE_VARIADIC 'v'
 #define PROARGMODE_TABLE	't'
 
+/* Is this a merge support function?  (Requires fmgroids.h) */
+#define IsMergeSupportFunction(funcid) \
+	((funcid) == F_PG_MERGE_ACTION || \
+	 (funcid) == F_PG_MERGE_WHEN_CLAUSE_NUMBER)
+
 #endif							/* EXPOSE_TO_CLIENT_CODE */
 
 
diff --git a/src/include/executor/spi.h b/src/include/executor/spi.h
new file mode 100644
index d1de139..ed8787d
--- a/src/include/executor/spi.h
+++ b/src/include/executor/spi.h
@@ -97,6 +97,7 @@ typedef struct _SPI_plan *SPIPlanPtr;
 #define SPI_OK_REL_UNREGISTER	16
 #define SPI_OK_TD_REGISTER		17
 #define SPI_OK_MERGE			18
+#define SPI_OK_MERGE_RETURNING	19
 
 #define SPI_OPT_NONATOMIC		(1 << 0)
 
diff --git a/src/include/nodes/execnodes.h b/src/include/nodes/execnodes.h
new file mode 100644
index cb714f4..6708b6b
--- a/src/include/nodes/execnodes.h
+++ b/src/include/nodes/execnodes.h
@@ -1319,6 +1319,9 @@ typedef struct ModifyTableState
 	/* Flags showing which subcommands are present INS/UPD/DEL/DO NOTHING */
 	int			mt_merge_subcommands;
 
+	/* For MERGE, the action currently being executed */
+	MergeActionState *mt_merge_action;
+
 	/* tuple counters for MERGE */
 	double		mt_merge_inserted;
 	double		mt_merge_updated;
diff --git a/src/include/nodes/parsenodes.h b/src/include/nodes/parsenodes.h
new file mode 100644
index 2565348..ac0a754
--- a/src/include/nodes/parsenodes.h
+++ b/src/include/nodes/parsenodes.h
@@ -1688,6 +1688,7 @@ typedef struct MergeWhenClause
 typedef struct MergeAction
 {
 	NodeTag		type;
+	int			index;			/* 1-based index of the clause */
 	bool		matched;		/* true=MATCHED, false=NOT MATCHED */
 	CmdType		commandType;	/* INSERT/UPDATE/DELETE/DO NOTHING */
 	/* OVERRIDING clause */
@@ -1951,6 +1952,7 @@ typedef struct MergeStmt
 	Node	   *sourceRelation; /* source relation */
 	Node	   *joinCondition;	/* join condition between source and target */
 	List	   *mergeWhenClauses;	/* list of MergeWhenClause(es) */
+	List	   *returningList;	/* list of expressions to return */
 	WithClause *withClause;		/* WITH clause */
 } MergeStmt;
 
diff --git a/src/include/optimizer/paramassign.h b/src/include/optimizer/paramassign.h
new file mode 100644
index 55c27a6..6f70c78
--- a/src/include/optimizer/paramassign.h
+++ b/src/include/optimizer/paramassign.h
@@ -20,6 +20,8 @@ extern Param *replace_outer_placeholderv
 										   PlaceHolderVar *phv);
 extern Param *replace_outer_agg(PlannerInfo *root, Aggref *agg);
 extern Param *replace_outer_grouping(PlannerInfo *root, GroupingFunc *grp);
+extern Param *replace_outer_merge_support_function(PlannerInfo *root,
+												   FuncExpr *func);
 extern Param *replace_nestloop_param_var(PlannerInfo *root, Var *var);
 extern Param *replace_nestloop_param_placeholdervar(PlannerInfo *root,
 													PlaceHolderVar *phv);
diff --git a/src/include/parser/analyze.h b/src/include/parser/analyze.h
new file mode 100644
index 1cef183..78eb55e
--- a/src/include/parser/analyze.h
+++ b/src/include/parser/analyze.h
@@ -44,6 +44,8 @@ extern List *transformInsertRow(ParseSta
 								bool strip_indirection);
 extern List *transformUpdateTargetList(ParseState *pstate,
 									   List *origTlist);
+extern List *transformReturningList(ParseState *pstate, List *returningList,
+									ParseExprKind exprKind);
 extern Query *transformTopLevelStmt(ParseState *pstate, RawStmt *parseTree);
 extern Query *transformStmt(ParseState *pstate, Node *parseTree);
 
diff --git a/src/include/parser/parse_node.h b/src/include/parser/parse_node.h
new file mode 100644
index f589112..836819d
--- a/src/include/parser/parse_node.h
+++ b/src/include/parser/parse_node.h
@@ -61,7 +61,8 @@ typedef enum ParseExprKind
 	EXPR_KIND_DISTINCT_ON,		/* DISTINCT ON */
 	EXPR_KIND_LIMIT,			/* LIMIT */
 	EXPR_KIND_OFFSET,			/* OFFSET */
-	EXPR_KIND_RETURNING,		/* RETURNING */
+	EXPR_KIND_RETURNING,		/* RETURNING in INSERT/UPDATE/DELETE */
+	EXPR_KIND_MERGE_RETURNING,	/* RETURNING in MERGE */
 	EXPR_KIND_VALUES,			/* VALUES */
 	EXPR_KIND_VALUES_SINGLE,	/* single-row VALUES (in INSERT only) */
 	EXPR_KIND_CHECK_CONSTRAINT, /* CHECK constraint for a table */
diff --git a/src/pl/plpgsql/src/pl_exec.c b/src/pl/plpgsql/src/pl_exec.c
new file mode 100644
index 4b76f76..fb6289c
--- a/src/pl/plpgsql/src/pl_exec.c
+++ b/src/pl/plpgsql/src/pl_exec.c
@@ -4267,9 +4267,9 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 	/*
 	 * If we have INTO, then we only need one row back ... but if we have INTO
 	 * STRICT or extra check too_many_rows, ask for two rows, so that we can
-	 * verify the statement returns only one.  INSERT/UPDATE/DELETE are always
-	 * treated strictly. Without INTO, just run the statement to completion
-	 * (tcount = 0).
+	 * verify the statement returns only one.  INSERT/UPDATE/DELETE/MERGE are
+	 * always treated strictly. Without INTO, just run the statement to
+	 * completion (tcount = 0).
 	 *
 	 * We could just ask for two rows always when using INTO, but there are
 	 * some cases where demanding the extra row costs significant time, eg by
@@ -4307,10 +4307,11 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 			Assert(stmt->mod_stmt);
 			exec_set_found(estate, (SPI_processed != 0));
 			break;
@@ -4489,10 +4490,11 @@ exec_stmt_dynexecute(PLpgSQL_execstate *
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 		case SPI_OK_UTILITY:
 		case SPI_OK_REWRITTEN:
 			break;
diff --git a/src/pl/tcl/pltcl.c b/src/pl/tcl/pltcl.c
new file mode 100644
index e8f9d7b..0c9f81e
--- a/src/pl/tcl/pltcl.c
+++ b/src/pl/tcl/pltcl.c
@@ -2456,6 +2456,7 @@ pltcl_process_SPI_result(Tcl_Interp *int
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
+		case SPI_OK_MERGE_RETURNING:
 
 			/*
 			 * Process the tuples we got
diff --git a/src/test/regress/expected/merge.out b/src/test/regress/expected/merge.out
new file mode 100644
index 133d421..5be3d2d
--- a/src/test/regress/expected/merge.out
+++ b/src/test/regress/expected/merge.out
@@ -123,20 +123,20 @@ ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
 ERROR:  name "target" specified more than once
 DETAIL:  The name is used both as MERGE target table and data source.
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
-ERROR:  MERGE not supported in WITH query
-LINE 1: WITH foo AS (
-             ^
--- used in COPY
+ERROR:  WITH query "foo" does not have a RETURNING clause
+LINE 4: ) SELECT * FROM foo;
+                        ^
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) TO stdout;
-ERROR:  MERGE not supported in COPY
+ERROR:  COPY query must have a RETURNING clause
 -- unsupported relation types
 -- view
 CREATE VIEW tv AS SELECT * FROM target;
@@ -1256,7 +1256,7 @@ BEGIN;
 MERGE INTO sq_target
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1271,7 +1271,7 @@ INSERT INTO sq_source (sid, balance, del
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1295,7 +1295,7 @@ WITH targq AS (
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1304,20 +1304,193 @@ WHEN MATCHED AND tid < 2 THEN
 ROLLBACK;
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
+CREATE TABLE merge_actions(action text, abbrev text);
+INSERT INTO merge_actions VALUES ('INSERT', 'ins'), ('UPDATE', 'upd'), ('DELETE', 'del');
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
-ERROR:  syntax error at or near "RETURNING"
-LINE 10: RETURNING *;
-         ^
+    DELETE
+RETURNING pg_merge_when_clause_number() AS clause_number,
+          (SELECT abbrev FROM merge_actions
+            WHERE action = pg_merge_action()) AS merge_action,
+          t.*,
+          CASE pg_merge_action()
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ clause_number | merge_action | tid | balance |     description     
+---------------+--------------+-----+---------+---------------------
+             3 | del          |   1 |     100 | Removed (1,100)
+             1 | upd          |   2 |     220 | Added 20 to balance
+             2 | ins          |   4 |      40 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- error when using MERGE support functions outside MERGE
+SELECT pg_merge_action() FROM sq_target;
+ERROR:  merge support function pg_merge_action can only be called from the RETURNING list of a MERGE command
+LINE 1: SELECT pg_merge_action() FROM sq_target;
+               ^
+UPDATE sq_target SET balance = balance + 1 RETURNING pg_merge_when_clause_number();
+ERROR:  merge support function pg_merge_when_clause_number can only be called from the RETURNING list of a MERGE command
+LINE 1: ...ATE sq_target SET balance = balance + 1 RETURNING pg_merge_w...
+                                                             ^
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_when_clause_number() AS clause_number,
+              pg_merge_action() AS merge_action,
+              t.*,
+              CASE pg_merge_action()
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING merge_action, pg_merge_action() AS merge_log_action, l.*
+)
+SELECT * FROM m2;
+ merge_action | merge_log_action | tid |     last_change     
+--------------+------------------+-----+---------------------
+ DELETE       | UPDATE           |   1 | Removed (1,100)
+ UPDATE       | INSERT           |   2 | Added 20 to balance
+ INSERT       | INSERT           |   4 | Inserted (4,40)
+(3 rows)
+
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ tid |     last_change     
+-----+---------------------
+   1 | Removed (1,100)
+   2 | Added 20 to balance
+   4 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*
+) TO stdout;
+DELETE	1	100
+UPDATE	2	220
+INSERT	4	40
+ROLLBACK;
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ action | tid | new_balance 
+--------+-----+-------------
+ DELETE |   1 |         100
+ UPDATE |   3 |         320
+ INSERT |   4 |         110
+(3 rows)
+
+ROLLBACK;
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ action | tid | balance 
+--------+-----+---------
+ DELETE |   1 |     100
+ UPDATE |   2 |     220
+ INSERT |   4 |      40
+(3 rows)
+
+ROLLBACK;
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ r_action | r_tid | r_balance 
+----------+-------+-----------
+ DELETE   |     1 |       100
+ UPDATE   |     3 |       320
+ INSERT   |     4 |       110
+(3 rows)
+
 ROLLBACK;
 -- EXPLAIN
 CREATE TABLE ex_mtarget (a int, b int)
@@ -1514,7 +1687,7 @@ SELECT * FROM sq_target WHERE tid = 1;
 (1 row)
 
 ROLLBACK;
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 NOTICE:  drop cascades to view v
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1640,6 +1813,32 @@ SELECT * FROM pa_target ORDER BY tid;
 (14 rows)
 
 ROLLBACK;
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING pg_merge_action(), pg_merge_when_clause_number(), t.*;
+ pg_merge_action | pg_merge_when_clause_number | tid | balance |           val            
+-----------------+-----------------------------+-----+---------+--------------------------
+ UPDATE          |                           1 |   2 |     110 | initial updated by merge
+(1 row)
+
+SELECT * FROM pa_target ORDER BY tid;
+ tid | balance |           val            
+-----+---------+--------------------------
+   2 |     110 | initial updated by merge
+   3 |     300 | initial
+   5 |     500 | initial
+   7 |     700 | initial
+   9 |     900 | initial
+  11 |    1100 | initial
+  13 |    1300 | initial
+(7 rows)
+
+ROLLBACK;
 DROP TABLE pa_target CASCADE;
 -- The target table is partitioned in the same way, but this time by attaching
 -- partitions which have columns in different order, dropped columns etc.
@@ -1731,7 +1930,26 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (sid, delta, 'inserted by merge');
+    INSERT VALUES (sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause_number(), t.*;
+ pg_merge_action | pg_merge_when_clause_number | tid | balance |           val            
+-----------------+-----------------------------+-----+---------+--------------------------
+ UPDATE          |                           1 |   2 |     110 | initial updated by merge
+ INSERT          |                           2 |   2 |      20 | inserted by merge
+ UPDATE          |                           1 |   4 |     330 | initial updated by merge
+ INSERT          |                           2 |   4 |      40 | inserted by merge
+ UPDATE          |                           1 |   6 |     550 | initial updated by merge
+ INSERT          |                           2 |   6 |      60 | inserted by merge
+ UPDATE          |                           1 |   8 |     770 | initial updated by merge
+ INSERT          |                           2 |   8 |      80 | inserted by merge
+ UPDATE          |                           1 |  10 |     990 | initial updated by merge
+ INSERT          |                           2 |  10 |     100 | inserted by merge
+ UPDATE          |                           1 |  12 |    1210 | initial updated by merge
+ INSERT          |                           2 |  12 |     120 | inserted by merge
+ UPDATE          |                           1 |  14 |    1430 | initial updated by merge
+ INSERT          |                           2 |  14 |     140 | inserted by merge
+(14 rows)
+
 SELECT * FROM pa_target ORDER BY tid;
  tid | balance |           val            
 -----+---------+--------------------------
@@ -1798,7 +2016,21 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause_number(), t.*;
+ pg_merge_action | pg_merge_when_clause_number |          logts           | tid | balance |           val            
+-----------------+-----------------------------+--------------------------+-----+---------+--------------------------
+ UPDATE          |                           1 | Tue Jan 31 00:00:00 2017 |   1 |     110 | initial updated by merge
+ UPDATE          |                           1 | Tue Feb 28 00:00:00 2017 |   2 |     220 | initial updated by merge
+ INSERT          |                           2 | Sun Jan 15 00:00:00 2017 |   3 |      30 | inserted by merge
+ UPDATE          |                           1 | Tue Jan 31 00:00:00 2017 |   4 |     440 | initial updated by merge
+ UPDATE          |                           1 | Tue Feb 28 00:00:00 2017 |   5 |     550 | initial updated by merge
+ INSERT          |                           2 | Sun Jan 15 00:00:00 2017 |   6 |      60 | inserted by merge
+ UPDATE          |                           1 | Tue Jan 31 00:00:00 2017 |   7 |     770 | initial updated by merge
+ UPDATE          |                           1 | Tue Feb 28 00:00:00 2017 |   8 |     880 | initial updated by merge
+ INSERT          |                           2 | Sun Jan 15 00:00:00 2017 |   9 |      90 | inserted by merge
+(9 rows)
+
 SELECT * FROM pa_target ORDER BY tid;
           logts           | tid | balance |           val            
 --------------------------+-----+---------+--------------------------
diff --git a/src/test/regress/expected/rowsecurity.out b/src/test/regress/expected/rowsecurity.out
new file mode 100644
index 97ca9bf..55250b3
--- a/src/test/regress/expected/rowsecurity.out
+++ b/src/test/regress/expected/rowsecurity.out
@@ -2300,6 +2300,35 @@ WHEN MATCHED THEN
 	UPDATE SET dnotes = dnotes || ' notes added by merge8 '
 WHEN NOT MATCHED THEN
 	INSERT VALUES (13, 44, 1, 'regress_rls_bob', 'new manga');
+SELECT * FROM document WHERE did = 13;
+ did | cid | dlevel | dauthor | dtitle | dnotes 
+-----+-----+--------+---------+--------+--------
+(0 rows)
+
+-- but not OK if RETURNING is used
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge9 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 44, 1, 'regress_rls_bob', 'new manga')
+RETURNING *;
+ERROR:  new row violates row-level security policy for table "document"
+-- but OK if new row is visible
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge10 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 11, 1, 'regress_rls_bob', 'new novel')
+RETURNING *;
+ sdid | did | cid | dlevel |     dauthor     |  dtitle   | dnotes 
+------+-----+-----+--------+-----------------+-----------+--------
+   14 |  14 |  11 |      1 | regress_rls_bob | new novel | 
+(1 row)
+
 RESET SESSION AUTHORIZATION;
 -- drop the restrictive SELECT policy so that we can look at the
 -- final state of the table
@@ -2323,7 +2352,8 @@ SELECT * FROM document;
   12 |  11 |      1 | regress_rls_bob   | another novel                    | 
    1 |  11 |      1 | regress_rls_bob   | my first novel                   |  notes added by merge2  notes added by merge3  notes added by merge4  notes added by merge7 
   13 |  44 |      1 | regress_rls_bob   | new manga                        | 
-(15 rows)
+  14 |  11 |      1 | regress_rls_bob   | new novel                        | 
+(16 rows)
 
 --
 -- ROLE/GROUP
diff --git a/src/test/regress/expected/rules.out b/src/test/regress/expected/rules.out
new file mode 100644
index 5058be5..2adfad7
--- a/src/test/regress/expected/rules.out
+++ b/src/test/regress/expected/rules.out
@@ -3596,7 +3596,8 @@ MERGE INTO rule_merge2 t USING (SELECT 1
 -- test deparsing
 CREATE TABLE sf_target(id int, data text, filling int[]);
 CREATE FUNCTION merge_sf_test()
- RETURNS void
+ RETURNS TABLE(pg_merge_action text, when_clause int,
+               a int, b text, id int, data text, filling int[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -3633,11 +3634,13 @@ WHEN NOT MATCHED
    VALUES (s.a, s.b, DEFAULT)
 WHEN NOT MATCHED
    THEN INSERT (filling[1], id)
-   VALUES (s.a, s.a);
+   VALUES (s.a, s.a)
+RETURNING
+   pg_merge_action(), pg_merge_when_clause_number() AS clause_number, *;
 END;
 \sf merge_sf_test
 CREATE OR REPLACE FUNCTION public.merge_sf_test()
- RETURNS void
+ RETURNS TABLE(pg_merge_action text, when_clause integer, a integer, b text, id integer, data text, filling integer[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -3674,7 +3677,14 @@ BEGIN ATOMIC
       VALUES (s.a, s.b, DEFAULT)
     WHEN NOT MATCHED
      THEN INSERT (filling[1], id)
-      VALUES (s.a, s.a);
+      VALUES (s.a, s.a)
+   RETURNING pg_merge_action() AS pg_merge_action,
+     pg_merge_when_clause_number() AS clause_number,
+     s.a,
+     s.b,
+     t.id,
+     t.data,
+     t.filling;
 END
 DROP FUNCTION merge_sf_test;
 DROP TABLE sf_target;
diff --git a/src/test/regress/sql/merge.sql b/src/test/regress/sql/merge.sql
new file mode 100644
index 4cf6db9..8394d01
--- a/src/test/regress/sql/merge.sql
+++ b/src/test/regress/sql/merge.sql
@@ -86,12 +86,12 @@ MERGE INTO target
 USING target
 ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
--- used in COPY
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
@@ -816,7 +816,7 @@ BEGIN;
 MERGE INTO sq_target
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -829,7 +829,7 @@ INSERT INTO sq_source (sid, balance, del
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -847,7 +847,7 @@ WITH targq AS (
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -857,17 +857,148 @@ ROLLBACK;
 
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
+CREATE TABLE merge_actions(action text, abbrev text);
+INSERT INTO merge_actions VALUES ('INSERT', 'ins'), ('UPDATE', 'upd'), ('DELETE', 'del');
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
+    DELETE
+RETURNING pg_merge_when_clause_number() AS clause_number,
+          (SELECT abbrev FROM merge_actions
+            WHERE action = pg_merge_action()) AS merge_action,
+          t.*,
+          CASE pg_merge_action()
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ROLLBACK;
+
+-- error when using MERGE support functions outside MERGE
+SELECT pg_merge_action() FROM sq_target;
+UPDATE sq_target SET balance = balance + 1 RETURNING pg_merge_when_clause_number();
+
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_when_clause_number() AS clause_number,
+              pg_merge_action() AS merge_action,
+              t.*,
+              CASE pg_merge_action()
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING merge_action, pg_merge_action() AS merge_log_action, l.*
+)
+SELECT * FROM m2;
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ROLLBACK;
+
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*
+) TO stdout;
+ROLLBACK;
+
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ROLLBACK;
+
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ROLLBACK;
+
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
 ROLLBACK;
 
 -- EXPLAIN
@@ -966,7 +1097,7 @@ WHEN MATCHED THEN
 SELECT * FROM sq_target WHERE tid = 1;
 ROLLBACK;
 
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1033,6 +1164,17 @@ SELECT merge_func();
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING pg_merge_action(), pg_merge_when_clause_number(), t.*;
+SELECT * FROM pa_target ORDER BY tid;
+ROLLBACK;
+
 DROP TABLE pa_target CASCADE;
 
 -- The target table is partitioned in the same way, but this time by attaching
@@ -1091,7 +1233,8 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (sid, delta, 'inserted by merge');
+    INSERT VALUES (sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause_number(), t.*;
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
@@ -1145,7 +1288,8 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause_number(), t.*;
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
diff --git a/src/test/regress/sql/rowsecurity.sql b/src/test/regress/sql/rowsecurity.sql
new file mode 100644
index dec7340..1d5ed0a
--- a/src/test/regress/sql/rowsecurity.sql
+++ b/src/test/regress/sql/rowsecurity.sql
@@ -992,6 +992,27 @@ WHEN MATCHED THEN
 	UPDATE SET dnotes = dnotes || ' notes added by merge8 '
 WHEN NOT MATCHED THEN
 	INSERT VALUES (13, 44, 1, 'regress_rls_bob', 'new manga');
+SELECT * FROM document WHERE did = 13;
+
+-- but not OK if RETURNING is used
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge9 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 44, 1, 'regress_rls_bob', 'new manga')
+RETURNING *;
+
+-- but OK if new row is visible
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge10 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 11, 1, 'regress_rls_bob', 'new novel')
+RETURNING *;
 
 RESET SESSION AUTHORIZATION;
 -- drop the restrictive SELECT policy so that we can look at the
diff --git a/src/test/regress/sql/rules.sql b/src/test/regress/sql/rules.sql
new file mode 100644
index 8b7e255..5dbeb6b
--- a/src/test/regress/sql/rules.sql
+++ b/src/test/regress/sql/rules.sql
@@ -1281,7 +1281,8 @@ MERGE INTO rule_merge2 t USING (SELECT 1
 CREATE TABLE sf_target(id int, data text, filling int[]);
 
 CREATE FUNCTION merge_sf_test()
- RETURNS void
+ RETURNS TABLE(pg_merge_action text, when_clause int,
+               a int, b text, id int, data text, filling int[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -1318,7 +1319,9 @@ WHEN NOT MATCHED
    VALUES (s.a, s.b, DEFAULT)
 WHEN NOT MATCHED
    THEN INSERT (filling[1], id)
-   VALUES (s.a, s.a);
+   VALUES (s.a, s.a)
+RETURNING
+   pg_merge_action(), pg_merge_when_clause_number() AS clause_number, *;
 END;
 
 \sf merge_sf_test

diff --git a/doc/src/sgml/dml.sgml b/doc/src/sgml/dml.sgml
new file mode 100644
index cbbc5e2..7f65f6e
--- a/doc/src/sgml/dml.sgml
+++ b/doc/src/sgml/dml.sgml
@@ -283,10 +283,15 @@ DELETE FROM products;
    <secondary>RETURNING</secondary>
   </indexterm>
 
+  <indexterm zone="dml-returning">
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
   <para>
    Sometimes it is useful to obtain data from modified rows while they are
    being manipulated.  The <command>INSERT</command>, <command>UPDATE</command>,
-   and <command>DELETE</command> commands all have an
+   <command>DELETE</command>, and <command>MERGE</command> commands all have an
    optional <literal>RETURNING</literal> clause that supports this.  Use
    of <literal>RETURNING</literal> avoids performing an extra database query to
    collect the data, and is especially valuable when it would otherwise be
@@ -339,6 +344,24 @@ DELETE FROM products
 </programlisting>
   </para>
 
+  <para>
+   In a <command>MERGE</command>, the data available to <literal>RETURNING</literal> is
+   the content of the source row plus the content of the inserted, updated, or
+   deleted target row.  In addition, the merge support functions
+   <xref linkend="pg-merge-action"/> and <xref linkend="pg-merge-when-clause-number"/>
+   may be used to return information about which merge action was executed for
+   each row.  Since it is quite common for the source and target to have many
+   of the same columns, specifying <literal>RETURNING *</literal> can lead to a
+   lot of duplicated columns, so it is often more useful to just return the
+   target row.  For example:
+<programlisting>
+MERGE INTO products p USING new_products n ON p.product_no = n.product_no
+  WHEN NOT MATCHED THEN INSERT VALUES (n.product_no, n.name, n.price)
+  WHEN MATCHED THEN UPDATE SET name = n.name, price = n.price
+  RETURNING pg_merge_action(), p.*;
+</programlisting>
+  </para>
+
   <para>
    If there are triggers (<xref linkend="triggers"/>) on the target table,
    the data available to <literal>RETURNING</literal> is the row as modified by
diff --git a/doc/src/sgml/func.sgml b/doc/src/sgml/func.sgml
new file mode 100644
index be2f54c..e79e2ad
--- a/doc/src/sgml/func.sgml
+++ b/doc/src/sgml/func.sgml
@@ -21772,6 +21772,100 @@ SELECT count(*) FROM sometable;
 
  </sect1>
 
+ <sect1 id="functions-merge-support">
+  <title>Merge Support Functions</title>
+
+  <indexterm>
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
+  <para>
+   The merge support functions shown in
+   <xref linkend="functions-merge-support-table"/> may be used in the
+   <literal>RETURNING</literal> list of a <xref linkend="sql-merge"/> command
+   to return additional information about the action taken for each row.
+  </para>
+
+  <table id="functions-merge-support-table">
+    <title>Merge Support Functions</title>
+
+    <tgroup cols="1">
+     <thead>
+      <row>
+       <entry role="func_table_entry"><para role="func_signature">
+        Function
+       </para>
+       <para>
+        Description
+       </para></entry>
+      </row>
+     </thead>
+
+     <tbody>
+      <row>
+       <entry role="func_table_entry"><para role="func_signature">
+        <indexterm id="pg-merge-action">
+         <primary>pg_merge_action</primary>
+        </indexterm>
+        <function>pg_merge_action</function> ( )
+        <returnvalue>text</returnvalue>
+       </para>
+       <para>
+        Returns the action performed on the current row (<literal>'INSERT'</literal>,
+        <literal>'UPDATE'</literal>, or <literal>'DELETE'</literal>).
+       </para></entry>
+      </row>
+
+      <row>
+       <entry role="func_table_entry"><para role="func_signature">
+        <indexterm id="pg-merge-when-clause-number">
+         <primary>pg_merge_when_clause_number</primary>
+        </indexterm>
+        <function>pg_merge_when_clause_number</function> ( )
+        <returnvalue>integer</returnvalue>
+       </para>
+       <para>
+        Returns the 1-based index of the <literal>WHEN</literal> clause executed
+        for the current row.
+       </para></entry>
+      </row>
+     </tbody>
+    </tgroup>
+  </table>
+
+  <para>
+   Example:
+<screen><![CDATA[
+MERGE INTO products p
+  USING stock s ON p.product_id = s.product_id
+  WHEN MATCHED AND s.quantity > 0 THEN
+    UPDATE SET in_stock = true, quantity = s.quantity
+  WHEN MATCHED THEN
+    UPDATE SET in_stock = false, quantity = 0
+  WHEN NOT MATCHED THEN
+    INSERT (product_id, in_stock, quantity)
+      VALUES (s.product_id, true, s.quantity)
+  RETURNING pg_merge_action() AS action,
+            pg_merge_when_clause_number() AS clause_number,
+            p.*;
+
+ action | clause_number | product_id | in_stock | quantity 
+--------+---------------+------------+----------+----------
+ UPDATE |             1 |       1001 | t        |       50
+ UPDATE |             2 |       1002 | f        |        0
+ INSERT |             3 |       1003 | t        |       10
+]]></screen>
+  </para>
+
+  <para>
+   Note that these functions can only be used in the <literal>RETURNING</literal>
+   list of a <command>MERGE</command> command.  It is an error to use them in
+   any other part of a query.
+  </para>
+
+ </sect1>
+
  <sect1 id="functions-subquery">
   <title>Subquery Expressions</title>
 
diff --git a/doc/src/sgml/glossary.sgml b/doc/src/sgml/glossary.sgml
new file mode 100644
index fe8def4..51a15ca
--- a/doc/src/sgml/glossary.sgml
+++ b/doc/src/sgml/glossary.sgml
@@ -1387,9 +1387,9 @@
      to a <glossterm linkend="glossary-client">client</glossterm> upon the
      completion of an <acronym>SQL</acronym> command, usually a
      <command>SELECT</command> but it can be an
-     <command>INSERT</command>, <command>UPDATE</command>, or
-     <command>DELETE</command> command if the <literal>RETURNING</literal>
-     clause is specified.
+     <command>INSERT</command>, <command>UPDATE</command>,
+     <command>DELETE</command>, or <command>MERGE</command> command if the
+     <literal>RETURNING</literal> clause is specified.
     </para>
     <para>
      The fact that a result set is a relation means that a query can be used
diff --git a/doc/src/sgml/plpgsql.sgml b/doc/src/sgml/plpgsql.sgml
new file mode 100644
index f55e901..6803240
--- a/doc/src/sgml/plpgsql.sgml
+++ b/doc/src/sgml/plpgsql.sgml
@@ -1020,8 +1020,8 @@ INSERT INTO mytable VALUES (1,'one'), (2
     </para>
 
     <para>
-     If the command does return rows (for example <command>SELECT</command>,
-     or <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>
+     If the command does return rows (for example <command>SELECT</command>, or
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
      with <literal>RETURNING</literal>), there are two ways to proceed.
      When the command will return at most one row, or you only care about
      the first row of output, write the command as usual but add
@@ -1149,6 +1149,7 @@ SELECT <replaceable>select_expressions</
 INSERT ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 UPDATE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 DELETE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
+MERGE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 </synopsis>
 
      where <replaceable>target</replaceable> can be a record variable, a row
@@ -1159,8 +1160,8 @@ DELETE ... RETURNING <replaceable>expres
      <literal>INTO</literal> clause) just as described above,
      and the plan is cached in the same way.
      This works for <command>SELECT</command>,
-     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
-     <literal>RETURNING</literal>, and certain utility commands
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
+     with <literal>RETURNING</literal>, and certain utility commands
      that return row sets, such as <command>EXPLAIN</command>.
      Except for the <literal>INTO</literal> clause, the SQL command is the same
      as it would be written outside <application>PL/pgSQL</application>.
@@ -1236,7 +1237,7 @@ END;
     </para>
 
     <para>
-     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
+     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command> with
      <literal>RETURNING</literal>, <application>PL/pgSQL</application> reports
      an error for more than one returned row, even when
      <literal>STRICT</literal> is not specified.  This is because there
diff --git a/doc/src/sgml/queries.sgml b/doc/src/sgml/queries.sgml
new file mode 100644
index 3f95849..9928373
--- a/doc/src/sgml/queries.sgml
+++ b/doc/src/sgml/queries.sgml
@@ -2062,9 +2062,10 @@ SELECT <replaceable>select_list</replace
    Table Expressions or <acronym>CTE</acronym>s, can be thought of as defining
    temporary tables that exist just for one query.  Each auxiliary statement
    in a <literal>WITH</literal> clause can be a <command>SELECT</command>,
-   <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>; and the
+   <command>INSERT</command>, <command>UPDATE</command>, <command>DELETE</command>,
+   or <command>MERGE</command>; and the
    <literal>WITH</literal> clause itself is attached to a primary statement that can
-   be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
+   also be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
    <command>DELETE</command>, or <command>MERGE</command>.
   </para>
 
@@ -2598,8 +2599,8 @@ SELECT * FROM w AS w1 JOIN w AS w2 ON w1
    <title>Data-Modifying Statements in <literal>WITH</literal></title>
 
    <para>
-    You can use most data-modifying statements (<command>INSERT</command>,
-    <command>UPDATE</command>, or <command>DELETE</command>, but not
+    You can use data-modifying statements (<command>INSERT</command>,
+    <command>UPDATE</command>, <command>DELETE</command>, or
     <command>MERGE</command>) in <literal>WITH</literal>.  This
     allows you to perform several different operations in the same query.
     An example is:
diff --git a/doc/src/sgml/ref/copy.sgml b/doc/src/sgml/ref/copy.sgml
new file mode 100644
index 4d614a0..efaacb9
--- a/doc/src/sgml/ref/copy.sgml
+++ b/doc/src/sgml/ref/copy.sgml
@@ -122,13 +122,16 @@ COPY { <replaceable class="parameter">ta
       A <link linkend="sql-select"><command>SELECT</command></link>,
       <link linkend="sql-values"><command>VALUES</command></link>,
       <link linkend="sql-insert"><command>INSERT</command></link>,
-      <link linkend="sql-update"><command>UPDATE</command></link>, or
-      <link linkend="sql-delete"><command>DELETE</command></link> command whose results are to be
-      copied.  Note that parentheses are required around the query.
+      <link linkend="sql-update"><command>UPDATE</command></link>,
+      <link linkend="sql-delete"><command>DELETE</command></link>, or
+      <link linkend="sql-merge"><command>MERGE</command></link> command
+      whose results are to be copied.  Note that parentheses are required
+      around the query.
      </para>
      <para>
-      For <command>INSERT</command>, <command>UPDATE</command> and
-      <command>DELETE</command> queries a RETURNING clause must be provided,
+      For <command>INSERT</command>, <command>UPDATE</command>,
+      <command>DELETE</command>, and <command>MERGE</command> queries a
+      RETURNING clause must be provided,
       and the target relation must not have a conditional rule, nor
       an <literal>ALSO</literal> rule, nor an <literal>INSTEAD</literal> rule
       that expands to multiple statements.
diff --git a/doc/src/sgml/ref/merge.sgml b/doc/src/sgml/ref/merge.sgml
new file mode 100644
index 0995fe0..d1aa937
--- a/doc/src/sgml/ref/merge.sgml
+++ b/doc/src/sgml/ref/merge.sgml
@@ -25,6 +25,7 @@ PostgreSQL documentation
 MERGE INTO [ ONLY ] <replaceable class="parameter">target_table_name</replaceable> [ * ] [ [ AS ] <replaceable class="parameter">target_alias</replaceable> ]
 USING <replaceable class="parameter">data_source</replaceable> ON <replaceable class="parameter">join_condition</replaceable>
 <replaceable class="parameter">when_clause</replaceable> [...]
+[ RETURNING * | <replaceable class="parameter">output_expression</replaceable> [ [ AS ] <replaceable class="parameter">output_name</replaceable> ] [, ...] ]
 
 <phrase>where <replaceable class="parameter">data_source</replaceable> is:</phrase>
 
@@ -95,6 +96,19 @@ DELETE
   </para>
 
   <para>
+   The optional <literal>RETURNING</literal> clause causes <command>MERGE</command>
+   to compute and return value(s) based on each row inserted, updated, or
+   deleted.  Any expression using the source or target table's columns, or the
+   merge support functions <xref linkend="pg-merge-action"/> and
+   <xref linkend="pg-merge-when-clause-number"/> can be computed.  When an
+   <command>INSERT</command> or <command>UPDATE</command> action is performed,
+   the new values of the target table's columns are used.  When a
+   <command>DELETE</command> is performed, the old values of the target table's
+   columns are used.  The syntax of the <literal>RETURNING</literal> list is
+   identical to that of the output list of <command>SELECT</command>.
+  </para>
+
+  <para>
    There is no separate <literal>MERGE</literal> privilege.
    If you specify an update action, you must have the
    <literal>UPDATE</literal> privilege on the column(s)
@@ -125,6 +139,18 @@ DELETE
 
   <variablelist>
    <varlistentry>
+    <term><replaceable class="parameter">with_query</replaceable></term>
+    <listitem>
+     <para>
+      The <literal>WITH</literal> clause allows you to specify one or more
+      subqueries that can be referenced by name in the <command>MERGE</command>
+      query. See <xref linkend="queries-with"/> and <xref linkend="sql-select"/>
+      for details.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
     <term><replaceable class="parameter">target_table_name</replaceable></term>
     <listitem>
      <para>
@@ -399,13 +425,31 @@ DELETE
    </varlistentry>
 
    <varlistentry>
-    <term><replaceable class="parameter">with_query</replaceable></term>
+    <term><replaceable class="parameter">output_expression</replaceable></term>
     <listitem>
      <para>
-      The <literal>WITH</literal> clause allows you to specify one or more
-      subqueries that can be referenced by name in the <command>MERGE</command>
-      query. See <xref linkend="queries-with"/> and <xref linkend="sql-select"/>
-      for details.
+      An expression to be computed and returned by the <command>MERGE</command>
+      command after each row is merged.  The expression can use any columns of
+      the source or target tables, or the merge support functions
+      <xref linkend="pg-merge-action"/> and
+      <xref linkend="pg-merge-when-clause-number"/>.
+     </para>
+     <para>
+      Writing <literal>*</literal> will return all columns from the source
+      table, followed by all columns from the target table.  Often this will
+      lead to a lot of duplication, since it is common for the source and
+      target tables to have a lot of the same columns.  This can be avoided by
+      qualifying the <literal>*</literal> with the name or alias of the source
+      or target table.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
+    <term><replaceable class="parameter">output_name</replaceable></term>
+    <listitem>
+     <para>
+      A name to use for a returned column.
      </para>
     </listitem>
    </varlistentry>
@@ -428,6 +472,13 @@ MERGE <replaceable class="parameter">tot
    were changed in any way.
   </para>
 
+  <para>
+   If the <command>MERGE</command> command contains a <literal>RETURNING</literal>
+   clause, the result will be similar to that of a <command>SELECT</command>
+   statement containing the columns and values defined in the
+   <literal>RETURNING</literal> list, computed over the row(s) inserted, updated,
+   or deleted by the command.
+  </para>
  </refsect1>
 
  <refsect1>
@@ -544,13 +595,6 @@ MERGE <replaceable class="parameter">tot
   </para>
 
   <para>
-   There is no <literal>RETURNING</literal> clause with
-   <command>MERGE</command>.  Actions of <command>INSERT</command>,
-   <command>UPDATE</command> and <command>DELETE</command> cannot contain
-   <literal>RETURNING</literal> or <literal>WITH</literal> clauses.
-  </para>
-
-  <para>
    When <command>MERGE</command> is run concurrently with other commands
    that modify the target table, the usual transaction isolation rules
    apply; see <xref linkend="transaction-iso"/> for an explanation
@@ -602,7 +646,8 @@ WHEN NOT MATCHED THEN
   <para>
    Attempt to insert a new stock item along with the quantity of stock. If
    the item already exists, instead update the stock count of the existing
-   item. Don't allow entries that have zero stock.
+   item. Don't allow entries that have zero stock. Return details of all
+   changes made.
 <programlisting>
 MERGE INTO wines w
 USING wine_stock_changes s
@@ -612,7 +657,8 @@ WHEN NOT MATCHED AND s.stock_delta > 0 T
 WHEN MATCHED AND w.stock + s.stock_delta > 0 THEN
   UPDATE SET stock = w.stock + s.stock_delta
 WHEN MATCHED THEN
-  DELETE;
+  DELETE
+RETURNING pg_merge_action(), w.*;
 </programlisting>
 
    The <literal>wine_stock_changes</literal> table might be, for example, a
@@ -627,7 +673,8 @@ WHEN MATCHED THEN
     This command conforms to the <acronym>SQL</acronym> standard.
   </para>
    <para>
-    The WITH clause and <literal>DO NOTHING</literal> action are extensions to
+    The WITH clause, <literal>DO NOTHING</literal> action, and
+    <literal>RETURNING</literal> clause are extensions to
     the <acronym>SQL</acronym> standard.
   </para>
  </refsect1>
diff --git a/doc/src/sgml/ref/select.sgml b/doc/src/sgml/ref/select.sgml
new file mode 100644
index 0ee0cc7..f81bdea
--- a/doc/src/sgml/ref/select.sgml
+++ b/doc/src/sgml/ref/select.sgml
@@ -74,7 +74,7 @@ SELECT [ ALL | DISTINCT [ ON ( <replacea
 
 <phrase>and <replaceable class="parameter">with_query</replaceable> is:</phrase>
 
-    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> )
+    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> | <replaceable class="parameter">merge</replaceable> )
         [ SEARCH { BREADTH | DEPTH } FIRST BY <replaceable>column_name</replaceable> [, ...] SET <replaceable>search_seq_col_name</replaceable> ]
         [ CYCLE <replaceable>column_name</replaceable> [, ...] SET <replaceable>cycle_mark_col_name</replaceable> [ TO <replaceable>cycle_mark_value</replaceable> DEFAULT <replaceable>cycle_mark_default</replaceable> ] USING <replaceable>cycle_path_col_name</replaceable> ]
 
@@ -227,10 +227,10 @@ TABLE [ ONLY ] <replaceable class="param
     The subqueries effectively act as temporary tables or views
     for the duration of the primary query.
     Each subquery can be a <command>SELECT</command>, <command>TABLE</command>, <command>VALUES</command>,
-    <command>INSERT</command>, <command>UPDATE</command> or
-    <command>DELETE</command> statement.
+    <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command> statement.
     When writing a data-modifying statement (<command>INSERT</command>,
-    <command>UPDATE</command> or <command>DELETE</command>) in
+    <command>UPDATE</command>, <command>DELETE</command>, or <command>MERGE</command>) in
     <literal>WITH</literal>, it is usual to include a <literal>RETURNING</literal> clause.
     It is the output of <literal>RETURNING</literal>, <emphasis>not</emphasis> the underlying
     table that the statement modifies, that forms the temporary table that is
@@ -2179,7 +2179,8 @@ SELECT 2+2;
 
    <para>
     <productname>PostgreSQL</productname> allows <command>INSERT</command>,
-    <command>UPDATE</command>, and <command>DELETE</command> to be used as <literal>WITH</literal>
+    <command>UPDATE</command>, <command>DELETE</command>, and
+    <command>MERGE</command> to be used as <literal>WITH</literal>
     queries.  This is not found in the SQL standard.
    </para>
   </refsect2>
diff --git a/doc/src/sgml/rowtypes.sgml b/doc/src/sgml/rowtypes.sgml
new file mode 100644
index 4d86f97..bbeac84
--- a/doc/src/sgml/rowtypes.sgml
+++ b/doc/src/sgml/rowtypes.sgml
@@ -348,7 +348,7 @@ SELECT m.* FROM some_table, LATERAL myfu
    column expansion of this kind when it appears at the top level of
    a <link linkend="queries-select-lists"><command>SELECT</command> output
    list</link>, a <link linkend="dml-returning"><literal>RETURNING</literal>
-   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>,
+   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>,
    a <link linkend="queries-values"><literal>VALUES</literal> clause</link>, or
    a <link linkend="sql-syntax-row-constructors">row constructor</link>.
    In all other contexts (including when nested inside one of those
diff --git a/doc/src/sgml/spi.sgml b/doc/src/sgml/spi.sgml
new file mode 100644
index 16c3ab7..6410862
--- a/doc/src/sgml/spi.sgml
+++ b/doc/src/sgml/spi.sgml
@@ -301,8 +301,9 @@ SPI_execute("INSERT INTO foo SELECT * FR
    is returned in the global variable <varname>SPI_processed</varname>.
    If the return value of the function is <symbol>SPI_OK_SELECT</symbol>,
    <symbol>SPI_OK_INSERT_RETURNING</symbol>,
-   <symbol>SPI_OK_DELETE_RETURNING</symbol>, or
-   <symbol>SPI_OK_UPDATE_RETURNING</symbol>,
+   <symbol>SPI_OK_DELETE_RETURNING</symbol>,
+   <symbol>SPI_OK_UPDATE_RETURNING</symbol>, or
+   <symbol>SPI_OK_MERGE_RETURNING</symbol>,
    then you can use the
    global pointer <literal>SPITupleTable *SPI_tuptable</literal> to
    access the result rows.  Some utility commands (such as
@@ -472,6 +473,15 @@ typedef struct SPITupleTable
       </para>
      </listitem>
     </varlistentry>
+
+    <varlistentry>
+     <term><symbol>SPI_OK_MERGE_RETURNING</symbol></term>
+     <listitem>
+      <para>
+       if a <command>MERGE RETURNING</command> was executed
+      </para>
+     </listitem>
+    </varlistentry>
 
     <varlistentry>
      <term><symbol>SPI_OK_UTILITY</symbol></term>
diff --git a/doc/src/sgml/xfunc.sgml b/doc/src/sgml/xfunc.sgml
new file mode 100644
index 281c178..3aec197
--- a/doc/src/sgml/xfunc.sgml
+++ b/doc/src/sgml/xfunc.sgml
@@ -177,7 +177,8 @@
     statements separated by semicolons.  A semicolon after the last
     statement is optional.  Unless the function is declared to return
     <type>void</type>, the last statement must be a <command>SELECT</command>,
-    or an <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>
+    or an <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command>
     that has a <literal>RETURNING</literal> clause.
    </para>
 
@@ -1236,8 +1237,9 @@ SELECT x, CASE WHEN x &gt; 0 THEN genera
 
     <note>
      <para>
-      If a function's last command is <command>INSERT</command>, <command>UPDATE</command>,
-      or <command>DELETE</command> with <literal>RETURNING</literal>, that command will
+      If a function's last command is <command>INSERT</command>,
+      <command>UPDATE</command>, <command>DELETE</command>, or
+      <command>MERGE</command> with <literal>RETURNING</literal>, that command will
       always be executed to completion, even if the function is not declared
       with <literal>SETOF</literal> or the calling query does not fetch all the
       result rows.  Any extra rows produced by the <literal>RETURNING</literal>
diff --git a/src/backend/commands/copy.c b/src/backend/commands/copy.c
new file mode 100644
index f14fae3..05f2e5b
--- a/src/backend/commands/copy.c
+++ b/src/backend/commands/copy.c
@@ -283,12 +283,6 @@ DoCopy(ParseState *pstate, const CopyStm
 	{
 		Assert(stmt->query);
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(stmt->query, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in COPY"));
-
 		query = makeNode(RawStmt);
 		query->stmt = stmt->query;
 		query->stmt_location = stmt_location;
diff --git a/src/backend/commands/copyto.c b/src/backend/commands/copyto.c
new file mode 100644
index eaa3172..a50cd30
--- a/src/backend/commands/copyto.c
+++ b/src/backend/commands/copyto.c
@@ -510,7 +510,8 @@ BeginCopyTo(ParseState *pstate,
 		{
 			Assert(query->commandType == CMD_INSERT ||
 				   query->commandType == CMD_UPDATE ||
-				   query->commandType == CMD_DELETE);
+				   query->commandType == CMD_DELETE ||
+				   query->commandType == CMD_MERGE);
 
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
diff --git a/src/backend/executor/execExpr.c b/src/backend/executor/execExpr.c
new file mode 100644
index 2c62b0c..f09d512
--- a/src/backend/executor/execExpr.c
+++ b/src/backend/executor/execExpr.c
@@ -48,6 +48,7 @@
 #include "utils/array.h"
 #include "utils/builtins.h"
 #include "utils/datum.h"
+#include "utils/fmgroids.h"
 #include "utils/jsonfuncs.h"
 #include "utils/lsyscache.h"
 #include "utils/typcache.h"
@@ -2620,6 +2621,22 @@ ExecInitFunc(ExprEvalStep *scratch, Expr
 	InitFunctionCallInfoData(*fcinfo, flinfo,
 							 nargs, inputcollid, NULL, NULL);
 
+	/*
+	 * Merge support functions should only be called from a MERGE command, and
+	 * need access to the parent ModifyTableState.  The parser should have
+	 * checked that such functions only appear in the RETURNING list of a
+	 * MERGE, so this should never fail.
+	 */
+	if (IsMergeSupportFunction(funcid))
+	{
+		if (!state->parent ||
+			!IsA(state->parent, ModifyTableState) ||
+			((ModifyTableState *) state->parent)->operation != CMD_MERGE)
+			elog(ERROR, "merge support function called in non-merge context");
+
+		fcinfo->context = (Node *) state->parent;
+	}
+
 	/* Keep extra copies of this info to save an indirection at runtime */
 	scratch->d.func.fn_addr = flinfo->fn_addr;
 	scratch->d.func.nargs = nargs;
diff --git a/src/backend/executor/execPartition.c b/src/backend/executor/execPartition.c
new file mode 100644
index eb8a87f..b9bd03c
--- a/src/backend/executor/execPartition.c
+++ b/src/backend/executor/execPartition.c
@@ -612,8 +612,8 @@ ExecInitPartitionInfo(ModifyTableState *
 	 * Build the RETURNING projection for the partition.  Note that we didn't
 	 * build the returningList for partitions within the planner, but simple
 	 * translation of varattnos will suffice.  This only occurs for the INSERT
-	 * case or in the case of UPDATE tuple routing where we didn't find a
-	 * result rel to reuse.
+	 * case or in the case of UPDATE/MERGE tuple routing where we didn't find
+	 * a result rel to reuse.
 	 */
 	if (node && node->returningLists != NIL)
 	{
@@ -622,12 +622,14 @@ ExecInitPartitionInfo(ModifyTableState *
 		List	   *returningList;
 
 		/* See the comment above for WCO lists. */
-		/* (except no RETURNING support for MERGE yet) */
 		Assert((node->operation == CMD_INSERT &&
 				list_length(node->returningLists) == 1 &&
 				list_length(node->resultRelations) == 1) ||
 			   (node->operation == CMD_UPDATE &&
 				list_length(node->returningLists) ==
+				list_length(node->resultRelations)) ||
+			   (node->operation == CMD_MERGE &&
+				list_length(node->returningLists) ==
 				list_length(node->resultRelations)));
 
 		/*
diff --git a/src/backend/executor/functions.c b/src/backend/executor/functions.c
new file mode 100644
index f55424e..97874a4
--- a/src/backend/executor/functions.c
+++ b/src/backend/executor/functions.c
@@ -1660,8 +1660,8 @@ check_sql_fn_retval(List *queryTreeLists
 
 	/*
 	 * If it's a plain SELECT, it returns whatever the targetlist says.
-	 * Otherwise, if it's INSERT/UPDATE/DELETE with RETURNING, it returns
-	 * that. Otherwise, the function return type must be VOID.
+	 * Otherwise, if it's INSERT/UPDATE/DELETE/MERGE with RETURNING, it
+	 * returns that. Otherwise, the function return type must be VOID.
 	 *
 	 * Note: eventually replace this test with QueryReturnsTuples?	We'd need
 	 * a more general method of determining the output type, though.  Also, it
@@ -1679,7 +1679,8 @@ check_sql_fn_retval(List *queryTreeLists
 	else if (parse &&
 			 (parse->commandType == CMD_INSERT ||
 			  parse->commandType == CMD_UPDATE ||
-			  parse->commandType == CMD_DELETE) &&
+			  parse->commandType == CMD_DELETE ||
+			  parse->commandType == CMD_MERGE) &&
 			 parse->returningList)
 	{
 		tlist = parse->returningList;
@@ -1693,7 +1694,7 @@ check_sql_fn_retval(List *queryTreeLists
 				(errcode(ERRCODE_INVALID_FUNCTION_DEFINITION),
 				 errmsg("return type mismatch in function declared to return %s",
 						format_type_be(rettype)),
-				 errdetail("Function's final statement must be SELECT or INSERT/UPDATE/DELETE RETURNING.")));
+				 errdetail("Function's final statement must be SELECT or INSERT/UPDATE/DELETE/MERGE RETURNING.")));
 		return false;			/* keep compiler quiet */
 	}
 
diff --git a/src/backend/executor/nodeModifyTable.c b/src/backend/executor/nodeModifyTable.c
new file mode 100644
index 5005d8c..63a8039
--- a/src/backend/executor/nodeModifyTable.c
+++ b/src/backend/executor/nodeModifyTable.c
@@ -36,8 +36,7 @@
  *		RETURNING tuple after completing each row insert, update, or delete.
  *		It must be called again to continue the operation.  Without RETURNING,
  *		we just loop within the node until all the work is done, then
- *		return NULL.  This avoids useless call/return overhead.  (MERGE does
- *		not support RETURNING.)
+ *		return NULL.  This avoids useless call/return overhead.
  */
 
 #include "postgres.h"
@@ -88,9 +87,6 @@ typedef struct ModifyTableContext
 	 */
 	TupleTableSlot *planSlot;
 
-	/* MERGE specific */
-	MergeActionState *relaction;	/* MERGE action in progress */
-
 	/*
 	 * Information about the changes that were made concurrently to a tuple
 	 * being updated or deleted
@@ -153,13 +149,14 @@ static TupleTableSlot *ExecMerge(ModifyT
 								 ItemPointer tupleid,
 								 bool canSetTag);
 static void ExecInitMerge(ModifyTableState *mtstate, EState *estate);
-static bool ExecMergeMatched(ModifyTableContext *context,
-							 ResultRelInfo *resultRelInfo,
-							 ItemPointer tupleid,
-							 bool canSetTag);
-static void ExecMergeNotMatched(ModifyTableContext *context,
-								ResultRelInfo *resultRelInfo,
-								bool canSetTag);
+static TupleTableSlot *ExecMergeMatched(ModifyTableContext *context,
+										ResultRelInfo *resultRelInfo,
+										ItemPointer tupleid,
+										bool canSetTag,
+										bool *matched);
+static TupleTableSlot *ExecMergeNotMatched(ModifyTableContext *context,
+										   ResultRelInfo *resultRelInfo,
+										   bool canSetTag);
 
 
 /*
@@ -979,7 +976,7 @@ ExecInsert(ModifyTableContext *context,
 		if (mtstate->operation == CMD_UPDATE)
 			wco_kind = WCO_RLS_UPDATE_CHECK;
 		else if (mtstate->operation == CMD_MERGE)
-			wco_kind = (context->relaction->mas_action->commandType == CMD_UPDATE) ?
+			wco_kind = (mtstate->mt_merge_action->mas_action->commandType == CMD_UPDATE) ?
 				WCO_RLS_UPDATE_CHECK : WCO_RLS_INSERT_CHECK;
 		else
 			wco_kind = WCO_RLS_INSERT_CHECK;
@@ -1828,7 +1825,7 @@ ExecCrossPartitionUpdate(ModifyTableCont
 		 * additional rechecking, and might end up executing a different
 		 * action entirely).
 		 */
-		if (context->relaction != NULL)
+		if (mtstate->mt_merge_action != NULL)
 			return false;
 		else if (TupIsNull(epqslot))
 			return true;
@@ -2069,7 +2066,7 @@ lreplace:
 		 * No luck, a retry is needed.  If running MERGE, we do not do so
 		 * here; instead let it handle that on its own rules.
 		 */
-		if (context->relaction != NULL)
+		if (context->mtstate->mt_merge_action != NULL)
 			return TM_Updated;
 
 		/*
@@ -2712,6 +2709,7 @@ static TupleTableSlot *
 ExecMerge(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 		  ItemPointer tupleid, bool canSetTag)
 {
+	TupleTableSlot *rslot = NULL;
 	bool		matched;
 
 	/*-----
@@ -2759,18 +2757,18 @@ ExecMerge(ModifyTableContext *context, R
 	 */
 	matched = tupleid != NULL;
 	if (matched)
-		matched = ExecMergeMatched(context, resultRelInfo, tupleid, canSetTag);
+		rslot = ExecMergeMatched(context, resultRelInfo, tupleid, canSetTag,
+								 &matched);
 
 	/*
-	 * Either we were dealing with a NOT MATCHED tuple or ExecMergeMatched()
-	 * returned "false", indicating the previously MATCHED tuple no longer
-	 * matches.
+	 * Deal with the NOT MATCHED case (either a NOT MATCHED tuple from the
+	 * join, or a previously MATCHED tuple for which ExecMergeMatched() set
+	 * "matched" to false, indicating that it no longer matches).
 	 */
 	if (!matched)
-		ExecMergeNotMatched(context, resultRelInfo, canSetTag);
+		rslot = ExecMergeNotMatched(context, resultRelInfo, canSetTag);
 
-	/* No RETURNING support yet */
-	return NULL;
+	return rslot;
 }
 
 /*
@@ -2780,8 +2778,8 @@ ExecMerge(ModifyTableContext *context, R
  * We start from the first WHEN MATCHED action and check if the WHEN quals
  * pass, if any. If the WHEN quals for the first action do not pass, we
  * check the second, then the third and so on. If we reach to the end, no
- * action is taken and we return true, indicating that no further action is
- * required for this tuple.
+ * action is taken and "matched" is set to true, indicating that no further
+ * action is required for this tuple.
  *
  * If we do find a qualifying action, then we attempt to execute the action.
  *
@@ -2790,16 +2788,17 @@ ExecMerge(ModifyTableContext *context, R
  * with individual actions are evaluated by this routine via ExecQual, while
  * EvalPlanQual checks for the join quals. If EvalPlanQual tells us that the
  * updated tuple still passes the join quals, then we restart from the first
- * action to look for a qualifying action. Otherwise, we return false --
- * meaning that a NOT MATCHED action must now be executed for the current
- * source tuple.
+ * action to look for a qualifying action. Otherwise, "matched" is set to
+ * false -- meaning that a NOT MATCHED action must now be executed for the
+ * current source tuple.
  */
-static bool
+static TupleTableSlot *
 ExecMergeMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
-				 ItemPointer tupleid, bool canSetTag)
+				 ItemPointer tupleid, bool canSetTag, bool *matched)
 {
 	ModifyTableState *mtstate = context->mtstate;
-	TupleTableSlot *newslot;
+	TupleTableSlot *rslot = NULL;
+	TupleTableSlot *newslot = NULL;
 	EState	   *estate = context->estate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
 	bool		isNull;
@@ -2810,7 +2809,10 @@ ExecMergeMatched(ModifyTableContext *con
 	 * If there are no WHEN MATCHED actions, we are done.
 	 */
 	if (resultRelInfo->ri_matchedMergeAction == NIL)
-		return true;
+	{
+		*matched = true;
+		return NULL;
+	}
 
 	/*
 	 * Make tuple and any needed join variables available to ExecQual and
@@ -2889,12 +2891,15 @@ lmerge_matched:
 				 */
 				newslot = ExecProject(relaction->mas_proj);
 
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				if (!ExecUpdatePrologue(context, resultRelInfo,
 										tupleid, NULL, newslot, &result))
 				{
 					if (result == TM_Ok)
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 					break;		/* concurrent update/delete */
 				}
 				result = ExecUpdateAct(context, resultRelInfo, tupleid, NULL,
@@ -2908,12 +2913,15 @@ lmerge_matched:
 				break;
 
 			case CMD_DELETE:
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				if (!ExecDeletePrologue(context, resultRelInfo, tupleid,
 										NULL, NULL, &result))
 				{
 					if (result == TM_Ok)
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 					break;		/* concurrent update/delete */
 				}
 				result = ExecDeleteAct(context, resultRelInfo, tupleid, false);
@@ -2969,7 +2977,8 @@ lmerge_matched:
 				 * If the tuple was already deleted, return to let caller
 				 * handle it under NOT MATCHED clauses.
 				 */
-				return false;
+				*matched = false;
+				return NULL;
 
 			case TM_Updated:
 				{
@@ -3015,13 +3024,19 @@ lmerge_matched:
 							 * NOT MATCHED actions.
 							 */
 							if (TupIsNull(epqslot))
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							(void) ExecGetJunkAttribute(epqslot,
 														resultRelInfo->ri_RowIdAttNo,
 														&isNull);
 							if (isNull)
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							/*
 							 * When a tuple was updated and migrated to
@@ -3056,7 +3071,8 @@ lmerge_matched:
 							 * tuple already deleted; tell caller to run NOT
 							 * MATCHED actions
 							 */
-							return false;
+							*matched = false;
+							return NULL;
 
 						case TM_SelfModified:
 
@@ -3076,13 +3092,14 @@ lmerge_matched:
 										(errcode(ERRCODE_TRIGGERED_DATA_CHANGE_VIOLATION),
 										 errmsg("tuple to be updated or deleted was already modified by an operation triggered by the current command"),
 										 errhint("Consider using an AFTER trigger instead of a BEFORE trigger to propagate changes to other rows.")));
-							return false;
+							*matched = false;
+							return NULL;
 
 						default:
 							/* see table_tuple_lock call in ExecDelete() */
 							elog(ERROR, "unexpected table_tuple_lock status: %u",
 								 result);
-							return false;
+							return NULL;
 					}
 				}
 
@@ -3094,6 +3111,30 @@ lmerge_matched:
 				break;
 		}
 
+		/* Process RETURNING if present */
+		if (resultRelInfo->ri_projectReturning)
+		{
+			switch (commandType)
+			{
+				case CMD_UPDATE:
+					rslot = ExecProcessReturning(resultRelInfo, newslot,
+												 context->planSlot);
+					break;
+
+				case CMD_DELETE:
+					rslot = ExecProcessReturning(resultRelInfo,
+												 resultRelInfo->ri_oldTupleSlot,
+												 context->planSlot);
+					break;
+
+				case CMD_NOTHING:
+					break;
+
+				default:
+					elog(ERROR, "unknown action in MERGE WHEN MATCHED clause");
+			}
+		}
+
 		/*
 		 * We've activated one of the WHEN clauses, so we don't search
 		 * further. This is required behaviour, not an optimization.
@@ -3104,19 +3145,22 @@ lmerge_matched:
 	/*
 	 * Successfully executed an action or no qualifying action was found.
 	 */
-	return true;
+	*matched = true;
+
+	return rslot;
 }
 
 /*
  * Execute the first qualifying NOT MATCHED action.
  */
-static void
+static TupleTableSlot *
 ExecMergeNotMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 					bool canSetTag)
 {
 	ModifyTableState *mtstate = context->mtstate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
 	List	   *actionStates = NIL;
+	TupleTableSlot *rslot = NULL;
 	ListCell   *l;
 
 	/*
@@ -3166,10 +3210,10 @@ ExecMergeNotMatched(ModifyTableContext *
 				 * so we don't need to map the tuple here.
 				 */
 				newslot = ExecProject(action->mas_proj);
-				context->relaction = action;
+				mtstate->mt_merge_action = action;
 
-				(void) ExecInsert(context, mtstate->rootResultRelInfo, newslot,
-								  canSetTag, NULL, NULL);
+				rslot = ExecInsert(context, mtstate->rootResultRelInfo,
+								   newslot, canSetTag, NULL, NULL);
 				mtstate->mt_merge_inserted += 1;
 				break;
 			case CMD_NOTHING:
@@ -3185,6 +3229,8 @@ ExecMergeNotMatched(ModifyTableContext *
 		 */
 		break;
 	}
+
+	return rslot;
 }
 
 /*
@@ -3362,6 +3408,65 @@ ExecInitMergeTupleSlots(ModifyTableState
 }
 
 /*
+ * pg_merge_action() -
+ *	  SQL merge support function to retrieve the currently executing merge
+ *	  action command string ("INSERT", "UPDATE", or "DELETE").
+ */
+Datum
+pg_merge_action(PG_FUNCTION_ARGS)
+{
+	ModifyTableState *mtstate = (ModifyTableState *) fcinfo->context;
+	MergeActionState *relaction;
+	CmdType		commandType;
+
+	if (!mtstate || mtstate->operation != CMD_MERGE)
+		elog(ERROR, "merge support function called in non-merge context");
+
+	relaction = mtstate->mt_merge_action;
+	if (!relaction)
+		elog(ERROR, "no merge action in progress");
+
+	commandType = relaction->mas_action->commandType;
+	switch (commandType)
+	{
+		case CMD_INSERT:
+			PG_RETURN_TEXT_P(cstring_to_text("INSERT"));
+		case CMD_UPDATE:
+			PG_RETURN_TEXT_P(cstring_to_text("UPDATE"));
+		case CMD_DELETE:
+			PG_RETURN_TEXT_P(cstring_to_text("DELETE"));
+		case CMD_NOTHING:
+			elog(ERROR, "unexpected merge action: DO NOTHING");
+			break;
+		default:
+			elog(ERROR, "unrecognized commandType: %d", (int) commandType);
+	}
+
+	PG_RETURN_NULL();			/* keep compiler quiet */
+}
+
+/*
+ * pg_merge_when_clause_number() -
+ *	  SQL merge support function to retrieve the 1-based index of the
+ *	  currently executing merge WHEN clause.
+ */
+Datum
+pg_merge_when_clause_number(PG_FUNCTION_ARGS)
+{
+	ModifyTableState *mtstate = (ModifyTableState *) fcinfo->context;
+	MergeActionState *relaction;
+
+	if (!mtstate || mtstate->operation != CMD_MERGE)
+		elog(ERROR, "merge support function called in non-merge context");
+
+	relaction = mtstate->mt_merge_action;
+	if (!relaction)
+		elog(ERROR, "no merge action in progress");
+
+	PG_RETURN_INT32((int32) relaction->mas_action->index);
+}
+
+/*
  * Process BEFORE EACH STATEMENT triggers
  */
 static void
@@ -3647,8 +3752,17 @@ ExecModifyTable(PlanState *pstate)
 				{
 					EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-					ExecMerge(&context, node->resultRelInfo, NULL, node->canSetTag);
-					continue;	/* no RETURNING support yet */
+					slot = ExecMerge(&context, node->resultRelInfo, NULL,
+									 node->canSetTag);
+
+					/*
+					 * If we got a RETURNING result, return it to the caller.
+					 * We'll continue the work on next call.
+					 */
+					if (slot)
+						return slot;
+
+					continue;	/* continue with the next tuple */
 				}
 
 				elog(ERROR, "tableoid is NULL");
@@ -3725,8 +3839,17 @@ ExecModifyTable(PlanState *pstate)
 					{
 						EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-						ExecMerge(&context, node->resultRelInfo, NULL, node->canSetTag);
-						continue;	/* no RETURNING support yet */
+						slot = ExecMerge(&context, node->resultRelInfo, NULL,
+										 node->canSetTag);
+
+						/*
+						 * If we got a RETURNING result, return it to the
+						 * caller.  We'll continue the work on next call.
+						 */
+						if (slot)
+							return slot;
+
+						continue;	/* continue with the next tuple */
 					}
 
 					elog(ERROR, "ctid is NULL");
@@ -3818,7 +3941,7 @@ ExecModifyTable(PlanState *pstate)
 				}
 				slot = ExecGetUpdateNewTuple(resultRelInfo, context.planSlot,
 											 oldSlot);
-				context.relaction = NULL;
+				node->mt_merge_action = NULL;
 
 				/* Now apply the update. */
 				slot = ExecUpdate(&context, resultRelInfo, tupleid, oldtuple,
diff --git a/src/backend/executor/spi.c b/src/backend/executor/spi.c
new file mode 100644
index 3397568..b9f443c
--- a/src/backend/executor/spi.c
+++ b/src/backend/executor/spi.c
@@ -2033,6 +2033,8 @@ SPI_result_code_string(int code)
 			return "SPI_OK_TD_REGISTER";
 		case SPI_OK_MERGE:
 			return "SPI_OK_MERGE";
+		case SPI_OK_MERGE_RETURNING:
+			return "SPI_OK_MERGE_RETURNING";
 	}
 	/* Unrecognized code ... return something useful ... */
 	sprintf(buf, "Unrecognized SPI code %d", code);
@@ -2886,7 +2888,10 @@ _SPI_pquery(QueryDesc *queryDesc, bool f
 				res = SPI_OK_UPDATE;
 			break;
 		case CMD_MERGE:
-			res = SPI_OK_MERGE;
+			if (queryDesc->plannedstmt->hasReturning)
+				res = SPI_OK_MERGE_RETURNING;
+			else
+				res = SPI_OK_MERGE;
 			break;
 		default:
 			return SPI_ERROR_OPUNKNOWN;
diff --git a/src/backend/optimizer/plan/subselect.c b/src/backend/optimizer/plan/subselect.c
new file mode 100644
index 7a9fe88..d3ee1a2
--- a/src/backend/optimizer/plan/subselect.c
+++ b/src/backend/optimizer/plan/subselect.c
@@ -18,6 +18,7 @@
 
 #include "access/htup_details.h"
 #include "catalog/pg_operator.h"
+#include "catalog/pg_proc.h"
 #include "catalog/pg_type.h"
 #include "executor/executor.h"
 #include "miscadmin.h"
@@ -35,6 +36,7 @@
 #include "parser/parse_relation.h"
 #include "rewrite/rewriteManip.h"
 #include "utils/builtins.h"
+#include "utils/fmgroids.h"
 #include "utils/lsyscache.h"
 #include "utils/syscache.h"
 
@@ -1846,7 +1848,8 @@ convert_EXISTS_to_ANY(PlannerInfo *root,
 /*
  * Replace correlation vars (uplevel vars) with Params.
  *
- * Uplevel PlaceHolderVars and aggregates are replaced, too.
+ * Uplevel PlaceHolderVars, aggregates, GROUPING() expressions, and merge
+ * support functions are replaced, too.
  *
  * Note: it is critical that this runs immediately after SS_process_sublinks.
  * Since we do not recurse into the arguments of uplevel PHVs and aggregates,
@@ -1900,6 +1903,13 @@ replace_correlation_vars_mutator(Node *n
 		if (((GroupingFunc *) node)->agglevelsup > 0)
 			return (Node *) replace_outer_grouping(root, (GroupingFunc *) node);
 	}
+	if (IsA(node, FuncExpr) &&
+		IsMergeSupportFunction(((FuncExpr *) node)->funcid))
+	{
+		if (root->parse->commandType != CMD_MERGE)
+			return (Node *) replace_outer_merge_support_function(root,
+																 (FuncExpr *) node);
+	}
 	return expression_tree_mutator(node,
 								   replace_correlation_vars_mutator,
 								   (void *) root);
diff --git a/src/backend/optimizer/util/paramassign.c b/src/backend/optimizer/util/paramassign.c
new file mode 100644
index d6a923b..5ecf502
--- a/src/backend/optimizer/util/paramassign.c
+++ b/src/backend/optimizer/util/paramassign.c
@@ -307,6 +307,57 @@ replace_outer_grouping(PlannerInfo *root
 }
 
 /*
+ * Generate a Param node to replace the given FuncExpr node, which is expected
+ * to be a merge support function in the RETURNING list of an upper-level
+ * MERGE query.  Record the need for the FuncExpr in the proper upper-level
+ * root->plan_params.
+ */
+Param *
+replace_outer_merge_support_function(PlannerInfo *root, FuncExpr *func)
+{
+	Param	   *retval;
+	PlannerParamItem *pitem;
+	Oid			ptype = exprType((Node *) func);
+
+	Assert(root->parse->commandType != CMD_MERGE);
+
+	/*
+	 * The parser should have ensured that the merge support function is in
+	 * the RETURNING list of an upper-level MERGE query, so find that query.
+	 */
+	do
+	{
+		root = root->parent_root;
+		if (root == NULL)
+			elog(ERROR, "merge support function used outside MERGE");
+	} while (root->parse->commandType != CMD_MERGE);
+
+	/*
+	 * It does not seem worthwhile to try to de-duplicate references to outer
+	 * merge support functions.  Just make a new slot every time.
+	 */
+	func = copyObject(func);
+
+	pitem = makeNode(PlannerParamItem);
+	pitem->item = (Node *) func;
+	pitem->paramId = list_length(root->glob->paramExecTypes);
+	root->glob->paramExecTypes = lappend_oid(root->glob->paramExecTypes,
+											 ptype);
+
+	root->plan_params = lappend(root->plan_params, pitem);
+
+	retval = makeNode(Param);
+	retval->paramkind = PARAM_EXEC;
+	retval->paramid = pitem->paramId;
+	retval->paramtype = ptype;
+	retval->paramtypmod = -1;
+	retval->paramcollid = InvalidOid;
+	retval->location = func->location;
+
+	return retval;
+}
+
+/*
  * Generate a Param node to replace the given Var,
  * which is expected to come from some upper NestLoop plan node.
  * Record the need for the Var in root->curOuterParams.
diff --git a/src/backend/parser/analyze.c b/src/backend/parser/analyze.c
new file mode 100644
index 4006632..6794511
--- a/src/backend/parser/analyze.c
+++ b/src/backend/parser/analyze.c
@@ -74,7 +74,6 @@ static void determineRecursiveColTypes(P
 									   Node *larg, List *nrtargetlist);
 static Query *transformReturnStmt(ParseState *pstate, ReturnStmt *stmt);
 static Query *transformUpdateStmt(ParseState *pstate, UpdateStmt *stmt);
-static List *transformReturningList(ParseState *pstate, List *returningList);
 static Query *transformPLAssignStmt(ParseState *pstate,
 									PLAssignStmt *stmt);
 static Query *transformDeclareCursorStmt(ParseState *pstate,
@@ -515,7 +514,8 @@ transformDeleteStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -942,7 +942,8 @@ transformInsertStmt(ParseState *pstate,
 	/* Process RETURNING, if any. */
 	if (stmt->returningList)
 		qry->returningList = transformReturningList(pstate,
-													stmt->returningList);
+													stmt->returningList,
+													EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -2407,7 +2408,8 @@ transformUpdateStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/*
 	 * Now we are done with SELECT-like processing, and can get on with
@@ -2501,10 +2503,11 @@ transformUpdateTargetList(ParseState *ps
 
 /*
  * transformReturningList -
- *	handle a RETURNING clause in INSERT/UPDATE/DELETE
+ *	handle a RETURNING clause in INSERT/UPDATE/DELETE/MERGE
  */
-static List *
-transformReturningList(ParseState *pstate, List *returningList)
+List *
+transformReturningList(ParseState *pstate, List *returningList,
+					   ParseExprKind exprKind)
 {
 	List	   *rlist;
 	int			save_next_resno;
@@ -2521,7 +2524,7 @@ transformReturningList(ParseState *pstat
 	pstate->p_next_resno = 1;
 
 	/* transform RETURNING identically to a SELECT targetlist */
-	rlist = transformTargetList(pstate, returningList, EXPR_KIND_RETURNING);
+	rlist = transformTargetList(pstate, returningList, exprKind);
 
 	/*
 	 * Complain if the nonempty tlist expanded to nothing (which is possible
diff --git a/src/backend/parser/gram.y b/src/backend/parser/gram.y
new file mode 100644
index b3bdf94..dd2a74f
--- a/src/backend/parser/gram.y
+++ b/src/backend/parser/gram.y
@@ -12270,6 +12270,7 @@ MergeStmt:
 			USING table_ref
 			ON a_expr
 			merge_when_list
+			returning_clause
 				{
 					MergeStmt  *m = makeNode(MergeStmt);
 
@@ -12278,6 +12279,7 @@ MergeStmt:
 					m->sourceRelation = $6;
 					m->joinCondition = $8;
 					m->mergeWhenClauses = $9;
+					m->returningList = $10;
 
 					$$ = (Node *) m;
 				}
diff --git a/src/backend/parser/parse_agg.c b/src/backend/parser/parse_agg.c
new file mode 100644
index 85cd47b..e72c37b
--- a/src/backend/parser/parse_agg.c
+++ b/src/backend/parser/parse_agg.c
@@ -456,6 +456,7 @@ check_agglevels_and_constraints(ParseSta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
@@ -903,6 +904,7 @@ transformWindowFuncCall(ParseState *psta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_cte.c b/src/backend/parser/parse_cte.c
new file mode 100644
index c5b1a49..ad3c525
--- a/src/backend/parser/parse_cte.c
+++ b/src/backend/parser/parse_cte.c
@@ -126,13 +126,6 @@ transformWithClause(ParseState *pstate,
 		CommonTableExpr *cte = (CommonTableExpr *) lfirst(lc);
 		ListCell   *rest;
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(cte->ctequery, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in WITH query"),
-					parser_errposition(pstate, cte->location));
-
 		for_each_cell(rest, withClause->ctes, lnext(withClause->ctes, lc))
 		{
 			CommonTableExpr *cte2 = (CommonTableExpr *) lfirst(rest);
@@ -153,7 +146,8 @@ transformWithClause(ParseState *pstate,
 			/* must be a data-modifying statement */
 			Assert(IsA(cte->ctequery, InsertStmt) ||
 				   IsA(cte->ctequery, UpdateStmt) ||
-				   IsA(cte->ctequery, DeleteStmt));
+				   IsA(cte->ctequery, DeleteStmt) ||
+				   IsA(cte->ctequery, MergeStmt));
 
 			pstate->p_hasModifyingCTE = true;
 		}
diff --git a/src/backend/parser/parse_expr.c b/src/backend/parser/parse_expr.c
new file mode 100644
index aa72665..3957d7c
--- a/src/backend/parser/parse_expr.c
+++ b/src/backend/parser/parse_expr.c
@@ -541,6 +541,7 @@ transformColumnRef(ParseState *pstate, C
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CHECK_CONSTRAINT:
@@ -1767,6 +1768,7 @@ transformSubLink(ParseState *pstate, Sub
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CYCLE_MARK:
@@ -3115,6 +3117,7 @@ ParseExprKindName(ParseExprKind exprKind
 		case EXPR_KIND_OFFSET:
 			return "OFFSET";
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			return "RETURNING";
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
diff --git a/src/backend/parser/parse_func.c b/src/backend/parser/parse_func.c
new file mode 100644
index b3f0b6a..cf7c0b9
--- a/src/backend/parser/parse_func.c
+++ b/src/backend/parser/parse_func.c
@@ -31,6 +31,7 @@
 #include "parser/parse_target.h"
 #include "parser/parse_type.h"
 #include "utils/builtins.h"
+#include "utils/fmgroids.h"
 #include "utils/lsyscache.h"
 #include "utils/syscache.h"
 
@@ -348,6 +349,26 @@ ParseFuncOrColumn(ParseState *pstate, Li
 					 parser_errposition(pstate, location)));
 	}
 
+	/* Merge support functions are only allowed in MERGE's RETURNING list */
+	if (IsMergeSupportFunction(funcid) &&
+		pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+	{
+		ParseState *parent_pstate = pstate->parentParseState;
+
+		/* May be in a subquery in RETURNING list */
+		while (parent_pstate &&
+			   parent_pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+			parent_pstate = parent_pstate->parentParseState;
+
+		if (!parent_pstate ||
+			parent_pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+			ereport(ERROR,
+					(errcode(ERRCODE_WRONG_OBJECT_TYPE),
+					 errmsg("merge support function %s can only be called from the RETURNING list of a MERGE command",
+							NameListToString(funcname)),
+					 parser_errposition(pstate, location)));
+	}
+
 	/*
 	 * So far so good, so do some fdresult-type-specific processing.
 	 */
@@ -2599,6 +2620,7 @@ check_srf_call_placement(ParseState *pst
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_merge.c b/src/backend/parser/parse_merge.c
new file mode 100644
index 91b1156..33852e9
--- a/src/backend/parser/parse_merge.c
+++ b/src/backend/parser/parse_merge.c
@@ -233,6 +233,10 @@ transformMergeStmt(ParseState *pstate, M
 	 */
 	qry->jointree = makeFromExpr(pstate->p_joinlist, joinExpr);
 
+	/* Transform the RETURNING list, if any */
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_MERGE_RETURNING);
+
 	/*
 	 * We now have a good query shape, so now look at the WHEN conditions and
 	 * action targetlists.
@@ -254,6 +258,7 @@ transformMergeStmt(ParseState *pstate, M
 		MergeAction *action;
 
 		action = makeNode(MergeAction);
+		action->index = foreach_current_index(l) + 1;
 		action->commandType = mergeWhenClause->commandType;
 		action->matched = mergeWhenClause->matched;
 
@@ -390,9 +395,6 @@ transformMergeStmt(ParseState *pstate, M
 
 	qry->mergeActionList = mergeActionList;
 
-	/* RETURNING could potentially be added in the future, but not in SQL std */
-	qry->returningList = NULL;
-
 	qry->hasTargetSRFs = false;
 	qry->hasSubLinks = pstate->p_hasSubLinks;
 
diff --git a/src/backend/parser/parse_relation.c b/src/backend/parser/parse_relation.c
new file mode 100644
index 864ea9b..cc19041
--- a/src/backend/parser/parse_relation.c
+++ b/src/backend/parser/parse_relation.c
@@ -2346,9 +2346,10 @@ addRangeTableEntryForCTE(ParseState *pst
 		cte->cterefcount++;
 
 	/*
-	 * We throw error if the CTE is INSERT/UPDATE/DELETE without RETURNING.
-	 * This won't get checked in case of a self-reference, but that's OK
-	 * because data-modifying CTEs aren't allowed to be recursive anyhow.
+	 * We throw error if the CTE is INSERT/UPDATE/DELETE/MERGE without
+	 * RETURNING.  This won't get checked in case of a self-reference, but
+	 * that's OK because data-modifying CTEs aren't allowed to be recursive
+	 * anyhow.
 	 */
 	if (IsA(cte->ctequery, Query))
 	{
diff --git a/src/backend/rewrite/rewriteHandler.c b/src/backend/rewrite/rewriteHandler.c
new file mode 100644
index b486ab5..1bd0ce2
--- a/src/backend/rewrite/rewriteHandler.c
+++ b/src/backend/rewrite/rewriteHandler.c
@@ -3629,9 +3629,9 @@ RewriteQuery(Query *parsetree, List *rew
 	ListCell   *lc1;
 
 	/*
-	 * First, recursively process any insert/update/delete statements in WITH
-	 * clauses.  (We have to do this first because the WITH clauses may get
-	 * copied into rule actions below.)
+	 * First, recursively process any insert/update/delete/merge statements in
+	 * WITH clauses.  (We have to do this first because the WITH clauses may
+	 * get copied into rule actions below.)
 	 */
 	foreach(lc1, parsetree->cteList)
 	{
@@ -3656,7 +3656,8 @@ RewriteQuery(Query *parsetree, List *rew
 			if (!(ctequery->commandType == CMD_SELECT ||
 				  ctequery->commandType == CMD_UPDATE ||
 				  ctequery->commandType == CMD_INSERT ||
-				  ctequery->commandType == CMD_DELETE))
+				  ctequery->commandType == CMD_DELETE ||
+				  ctequery->commandType == CMD_MERGE))
 			{
 				/*
 				 * Currently it could only be NOTIFY; this error message will
diff --git a/src/backend/rewrite/rowsecurity.c b/src/backend/rewrite/rowsecurity.c
new file mode 100644
index b1620e4..966dcf0
--- a/src/backend/rewrite/rowsecurity.c
+++ b/src/backend/rewrite/rowsecurity.c
@@ -395,10 +395,10 @@ get_row_security_policies(Query *root, R
 	 * on the final action we take.
 	 *
 	 * We already fetched the SELECT policies above, to check existing rows,
-	 * but we must also check that new rows created by UPDATE actions are
-	 * visible, if SELECT rights are required for this relation. We don't do
-	 * this for INSERT actions, since an INSERT command would only do this
-	 * check if it had a RETURNING list, and MERGE does not support RETURNING.
+	 * but we must also check that new rows created by INSERT/UPDATE actions
+	 * are visible, if SELECT rights are required. For INSERT actions, we only
+	 * do this if RETURNING is specified, to be consistent with a plain INSERT
+	 * command, which can only require SELECT rights when RETURNING is used.
 	 *
 	 * We don't push the UPDATE/DELETE USING quals to the RTE because we don't
 	 * really want to apply them while scanning the relation since we don't
@@ -420,6 +420,8 @@ get_row_security_policies(Query *root, R
 		List	   *merge_delete_restrictive_policies;
 		List	   *merge_insert_permissive_policies;
 		List	   *merge_insert_restrictive_policies;
+		List	   *merge_select_permissive_policies = NIL;
+		List	   *merge_select_restrictive_policies = NIL;
 
 		/*
 		 * Fetch the UPDATE policies and set them up to execute on the
@@ -457,9 +459,6 @@ get_row_security_policies(Query *root, R
 		 */
 		if (perminfo->requiredPerms & ACL_SELECT)
 		{
-			List	   *merge_select_permissive_policies;
-			List	   *merge_select_restrictive_policies;
-
 			get_policies_for_relation(rel, CMD_SELECT, user_id,
 									  &merge_select_permissive_policies,
 									  &merge_select_restrictive_policies);
@@ -508,6 +507,21 @@ get_row_security_policies(Query *root, R
 							   withCheckOptions,
 							   hasSubLinks,
 							   false);
+
+		/*
+		 * Add ALL/SELECT policies as WCO_RLS_INSERT_CHECK WCOs, to ensure
+		 * that the inserted row is visible when executing an INSERT action,
+		 * if RETURNING is specified and SELECT rights are required for this
+		 * relation.
+		 */
+		if (perminfo->requiredPerms & ACL_SELECT && root->returningList)
+			add_with_check_options(rel, rt_index,
+								   WCO_RLS_INSERT_CHECK,
+								   merge_select_permissive_policies,
+								   merge_select_restrictive_policies,
+								   withCheckOptions,
+								   hasSubLinks,
+								   true);
 	}
 
 	table_close(rel, NoLock);
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
new file mode 100644
index e3ccf6c..6f3dc89
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -2142,11 +2142,10 @@ QueryReturnsTuples(Query *parsetree)
 		case CMD_SELECT:
 			/* returns tuples */
 			return true;
-		case CMD_MERGE:
-			return false;
 		case CMD_INSERT:
 		case CMD_UPDATE:
 		case CMD_DELETE:
+		case CMD_MERGE:
 			/* the forms with RETURNING return tuples */
 			if (parsetree->returningList)
 				return true;
diff --git a/src/backend/utils/adt/ruleutils.c b/src/backend/utils/adt/ruleutils.c
new file mode 100644
index 03f2835..25f49fa
--- a/src/backend/utils/adt/ruleutils.c
+++ b/src/backend/utils/adt/ruleutils.c
@@ -7183,8 +7183,13 @@ get_merge_query_def(Query *query, depars
 			appendStringInfoString(buf, "DO NOTHING");
 	}
 
-	/* No RETURNING support in MERGE yet */
-	Assert(query->returningList == NIL);
+	/* Add RETURNING if present */
+	if (query->returningList)
+	{
+		appendContextKeyword(context, " RETURNING",
+							 -PRETTYINDENT_STD, PRETTYINDENT_STD, 1);
+		get_target_list(query->returningList, context, NULL, colNamesVisible);
+	}
 }
 
 
diff --git a/src/bin/psql/common.c b/src/bin/psql/common.c
new file mode 100644
index 10ad1f2..3f42ca4
--- a/src/bin/psql/common.c
+++ b/src/bin/psql/common.c
@@ -982,13 +982,17 @@ PrintQueryResult(PGresult *result, bool
 			else
 				success = true;
 
-			/* if it's INSERT/UPDATE/DELETE RETURNING, also print status */
+			/*
+			 * If it's INSERT/UPDATE/DELETE/MERGE RETURNING, also print
+			 * status.
+			 */
 			if (last || pset.show_all_results)
 			{
 				cmdstatus = PQcmdStatus(result);
 				if (strncmp(cmdstatus, "INSERT", 6) == 0 ||
 					strncmp(cmdstatus, "UPDATE", 6) == 0 ||
-					strncmp(cmdstatus, "DELETE", 6) == 0)
+					strncmp(cmdstatus, "DELETE", 6) == 0 ||
+					strncmp(cmdstatus, "MERGE", 5) == 0)
 					PrintQueryStatus(result, printStatusFout);
 			}
 
diff --git a/src/include/catalog/pg_proc.dat b/src/include/catalog/pg_proc.dat
new file mode 100644
index 12fac15..88a3b29
--- a/src/include/catalog/pg_proc.dat
+++ b/src/include/catalog/pg_proc.dat
@@ -12049,4 +12049,14 @@
   proname => 'any_value_transfn', prorettype => 'anyelement',
   proargtypes => 'anyelement anyelement', prosrc => 'any_value_transfn' },
 
+# MERGE support functions
+{ oid => '9499', descr => 'command type of current MERGE action',
+  proname => 'pg_merge_action',  provolatile => 'v', proparallel => 'r',
+  prorettype => 'text', proargtypes => '',
+  prosrc => 'pg_merge_action' },
+{ oid => '9500', descr => 'index of current MERGE WHEN clause',
+  proname => 'pg_merge_when_clause_number',  provolatile => 'v',
+  proparallel => 'r', prorettype => 'int4', proargtypes => '',
+  prosrc => 'pg_merge_when_clause_number' },
+
 ]
diff --git a/src/include/catalog/pg_proc.h b/src/include/catalog/pg_proc.h
new file mode 100644
index e7abe0b..eb6cad9
--- a/src/include/catalog/pg_proc.h
+++ b/src/include/catalog/pg_proc.h
@@ -182,6 +182,11 @@ DECLARE_UNIQUE_INDEX(pg_proc_proname_arg
 #define PROARGMODE_VARIADIC 'v'
 #define PROARGMODE_TABLE	't'
 
+/* Is this a merge support function?  (Requires fmgroids.h) */
+#define IsMergeSupportFunction(funcid) \
+	((funcid) == F_PG_MERGE_ACTION || \
+	 (funcid) == F_PG_MERGE_WHEN_CLAUSE_NUMBER)
+
 #endif							/* EXPOSE_TO_CLIENT_CODE */
 
 
diff --git a/src/include/executor/spi.h b/src/include/executor/spi.h
new file mode 100644
index d1de139..ed8787d
--- a/src/include/executor/spi.h
+++ b/src/include/executor/spi.h
@@ -97,6 +97,7 @@ typedef struct _SPI_plan *SPIPlanPtr;
 #define SPI_OK_REL_UNREGISTER	16
 #define SPI_OK_TD_REGISTER		17
 #define SPI_OK_MERGE			18
+#define SPI_OK_MERGE_RETURNING	19
 
 #define SPI_OPT_NONATOMIC		(1 << 0)
 
diff --git a/src/include/nodes/execnodes.h b/src/include/nodes/execnodes.h
new file mode 100644
index cb714f4..6708b6b
--- a/src/include/nodes/execnodes.h
+++ b/src/include/nodes/execnodes.h
@@ -1319,6 +1319,9 @@ typedef struct ModifyTableState
 	/* Flags showing which subcommands are present INS/UPD/DEL/DO NOTHING */
 	int			mt_merge_subcommands;
 
+	/* For MERGE, the action currently being executed */
+	MergeActionState *mt_merge_action;
+
 	/* tuple counters for MERGE */
 	double		mt_merge_inserted;
 	double		mt_merge_updated;
diff --git a/src/include/nodes/parsenodes.h b/src/include/nodes/parsenodes.h
new file mode 100644
index 2565348..ac0a754
--- a/src/include/nodes/parsenodes.h
+++ b/src/include/nodes/parsenodes.h
@@ -1688,6 +1688,7 @@ typedef struct MergeWhenClause
 typedef struct MergeAction
 {
 	NodeTag		type;
+	int			index;			/* 1-based index of the clause */
 	bool		matched;		/* true=MATCHED, false=NOT MATCHED */
 	CmdType		commandType;	/* INSERT/UPDATE/DELETE/DO NOTHING */
 	/* OVERRIDING clause */
@@ -1951,6 +1952,7 @@ typedef struct MergeStmt
 	Node	   *sourceRelation; /* source relation */
 	Node	   *joinCondition;	/* join condition between source and target */
 	List	   *mergeWhenClauses;	/* list of MergeWhenClause(es) */
+	List	   *returningList;	/* list of expressions to return */
 	WithClause *withClause;		/* WITH clause */
 } MergeStmt;
 
diff --git a/src/include/optimizer/paramassign.h b/src/include/optimizer/paramassign.h
new file mode 100644
index 55c27a6..6f70c78
--- a/src/include/optimizer/paramassign.h
+++ b/src/include/optimizer/paramassign.h
@@ -20,6 +20,8 @@ extern Param *replace_outer_placeholderv
 										   PlaceHolderVar *phv);
 extern Param *replace_outer_agg(PlannerInfo *root, Aggref *agg);
 extern Param *replace_outer_grouping(PlannerInfo *root, GroupingFunc *grp);
+extern Param *replace_outer_merge_support_function(PlannerInfo *root,
+												   FuncExpr *func);
 extern Param *replace_nestloop_param_var(PlannerInfo *root, Var *var);
 extern Param *replace_nestloop_param_placeholdervar(PlannerInfo *root,
 													PlaceHolderVar *phv);
diff --git a/src/include/parser/analyze.h b/src/include/parser/analyze.h
new file mode 100644
index 1cef183..78eb55e
--- a/src/include/parser/analyze.h
+++ b/src/include/parser/analyze.h
@@ -44,6 +44,8 @@ extern List *transformInsertRow(ParseSta
 								bool strip_indirection);
 extern List *transformUpdateTargetList(ParseState *pstate,
 									   List *origTlist);
+extern List *transformReturningList(ParseState *pstate, List *returningList,
+									ParseExprKind exprKind);
 extern Query *transformTopLevelStmt(ParseState *pstate, RawStmt *parseTree);
 extern Query *transformStmt(ParseState *pstate, Node *parseTree);
 
diff --git a/src/include/parser/parse_node.h b/src/include/parser/parse_node.h
new file mode 100644
index f589112..836819d
--- a/src/include/parser/parse_node.h
+++ b/src/include/parser/parse_node.h
@@ -61,7 +61,8 @@ typedef enum ParseExprKind
 	EXPR_KIND_DISTINCT_ON,		/* DISTINCT ON */
 	EXPR_KIND_LIMIT,			/* LIMIT */
 	EXPR_KIND_OFFSET,			/* OFFSET */
-	EXPR_KIND_RETURNING,		/* RETURNING */
+	EXPR_KIND_RETURNING,		/* RETURNING in INSERT/UPDATE/DELETE */
+	EXPR_KIND_MERGE_RETURNING,	/* RETURNING in MERGE */
 	EXPR_KIND_VALUES,			/* VALUES */
 	EXPR_KIND_VALUES_SINGLE,	/* single-row VALUES (in INSERT only) */
 	EXPR_KIND_CHECK_CONSTRAINT, /* CHECK constraint for a table */
diff --git a/src/pl/plpgsql/src/pl_exec.c b/src/pl/plpgsql/src/pl_exec.c
new file mode 100644
index 4b76f76..fb6289c
--- a/src/pl/plpgsql/src/pl_exec.c
+++ b/src/pl/plpgsql/src/pl_exec.c
@@ -4267,9 +4267,9 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 	/*
 	 * If we have INTO, then we only need one row back ... but if we have INTO
 	 * STRICT or extra check too_many_rows, ask for two rows, so that we can
-	 * verify the statement returns only one.  INSERT/UPDATE/DELETE are always
-	 * treated strictly. Without INTO, just run the statement to completion
-	 * (tcount = 0).
+	 * verify the statement returns only one.  INSERT/UPDATE/DELETE/MERGE are
+	 * always treated strictly. Without INTO, just run the statement to
+	 * completion (tcount = 0).
 	 *
 	 * We could just ask for two rows always when using INTO, but there are
 	 * some cases where demanding the extra row costs significant time, eg by
@@ -4307,10 +4307,11 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 			Assert(stmt->mod_stmt);
 			exec_set_found(estate, (SPI_processed != 0));
 			break;
@@ -4489,10 +4490,11 @@ exec_stmt_dynexecute(PLpgSQL_execstate *
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 		case SPI_OK_UTILITY:
 		case SPI_OK_REWRITTEN:
 			break;
diff --git a/src/pl/tcl/pltcl.c b/src/pl/tcl/pltcl.c
new file mode 100644
index e8f9d7b..0c9f81e
--- a/src/pl/tcl/pltcl.c
+++ b/src/pl/tcl/pltcl.c
@@ -2456,6 +2456,7 @@ pltcl_process_SPI_result(Tcl_Interp *int
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
+		case SPI_OK_MERGE_RETURNING:
 
 			/*
 			 * Process the tuples we got
diff --git a/src/test/regress/expected/merge.out b/src/test/regress/expected/merge.out
new file mode 100644
index 133d421..5be3d2d
--- a/src/test/regress/expected/merge.out
+++ b/src/test/regress/expected/merge.out
@@ -123,20 +123,20 @@ ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
 ERROR:  name "target" specified more than once
 DETAIL:  The name is used both as MERGE target table and data source.
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
-ERROR:  MERGE not supported in WITH query
-LINE 1: WITH foo AS (
-             ^
--- used in COPY
+ERROR:  WITH query "foo" does not have a RETURNING clause
+LINE 4: ) SELECT * FROM foo;
+                        ^
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) TO stdout;
-ERROR:  MERGE not supported in COPY
+ERROR:  COPY query must have a RETURNING clause
 -- unsupported relation types
 -- view
 CREATE VIEW tv AS SELECT * FROM target;
@@ -1256,7 +1256,7 @@ BEGIN;
 MERGE INTO sq_target
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1271,7 +1271,7 @@ INSERT INTO sq_source (sid, balance, del
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1295,7 +1295,7 @@ WITH targq AS (
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1304,20 +1304,193 @@ WHEN MATCHED AND tid < 2 THEN
 ROLLBACK;
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
+CREATE TABLE merge_actions(action text, abbrev text);
+INSERT INTO merge_actions VALUES ('INSERT', 'ins'), ('UPDATE', 'upd'), ('DELETE', 'del');
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
-ERROR:  syntax error at or near "RETURNING"
-LINE 10: RETURNING *;
-         ^
+    DELETE
+RETURNING pg_merge_when_clause_number() AS clause_number,
+          (SELECT abbrev FROM merge_actions
+            WHERE action = pg_merge_action()) AS merge_action,
+          t.*,
+          CASE pg_merge_action()
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ clause_number | merge_action | tid | balance |     description     
+---------------+--------------+-----+---------+---------------------
+             3 | del          |   1 |     100 | Removed (1,100)
+             1 | upd          |   2 |     220 | Added 20 to balance
+             2 | ins          |   4 |      40 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- error when using MERGE support functions outside MERGE
+SELECT pg_merge_action() FROM sq_target;
+ERROR:  merge support function pg_merge_action can only be called from the RETURNING list of a MERGE command
+LINE 1: SELECT pg_merge_action() FROM sq_target;
+               ^
+UPDATE sq_target SET balance = balance + 1 RETURNING pg_merge_when_clause_number();
+ERROR:  merge support function pg_merge_when_clause_number can only be called from the RETURNING list of a MERGE command
+LINE 1: ...ATE sq_target SET balance = balance + 1 RETURNING pg_merge_w...
+                                                             ^
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_when_clause_number() AS clause_number,
+              pg_merge_action() AS merge_action,
+              t.*,
+              CASE pg_merge_action()
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING merge_action, pg_merge_action() AS merge_log_action, l.*
+)
+SELECT * FROM m2;
+ merge_action | merge_log_action | tid |     last_change     
+--------------+------------------+-----+---------------------
+ DELETE       | UPDATE           |   1 | Removed (1,100)
+ UPDATE       | INSERT           |   2 | Added 20 to balance
+ INSERT       | INSERT           |   4 | Inserted (4,40)
+(3 rows)
+
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ tid |     last_change     
+-----+---------------------
+   1 | Removed (1,100)
+   2 | Added 20 to balance
+   4 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*
+) TO stdout;
+DELETE	1	100
+UPDATE	2	220
+INSERT	4	40
+ROLLBACK;
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ action | tid | new_balance 
+--------+-----+-------------
+ DELETE |   1 |         100
+ UPDATE |   3 |         320
+ INSERT |   4 |         110
+(3 rows)
+
+ROLLBACK;
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ action | tid | balance 
+--------+-----+---------
+ DELETE |   1 |     100
+ UPDATE |   2 |     220
+ INSERT |   4 |      40
+(3 rows)
+
+ROLLBACK;
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ r_action | r_tid | r_balance 
+----------+-------+-----------
+ DELETE   |     1 |       100
+ UPDATE   |     3 |       320
+ INSERT   |     4 |       110
+(3 rows)
+
 ROLLBACK;
 -- EXPLAIN
 CREATE TABLE ex_mtarget (a int, b int)
@@ -1514,7 +1687,7 @@ SELECT * FROM sq_target WHERE tid = 1;
 (1 row)
 
 ROLLBACK;
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 NOTICE:  drop cascades to view v
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1640,6 +1813,32 @@ SELECT * FROM pa_target ORDER BY tid;
 (14 rows)
 
 ROLLBACK;
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING pg_merge_action(), pg_merge_when_clause_number(), t.*;
+ pg_merge_action | pg_merge_when_clause_number | tid | balance |           val            
+-----------------+-----------------------------+-----+---------+--------------------------
+ UPDATE          |                           1 |   2 |     110 | initial updated by merge
+(1 row)
+
+SELECT * FROM pa_target ORDER BY tid;
+ tid | balance |           val            
+-----+---------+--------------------------
+   2 |     110 | initial updated by merge
+   3 |     300 | initial
+   5 |     500 | initial
+   7 |     700 | initial
+   9 |     900 | initial
+  11 |    1100 | initial
+  13 |    1300 | initial
+(7 rows)
+
+ROLLBACK;
 DROP TABLE pa_target CASCADE;
 -- The target table is partitioned in the same way, but this time by attaching
 -- partitions which have columns in different order, dropped columns etc.
@@ -1731,7 +1930,26 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (sid, delta, 'inserted by merge');
+    INSERT VALUES (sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause_number(), t.*;
+ pg_merge_action | pg_merge_when_clause_number | tid | balance |           val            
+-----------------+-----------------------------+-----+---------+--------------------------
+ UPDATE          |                           1 |   2 |     110 | initial updated by merge
+ INSERT          |                           2 |   2 |      20 | inserted by merge
+ UPDATE          |                           1 |   4 |     330 | initial updated by merge
+ INSERT          |                           2 |   4 |      40 | inserted by merge
+ UPDATE          |                           1 |   6 |     550 | initial updated by merge
+ INSERT          |                           2 |   6 |      60 | inserted by merge
+ UPDATE          |                           1 |   8 |     770 | initial updated by merge
+ INSERT          |                           2 |   8 |      80 | inserted by merge
+ UPDATE          |                           1 |  10 |     990 | initial updated by merge
+ INSERT          |                           2 |  10 |     100 | inserted by merge
+ UPDATE          |                           1 |  12 |    1210 | initial updated by merge
+ INSERT          |                           2 |  12 |     120 | inserted by merge
+ UPDATE          |                           1 |  14 |    1430 | initial updated by merge
+ INSERT          |                           2 |  14 |     140 | inserted by merge
+(14 rows)
+
 SELECT * FROM pa_target ORDER BY tid;
  tid | balance |           val            
 -----+---------+--------------------------
@@ -1798,7 +2016,21 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause_number(), t.*;
+ pg_merge_action | pg_merge_when_clause_number |          logts           | tid | balance |           val            
+-----------------+-----------------------------+--------------------------+-----+---------+--------------------------
+ UPDATE          |                           1 | Tue Jan 31 00:00:00 2017 |   1 |     110 | initial updated by merge
+ UPDATE          |                           1 | Tue Feb 28 00:00:00 2017 |   2 |     220 | initial updated by merge
+ INSERT          |                           2 | Sun Jan 15 00:00:00 2017 |   3 |      30 | inserted by merge
+ UPDATE          |                           1 | Tue Jan 31 00:00:00 2017 |   4 |     440 | initial updated by merge
+ UPDATE          |                           1 | Tue Feb 28 00:00:00 2017 |   5 |     550 | initial updated by merge
+ INSERT          |                           2 | Sun Jan 15 00:00:00 2017 |   6 |      60 | inserted by merge
+ UPDATE          |                           1 | Tue Jan 31 00:00:00 2017 |   7 |     770 | initial updated by merge
+ UPDATE          |                           1 | Tue Feb 28 00:00:00 2017 |   8 |     880 | initial updated by merge
+ INSERT          |                           2 | Sun Jan 15 00:00:00 2017 |   9 |      90 | inserted by merge
+(9 rows)
+
 SELECT * FROM pa_target ORDER BY tid;
           logts           | tid | balance |           val            
 --------------------------+-----+---------+--------------------------
diff --git a/src/test/regress/expected/rowsecurity.out b/src/test/regress/expected/rowsecurity.out
new file mode 100644
index 97ca9bf..55250b3
--- a/src/test/regress/expected/rowsecurity.out
+++ b/src/test/regress/expected/rowsecurity.out
@@ -2300,6 +2300,35 @@ WHEN MATCHED THEN
 	UPDATE SET dnotes = dnotes || ' notes added by merge8 '
 WHEN NOT MATCHED THEN
 	INSERT VALUES (13, 44, 1, 'regress_rls_bob', 'new manga');
+SELECT * FROM document WHERE did = 13;
+ did | cid | dlevel | dauthor | dtitle | dnotes 
+-----+-----+--------+---------+--------+--------
+(0 rows)
+
+-- but not OK if RETURNING is used
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge9 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 44, 1, 'regress_rls_bob', 'new manga')
+RETURNING *;
+ERROR:  new row violates row-level security policy for table "document"
+-- but OK if new row is visible
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge10 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 11, 1, 'regress_rls_bob', 'new novel')
+RETURNING *;
+ sdid | did | cid | dlevel |     dauthor     |  dtitle   | dnotes 
+------+-----+-----+--------+-----------------+-----------+--------
+   14 |  14 |  11 |      1 | regress_rls_bob | new novel | 
+(1 row)
+
 RESET SESSION AUTHORIZATION;
 -- drop the restrictive SELECT policy so that we can look at the
 -- final state of the table
@@ -2323,7 +2352,8 @@ SELECT * FROM document;
   12 |  11 |      1 | regress_rls_bob   | another novel                    | 
    1 |  11 |      1 | regress_rls_bob   | my first novel                   |  notes added by merge2  notes added by merge3  notes added by merge4  notes added by merge7 
   13 |  44 |      1 | regress_rls_bob   | new manga                        | 
-(15 rows)
+  14 |  11 |      1 | regress_rls_bob   | new novel                        | 
+(16 rows)
 
 --
 -- ROLE/GROUP
diff --git a/src/test/regress/expected/rules.out b/src/test/regress/expected/rules.out
new file mode 100644
index 5058be5..2adfad7
--- a/src/test/regress/expected/rules.out
+++ b/src/test/regress/expected/rules.out
@@ -3596,7 +3596,8 @@ MERGE INTO rule_merge2 t USING (SELECT 1
 -- test deparsing
 CREATE TABLE sf_target(id int, data text, filling int[]);
 CREATE FUNCTION merge_sf_test()
- RETURNS void
+ RETURNS TABLE(pg_merge_action text, when_clause int,
+               a int, b text, id int, data text, filling int[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -3633,11 +3634,13 @@ WHEN NOT MATCHED
    VALUES (s.a, s.b, DEFAULT)
 WHEN NOT MATCHED
    THEN INSERT (filling[1], id)
-   VALUES (s.a, s.a);
+   VALUES (s.a, s.a)
+RETURNING
+   pg_merge_action(), pg_merge_when_clause_number() AS clause_number, *;
 END;
 \sf merge_sf_test
 CREATE OR REPLACE FUNCTION public.merge_sf_test()
- RETURNS void
+ RETURNS TABLE(pg_merge_action text, when_clause integer, a integer, b text, id integer, data text, filling integer[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -3674,7 +3677,14 @@ BEGIN ATOMIC
       VALUES (s.a, s.b, DEFAULT)
     WHEN NOT MATCHED
      THEN INSERT (filling[1], id)
-      VALUES (s.a, s.a);
+      VALUES (s.a, s.a)
+   RETURNING pg_merge_action() AS pg_merge_action,
+     pg_merge_when_clause_number() AS clause_number,
+     s.a,
+     s.b,
+     t.id,
+     t.data,
+     t.filling;
 END
 DROP FUNCTION merge_sf_test;
 DROP TABLE sf_target;
diff --git a/src/test/regress/sql/merge.sql b/src/test/regress/sql/merge.sql
new file mode 100644
index 4cf6db9..8394d01
--- a/src/test/regress/sql/merge.sql
+++ b/src/test/regress/sql/merge.sql
@@ -86,12 +86,12 @@ MERGE INTO target
 USING target
 ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
--- used in COPY
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
@@ -816,7 +816,7 @@ BEGIN;
 MERGE INTO sq_target
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -829,7 +829,7 @@ INSERT INTO sq_source (sid, balance, del
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -847,7 +847,7 @@ WITH targq AS (
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -857,17 +857,148 @@ ROLLBACK;
 
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
+CREATE TABLE merge_actions(action text, abbrev text);
+INSERT INTO merge_actions VALUES ('INSERT', 'ins'), ('UPDATE', 'upd'), ('DELETE', 'del');
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
+    DELETE
+RETURNING pg_merge_when_clause_number() AS clause_number,
+          (SELECT abbrev FROM merge_actions
+            WHERE action = pg_merge_action()) AS merge_action,
+          t.*,
+          CASE pg_merge_action()
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ROLLBACK;
+
+-- error when using MERGE support functions outside MERGE
+SELECT pg_merge_action() FROM sq_target;
+UPDATE sq_target SET balance = balance + 1 RETURNING pg_merge_when_clause_number();
+
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_when_clause_number() AS clause_number,
+              pg_merge_action() AS merge_action,
+              t.*,
+              CASE pg_merge_action()
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING merge_action, pg_merge_action() AS merge_log_action, l.*
+)
+SELECT * FROM m2;
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ROLLBACK;
+
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*
+) TO stdout;
+ROLLBACK;
+
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ROLLBACK;
+
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ROLLBACK;
+
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING pg_merge_action(), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
 ROLLBACK;
 
 -- EXPLAIN
@@ -966,7 +1097,7 @@ WHEN MATCHED THEN
 SELECT * FROM sq_target WHERE tid = 1;
 ROLLBACK;
 
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1033,6 +1164,17 @@ SELECT merge_func();
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING pg_merge_action(), pg_merge_when_clause_number(), t.*;
+SELECT * FROM pa_target ORDER BY tid;
+ROLLBACK;
+
 DROP TABLE pa_target CASCADE;
 
 -- The target table is partitioned in the same way, but this time by attaching
@@ -1091,7 +1233,8 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (sid, delta, 'inserted by merge');
+    INSERT VALUES (sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause_number(), t.*;
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
@@ -1145,7 +1288,8 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING pg_merge_action(), pg_merge_when_clause_number(), t.*;
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
diff --git a/src/test/regress/sql/rowsecurity.sql b/src/test/regress/sql/rowsecurity.sql
new file mode 100644
index dec7340..1d5ed0a
--- a/src/test/regress/sql/rowsecurity.sql
+++ b/src/test/regress/sql/rowsecurity.sql
@@ -992,6 +992,27 @@ WHEN MATCHED THEN
 	UPDATE SET dnotes = dnotes || ' notes added by merge8 '
 WHEN NOT MATCHED THEN
 	INSERT VALUES (13, 44, 1, 'regress_rls_bob', 'new manga');
+SELECT * FROM document WHERE did = 13;
+
+-- but not OK if RETURNING is used
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge9 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 44, 1, 'regress_rls_bob', 'new manga')
+RETURNING *;
+
+-- but OK if new row is visible
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge10 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 11, 1, 'regress_rls_bob', 'new novel')
+RETURNING *;
 
 RESET SESSION AUTHORIZATION;
 -- drop the restrictive SELECT policy so that we can look at the
diff --git a/src/test/regress/sql/rules.sql b/src/test/regress/sql/rules.sql
new file mode 100644
index 8b7e255..5dbeb6b
--- a/src/test/regress/sql/rules.sql
+++ b/src/test/regress/sql/rules.sql
@@ -1281,7 +1281,8 @@ MERGE INTO rule_merge2 t USING (SELECT 1
 CREATE TABLE sf_target(id int, data text, filling int[]);
 
 CREATE FUNCTION merge_sf_test()
- RETURNS void
+ RETURNS TABLE(pg_merge_action text, when_clause int,
+               a int, b text, id int, data text, filling int[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -1318,7 +1319,9 @@ WHEN NOT MATCHED
    VALUES (s.a, s.b, DEFAULT)
 WHEN NOT MATCHED
    THEN INSERT (filling[1], id)
-   VALUES (s.a, s.a);
+   VALUES (s.a, s.a)
+RETURNING
+   pg_merge_action(), pg_merge_when_clause_number() AS clause_number, *;
 END;
 
 \sf merge_sf_test

diff --git a/src/backend/commands/copy.c b/src/backend/commands/copy.c
new file mode 100644
index c5d7d78..7977873
--- a/src/backend/commands/copy.c
+++ b/src/backend/commands/copy.c
@@ -283,12 +283,6 @@ DoCopy(ParseState *pstate, const CopyStm
 	{
 		Assert(stmt->query);
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(stmt->query, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in COPY"));
-
 		query = makeNode(RawStmt);
 		query->stmt = stmt->query;
 		query->stmt_location = stmt_location;
diff --git a/src/backend/commands/copyto.c b/src/backend/commands/copyto.c
new file mode 100644
index c66a047..1145aaf
--- a/src/backend/commands/copyto.c
+++ b/src/backend/commands/copyto.c
@@ -510,7 +510,8 @@ BeginCopyTo(ParseState *pstate,
 		{
 			Assert(query->commandType == CMD_INSERT ||
 				   query->commandType == CMD_UPDATE ||
-				   query->commandType == CMD_DELETE);
+				   query->commandType == CMD_DELETE ||
+				   query->commandType == CMD_MERGE);
 
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
diff --git a/src/backend/executor/execPartition.c b/src/backend/executor/execPartition.c
new file mode 100644
index f6c3432..f0e75c3
--- a/src/backend/executor/execPartition.c
+++ b/src/backend/executor/execPartition.c
@@ -613,16 +613,16 @@ ExecInitPartitionInfo(ModifyTableState *
 	 * build the returningList for partitions within the planner, but simple
 	 * translation of varattnos will suffice.  This only occurs for the INSERT
 	 * case or in the case of UPDATE tuple routing where we didn't find a
-	 * result rel to reuse.
+	 * result rel to reuse.  We skip this for MERGE, since it uses per-action
+	 * RETURNING lists, which are handled below.
 	 */
-	if (node && node->returningLists != NIL)
+	if (node && node->returningLists != NIL && node->operation != CMD_MERGE)
 	{
 		TupleTableSlot *slot;
 		ExprContext *econtext;
 		List	   *returningList;
 
 		/* See the comment above for WCO lists. */
-		/* (except no RETURNING support for MERGE yet) */
 		Assert((node->operation == CMD_INSERT &&
 				list_length(node->returningLists) == 1 &&
 				list_length(node->resultRelations) == 1) ||
@@ -959,6 +959,25 @@ ExecInitPartitionInfo(ModifyTableState *
 									&found_whole_row);
 			action_state->mas_whenqual =
 				ExecInitQual((List *) action->qual, &mtstate->ps);
+
+			/* Build a projection for the action's RETURNING list, if any */
+			if (action->returningList)
+			{
+				/* found_whole_row intentionally ignored. */
+				action->returningList = (List *)
+					map_variable_attnos((Node *) action->returningList,
+										firstVarno, 0,
+										part_attmap,
+										RelationGetForm(partrel)->reltype,
+										&found_whole_row);
+
+				action_state->mas_proj_returning =
+					ExecBuildProjectionInfo(action->returningList,
+											econtext,
+											mtstate->ps.ps_ResultTupleSlot,
+											&mtstate->ps,
+											RelationGetDescr(partrel));
+			}
 		}
 	}
 	MemoryContextSwitchTo(oldcxt);
diff --git a/src/backend/executor/functions.c b/src/backend/executor/functions.c
new file mode 100644
index bace252..68cb4fa
--- a/src/backend/executor/functions.c
+++ b/src/backend/executor/functions.c
@@ -1660,8 +1660,8 @@ check_sql_fn_retval(List *queryTreeLists
 
 	/*
 	 * If it's a plain SELECT, it returns whatever the targetlist says.
-	 * Otherwise, if it's INSERT/UPDATE/DELETE with RETURNING, it returns
-	 * that. Otherwise, the function return type must be VOID.
+	 * Otherwise, if it's INSERT/UPDATE/DELETE/MERGE with RETURNING, it
+	 * returns that. Otherwise, the function return type must be VOID.
 	 *
 	 * Note: eventually replace this test with QueryReturnsTuples?	We'd need
 	 * a more general method of determining the output type, though.  Also, it
@@ -1679,7 +1679,8 @@ check_sql_fn_retval(List *queryTreeLists
 	else if (parse &&
 			 (parse->commandType == CMD_INSERT ||
 			  parse->commandType == CMD_UPDATE ||
-			  parse->commandType == CMD_DELETE) &&
+			  parse->commandType == CMD_DELETE ||
+			  parse->commandType == CMD_MERGE) &&
 			 parse->returningList)
 	{
 		tlist = parse->returningList;
@@ -1693,7 +1694,7 @@ check_sql_fn_retval(List *queryTreeLists
 				(errcode(ERRCODE_INVALID_FUNCTION_DEFINITION),
 				 errmsg("return type mismatch in function declared to return %s",
 						format_type_be(rettype)),
-				 errdetail("Function's final statement must be SELECT or INSERT/UPDATE/DELETE RETURNING.")));
+				 errdetail("Function's final statement must be SELECT or INSERT/UPDATE/DELETE/MERGE RETURNING.")));
 		return false;			/* keep compiler quiet */
 	}
 
diff --git a/src/backend/executor/nodeModifyTable.c b/src/backend/executor/nodeModifyTable.c
new file mode 100644
index 299c2c7..b8a3500
--- a/src/backend/executor/nodeModifyTable.c
+++ b/src/backend/executor/nodeModifyTable.c
@@ -36,8 +36,7 @@
  *		RETURNING tuple after completing each row insert, update, or delete.
  *		It must be called again to continue the operation.  Without RETURNING,
  *		we just loop within the node until all the work is done, then
- *		return NULL.  This avoids useless call/return overhead.  (MERGE does
- *		not support RETURNING.)
+ *		return NULL.  This avoids useless call/return overhead.
  */
 
 #include "postgres.h"
@@ -90,6 +89,7 @@ typedef struct ModifyTableContext
 
 	/* MERGE specific */
 	MergeActionState *relaction;	/* MERGE action in progress */
+	int			relaction_idx;		/* its position in the list */
 
 	/*
 	 * Information about the changes that were made concurrently to a tuple
@@ -153,13 +153,14 @@ static TupleTableSlot *ExecMerge(ModifyT
 								 ItemPointer tupleid,
 								 bool canSetTag);
 static void ExecInitMerge(ModifyTableState *mtstate, EState *estate);
-static bool ExecMergeMatched(ModifyTableContext *context,
-							 ResultRelInfo *resultRelInfo,
-							 ItemPointer tupleid,
-							 bool canSetTag);
-static void ExecMergeNotMatched(ModifyTableContext *context,
-								ResultRelInfo *resultRelInfo,
-								bool canSetTag);
+static TupleTableSlot *ExecMergeMatched(ModifyTableContext *context,
+										ResultRelInfo *resultRelInfo,
+										ItemPointer tupleid,
+										bool canSetTag,
+										bool *matched);
+static TupleTableSlot *ExecMergeNotMatched(ModifyTableContext *context,
+										   ResultRelInfo *resultRelInfo,
+										   bool canSetTag);
 
 
 /*
@@ -779,6 +780,19 @@ ExecInsert(ModifyTableContext *context,
 		slot = ExecPrepareTupleRouting(mtstate, estate, proute,
 									   resultRelInfo, slot,
 									   &partRelInfo);
+
+		/*
+		 * If we're running MERGE, update the context's merge action to the
+		 * corresponding action for this partition, so that we have the
+		 * correct projection for RETURNING below.
+		 */
+		if (mtstate->operation == CMD_MERGE)
+			context->relaction = (MergeActionState *)
+				list_nth(context->relaction->mas_action->matched ?
+						 partRelInfo->ri_matchedMergeAction :
+						 partRelInfo->ri_notMatchedMergeAction,
+						 context->relaction_idx);
+
 		resultRelInfo = partRelInfo;
 	}
 
@@ -1193,7 +1207,12 @@ ExecInsert(ModifyTableContext *context,
 	if (resultRelInfo->ri_WithCheckOptions != NIL)
 		ExecWithCheckOptions(WCO_VIEW_CHECK, resultRelInfo, slot, estate);
 
-	/* Process RETURNING if present */
+	/*
+	 * Process RETURNING if present.  If we're running MERGE, be sure to use
+	 * the projection for the current merge action's RETURNING list, if any.
+	 */
+	if (mtstate->operation == CMD_MERGE)
+		resultRelInfo->ri_projectReturning = context->relaction->mas_proj_returning;
 	if (resultRelInfo->ri_projectReturning)
 		result = ExecProcessReturning(resultRelInfo, slot, planSlot);
 
@@ -2712,6 +2731,7 @@ static TupleTableSlot *
 ExecMerge(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 		  ItemPointer tupleid, bool canSetTag)
 {
+	TupleTableSlot *rslot = NULL;
 	bool		matched;
 
 	/*-----
@@ -2759,18 +2779,18 @@ ExecMerge(ModifyTableContext *context, R
 	 */
 	matched = tupleid != NULL;
 	if (matched)
-		matched = ExecMergeMatched(context, resultRelInfo, tupleid, canSetTag);
+		rslot = ExecMergeMatched(context, resultRelInfo, tupleid, canSetTag,
+								 &matched);
 
 	/*
-	 * Either we were dealing with a NOT MATCHED tuple or ExecMergeMatched()
-	 * returned "false", indicating the previously MATCHED tuple no longer
-	 * matches.
+	 * Deal with the NOT MATCHED case (either a NOT MATCHED tuple from the
+	 * join, or a previously MATCHED tuple for which ExecMergeMatched() set
+	 * "matched" to false, indicating that it no longer matches).
 	 */
 	if (!matched)
-		ExecMergeNotMatched(context, resultRelInfo, canSetTag);
+		rslot = ExecMergeNotMatched(context, resultRelInfo, canSetTag);
 
-	/* No RETURNING support yet */
-	return NULL;
+	return rslot;
 }
 
 /*
@@ -2780,8 +2800,8 @@ ExecMerge(ModifyTableContext *context, R
  * We start from the first WHEN MATCHED action and check if the WHEN quals
  * pass, if any. If the WHEN quals for the first action do not pass, we
  * check the second, then the third and so on. If we reach to the end, no
- * action is taken and we return true, indicating that no further action is
- * required for this tuple.
+ * action is taken and "matched" is set to true, indicating that no further
+ * action is required for this tuple.
  *
  * If we do find a qualifying action, then we attempt to execute the action.
  *
@@ -2790,16 +2810,17 @@ ExecMerge(ModifyTableContext *context, R
  * with individual actions are evaluated by this routine via ExecQual, while
  * EvalPlanQual checks for the join quals. If EvalPlanQual tells us that the
  * updated tuple still passes the join quals, then we restart from the first
- * action to look for a qualifying action. Otherwise, we return false --
- * meaning that a NOT MATCHED action must now be executed for the current
- * source tuple.
+ * action to look for a qualifying action. Otherwise, "matched" is set to
+ * false -- meaning that a NOT MATCHED action must now be executed for the
+ * current source tuple.
  */
-static bool
+static TupleTableSlot *
 ExecMergeMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
-				 ItemPointer tupleid, bool canSetTag)
+				 ItemPointer tupleid, bool canSetTag, bool *matched)
 {
 	ModifyTableState *mtstate = context->mtstate;
-	TupleTableSlot *newslot;
+	TupleTableSlot *rslot = NULL;
+	TupleTableSlot *newslot = NULL;
 	EState	   *estate = context->estate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
 	bool		isNull;
@@ -2810,7 +2831,10 @@ ExecMergeMatched(ModifyTableContext *con
 	 * If there are no WHEN MATCHED actions, we are done.
 	 */
 	if (resultRelInfo->ri_matchedMergeAction == NIL)
-		return true;
+	{
+		*matched = true;
+		return NULL;
+	}
 
 	/*
 	 * Make tuple and any needed join variables available to ExecQual and
@@ -2890,15 +2914,38 @@ lmerge_matched:
 				newslot = ExecProject(relaction->mas_proj);
 
 				context->relaction = relaction;
+				context->relaction_idx = foreach_current_index(l);
 				if (!ExecUpdatePrologue(context, resultRelInfo,
 										tupleid, NULL, newslot, &result))
 				{
 					if (result == TM_Ok)
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 					break;		/* concurrent update/delete */
 				}
 				result = ExecUpdateAct(context, resultRelInfo, tupleid, NULL,
 									   newslot, false, &updateCxt);
+
+				/*
+				 * If ExecUpdateAct reports that a cross-partition update was
+				 * done, then the RETURNING tuple (if any) has been projected
+				 * and there's nothing else for us to do.
+				 *
+				 * XXX: Pre-existing bug here? For a cross-partition update,
+				 * this should not be calling ExecUpdateEpilogue(). Row update
+				 * triggers should not be fired, etc.
+				 */
+				if (updateCxt.crossPartUpdate)
+				{
+					*matched = true;
+					mtstate->mt_merge_updated += 1;
+					if (canSetTag)
+						(estate->es_processed)++;
+					return context->cpUpdateReturningSlot;
+				}
+
 				if (result == TM_Ok && updateCxt.updated)
 				{
 					ExecUpdateEpilogue(context, &updateCxt, resultRelInfo,
@@ -2909,11 +2956,15 @@ lmerge_matched:
 
 			case CMD_DELETE:
 				context->relaction = relaction;
+				context->relaction_idx = foreach_current_index(l);
 				if (!ExecDeletePrologue(context, resultRelInfo, tupleid,
 										NULL, NULL, &result))
 				{
 					if (result == TM_Ok)
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 					break;		/* concurrent update/delete */
 				}
 				result = ExecDeleteAct(context, resultRelInfo, tupleid, false);
@@ -2969,7 +3020,8 @@ lmerge_matched:
 				 * If the tuple was already deleted, return to let caller
 				 * handle it under NOT MATCHED clauses.
 				 */
-				return false;
+				*matched = false;
+				return NULL;
 
 			case TM_Updated:
 				{
@@ -3015,13 +3067,19 @@ lmerge_matched:
 							 * NOT MATCHED actions.
 							 */
 							if (TupIsNull(epqslot))
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							(void) ExecGetJunkAttribute(epqslot,
 														resultRelInfo->ri_RowIdAttNo,
 														&isNull);
 							if (isNull)
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							/*
 							 * When a tuple was updated and migrated to
@@ -3056,7 +3114,8 @@ lmerge_matched:
 							 * tuple already deleted; tell caller to run NOT
 							 * MATCHED actions
 							 */
-							return false;
+							*matched = false;
+							return NULL;
 
 						case TM_SelfModified:
 
@@ -3076,13 +3135,14 @@ lmerge_matched:
 										(errcode(ERRCODE_TRIGGERED_DATA_CHANGE_VIOLATION),
 										 errmsg("tuple to be updated or deleted was already modified by an operation triggered by the current command"),
 										 errhint("Consider using an AFTER trigger instead of a BEFORE trigger to propagate changes to other rows.")));
-							return false;
+							*matched = false;
+							return NULL;
 
 						default:
 							/* see table_tuple_lock call in ExecDelete() */
 							elog(ERROR, "unexpected table_tuple_lock status: %u",
 								 result);
-							return false;
+							return NULL;
 					}
 				}
 
@@ -3095,6 +3155,36 @@ lmerge_matched:
 		}
 
 		/*
+		 * Process RETURNING if present, using the action's RETURNING list.
+		 * Using the root relation's action is correct here, since we know
+		 * that this wasn't a cross-partition update at this point.
+		 */
+		if (relaction->mas_proj_returning)
+		{
+			resultRelInfo->ri_projectReturning = relaction->mas_proj_returning;
+
+			switch (commandType)
+			{
+				case CMD_UPDATE:
+					rslot = ExecProcessReturning(resultRelInfo, newslot,
+												 context->planSlot);
+					break;
+
+				case CMD_DELETE:
+					rslot = ExecProcessReturning(resultRelInfo,
+												 resultRelInfo->ri_oldTupleSlot,
+												 context->planSlot);
+					break;
+
+				case CMD_NOTHING:
+					break;
+
+				default:
+					elog(ERROR, "unknown action in MERGE WHEN MATCHED clause");
+			}
+		}
+
+		/*
 		 * We've activated one of the WHEN clauses, so we don't search
 		 * further. This is required behaviour, not an optimization.
 		 */
@@ -3104,19 +3194,22 @@ lmerge_matched:
 	/*
 	 * Successfully executed an action or no qualifying action was found.
 	 */
-	return true;
+	*matched = true;
+
+	return rslot;
 }
 
 /*
  * Execute the first qualifying NOT MATCHED action.
  */
-static void
+static TupleTableSlot *
 ExecMergeNotMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 					bool canSetTag)
 {
 	ModifyTableState *mtstate = context->mtstate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
 	List	   *actionStates = NIL;
+	TupleTableSlot *rslot = NULL;
 	ListCell   *l;
 
 	/*
@@ -3167,9 +3260,10 @@ ExecMergeNotMatched(ModifyTableContext *
 				 */
 				newslot = ExecProject(action->mas_proj);
 				context->relaction = action;
+				context->relaction_idx = foreach_current_index(l);
 
-				(void) ExecInsert(context, mtstate->rootResultRelInfo, newslot,
-								  canSetTag, NULL, NULL);
+				rslot = ExecInsert(context, mtstate->rootResultRelInfo, newslot,
+								   canSetTag, NULL, NULL);
 				mtstate->mt_merge_inserted += 1;
 				break;
 			case CMD_NOTHING:
@@ -3185,6 +3279,8 @@ ExecMergeNotMatched(ModifyTableContext *
 		 */
 		break;
 	}
+
+	return rslot;
 }
 
 /*
@@ -3334,6 +3430,15 @@ ExecInitMerge(ModifyTableState *mtstate,
 					elog(ERROR, "unknown operation");
 					break;
 			}
+
+			/* Build a projection for the action's RETURNING list, if any */
+			if (action->returningList)
+				action_state->mas_proj_returning =
+					ExecBuildProjectionInfo(action->returningList,
+											econtext,
+											mtstate->ps.ps_ResultTupleSlot,
+											&mtstate->ps,
+											relationDesc);
 		}
 	}
 }
@@ -3647,8 +3752,17 @@ ExecModifyTable(PlanState *pstate)
 				{
 					EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-					ExecMerge(&context, node->resultRelInfo, NULL, node->canSetTag);
-					continue;	/* no RETURNING support yet */
+					slot = ExecMerge(&context, node->resultRelInfo, NULL,
+									 node->canSetTag);
+
+					/*
+					 * If we got a RETURNING result, return it to the caller.
+					 * We'll continue the work on next call.
+					 */
+					if (slot)
+						return slot;
+
+					continue;	/* continue with the next tuple */
 				}
 
 				elog(ERROR, "tableoid is NULL");
@@ -3725,8 +3839,17 @@ ExecModifyTable(PlanState *pstate)
 					{
 						EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-						ExecMerge(&context, node->resultRelInfo, NULL, node->canSetTag);
-						continue;	/* no RETURNING support yet */
+						slot = ExecMerge(&context, node->resultRelInfo, NULL,
+										 node->canSetTag);
+
+						/*
+						 * If we got a RETURNING result, return it to the
+						 * caller.  We'll continue the work on next call.
+						 */
+						if (slot)
+							return slot;
+
+						continue;	/* continue with the next tuple */
 					}
 
 					elog(ERROR, "ctid is NULL");
@@ -4192,18 +4315,23 @@ ExecInitModifyTable(ModifyTable *node, E
 		econtext = mtstate->ps.ps_ExprContext;
 
 		/*
-		 * Build a projection for each result rel.
+		 * Build a projection for each result rel.  We skip this for MERGE,
+		 * since it uses per-action RETURNING lists, which each have their own
+		 * projections, built in ExecInitMerge().
 		 */
-		resultRelInfo = mtstate->resultRelInfo;
-		foreach(l, node->returningLists)
+		if (operation != CMD_MERGE)
 		{
-			List	   *rlist = (List *) lfirst(l);
+			resultRelInfo = mtstate->resultRelInfo;
+			foreach(l, node->returningLists)
+			{
+				List	   *rlist = (List *) lfirst(l);
 
-			resultRelInfo->ri_returningList = rlist;
-			resultRelInfo->ri_projectReturning =
-				ExecBuildProjectionInfo(rlist, econtext, slot, &mtstate->ps,
-										resultRelInfo->ri_RelationDesc->rd_att);
-			resultRelInfo++;
+				resultRelInfo->ri_returningList = rlist;
+				resultRelInfo->ri_projectReturning =
+					ExecBuildProjectionInfo(rlist, econtext, slot, &mtstate->ps,
+											resultRelInfo->ri_RelationDesc->rd_att);
+				resultRelInfo++;
+			}
 		}
 	}
 	else
diff --git a/src/backend/executor/spi.c b/src/backend/executor/spi.c
new file mode 100644
index 3397568..b9f443c
--- a/src/backend/executor/spi.c
+++ b/src/backend/executor/spi.c
@@ -2033,6 +2033,8 @@ SPI_result_code_string(int code)
 			return "SPI_OK_TD_REGISTER";
 		case SPI_OK_MERGE:
 			return "SPI_OK_MERGE";
+		case SPI_OK_MERGE_RETURNING:
+			return "SPI_OK_MERGE_RETURNING";
 	}
 	/* Unrecognized code ... return something useful ... */
 	sprintf(buf, "Unrecognized SPI code %d", code);
@@ -2886,7 +2888,10 @@ _SPI_pquery(QueryDesc *queryDesc, bool f
 				res = SPI_OK_UPDATE;
 			break;
 		case CMD_MERGE:
-			res = SPI_OK_MERGE;
+			if (queryDesc->plannedstmt->hasReturning)
+				res = SPI_OK_MERGE_RETURNING;
+			else
+				res = SPI_OK_MERGE;
 			break;
 		default:
 			return SPI_ERROR_OPUNKNOWN;
diff --git a/src/backend/nodes/nodeFuncs.c b/src/backend/nodes/nodeFuncs.c
new file mode 100644
index c03f4f2..5ca63ae
--- a/src/backend/nodes/nodeFuncs.c
+++ b/src/backend/nodes/nodeFuncs.c
@@ -2440,6 +2440,8 @@ expression_tree_walker_impl(Node *node,
 					return true;
 				if (WALK(action->targetList))
 					return true;
+				if (WALK(action->returningList))
+					return true;
 			}
 			break;
 		case T_PartitionPruneStepOp:
@@ -3421,6 +3423,7 @@ expression_tree_mutator_impl(Node *node,
 				FLATCOPY(newnode, action, MergeAction);
 				MUTATE(newnode->qual, action->qual, Node *);
 				MUTATE(newnode->targetList, action->targetList, List *);
+				MUTATE(newnode->returningList, action->returningList, List *);
 
 				return (Node *) newnode;
 			}
@@ -4061,6 +4064,8 @@ raw_expression_tree_walker_impl(Node *no
 					return true;
 				if (WALK(mergeWhenClause->values))
 					return true;
+				if (WALK(mergeWhenClause->returningList))
+					return true;
 			}
 			break;
 		case T_SelectStmt:
diff --git a/src/backend/optimizer/plan/planner.c b/src/backend/optimizer/plan/planner.c
new file mode 100644
index a8cea5e..07d4f20
--- a/src/backend/optimizer/plan/planner.c
+++ b/src/backend/optimizer/plan/planner.c
@@ -908,6 +908,11 @@ subquery_planner(PlannerGlobal *glob, Qu
 			preprocess_expression(root,
 								  (Node *) action->qual,
 								  EXPRKIND_QUAL);
+
+		action->returningList = (List *)
+			preprocess_expression(root,
+								  (Node *) action->returningList,
+								  EXPRKIND_TARGET);
 	}
 
 	root->append_rel_list = (List *)
@@ -1884,6 +1889,11 @@ grouping_planner(PlannerInfo *root, doub
 																  (Node *) action->targetList,
 																  this_result_rel,
 																  top_result_rel);
+							leaf_action->returningList = (List *)
+								adjust_appendrel_attrs_multilevel(root,
+																  (Node *) action->returningList,
+																  this_result_rel,
+																  top_result_rel);
 							if (leaf_action->commandType == CMD_UPDATE)
 								leaf_action->updateColnos =
 									adjust_inherited_attnums_multilevel(root,
diff --git a/src/backend/optimizer/plan/setrefs.c b/src/backend/optimizer/plan/setrefs.c
new file mode 100644
index fc37095..000e204
--- a/src/backend/optimizer/plan/setrefs.c
+++ b/src/backend/optimizer/plan/setrefs.c
@@ -1192,6 +1192,15 @@ set_plan_refs(PlannerInfo *root, Plan *p
 																  rtoffset,
 																  NRM_EQUAL,
 																  NUM_EXEC_QUAL(plan));
+
+							/* Fix returningList too. */
+							action->returningList =
+								set_returning_clause_references(root,
+																action->returningList,
+																subplan,
+																resultrel,
+																rtoffset);
+
 						}
 					}
 				}
diff --git a/src/backend/optimizer/prep/prepjointree.c b/src/backend/optimizer/prep/prepjointree.c
new file mode 100644
index 73ff407..ca2b176
--- a/src/backend/optimizer/prep/prepjointree.c
+++ b/src/backend/optimizer/prep/prepjointree.c
@@ -2153,6 +2153,8 @@ perform_pullup_replace_vars(PlannerInfo
 			action->qual = pullup_replace_vars(action->qual, rvcontext);
 			action->targetList = (List *)
 				pullup_replace_vars((Node *) action->targetList, rvcontext);
+			action->returningList = (List *)
+				pullup_replace_vars((Node *) action->returningList, rvcontext);
 		}
 	}
 	replace_vars_in_jointree((Node *) parse->jointree, rvcontext);
diff --git a/src/backend/optimizer/prep/preptlist.c b/src/backend/optimizer/prep/preptlist.c
new file mode 100644
index 9d46488..d209371
--- a/src/backend/optimizer/prep/preptlist.c
+++ b/src/backend/optimizer/prep/preptlist.c
@@ -144,6 +144,7 @@ preprocess_targetlist(PlannerInfo *root)
 		foreach(l, parse->mergeActionList)
 		{
 			MergeAction *action = (MergeAction *) lfirst(l);
+			List	   *exprs;
 			List	   *vars;
 			ListCell   *l2;
 
@@ -156,14 +157,14 @@ preprocess_targetlist(PlannerInfo *root)
 
 			/*
 			 * Add resjunk entries for any Vars and PlaceHolderVars used in
-			 * each action's targetlist and WHEN condition that belong to
-			 * relations other than the target.  We don't expect to see any
-			 * aggregates or window functions here.
+			 * each action's targetlist, WHEN condition, and returningList
+			 * that belong to relations other than the target.  We don't
+			 * expect to see any aggregates or window functions here.
 			 */
-			vars = pull_var_clause((Node *)
-								   list_concat_copy((List *) action->qual,
-													action->targetList),
-								   PVC_INCLUDE_PLACEHOLDERS);
+			exprs = list_concat_copy((List *) action->qual, action->targetList);
+			exprs = list_concat(exprs, action->returningList);
+			vars = pull_var_clause((Node *) exprs, PVC_INCLUDE_PLACEHOLDERS);
+
 			foreach(l2, vars)
 			{
 				Var		   *var = (Var *) lfirst(l2);
diff --git a/src/backend/parser/analyze.c b/src/backend/parser/analyze.c
new file mode 100644
index 7a1dfb6..a7021d2
--- a/src/backend/parser/analyze.c
+++ b/src/backend/parser/analyze.c
@@ -74,7 +74,6 @@ static void determineRecursiveColTypes(P
 									   Node *larg, List *nrtargetlist);
 static Query *transformReturnStmt(ParseState *pstate, ReturnStmt *stmt);
 static Query *transformUpdateStmt(ParseState *pstate, UpdateStmt *stmt);
-static List *transformReturningList(ParseState *pstate, List *returningList);
 static Query *transformPLAssignStmt(ParseState *pstate,
 									PLAssignStmt *stmt);
 static Query *transformDeclareCursorStmt(ParseState *pstate,
@@ -2539,9 +2538,9 @@ transformUpdateTargetList(ParseState *ps
 
 /*
  * transformReturningList -
- *	handle a RETURNING clause in INSERT/UPDATE/DELETE
+ *	handle a RETURNING clause in INSERT/UPDATE/DELETE/MERGE
  */
-static List *
+List *
 transformReturningList(ParseState *pstate, List *returningList)
 {
 	List	   *rlist;
diff --git a/src/backend/parser/gram.y b/src/backend/parser/gram.y
new file mode 100644
index c224df4..b4ae1c5
--- a/src/backend/parser/gram.y
+++ b/src/backend/parser/gram.y
@@ -438,7 +438,7 @@ static Node *makeRecursiveViewSelect(cha
 				any_operator expr_list attrs
 				distinct_clause opt_distinct_clause
 				target_list opt_target_list insert_column_list set_target_list
-				merge_values_clause
+				merge_values_clause merge_returning_clause merge_target_list
 				set_clause_list set_clause
 				def_list operator_def_list indirection opt_indirection
 				reloption_list TriggerFuncArgs opclass_item_list opclass_drop_list
@@ -550,7 +550,7 @@ static Node *makeRecursiveViewSelect(cha
 %type <range>	extended_relation_expr
 %type <range>	relation_expr_opt_alias
 %type <node>	tablesample_clause opt_repeatable_clause
-%type <target>	target_el set_target insert_column_item
+%type <target>	target_el set_target insert_column_item merge_target_el
 
 %type <str>		generic_option_name
 %type <node>	generic_option_arg
@@ -835,7 +835,8 @@ static Node *makeRecursiveViewSelect(cha
  * the same as non-keywords, reducing the risk of unwanted precedence effects.
  *
  * We need to do this for PARTITION, RANGE, ROWS, and GROUPS to support
- * opt_existing_window_name (see comment there).
+ * opt_existing_window_name (see comment there), and for WHEN to support
+ * merge_target_el.
  *
  * The frame_bound productions UNBOUNDED PRECEDING and UNBOUNDED FOLLOWING
  * are even messier: since UNBOUNDED is an unreserved keyword (per spec!),
@@ -852,7 +853,7 @@ static Node *makeRecursiveViewSelect(cha
  * Using the same precedence as IDENT seems right for the reasons given above.
  */
 %nonassoc	UNBOUNDED		/* ideally would have same precedence as IDENT */
-%nonassoc	IDENT PARTITION RANGE ROWS GROUPS PRECEDING FOLLOWING CUBE ROLLUP
+%nonassoc	IDENT PARTITION RANGE ROWS GROUPS PRECEDING FOLLOWING CUBE ROLLUP WHEN
 %left		Op OPERATOR		/* multi-character ops and user-defined operators */
 %left		'+' '-'
 %left		'*' '/' '%'
@@ -12379,6 +12380,7 @@ merge_when_clause:
 					m->matched = true;
 					m->commandType = CMD_NOTHING;
 					m->condition = $3;
+					m->location = @1;
 
 					$$ = (Node *) m;
 				}
@@ -12389,6 +12391,7 @@ merge_when_clause:
 					m->matched = false;
 					m->commandType = CMD_NOTHING;
 					m->condition = $4;
+					m->location = @1;
 
 					$$ = (Node *) m;
 				}
@@ -12400,75 +12403,89 @@ opt_merge_when_condition:
 		;
 
 merge_update:
-			UPDATE SET set_clause_list
+			UPDATE SET set_clause_list merge_returning_clause
 				{
 					MergeWhenClause *n = makeNode(MergeWhenClause);
 					n->commandType = CMD_UPDATE;
 					n->override = OVERRIDING_NOT_SET;
 					n->targetList = $3;
 					n->values = NIL;
+					n->returningList = $4;
+					n->location = @1;
 
 					$$ = n;
 				}
 		;
 
 merge_delete:
-			DELETE_P
+			DELETE_P merge_returning_clause
 				{
 					MergeWhenClause *n = makeNode(MergeWhenClause);
 					n->commandType = CMD_DELETE;
 					n->override = OVERRIDING_NOT_SET;
 					n->targetList = NIL;
 					n->values = NIL;
+					n->returningList = $2;
+					n->location = @1;
 
 					$$ = n;
 				}
 		;
 
 merge_insert:
-			INSERT merge_values_clause
+			INSERT merge_values_clause merge_returning_clause
 				{
 					MergeWhenClause *n = makeNode(MergeWhenClause);
 					n->commandType = CMD_INSERT;
 					n->override = OVERRIDING_NOT_SET;
 					n->targetList = NIL;
 					n->values = $2;
+					n->returningList = $3;
+					n->location = @1;
 					$$ = n;
 				}
-			| INSERT OVERRIDING override_kind VALUE_P merge_values_clause
+			| INSERT OVERRIDING override_kind VALUE_P merge_values_clause merge_returning_clause
 				{
 					MergeWhenClause *n = makeNode(MergeWhenClause);
 					n->commandType = CMD_INSERT;
 					n->override = $3;
 					n->targetList = NIL;
 					n->values = $5;
+					n->returningList = $6;
+					n->location = @1;
 					$$ = n;
 				}
-			| INSERT '(' insert_column_list ')' merge_values_clause
+			| INSERT '(' insert_column_list ')' merge_values_clause merge_returning_clause
 				{
 					MergeWhenClause *n = makeNode(MergeWhenClause);
 					n->commandType = CMD_INSERT;
 					n->override = OVERRIDING_NOT_SET;
 					n->targetList = $3;
 					n->values = $5;
+					n->returningList = $6;
+					n->location = @1;
 					$$ = n;
 				}
-			| INSERT '(' insert_column_list ')' OVERRIDING override_kind VALUE_P merge_values_clause
+			| INSERT '(' insert_column_list ')' OVERRIDING override_kind VALUE_P merge_values_clause merge_returning_clause
 				{
 					MergeWhenClause *n = makeNode(MergeWhenClause);
 					n->commandType = CMD_INSERT;
 					n->override = $6;
 					n->targetList = $3;
 					n->values = $8;
+					n->returningList = $9;
+					n->location = @1;
 					$$ = n;
 				}
-			| INSERT DEFAULT VALUES
+			| INSERT DEFAULT VALUES merge_returning_clause
 				{
 					MergeWhenClause *n = makeNode(MergeWhenClause);
 					n->commandType = CMD_INSERT;
 					n->override = OVERRIDING_NOT_SET;
 					n->targetList = NIL;
 					n->values = NIL;
+					n->returningList = $4;
+					n->location = @1;
 					$$ = n;
 				}
 		;
@@ -12480,6 +12497,64 @@ merge_values_clause:
 				}
 		;
 
+merge_returning_clause:
+			RETURNING merge_target_list		{ $$ = $2; }
+			| /* EMPTY */					{ $$ = NIL; }
+		;
+
+merge_target_list:
+			merge_target_el								{ $$ = list_make1($1); }
+			| merge_target_list ',' merge_target_el		{ $$ = lappend($1, $3); }
+		;
+
+/*
+ * Given "RETURNING expr when ...", we have to decide without looking any
+ * further ahead whether the "when" is an alias or the start of the next WHEN
+ * clause.  We resolve the shift/reduce conflict by giving the first
+ * merge_target_el production a slightly higher precedence than the WHEN token
+ * has, causing the parser to prefer to reduce, in effect assuming that the
+ * WHEN is not an alias.
+ */
+merge_target_el:
+			a_expr			%prec Op
+				{
+					$$ = makeNode(ResTarget);
+					$$->name = NULL;
+					$$->indirection = NIL;
+					$$->val = (Node *) $1;
+					$$->location = @1;
+				}
+			| a_expr BareColLabel
+				{
+					$$ = makeNode(ResTarget);
+					$$->name = $2;
+					$$->indirection = NIL;
+					$$->val = (Node *) $1;
+					$$->location = @1;
+				}
+			| a_expr AS ColLabel
+				{
+					$$ = makeNode(ResTarget);
+					$$->name = $3;
+					$$->indirection = NIL;
+					$$->val = (Node *) $1;
+					$$->location = @1;
+				}
+			| '*'
+				{
+					ColumnRef  *n = makeNode(ColumnRef);
+
+					n->fields = list_make1(makeNode(A_Star));
+					n->location = @1;
+
+					$$ = makeNode(ResTarget);
+					$$->name = NULL;
+					$$->indirection = NIL;
+					$$->val = (Node *) n;
+					$$->location = @1;
+				}
+		;
+
 /*****************************************************************************
  *
  *		QUERY:
diff --git a/src/backend/parser/parse_cte.c b/src/backend/parser/parse_cte.c
new file mode 100644
index 6992a78..5cad6d8
--- a/src/backend/parser/parse_cte.c
+++ b/src/backend/parser/parse_cte.c
@@ -126,13 +126,6 @@ transformWithClause(ParseState *pstate,
 		CommonTableExpr *cte = (CommonTableExpr *) lfirst(lc);
 		ListCell   *rest;
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(cte->ctequery, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in WITH query"),
-					parser_errposition(pstate, cte->location));
-
 		for_each_cell(rest, withClause->ctes, lnext(withClause->ctes, lc))
 		{
 			CommonTableExpr *cte2 = (CommonTableExpr *) lfirst(rest);
@@ -153,7 +146,8 @@ transformWithClause(ParseState *pstate,
 			/* must be a data-modifying statement */
 			Assert(IsA(cte->ctequery, InsertStmt) ||
 				   IsA(cte->ctequery, UpdateStmt) ||
-				   IsA(cte->ctequery, DeleteStmt));
+				   IsA(cte->ctequery, DeleteStmt) ||
+				   IsA(cte->ctequery, MergeStmt));
 
 			pstate->p_hasModifyingCTE = true;
 		}
diff --git a/src/backend/parser/parse_merge.c b/src/backend/parser/parse_merge.c
new file mode 100644
index 91b1156..367857a
--- a/src/backend/parser/parse_merge.c
+++ b/src/backend/parser/parse_merge.c
@@ -18,11 +18,13 @@
 #include "access/sysattr.h"
 #include "miscadmin.h"
 #include "nodes/makefuncs.h"
+#include "nodes/nodeFuncs.h"
 #include "parser/analyze.h"
 #include "parser/parse_collate.h"
 #include "parser/parsetree.h"
 #include "parser/parser.h"
 #include "parser/parse_clause.h"
+#include "parser/parse_coerce.h"
 #include "parser/parse_cte.h"
 #include "parser/parse_expr.h"
 #include "parser/parse_merge.h"
@@ -103,6 +105,8 @@ transformMergeStmt(ParseState *pstate, M
 	List	   *mergeActionList;
 	Node	   *joinExpr;
 	ParseNamespaceItem *nsitem;
+	int			returning_list_length = -1;
+	List	  **returning_colexprs = NULL;
 
 	/* There can't be any outer WITH to worry about */
 	Assert(pstate->p_ctenamespace == NIL);
@@ -385,14 +389,146 @@ transformMergeStmt(ParseState *pstate, M
 				elog(ERROR, "unknown action in MERGE WHEN clause");
 		}
 
+		/*
+		 * Transform the RETURNING list, if any.  All action RETURNING lists
+		 * must be the same length, *after* transformation (which might expand
+		 * '*' into multiple items).  At this stage, we prevent resolving any
+		 * unknown-type outputs as TEXT, allowing us to resolve such outputs
+		 * using select_common_type() below.
+		 */
+		if (mergeWhenClause->returningList)
+		{
+			bool		save_resolve_unknowns = pstate->p_resolve_unknowns;
+			int			i;
+			ListCell   *lc2;
+
+			/* The RETURNING list can see both target and source relations */
+			setNamespaceVisibilityForRTE(pstate->p_namespace,
+										 rt_fetch(qry->resultRelation,
+												  pstate->p_rtable),
+										 true, true);
+
+			setNamespaceVisibilityForRTE(pstate->p_namespace,
+										 rt_fetch(sourceRTI,
+												  pstate->p_rtable),
+										 true, true);
+
+			pstate->p_resolve_unknowns = false;
+
+			action->returningList = transformReturningList(pstate,
+														   mergeWhenClause->returningList);
+
+			pstate->p_resolve_unknowns = save_resolve_unknowns;
+
+			if (returning_list_length < 0)
+			{
+				/* Remember post-transformation length of first list */
+				returning_list_length = list_length(action->returningList);
+				/* and allocate array for per-column lists */
+				returning_colexprs = (List **) palloc0(returning_list_length * sizeof(List *));
+			}
+			else if (list_length(action->returningList) != returning_list_length)
+				ereport(ERROR,
+						errcode(ERRCODE_SYNTAX_ERROR),
+						errmsg("RETURNING lists must all be the same length"),
+						parser_errposition(pstate,
+										   exprLocation((Node *) action->returningList)));
+
+			/* Build per-column RETURNING expression lists */
+			i = 0;
+			foreach(lc2, action->returningList)
+			{
+				TargetEntry *tle = lfirst_node(TargetEntry, lc2);
+
+				returning_colexprs[i] = lappend(returning_colexprs[i], tle->expr);
+				i++;
+			}
+		}
+		else
+			action->returningList = NIL;
+
 		mergeActionList = lappend(mergeActionList, action);
 	}
 
-	qry->mergeActionList = mergeActionList;
+	/*
+	 * Enforce the rule that if any action has a RETURNING list, all must
+	 * have (except for DO NOTHING actions).  We do this using a second pass
+	 * over the merge actions to generate a more useful error message.
+	 */
+	if (returning_list_length > 0)
+	{
+		ListCell   *l2;
 
-	/* RETURNING could potentially be added in the future, but not in SQL std */
+		forboth(l, stmt->mergeWhenClauses, l2, mergeActionList)
+		{
+			MergeWhenClause *mergeWhenClause = lfirst_node(MergeWhenClause, l);
+			MergeAction *action = lfirst_node(MergeAction, l2);
+
+			if (action->commandType != CMD_NOTHING && !action->returningList)
+				ereport(ERROR,
+						errcode(ERRCODE_SYNTAX_ERROR),
+						errmsg("missing RETURNING list for MERGE action"),
+						errhint("To enable RETURNING for MERGE, add a RETURNING list to every INSERT/UPDATE/DELETE action."),
+						parser_errposition(pstate, mergeWhenClause->location));
+		}
+	}
+
+	/*
+	 * Resolve the common types of the columns of any RETURNING lists, and
+	 * coerce everything to those types.
+	 *
+	 * Then identify the common typmod and common collation, if any, of each
+	 * column, and build a top-level RETURNING list for the Query node, for
+	 * use in outer queries.  This RETURNING list is never evaluated, and just
+	 * contains dummy expressions, to hold the common type, typmod, and
+	 * collation information returned by this query.
+	 */
 	qry->returningList = NULL;
 
+	for (int i = 0; i < returning_list_length; i++)
+	{
+		TargetEntry *first_tle = NULL;
+		Oid			coltype;
+		int32		coltypmod;
+		Oid			colcoll;
+		Const	   *con;
+		TargetEntry *new_tle;
+
+		coltype = select_common_type(pstate, returning_colexprs[i],
+									 "RETURNING", NULL);
+
+		foreach(l, mergeActionList)
+		{
+			MergeAction *action = lfirst_node(MergeAction, l);
+
+			if (action->returningList)
+			{
+				TargetEntry *tle;
+
+				tle = list_nth_node(TargetEntry, action->returningList, i);
+				tle->expr = (Expr *) coerce_to_common_type(pstate,
+														   (Node *) tle->expr,
+														   coltype,
+														   "RETURNING");
+				if (first_tle == NULL)
+					first_tle = tle;
+			}
+		}
+
+		coltypmod = select_common_typmod(pstate, returning_colexprs[i], coltype);
+		colcoll = select_common_collation(pstate, returning_colexprs[i], true);
+
+		con = makeNullConst(coltype, coltypmod, colcoll);
+		new_tle = makeTargetEntry((Expr *) con,
+								  first_tle->resno,
+								  pstrdup(first_tle->resname),
+								  first_tle->resjunk);
+
+		qry->returningList = lappend(qry->returningList, new_tle);
+	}
+
+	qry->mergeActionList = mergeActionList;
+
 	qry->hasTargetSRFs = false;
 	qry->hasSubLinks = pstate->p_hasSubLinks;
 
diff --git a/src/backend/parser/parse_relation.c b/src/backend/parser/parse_relation.c
new file mode 100644
index 864ea9b..cc19041
--- a/src/backend/parser/parse_relation.c
+++ b/src/backend/parser/parse_relation.c
@@ -2346,9 +2346,10 @@ addRangeTableEntryForCTE(ParseState *pst
 		cte->cterefcount++;
 
 	/*
-	 * We throw error if the CTE is INSERT/UPDATE/DELETE without RETURNING.
-	 * This won't get checked in case of a self-reference, but that's OK
-	 * because data-modifying CTEs aren't allowed to be recursive anyhow.
+	 * We throw error if the CTE is INSERT/UPDATE/DELETE/MERGE without
+	 * RETURNING.  This won't get checked in case of a self-reference, but
+	 * that's OK because data-modifying CTEs aren't allowed to be recursive
+	 * anyhow.
 	 */
 	if (IsA(cte->ctequery, Query))
 	{
diff --git a/src/backend/rewrite/rewriteHandler.c b/src/backend/rewrite/rewriteHandler.c
new file mode 100644
index 41a3623..3786a7c
--- a/src/backend/rewrite/rewriteHandler.c
+++ b/src/backend/rewrite/rewriteHandler.c
@@ -3615,9 +3615,9 @@ RewriteQuery(Query *parsetree, List *rew
 	ListCell   *lc1;
 
 	/*
-	 * First, recursively process any insert/update/delete statements in WITH
-	 * clauses.  (We have to do this first because the WITH clauses may get
-	 * copied into rule actions below.)
+	 * First, recursively process any insert/update/delete/merge statements in
+	 * WITH clauses.  (We have to do this first because the WITH clauses may
+	 * get copied into rule actions below.)
 	 */
 	foreach(lc1, parsetree->cteList)
 	{
@@ -3642,7 +3642,8 @@ RewriteQuery(Query *parsetree, List *rew
 			if (!(ctequery->commandType == CMD_SELECT ||
 				  ctequery->commandType == CMD_UPDATE ||
 				  ctequery->commandType == CMD_INSERT ||
-				  ctequery->commandType == CMD_DELETE))
+				  ctequery->commandType == CMD_DELETE ||
+				  ctequery->commandType == CMD_MERGE))
 			{
 				/*
 				 * Currently it could only be NOTIFY; this error message will
diff --git a/src/backend/rewrite/rowsecurity.c b/src/backend/rewrite/rowsecurity.c
new file mode 100644
index b1620e4..966dcf0
--- a/src/backend/rewrite/rowsecurity.c
+++ b/src/backend/rewrite/rowsecurity.c
@@ -395,10 +395,10 @@ get_row_security_policies(Query *root, R
 	 * on the final action we take.
 	 *
 	 * We already fetched the SELECT policies above, to check existing rows,
-	 * but we must also check that new rows created by UPDATE actions are
-	 * visible, if SELECT rights are required for this relation. We don't do
-	 * this for INSERT actions, since an INSERT command would only do this
-	 * check if it had a RETURNING list, and MERGE does not support RETURNING.
+	 * but we must also check that new rows created by INSERT/UPDATE actions
+	 * are visible, if SELECT rights are required. For INSERT actions, we only
+	 * do this if RETURNING is specified, to be consistent with a plain INSERT
+	 * command, which can only require SELECT rights when RETURNING is used.
 	 *
 	 * We don't push the UPDATE/DELETE USING quals to the RTE because we don't
 	 * really want to apply them while scanning the relation since we don't
@@ -420,6 +420,8 @@ get_row_security_policies(Query *root, R
 		List	   *merge_delete_restrictive_policies;
 		List	   *merge_insert_permissive_policies;
 		List	   *merge_insert_restrictive_policies;
+		List	   *merge_select_permissive_policies = NIL;
+		List	   *merge_select_restrictive_policies = NIL;
 
 		/*
 		 * Fetch the UPDATE policies and set them up to execute on the
@@ -457,9 +459,6 @@ get_row_security_policies(Query *root, R
 		 */
 		if (perminfo->requiredPerms & ACL_SELECT)
 		{
-			List	   *merge_select_permissive_policies;
-			List	   *merge_select_restrictive_policies;
-
 			get_policies_for_relation(rel, CMD_SELECT, user_id,
 									  &merge_select_permissive_policies,
 									  &merge_select_restrictive_policies);
@@ -508,6 +507,21 @@ get_row_security_policies(Query *root, R
 							   withCheckOptions,
 							   hasSubLinks,
 							   false);
+
+		/*
+		 * Add ALL/SELECT policies as WCO_RLS_INSERT_CHECK WCOs, to ensure
+		 * that the inserted row is visible when executing an INSERT action,
+		 * if RETURNING is specified and SELECT rights are required for this
+		 * relation.
+		 */
+		if (perminfo->requiredPerms & ACL_SELECT && root->returningList)
+			add_with_check_options(rel, rt_index,
+								   WCO_RLS_INSERT_CHECK,
+								   merge_select_permissive_policies,
+								   merge_select_restrictive_policies,
+								   withCheckOptions,
+								   hasSubLinks,
+								   true);
 	}
 
 	table_close(rel, NoLock);
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
new file mode 100644
index e3ccf6c..6f3dc89
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -2142,11 +2142,10 @@ QueryReturnsTuples(Query *parsetree)
 		case CMD_SELECT:
 			/* returns tuples */
 			return true;
-		case CMD_MERGE:
-			return false;
 		case CMD_INSERT:
 		case CMD_UPDATE:
 		case CMD_DELETE:
+		case CMD_MERGE:
 			/* the forms with RETURNING return tuples */
 			if (parsetree->returningList)
 				return true;
diff --git a/src/backend/utils/adt/ruleutils.c b/src/backend/utils/adt/ruleutils.c
new file mode 100644
index ed7f40f..b99118c
--- a/src/backend/utils/adt/ruleutils.c
+++ b/src/backend/utils/adt/ruleutils.c
@@ -7195,10 +7195,16 @@ get_merge_query_def(Query *query, depars
 			appendStringInfoString(buf, "DELETE");
 		else if (action->commandType == CMD_NOTHING)
 			appendStringInfoString(buf, "DO NOTHING");
-	}
 
-	/* No RETURNING support in MERGE yet */
-	Assert(query->returningList == NIL);
+		/* Add RETURNING if present */
+		if (action->returningList)
+		{
+			appendContextKeyword(context, " RETURNING",
+								 -PRETTYINDENT_STD, PRETTYINDENT_STD, 4);
+			get_target_list(action->returningList, context, NULL,
+							colNamesVisible);
+		}
+	}
 }
 
 
diff --git a/src/bin/psql/common.c b/src/bin/psql/common.c
new file mode 100644
index daabf6f..6b63214
--- a/src/bin/psql/common.c
+++ b/src/bin/psql/common.c
@@ -983,13 +983,17 @@ PrintQueryResult(PGresult *result, bool
 			else
 				success = true;
 
-			/* if it's INSERT/UPDATE/DELETE RETURNING, also print status */
+			/*
+			 * If it's INSERT/UPDATE/DELETE/MERGE RETURNING, also print
+			 * status.
+			 */
 			if (last || pset.show_all_results)
 			{
 				cmdstatus = PQcmdStatus(result);
 				if (strncmp(cmdstatus, "INSERT", 6) == 0 ||
 					strncmp(cmdstatus, "UPDATE", 6) == 0 ||
-					strncmp(cmdstatus, "DELETE", 6) == 0)
+					strncmp(cmdstatus, "DELETE", 6) == 0 ||
+					strncmp(cmdstatus, "MERGE", 5) == 0)
 					PrintQueryStatus(result, printStatusFout);
 			}
 
diff --git a/src/include/executor/spi.h b/src/include/executor/spi.h
new file mode 100644
index d1de139..ed8787d
--- a/src/include/executor/spi.h
+++ b/src/include/executor/spi.h
@@ -97,6 +97,7 @@ typedef struct _SPI_plan *SPIPlanPtr;
 #define SPI_OK_REL_UNREGISTER	16
 #define SPI_OK_TD_REGISTER		17
 #define SPI_OK_MERGE			18
+#define SPI_OK_MERGE_RETURNING	19
 
 #define SPI_OPT_NONATOMIC		(1 << 0)
 
diff --git a/src/include/nodes/execnodes.h b/src/include/nodes/execnodes.h
new file mode 100644
index 5d7f17d..6c27fee
--- a/src/include/nodes/execnodes.h
+++ b/src/include/nodes/execnodes.h
@@ -419,6 +419,8 @@ typedef struct MergeActionState
 	ProjectionInfo *mas_proj;	/* projection of the action's targetlist for
 								 * this rel */
 	ExprState  *mas_whenqual;	/* WHEN [NOT] MATCHED AND conditions */
+	/* projection of the action's returningList for this rel */
+	ProjectionInfo *mas_proj_returning;
 } MergeActionState;
 
 /*
diff --git a/src/include/nodes/parsenodes.h b/src/include/nodes/parsenodes.h
new file mode 100644
index cf7e790..13bc017
--- a/src/include/nodes/parsenodes.h
+++ b/src/include/nodes/parsenodes.h
@@ -1679,6 +1679,8 @@ typedef struct MergeWhenClause
 	List	   *targetList;		/* INSERT/UPDATE targetlist */
 	/* the following members are only used in INSERT actions */
 	List	   *values;			/* VALUES to INSERT, or NULL */
+	List	   *returningList;	/* list of expressions to return */
+	int			location;		/* token location, or -1 if unknown */
 } MergeWhenClause;
 
 /*
@@ -1696,6 +1698,7 @@ typedef struct MergeAction
 	List	   *targetList;		/* the target list (of TargetEntry) */
 	/* target attribute numbers of an UPDATE */
 	List	   *updateColnos pg_node_attr(query_jumble_ignore);
+	List	   *returningList;	/* list of expressions to return */
 } MergeAction;
 
 /*
diff --git a/src/include/parser/analyze.h b/src/include/parser/analyze.h
new file mode 100644
index c96483a..14e6877
--- a/src/include/parser/analyze.h
+++ b/src/include/parser/analyze.h
@@ -44,6 +44,7 @@ extern List *transformInsertRow(ParseSta
 								bool strip_indirection);
 extern List *transformUpdateTargetList(ParseState *pstate,
 									   List *origTlist);
+extern List *transformReturningList(ParseState *pstate, List *returningList);
 extern Query *transformTopLevelStmt(ParseState *pstate, RawStmt *parseTree);
 extern Query *transformStmt(ParseState *pstate, Node *parseTree);
 
diff --git a/src/pl/plpgsql/src/pl_exec.c b/src/pl/plpgsql/src/pl_exec.c
new file mode 100644
index 4b76f76..fb6289c
--- a/src/pl/plpgsql/src/pl_exec.c
+++ b/src/pl/plpgsql/src/pl_exec.c
@@ -4267,9 +4267,9 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 	/*
 	 * If we have INTO, then we only need one row back ... but if we have INTO
 	 * STRICT or extra check too_many_rows, ask for two rows, so that we can
-	 * verify the statement returns only one.  INSERT/UPDATE/DELETE are always
-	 * treated strictly. Without INTO, just run the statement to completion
-	 * (tcount = 0).
+	 * verify the statement returns only one.  INSERT/UPDATE/DELETE/MERGE are
+	 * always treated strictly. Without INTO, just run the statement to
+	 * completion (tcount = 0).
 	 *
 	 * We could just ask for two rows always when using INTO, but there are
 	 * some cases where demanding the extra row costs significant time, eg by
@@ -4307,10 +4307,11 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 			Assert(stmt->mod_stmt);
 			exec_set_found(estate, (SPI_processed != 0));
 			break;
@@ -4489,10 +4490,11 @@ exec_stmt_dynexecute(PLpgSQL_execstate *
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 		case SPI_OK_UTILITY:
 		case SPI_OK_REWRITTEN:
 			break;
diff --git a/src/pl/tcl/pltcl.c b/src/pl/tcl/pltcl.c
new file mode 100644
index e8f9d7b..0c9f81e
--- a/src/pl/tcl/pltcl.c
+++ b/src/pl/tcl/pltcl.c
@@ -2456,6 +2456,7 @@ pltcl_process_SPI_result(Tcl_Interp *int
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
+		case SPI_OK_MERGE_RETURNING:
 
 			/*
 			 * Process the tuples we got
diff --git a/src/test/regress/expected/merge.out b/src/test/regress/expected/merge.out
new file mode 100644
index 28a6d0b..8dafe10
--- a/src/test/regress/expected/merge.out
+++ b/src/test/regress/expected/merge.out
@@ -123,20 +123,20 @@ ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
 ERROR:  name "target" specified more than once
 DETAIL:  The name is used both as MERGE target table and data source.
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
-ERROR:  MERGE not supported in WITH query
-LINE 1: WITH foo AS (
-             ^
--- used in COPY
+ERROR:  WITH query "foo" does not have a RETURNING clause
+LINE 4: ) SELECT * FROM foo;
+                        ^
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) TO stdout;
-ERROR:  MERGE not supported in COPY
+ERROR:  COPY query must have a RETURNING clause
 -- unsupported relation types
 -- view
 CREATE VIEW tv AS SELECT * FROM target;
@@ -1256,7 +1256,7 @@ BEGIN;
 MERGE INTO sq_target
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1271,7 +1271,7 @@ INSERT INTO sq_source (sid, balance, del
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1295,7 +1295,7 @@ WITH targq AS (
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1304,20 +1304,183 @@ WHEN MATCHED AND tid < 2 THEN
 ROLLBACK;
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
+	RETURNING 1 AS clause_number, 'upd' AS merge_action, t.*,
+				'Added '||delta||' to balance' AS description
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
+	RETURNING 2, 'ins', t.*, 'Inserted '||t
 WHEN MATCHED AND tid < 2 THEN
 	DELETE
-RETURNING *;
-ERROR:  syntax error at or near "RETURNING"
-LINE 10: RETURNING *;
-         ^
+	RETURNING 3, 'del', t.*, 'Removed '||t;
+ clause_number | merge_action | tid | balance |     description     
+---------------+--------------+-----+---------+---------------------
+             3 | del          |   1 |     100 | Removed (1,100)
+             1 | upd          |   2 |     220 | Added 20 to balance
+             2 | ins          |   4 |      40 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+        RETURNING 1 AS clause_number, 'UPDATE' AS merge_action, t.*,
+                  'Added '||delta||' to balance' AS description
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+        RETURNING 2, 'INSERT', t.*, 'Inserted '||t AS colname_ignored
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+        RETURNING 3, 'DELETE', t.*, 'Removed '||t
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+        RETURNING merge_action, 'UPDATE' AS merge_log_action, l.*
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+        RETURNING merge_action, 'INSERT', l.*
+)
+SELECT * FROM m2;
+ merge_action | merge_log_action | tid |     last_change     
+--------------+------------------+-----+---------------------
+ DELETE       | UPDATE           |   1 | Removed (1,100)
+ UPDATE       | INSERT           |   2 | Added 20 to balance
+ INSERT       | INSERT           |   4 | Inserted (4,40)
+(3 rows)
+
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ tid |     last_change     
+-----+---------------------
+   1 | Removed (1,100)
+   2 | Added 20 to balance
+   4 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+        RETURNING 'UPDATE', t.*
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+        RETURNING 'INSERT', t.*
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+        RETURNING 'DELETE', t.*
+) TO stdout;
+DELETE	1	100
+UPDATE	2	220
+INSERT	4	40
+ROLLBACK;
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+        RETURNING 'UPDATE', t.*
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+        RETURNING 'INSERT', t.*
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+        RETURNING 'DELETE', t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ action | tid | new_balance 
+--------+-----+-------------
+ DELETE |   1 |         100
+ UPDATE |   3 |         320
+ INSERT |   4 |         110
+(3 rows)
+
+ROLLBACK;
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+        RETURNING 'UPDATE', t.*
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+        RETURNING 'INSERT', t.*
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+        RETURNING 'DELETE', t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ action | tid | balance 
+--------+-----+---------
+ DELETE |   1 |     100
+ UPDATE |   2 |     220
+ INSERT |   4 |      40
+(3 rows)
+
+ROLLBACK;
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+        RETURNING 'UPDATE', t.*
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+        RETURNING 'INSERT', t.*
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+        RETURNING 'DELETE', t.*
+    INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ r_action | r_tid | r_balance 
+----------+-------+-----------
+ DELETE   |     1 |       100
+ UPDATE   |     3 |       320
+ INSERT   |     4 |       110
+(3 rows)
+
 ROLLBACK;
 -- EXPLAIN
 CREATE TABLE ex_mtarget (a int, b int)
@@ -1514,7 +1677,7 @@ SELECT * FROM sq_target WHERE tid = 1;
 (1 row)
 
 ROLLBACK;
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 NOTICE:  drop cascades to view v
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1640,6 +1803,32 @@ SELECT * FROM pa_target ORDER BY tid;
 (14 rows)
 
 ROLLBACK;
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+    RETURNING t.*;
+ tid | balance |           val            
+-----+---------+--------------------------
+   2 |     110 | initial updated by merge
+(1 row)
+
+SELECT * FROM pa_target ORDER BY tid;
+ tid | balance |           val            
+-----+---------+--------------------------
+   2 |     110 | initial updated by merge
+   3 |     300 | initial
+   5 |     500 | initial
+   7 |     700 | initial
+   9 |     900 | initial
+  11 |    1100 | initial
+  13 |    1300 | initial
+(7 rows)
+
+ROLLBACK;
 DROP TABLE pa_target CASCADE;
 -- The target table is partitioned in the same way, but this time by attaching
 -- partitions which have columns in different order, dropped columns etc.
@@ -1730,8 +1919,28 @@ MERGE INTO pa_target t
   ON t.tid = s.sid
   WHEN MATCHED THEN
     UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+    RETURNING 'UPDATE' AS merge_action, t.*
   WHEN NOT MATCHED THEN
-    INSERT VALUES (sid, delta, 'inserted by merge');
+    INSERT VALUES (sid, delta, 'inserted by merge')
+    RETURNING 'INSERT', t.*;
+ merge_action | tid | balance |           val            
+--------------+-----+---------+--------------------------
+ UPDATE       |   2 |     110 | initial updated by merge
+ INSERT       |   2 |      20 | inserted by merge
+ UPDATE       |   4 |     330 | initial updated by merge
+ INSERT       |   4 |      40 | inserted by merge
+ UPDATE       |   6 |     550 | initial updated by merge
+ INSERT       |   6 |      60 | inserted by merge
+ UPDATE       |   8 |     770 | initial updated by merge
+ INSERT       |   8 |      80 | inserted by merge
+ UPDATE       |  10 |     990 | initial updated by merge
+ INSERT       |  10 |     100 | inserted by merge
+ UPDATE       |  12 |    1210 | initial updated by merge
+ INSERT       |  12 |     120 | inserted by merge
+ UPDATE       |  14 |    1430 | initial updated by merge
+ INSERT       |  14 |     140 | inserted by merge
+(14 rows)
+
 SELECT * FROM pa_target ORDER BY tid;
  tid | balance |           val            
 -----+---------+--------------------------
@@ -1797,8 +2006,23 @@ MERGE INTO pa_target t
   ON t.tid = s.sid
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
+    RETURNING 'UPDATE' AS merge_action, t.*
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+    RETURNING 'INSERT' AS merge_action, t.*;
+ merge_action |          logts           | tid | balance |           val            
+--------------+--------------------------+-----+---------+--------------------------
+ UPDATE       | Tue Jan 31 00:00:00 2017 |   1 |     110 | initial updated by merge
+ UPDATE       | Tue Feb 28 00:00:00 2017 |   2 |     220 | initial updated by merge
+ INSERT       | Sun Jan 15 00:00:00 2017 |   3 |      30 | inserted by merge
+ UPDATE       | Tue Jan 31 00:00:00 2017 |   4 |     440 | initial updated by merge
+ UPDATE       | Tue Feb 28 00:00:00 2017 |   5 |     550 | initial updated by merge
+ INSERT       | Sun Jan 15 00:00:00 2017 |   6 |      60 | inserted by merge
+ UPDATE       | Tue Jan 31 00:00:00 2017 |   7 |     770 | initial updated by merge
+ UPDATE       | Tue Feb 28 00:00:00 2017 |   8 |     880 | initial updated by merge
+ INSERT       | Sun Jan 15 00:00:00 2017 |   9 |      90 | inserted by merge
+(9 rows)
+
 SELECT * FROM pa_target ORDER BY tid;
           logts           | tid | balance |           val            
 --------------------------+-----+---------+--------------------------
diff --git a/src/test/regress/expected/rowsecurity.out b/src/test/regress/expected/rowsecurity.out
new file mode 100644
index 6988128..d0c385d
--- a/src/test/regress/expected/rowsecurity.out
+++ b/src/test/regress/expected/rowsecurity.out
@@ -2302,6 +2302,37 @@ WHEN MATCHED THEN
 	UPDATE SET dnotes = dnotes || ' notes added by merge8 '
 WHEN NOT MATCHED THEN
 	INSERT VALUES (13, 44, 1, 'regress_rls_bob', 'new manga');
+SELECT * FROM document WHERE did = 13;
+ did | cid | dlevel | dauthor | dtitle | dnotes 
+-----+-----+--------+---------+--------+--------
+(0 rows)
+
+-- but not OK if RETURNING is used
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge9 '
+	RETURNING *
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 44, 1, 'regress_rls_bob', 'new manga')
+	RETURNING *;
+ERROR:  new row violates row-level security policy for table "document"
+-- but OK if new row is visible
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge10 '
+	RETURNING *
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 11, 1, 'regress_rls_bob', 'new novel')
+	RETURNING *;
+ sdid | did | cid | dlevel |     dauthor     |  dtitle   | dnotes 
+------+-----+-----+--------+-----------------+-----------+--------
+   14 |  14 |  11 |      1 | regress_rls_bob | new novel | 
+(1 row)
+
 RESET SESSION AUTHORIZATION;
 -- drop the restrictive SELECT policy so that we can look at the
 -- final state of the table
@@ -2325,7 +2356,8 @@ SELECT * FROM document;
   12 |  11 |      1 | regress_rls_bob   | another novel                    | 
    1 |  11 |      1 | regress_rls_bob   | my first novel                   |  notes added by merge2  notes added by merge3  notes added by merge4  notes added by merge7 
   13 |  44 |      1 | regress_rls_bob   | new manga                        | 
-(15 rows)
+  14 |  11 |      1 | regress_rls_bob   | new novel                        | 
+(16 rows)
 
 --
 -- ROLE/GROUP
diff --git a/src/test/regress/expected/rules.out b/src/test/regress/expected/rules.out
new file mode 100644
index 2c60400..e8b1b0d
--- a/src/test/regress/expected/rules.out
+++ b/src/test/regress/expected/rules.out
@@ -3597,7 +3597,8 @@ MERGE INTO rule_merge2 t USING (SELECT 1
 -- test deparsing
 CREATE TABLE sf_target(id int, data text, filling int[]);
 CREATE FUNCTION merge_sf_test()
- RETURNS void
+ RETURNS TABLE(merge_action text, when_clause int,
+               a int, b text, id int, data text, filling int[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -3606,39 +3607,48 @@ BEGIN ATOMIC
 WHEN MATCHED
    AND (s.a + t.id) = 42
    THEN UPDATE SET data = repeat(t.data, s.a) || s.b, id = length(s.b)
+   RETURNING 'UPDATE', 1, *
 WHEN NOT MATCHED
    AND (s.b IS NOT NULL)
    THEN INSERT (data, id)
    VALUES (s.b, s.a)
+   RETURNING 'INSERT', 2, *
 WHEN MATCHED
    AND length(s.b || t.data) > 10
    THEN UPDATE SET data = s.b
+   RETURNING 'UPDATE', 3, *
 WHEN MATCHED
    AND s.a > 200
    THEN UPDATE SET filling[s.a] = t.id
+   RETURNING 'UPDATE', 4, *
 WHEN MATCHED
    AND s.a > 100
    THEN DELETE
+   RETURNING 'DELETE', 5, *
 WHEN MATCHED
    THEN DO NOTHING
 WHEN NOT MATCHED
    AND s.a > 200
    THEN INSERT DEFAULT VALUES
+   RETURNING 'INSERT', 6, *
 WHEN NOT MATCHED
    AND s.a > 100
    THEN INSERT (id, data) OVERRIDING USER VALUE
    VALUES (s.a, DEFAULT)
+   RETURNING 'INSERT', 7, *
 WHEN NOT MATCHED
    AND s.a > 0
    THEN INSERT
    VALUES (s.a, s.b, DEFAULT)
+   RETURNING 'INSERT', 8, *
 WHEN NOT MATCHED
    THEN INSERT (filling[1], id)
-   VALUES (s.a, s.a);
+   VALUES (s.a, s.a)
+   RETURNING 'INSERT', 9, *;
 END;
 \sf merge_sf_test
 CREATE OR REPLACE FUNCTION public.merge_sf_test()
- RETURNS void
+ RETURNS TABLE(merge_action text, when_clause integer, a integer, b text, id integer, data text, filling integer[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -3647,35 +3657,98 @@ BEGIN ATOMIC
     WHEN MATCHED
      AND ((s.a + t.id) = 42)
      THEN UPDATE SET data = (repeat(t.data, s.a) || s.b), id = length(s.b)
+      RETURNING 'UPDATE'::text,
+     1,
+     s.a,
+     s.b,
+     t.id,
+     t.data,
+     t.filling
     WHEN NOT MATCHED
      AND (s.b IS NOT NULL)
      THEN INSERT (data, id)
       VALUES (s.b, s.a)
+      RETURNING 'INSERT'::text,
+     2,
+     s.a,
+     s.b,
+     t.id,
+     t.data,
+     t.filling
     WHEN MATCHED
      AND (length((s.b || t.data)) > 10)
      THEN UPDATE SET data = s.b
+      RETURNING 'UPDATE'::text,
+     3,
+     s.a,
+     s.b,
+     t.id,
+     t.data,
+     t.filling
     WHEN MATCHED
      AND (s.a > 200)
      THEN UPDATE SET filling[s.a] = t.id
+      RETURNING 'UPDATE'::text,
+     4,
+     s.a,
+     s.b,
+     t.id,
+     t.data,
+     t.filling
     WHEN MATCHED
      AND (s.a > 100)
      THEN DELETE
+      RETURNING 'DELETE'::text,
+     5,
+     s.a,
+     s.b,
+     t.id,
+     t.data,
+     t.filling
     WHEN MATCHED
      THEN DO NOTHING
     WHEN NOT MATCHED
      AND (s.a > 200)
      THEN INSERT DEFAULT VALUES
+      RETURNING 'INSERT'::text,
+     6,
+     s.a,
+     s.b,
+     t.id,
+     t.data,
+     t.filling
     WHEN NOT MATCHED
      AND (s.a > 100)
      THEN INSERT (id, data) OVERRIDING USER VALUE
       VALUES (s.a, DEFAULT)
+      RETURNING 'INSERT'::text,
+     7,
+     s.a,
+     s.b,
+     t.id,
+     t.data,
+     t.filling
     WHEN NOT MATCHED
      AND (s.a > 0)
      THEN INSERT (id, data, filling)
       VALUES (s.a, s.b, DEFAULT)
+      RETURNING 'INSERT'::text,
+     8,
+     s.a,
+     s.b,
+     t.id,
+     t.data,
+     t.filling
     WHEN NOT MATCHED
      THEN INSERT (filling[1], id)
-      VALUES (s.a, s.a);
+      VALUES (s.a, s.a)
+      RETURNING 'INSERT'::text,
+     9,
+     s.a,
+     s.b,
+     t.id,
+     t.data,
+     t.filling;
 END
 DROP FUNCTION merge_sf_test;
 DROP TABLE sf_target;
diff --git a/src/test/regress/sql/merge.sql b/src/test/regress/sql/merge.sql
new file mode 100644
index 4cf6db9..32a70e7
--- a/src/test/regress/sql/merge.sql
+++ b/src/test/regress/sql/merge.sql
@@ -86,12 +86,12 @@ MERGE INTO target
 USING target
 ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
--- used in COPY
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
@@ -816,7 +816,7 @@ BEGIN;
 MERGE INTO sq_target
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -829,7 +829,7 @@ INSERT INTO sq_source (sid, balance, del
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -847,7 +847,7 @@ WITH targq AS (
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -857,17 +857,143 @@ ROLLBACK;
 
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
+	RETURNING 1 AS clause_number, 'upd' AS merge_action, t.*,
+				'Added '||delta||' to balance' AS description
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
+	RETURNING 2, 'ins', t.*, 'Inserted '||t
 WHEN MATCHED AND tid < 2 THEN
 	DELETE
-RETURNING *;
+	RETURNING 3, 'del', t.*, 'Removed '||t;
+ROLLBACK;
+
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+        RETURNING 1 AS clause_number, 'UPDATE' AS merge_action, t.*,
+                  'Added '||delta||' to balance' AS description
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+        RETURNING 2, 'INSERT', t.*, 'Inserted '||t AS colname_ignored
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+        RETURNING 3, 'DELETE', t.*, 'Removed '||t
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+        RETURNING merge_action, 'UPDATE' AS merge_log_action, l.*
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+        RETURNING merge_action, 'INSERT', l.*
+)
+SELECT * FROM m2;
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ROLLBACK;
+
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+        RETURNING 'UPDATE', t.*
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+        RETURNING 'INSERT', t.*
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+        RETURNING 'DELETE', t.*
+) TO stdout;
+ROLLBACK;
+
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+        RETURNING 'UPDATE', t.*
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+        RETURNING 'INSERT', t.*
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+        RETURNING 'DELETE', t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ROLLBACK;
+
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+        RETURNING 'UPDATE', t.*
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+        RETURNING 'INSERT', t.*
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+        RETURNING 'DELETE', t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ROLLBACK;
+
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+        RETURNING 'UPDATE', t.*
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+        RETURNING 'INSERT', t.*
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+        RETURNING 'DELETE', t.*
+    INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
 ROLLBACK;
 
 -- EXPLAIN
@@ -966,7 +1092,7 @@ WHEN MATCHED THEN
 SELECT * FROM sq_target WHERE tid = 1;
 ROLLBACK;
 
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1033,6 +1159,17 @@ SELECT merge_func();
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+    RETURNING t.*;
+SELECT * FROM pa_target ORDER BY tid;
+ROLLBACK;
+
 DROP TABLE pa_target CASCADE;
 
 -- The target table is partitioned in the same way, but this time by attaching
@@ -1090,8 +1227,10 @@ MERGE INTO pa_target t
   ON t.tid = s.sid
   WHEN MATCHED THEN
     UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+    RETURNING 'UPDATE' AS merge_action, t.*
   WHEN NOT MATCHED THEN
-    INSERT VALUES (sid, delta, 'inserted by merge');
+    INSERT VALUES (sid, delta, 'inserted by merge')
+    RETURNING 'INSERT', t.*;
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
@@ -1144,8 +1283,10 @@ MERGE INTO pa_target t
   ON t.tid = s.sid
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
+    RETURNING 'UPDATE' AS merge_action, t.*
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+    RETURNING 'INSERT' AS merge_action, t.*;
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
diff --git a/src/test/regress/sql/rowsecurity.sql b/src/test/regress/sql/rowsecurity.sql
new file mode 100644
index dec7340..8c093ee
--- a/src/test/regress/sql/rowsecurity.sql
+++ b/src/test/regress/sql/rowsecurity.sql
@@ -992,6 +992,29 @@ WHEN MATCHED THEN
 	UPDATE SET dnotes = dnotes || ' notes added by merge8 '
 WHEN NOT MATCHED THEN
 	INSERT VALUES (13, 44, 1, 'regress_rls_bob', 'new manga');
+SELECT * FROM document WHERE did = 13;
+
+-- but not OK if RETURNING is used
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge9 '
+	RETURNING *
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 44, 1, 'regress_rls_bob', 'new manga')
+	RETURNING *;
+
+-- but OK if new row is visible
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge10 '
+	RETURNING *
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 11, 1, 'regress_rls_bob', 'new novel')
+	RETURNING *;
 
 RESET SESSION AUTHORIZATION;
 -- drop the restrictive SELECT policy so that we can look at the
diff --git a/src/test/regress/sql/rules.sql b/src/test/regress/sql/rules.sql
new file mode 100644
index 8b7e255..92d2ccd
--- a/src/test/regress/sql/rules.sql
+++ b/src/test/regress/sql/rules.sql
@@ -1281,7 +1281,8 @@ MERGE INTO rule_merge2 t USING (SELECT 1
 CREATE TABLE sf_target(id int, data text, filling int[]);
 
 CREATE FUNCTION merge_sf_test()
- RETURNS void
+ RETURNS TABLE(merge_action text, when_clause int,
+               a int, b text, id int, data text, filling int[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -1290,35 +1291,44 @@ BEGIN ATOMIC
 WHEN MATCHED
    AND (s.a + t.id) = 42
    THEN UPDATE SET data = repeat(t.data, s.a) || s.b, id = length(s.b)
+   RETURNING 'UPDATE', 1, *
 WHEN NOT MATCHED
    AND (s.b IS NOT NULL)
    THEN INSERT (data, id)
    VALUES (s.b, s.a)
+   RETURNING 'INSERT', 2, *
 WHEN MATCHED
    AND length(s.b || t.data) > 10
    THEN UPDATE SET data = s.b
+   RETURNING 'UPDATE', 3, *
 WHEN MATCHED
    AND s.a > 200
    THEN UPDATE SET filling[s.a] = t.id
+   RETURNING 'UPDATE', 4, *
 WHEN MATCHED
    AND s.a > 100
    THEN DELETE
+   RETURNING 'DELETE', 5, *
 WHEN MATCHED
    THEN DO NOTHING
 WHEN NOT MATCHED
    AND s.a > 200
    THEN INSERT DEFAULT VALUES
+   RETURNING 'INSERT', 6, *
 WHEN NOT MATCHED
    AND s.a > 100
    THEN INSERT (id, data) OVERRIDING USER VALUE
    VALUES (s.a, DEFAULT)
+   RETURNING 'INSERT', 7, *
 WHEN NOT MATCHED
    AND s.a > 0
    THEN INSERT
    VALUES (s.a, s.b, DEFAULT)
+   RETURNING 'INSERT', 8, *
 WHEN NOT MATCHED
    THEN INSERT (filling[1], id)
-   VALUES (s.a, s.a);
+   VALUES (s.a, s.a)
+   RETURNING 'INSERT', 9, *;
 END;
 
 \sf merge_sf_test

diff --git a/doc/src/sgml/dml.sgml b/doc/src/sgml/dml.sgml
new file mode 100644
index cbbc5e2..3d95bdb
--- a/doc/src/sgml/dml.sgml
+++ b/doc/src/sgml/dml.sgml
@@ -283,10 +283,15 @@ DELETE FROM products;
    <secondary>RETURNING</secondary>
   </indexterm>
 
+  <indexterm zone="dml-returning">
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
   <para>
    Sometimes it is useful to obtain data from modified rows while they are
    being manipulated.  The <command>INSERT</command>, <command>UPDATE</command>,
-   and <command>DELETE</command> commands all have an
+   <command>DELETE</command>, and <command>MERGE</command> commands all have an
    optional <literal>RETURNING</literal> clause that supports this.  Use
    of <literal>RETURNING</literal> avoids performing an extra database query to
    collect the data, and is especially valuable when it would otherwise be
@@ -339,6 +344,21 @@ DELETE FROM products
 </programlisting>
   </para>
 
+  <para>
+   In a <command>MERGE</command>, the data available to <literal>RETURNING</literal> is
+   the content of the source row plus the content of the inserted, updated, or
+   deleted target row.  Since it is quite common for the source and target to
+   have many of the same columns, specifying <literal>RETURNING *</literal>
+   can lead to a lot of duplicated columns, so it is often more useful to
+   qualify it so as to return just the source or target row.  For example:
+<programlisting>
+MERGE INTO products p USING new_products n ON p.product_no = n.product_no
+  WHEN NOT MATCHED THEN INSERT VALUES (n.product_no, n.name, n.price)
+  WHEN MATCHED THEN UPDATE SET name = n.name, price = n.price
+  RETURNING p.*;
+</programlisting>
+  </para>
+
   <para>
    If there are triggers (<xref linkend="triggers"/>) on the target table,
    the data available to <literal>RETURNING</literal> is the row as modified by
diff --git a/doc/src/sgml/func.sgml b/doc/src/sgml/func.sgml
new file mode 100644
index a6fcac0..06d5fe8
--- a/doc/src/sgml/func.sgml
+++ b/doc/src/sgml/func.sgml
@@ -21960,6 +21960,84 @@ SELECT count(*) FROM sometable;
 
  </sect1>
 
+ <sect1 id="functions-merge-support">
+  <title>Merge Support Functions</title>
+
+  <indexterm>
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
+  <indexterm>
+   <primary>MERGING</primary>
+  </indexterm>
+
+  <para>
+   <productname>PostgreSQL</productname> includes one merge support function
+   that may be used in the <literal>RETURNING</literal> list of a
+   <xref linkend="sql-merge"/> command to return additional information about
+   the action taken for each row:
+<synopsis>
+<function id="function-merging">MERGING</function> ( <replaceable>property</replaceable> )
+</synopsis>
+   The following are valid property values specifying what to return:
+
+   <variablelist>
+    <varlistentry>
+     <term><literal>ACTION</literal></term>
+     <listitem>
+      <para>
+       The merge action command executed for the current row
+       (<literal>'INSERT'</literal>, <literal>'UPDATE'</literal>, or
+       <literal>'DELETE'</literal>).
+      </para>
+     </listitem>
+    </varlistentry>
+
+    <varlistentry>
+     <term><literal>CLAUSE_NUMBER</literal></term>
+     <listitem>
+      <para>
+       The 1-based index of the <literal>WHEN</literal> clause executed for
+       the current row.
+      </para>
+     </listitem>
+    </varlistentry>
+   </variablelist>
+  </para>
+
+  <para>
+   The following example illustrates how this may be used to determine
+   which actions were performed for each row affected by the
+   <command>MERGE</command> command:
+<screen><![CDATA[
+MERGE INTO products p
+  USING stock s ON p.product_id = s.product_id
+  WHEN MATCHED AND s.quantity > 0 THEN
+    UPDATE SET in_stock = true, quantity = s.quantity
+  WHEN MATCHED THEN
+    UPDATE SET in_stock = false, quantity = 0
+  WHEN NOT MATCHED THEN
+    INSERT (product_id, in_stock, quantity)
+      VALUES (s.product_id, true, s.quantity)
+  RETURNING MERGING(ACTION), MERGING(CLAUSE_NUMBER), p.*;
+
+ action | clause_number | product_id | in_stock | quantity 
+--------+---------------+------------+----------+----------
+ UPDATE |             1 |       1001 | t        |       50
+ UPDATE |             2 |       1002 | f        |        0
+ INSERT |             3 |       1003 | t        |       10
+]]></screen>
+  </para>
+
+  <para>
+   Note that this function can only be used in the <literal>RETURNING</literal>
+   list of a <command>MERGE</command> command.  It is an error to use it in any
+   other part of a query.
+  </para>
+
+ </sect1>
+
  <sect1 id="functions-subquery">
   <title>Subquery Expressions</title>
 
diff --git a/doc/src/sgml/glossary.sgml b/doc/src/sgml/glossary.sgml
new file mode 100644
index fe8def4..51a15ca
--- a/doc/src/sgml/glossary.sgml
+++ b/doc/src/sgml/glossary.sgml
@@ -1387,9 +1387,9 @@
      to a <glossterm linkend="glossary-client">client</glossterm> upon the
      completion of an <acronym>SQL</acronym> command, usually a
      <command>SELECT</command> but it can be an
-     <command>INSERT</command>, <command>UPDATE</command>, or
-     <command>DELETE</command> command if the <literal>RETURNING</literal>
-     clause is specified.
+     <command>INSERT</command>, <command>UPDATE</command>,
+     <command>DELETE</command>, or <command>MERGE</command> command if the
+     <literal>RETURNING</literal> clause is specified.
     </para>
     <para>
      The fact that a result set is a relation means that a query can be used
diff --git a/doc/src/sgml/plpgsql.sgml b/doc/src/sgml/plpgsql.sgml
new file mode 100644
index 5977534..4ad9894
--- a/doc/src/sgml/plpgsql.sgml
+++ b/doc/src/sgml/plpgsql.sgml
@@ -1020,8 +1020,8 @@ INSERT INTO mytable VALUES (1,'one'), (2
     </para>
 
     <para>
-     If the command does return rows (for example <command>SELECT</command>,
-     or <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>
+     If the command does return rows (for example <command>SELECT</command>, or
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
      with <literal>RETURNING</literal>), there are two ways to proceed.
      When the command will return at most one row, or you only care about
      the first row of output, write the command as usual but add
@@ -1149,6 +1149,7 @@ SELECT <replaceable>select_expressions</
 INSERT ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 UPDATE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 DELETE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
+MERGE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 </synopsis>
 
      where <replaceable>target</replaceable> can be a record variable, a row
@@ -1159,8 +1160,8 @@ DELETE ... RETURNING <replaceable>expres
      <literal>INTO</literal> clause) just as described above,
      and the plan is cached in the same way.
      This works for <command>SELECT</command>,
-     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
-     <literal>RETURNING</literal>, and certain utility commands
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
+     with <literal>RETURNING</literal>, and certain utility commands
      that return row sets, such as <command>EXPLAIN</command>.
      Except for the <literal>INTO</literal> clause, the SQL command is the same
      as it would be written outside <application>PL/pgSQL</application>.
@@ -1236,7 +1237,7 @@ END;
     </para>
 
     <para>
-     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
+     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command> with
      <literal>RETURNING</literal>, <application>PL/pgSQL</application> reports
      an error for more than one returned row, even when
      <literal>STRICT</literal> is not specified.  This is because there
diff --git a/doc/src/sgml/queries.sgml b/doc/src/sgml/queries.sgml
new file mode 100644
index 8a4674e..ae1f738
--- a/doc/src/sgml/queries.sgml
+++ b/doc/src/sgml/queries.sgml
@@ -2062,9 +2062,10 @@ SELECT <replaceable>select_list</replace
    Table Expressions or <acronym>CTE</acronym>s, can be thought of as defining
    temporary tables that exist just for one query.  Each auxiliary statement
    in a <literal>WITH</literal> clause can be a <command>SELECT</command>,
-   <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>; and the
+   <command>INSERT</command>, <command>UPDATE</command>, <command>DELETE</command>,
+   or <command>MERGE</command>; and the
    <literal>WITH</literal> clause itself is attached to a primary statement that can
-   be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
+   also be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
    <command>DELETE</command>, or <command>MERGE</command>.
   </para>
 
@@ -2598,8 +2599,8 @@ SELECT * FROM w AS w1 JOIN w AS w2 ON w1
    <title>Data-Modifying Statements in <literal>WITH</literal></title>
 
    <para>
-    You can use most data-modifying statements (<command>INSERT</command>,
-    <command>UPDATE</command>, or <command>DELETE</command>, but not
+    You can use data-modifying statements (<command>INSERT</command>,
+    <command>UPDATE</command>, <command>DELETE</command>, or
     <command>MERGE</command>) in <literal>WITH</literal>.  This
     allows you to perform several different operations in the same query.
     An example is:
diff --git a/doc/src/sgml/ref/copy.sgml b/doc/src/sgml/ref/copy.sgml
new file mode 100644
index d12ba96..9b9536b
--- a/doc/src/sgml/ref/copy.sgml
+++ b/doc/src/sgml/ref/copy.sgml
@@ -122,13 +122,16 @@ COPY { <replaceable class="parameter">ta
       A <link linkend="sql-select"><command>SELECT</command></link>,
       <link linkend="sql-values"><command>VALUES</command></link>,
       <link linkend="sql-insert"><command>INSERT</command></link>,
-      <link linkend="sql-update"><command>UPDATE</command></link>, or
-      <link linkend="sql-delete"><command>DELETE</command></link> command whose results are to be
-      copied.  Note that parentheses are required around the query.
+      <link linkend="sql-update"><command>UPDATE</command></link>,
+      <link linkend="sql-delete"><command>DELETE</command></link>, or
+      <link linkend="sql-merge"><command>MERGE</command></link> command
+      whose results are to be copied.  Note that parentheses are required
+      around the query.
      </para>
      <para>
-      For <command>INSERT</command>, <command>UPDATE</command> and
-      <command>DELETE</command> queries a RETURNING clause must be provided,
+      For <command>INSERT</command>, <command>UPDATE</command>,
+      <command>DELETE</command>, and <command>MERGE</command> queries a
+      RETURNING clause must be provided,
       and the target relation must not have a conditional rule, nor
       an <literal>ALSO</literal> rule, nor an <literal>INSTEAD</literal> rule
       that expands to multiple statements.
diff --git a/doc/src/sgml/ref/merge.sgml b/doc/src/sgml/ref/merge.sgml
new file mode 100644
index 0f61d47..1ef29ab
--- a/doc/src/sgml/ref/merge.sgml
+++ b/doc/src/sgml/ref/merge.sgml
@@ -25,6 +25,7 @@ PostgreSQL documentation
 MERGE INTO [ ONLY ] <replaceable class="parameter">target_table_name</replaceable> [ * ] [ [ AS ] <replaceable class="parameter">target_alias</replaceable> ]
 USING <replaceable class="parameter">data_source</replaceable> ON <replaceable class="parameter">join_condition</replaceable>
 <replaceable class="parameter">when_clause</replaceable> [...]
+[ RETURNING * | <replaceable class="parameter">output_expression</replaceable> [ [ AS ] <replaceable class="parameter">output_name</replaceable> ] [, ...] ]
 
 <phrase>where <replaceable class="parameter">data_source</replaceable> is:</phrase>
 
@@ -95,6 +96,18 @@ DELETE
   </para>
 
   <para>
+   The optional <literal>RETURNING</literal> clause causes <command>MERGE</command>
+   to compute and return value(s) based on each row inserted, updated, or
+   deleted.  Any expression using the source or target table's columns, or
+   the <xref linkend="function-merging"/> function can be computed.  When an
+   <command>INSERT</command> or <command>UPDATE</command> action is performed,
+   the new values of the target table's columns are used.  When a
+   <command>DELETE</command> is performed, the old values of the target table's
+   columns are used.  The syntax of the <literal>RETURNING</literal> list is
+   identical to that of the output list of <command>SELECT</command>.
+  </para>
+
+  <para>
    There is no separate <literal>MERGE</literal> privilege.
    If you specify an update action, you must have the
    <literal>UPDATE</literal> privilege on the column(s)
@@ -125,6 +138,18 @@ DELETE
 
   <variablelist>
    <varlistentry>
+    <term><replaceable class="parameter">with_query</replaceable></term>
+    <listitem>
+     <para>
+      The <literal>WITH</literal> clause allows you to specify one or more
+      subqueries that can be referenced by name in the <command>MERGE</command>
+      query. See <xref linkend="queries-with"/> and <xref linkend="sql-select"/>
+      for details.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
     <term><replaceable class="parameter">target_table_name</replaceable></term>
     <listitem>
      <para>
@@ -399,13 +424,30 @@ DELETE
    </varlistentry>
 
    <varlistentry>
-    <term><replaceable class="parameter">with_query</replaceable></term>
+    <term><replaceable class="parameter">output_expression</replaceable></term>
     <listitem>
      <para>
-      The <literal>WITH</literal> clause allows you to specify one or more
-      subqueries that can be referenced by name in the <command>MERGE</command>
-      query. See <xref linkend="queries-with"/> and <xref linkend="sql-select"/>
-      for details.
+      An expression to be computed and returned by the <command>MERGE</command>
+      command after each row is merged.  The expression can use any columns of
+      the source or target tables, or the <xref linkend="function-merging"/>
+      function to return additional information about the action executed.
+     </para>
+     <para>
+      Writing <literal>*</literal> will return all columns from the source
+      table, followed by all columns from the target table.  Often this will
+      lead to a lot of duplication, since it is common for the source and
+      target tables to have a lot of the same columns.  This can be avoided by
+      qualifying the <literal>*</literal> with the name or alias of the source
+      or target table.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
+    <term><replaceable class="parameter">output_name</replaceable></term>
+    <listitem>
+     <para>
+      A name to use for a returned column.
      </para>
     </listitem>
    </varlistentry>
@@ -428,6 +470,13 @@ MERGE <replaceable class="parameter">tot
    were changed in any way.
   </para>
 
+  <para>
+   If the <command>MERGE</command> command contains a <literal>RETURNING</literal>
+   clause, the result will be similar to that of a <command>SELECT</command>
+   statement containing the columns and values defined in the
+   <literal>RETURNING</literal> list, computed over the row(s) inserted, updated,
+   or deleted by the command.
+  </para>
  </refsect1>
 
  <refsect1>
@@ -544,13 +593,6 @@ MERGE <replaceable class="parameter">tot
   </para>
 
   <para>
-   There is no <literal>RETURNING</literal> clause with
-   <command>MERGE</command>.  Actions of <command>INSERT</command>,
-   <command>UPDATE</command> and <command>DELETE</command> cannot contain
-   <literal>RETURNING</literal> or <literal>WITH</literal> clauses.
-  </para>
-
-  <para>
    When <command>MERGE</command> is run concurrently with other commands
    that modify the target table, the usual transaction isolation rules
    apply; see <xref linkend="transaction-iso"/> for an explanation
@@ -602,7 +644,8 @@ WHEN NOT MATCHED THEN
   <para>
    Attempt to insert a new stock item along with the quantity of stock. If
    the item already exists, instead update the stock count of the existing
-   item. Don't allow entries that have zero stock.
+   item. Don't allow entries that have zero stock. Return details of all
+   changes made.
 <programlisting>
 MERGE INTO wines w
 USING wine_stock_changes s
@@ -612,7 +655,8 @@ WHEN NOT MATCHED AND s.stock_delta > 0 T
 WHEN MATCHED AND w.stock + s.stock_delta > 0 THEN
   UPDATE SET stock = w.stock + s.stock_delta
 WHEN MATCHED THEN
-  DELETE;
+  DELETE
+RETURNING MERGING(ACTION), w.*;
 </programlisting>
 
    The <literal>wine_stock_changes</literal> table might be, for example, a
@@ -627,7 +671,8 @@ WHEN MATCHED THEN
     This command conforms to the <acronym>SQL</acronym> standard.
   </para>
    <para>
-    The WITH clause and <literal>DO NOTHING</literal> action are extensions to
+    The WITH clause, <literal>DO NOTHING</literal> action, and
+    <literal>RETURNING</literal> clause are extensions to
     the <acronym>SQL</acronym> standard.
   </para>
  </refsect1>
diff --git a/doc/src/sgml/ref/select.sgml b/doc/src/sgml/ref/select.sgml
new file mode 100644
index 42d7891..8513ca6
--- a/doc/src/sgml/ref/select.sgml
+++ b/doc/src/sgml/ref/select.sgml
@@ -74,7 +74,7 @@ SELECT [ ALL | DISTINCT [ ON ( <replacea
 
 <phrase>and <replaceable class="parameter">with_query</replaceable> is:</phrase>
 
-    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> )
+    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> | <replaceable class="parameter">merge</replaceable> )
         [ SEARCH { BREADTH | DEPTH } FIRST BY <replaceable>column_name</replaceable> [, ...] SET <replaceable>search_seq_col_name</replaceable> ]
         [ CYCLE <replaceable>column_name</replaceable> [, ...] SET <replaceable>cycle_mark_col_name</replaceable> [ TO <replaceable>cycle_mark_value</replaceable> DEFAULT <replaceable>cycle_mark_default</replaceable> ] USING <replaceable>cycle_path_col_name</replaceable> ]
 
@@ -230,10 +230,10 @@ TABLE [ ONLY ] <replaceable class="param
     The subqueries effectively act as temporary tables or views
     for the duration of the primary query.
     Each subquery can be a <command>SELECT</command>, <command>TABLE</command>, <command>VALUES</command>,
-    <command>INSERT</command>, <command>UPDATE</command> or
-    <command>DELETE</command> statement.
+    <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command> statement.
     When writing a data-modifying statement (<command>INSERT</command>,
-    <command>UPDATE</command> or <command>DELETE</command>) in
+    <command>UPDATE</command>, <command>DELETE</command>, or <command>MERGE</command>) in
     <literal>WITH</literal>, it is usual to include a <literal>RETURNING</literal> clause.
     It is the output of <literal>RETURNING</literal>, <emphasis>not</emphasis> the underlying
     table that the statement modifies, that forms the temporary table that is
@@ -2182,7 +2182,8 @@ SELECT 2+2;
 
    <para>
     <productname>PostgreSQL</productname> allows <command>INSERT</command>,
-    <command>UPDATE</command>, and <command>DELETE</command> to be used as <literal>WITH</literal>
+    <command>UPDATE</command>, <command>DELETE</command>, and
+    <command>MERGE</command> to be used as <literal>WITH</literal>
     queries.  This is not found in the SQL standard.
    </para>
   </refsect2>
diff --git a/doc/src/sgml/rowtypes.sgml b/doc/src/sgml/rowtypes.sgml
new file mode 100644
index 4d86f97..bbeac84
--- a/doc/src/sgml/rowtypes.sgml
+++ b/doc/src/sgml/rowtypes.sgml
@@ -348,7 +348,7 @@ SELECT m.* FROM some_table, LATERAL myfu
    column expansion of this kind when it appears at the top level of
    a <link linkend="queries-select-lists"><command>SELECT</command> output
    list</link>, a <link linkend="dml-returning"><literal>RETURNING</literal>
-   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>,
+   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>,
    a <link linkend="queries-values"><literal>VALUES</literal> clause</link>, or
    a <link linkend="sql-syntax-row-constructors">row constructor</link>.
    In all other contexts (including when nested inside one of those
diff --git a/doc/src/sgml/spi.sgml b/doc/src/sgml/spi.sgml
new file mode 100644
index 47f4b5b..bb37786
--- a/doc/src/sgml/spi.sgml
+++ b/doc/src/sgml/spi.sgml
@@ -301,8 +301,9 @@ SPI_execute("INSERT INTO foo SELECT * FR
    is returned in the global variable <varname>SPI_processed</varname>.
    If the return value of the function is <symbol>SPI_OK_SELECT</symbol>,
    <symbol>SPI_OK_INSERT_RETURNING</symbol>,
-   <symbol>SPI_OK_DELETE_RETURNING</symbol>, or
-   <symbol>SPI_OK_UPDATE_RETURNING</symbol>,
+   <symbol>SPI_OK_DELETE_RETURNING</symbol>,
+   <symbol>SPI_OK_UPDATE_RETURNING</symbol>, or
+   <symbol>SPI_OK_MERGE_RETURNING</symbol>,
    then you can use the
    global pointer <literal>SPITupleTable *SPI_tuptable</literal> to
    access the result rows.  Some utility commands (such as
@@ -472,6 +473,15 @@ typedef struct SPITupleTable
       </para>
      </listitem>
     </varlistentry>
+
+    <varlistentry>
+     <term><symbol>SPI_OK_MERGE_RETURNING</symbol></term>
+     <listitem>
+      <para>
+       if a <command>MERGE RETURNING</command> was executed
+      </para>
+     </listitem>
+    </varlistentry>
 
     <varlistentry>
      <term><symbol>SPI_OK_UTILITY</symbol></term>
diff --git a/doc/src/sgml/xfunc.sgml b/doc/src/sgml/xfunc.sgml
new file mode 100644
index 89116ae..cb0aa55
--- a/doc/src/sgml/xfunc.sgml
+++ b/doc/src/sgml/xfunc.sgml
@@ -177,7 +177,8 @@
     statements separated by semicolons.  A semicolon after the last
     statement is optional.  Unless the function is declared to return
     <type>void</type>, the last statement must be a <command>SELECT</command>,
-    or an <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>
+    or an <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command>
     that has a <literal>RETURNING</literal> clause.
    </para>
 
@@ -1236,8 +1237,9 @@ SELECT x, CASE WHEN x &gt; 0 THEN genera
 
     <note>
      <para>
-      If a function's last command is <command>INSERT</command>, <command>UPDATE</command>,
-      or <command>DELETE</command> with <literal>RETURNING</literal>, that command will
+      If a function's last command is <command>INSERT</command>,
+      <command>UPDATE</command>, <command>DELETE</command>, or
+      <command>MERGE</command> with <literal>RETURNING</literal>, that command will
       always be executed to completion, even if the function is not declared
       with <literal>SETOF</literal> or the calling query does not fetch all the
       result rows.  Any extra rows produced by the <literal>RETURNING</literal>
diff --git a/src/backend/commands/copy.c b/src/backend/commands/copy.c
new file mode 100644
index c5d7d78..7977873
--- a/src/backend/commands/copy.c
+++ b/src/backend/commands/copy.c
@@ -283,12 +283,6 @@ DoCopy(ParseState *pstate, const CopyStm
 	{
 		Assert(stmt->query);
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(stmt->query, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in COPY"));
-
 		query = makeNode(RawStmt);
 		query->stmt = stmt->query;
 		query->stmt_location = stmt_location;
diff --git a/src/backend/commands/copyto.c b/src/backend/commands/copyto.c
new file mode 100644
index c66a047..1145aaf
--- a/src/backend/commands/copyto.c
+++ b/src/backend/commands/copyto.c
@@ -510,7 +510,8 @@ BeginCopyTo(ParseState *pstate,
 		{
 			Assert(query->commandType == CMD_INSERT ||
 				   query->commandType == CMD_UPDATE ||
-				   query->commandType == CMD_DELETE);
+				   query->commandType == CMD_DELETE ||
+				   query->commandType == CMD_MERGE);
 
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
diff --git a/src/backend/executor/execExpr.c b/src/backend/executor/execExpr.c
new file mode 100644
index 2c62b0c..6414c0d
--- a/src/backend/executor/execExpr.c
+++ b/src/backend/executor/execExpr.c
@@ -1111,6 +1111,21 @@ ExecInitExprRec(Expr *node, ExprState *s
 				break;
 			}
 
+		case T_MergingFunc:
+			{
+				MergingFunc *mf_node = (MergingFunc *) node;
+
+				if (!state->parent ||
+					!IsA(state->parent, ModifyTableState) ||
+					((ModifyTableState *) state->parent)->operation != CMD_MERGE)
+					elog(ERROR, "MergingFunc found in non-merge plan node");
+
+				scratch.opcode = EEOP_MERGING_FUNC;
+				scratch.d.merging_func.mfop = mf_node->mfop;
+				ExprEvalPushStep(state, &scratch);
+				break;
+			}
+
 		case T_SubscriptingRef:
 			{
 				SubscriptingRef *sbsref = (SubscriptingRef *) node;
diff --git a/src/backend/executor/execExprInterp.c b/src/backend/executor/execExprInterp.c
new file mode 100644
index 24c2b60..b176ab6
--- a/src/backend/executor/execExprInterp.c
+++ b/src/backend/executor/execExprInterp.c
@@ -483,6 +483,7 @@ ExecInterpExpr(ExprState *state, ExprCon
 		&&CASE_EEOP_AGGREF,
 		&&CASE_EEOP_GROUPING_FUNC,
 		&&CASE_EEOP_WINDOW_FUNC,
+		&&CASE_EEOP_MERGING_FUNC,
 		&&CASE_EEOP_SUBPLAN,
 		&&CASE_EEOP_AGG_STRICT_DESERIALIZE,
 		&&CASE_EEOP_AGG_DESERIALIZE,
@@ -1582,6 +1583,14 @@ ExecInterpExpr(ExprState *state, ExprCon
 			EEO_NEXT();
 		}
 
+		EEO_CASE(EEOP_MERGING_FUNC)
+		{
+			/* too complex/uncommon for an inline implementation */
+			ExecEvalMergingFunc(state, op, econtext);
+
+			EEO_NEXT();
+		}
+
 		EEO_CASE(EEOP_SUBPLAN)
 		{
 			/* too complex for an inline implementation */
@@ -4171,6 +4180,59 @@ ExecEvalGroupingFunc(ExprState *state, E
 }
 
 /*
+ * ExecEvalMergingFunc
+ *
+ * Returns information about the current MERGE action for its RETURNING list.
+ */
+void
+ExecEvalMergingFunc(ExprState *state, ExprEvalStep *op, ExprContext *econtext)
+{
+	MergingFuncOp mfop = op->d.merging_func.mfop;
+	ModifyTableState *mtstate = castNode(ModifyTableState, state->parent);
+	MergeActionState *relaction = mtstate->mt_merge_action;
+
+	if (!relaction)
+		elog(ERROR, "no merge action in progress");
+
+	switch (mfop)
+	{
+		case MERGING_ACTION:
+			/* Return the MERGE action ("INSERT", "UPDATE", or "DELETE") */
+			switch (relaction->mas_action->commandType)
+			{
+				case CMD_INSERT:
+					*op->resvalue = PointerGetDatum(cstring_to_text_with_len("INSERT", 6));
+					*op->resnull = false;
+					break;
+				case CMD_UPDATE:
+					*op->resvalue = PointerGetDatum(cstring_to_text_with_len("UPDATE", 6));
+					*op->resnull = false;
+					break;
+				case CMD_DELETE:
+					*op->resvalue = PointerGetDatum(cstring_to_text_with_len("DELETE", 6));
+					*op->resnull = false;
+					break;
+				case CMD_NOTHING:
+					elog(ERROR, "unexpected merge action: DO NOTHING");
+					break;
+				default:
+					elog(ERROR, "unrecognized commandType: %d",
+						 (int) relaction->mas_action->commandType);
+			}
+			break;
+
+		case MERGING_CLAUSE_NUMBER:
+			/* Return the 1-based index of the MERGE action */
+			*op->resvalue = Int32GetDatum(relaction->mas_action->index);
+			*op->resnull = false;
+			break;
+
+		default:
+			elog(ERROR, "unrecognized MergingFuncOp: %d", (int) mfop);
+	}
+}
+
+/*
  * Hand off evaluation of a subplan to nodeSubplan.c
  */
 void
diff --git a/src/backend/executor/execPartition.c b/src/backend/executor/execPartition.c
new file mode 100644
index f6c3432..34d8231
--- a/src/backend/executor/execPartition.c
+++ b/src/backend/executor/execPartition.c
@@ -612,8 +612,8 @@ ExecInitPartitionInfo(ModifyTableState *
 	 * Build the RETURNING projection for the partition.  Note that we didn't
 	 * build the returningList for partitions within the planner, but simple
 	 * translation of varattnos will suffice.  This only occurs for the INSERT
-	 * case or in the case of UPDATE tuple routing where we didn't find a
-	 * result rel to reuse.
+	 * case or in the case of UPDATE/MERGE tuple routing where we didn't find
+	 * a result rel to reuse.
 	 */
 	if (node && node->returningLists != NIL)
 	{
@@ -622,12 +622,14 @@ ExecInitPartitionInfo(ModifyTableState *
 		List	   *returningList;
 
 		/* See the comment above for WCO lists. */
-		/* (except no RETURNING support for MERGE yet) */
 		Assert((node->operation == CMD_INSERT &&
 				list_length(node->returningLists) == 1 &&
 				list_length(node->resultRelations) == 1) ||
 			   (node->operation == CMD_UPDATE &&
 				list_length(node->returningLists) ==
+				list_length(node->resultRelations)) ||
+			   (node->operation == CMD_MERGE &&
+				list_length(node->returningLists) ==
 				list_length(node->resultRelations)));
 
 		/*
diff --git a/src/backend/executor/functions.c b/src/backend/executor/functions.c
new file mode 100644
index bace252..68cb4fa
--- a/src/backend/executor/functions.c
+++ b/src/backend/executor/functions.c
@@ -1660,8 +1660,8 @@ check_sql_fn_retval(List *queryTreeLists
 
 	/*
 	 * If it's a plain SELECT, it returns whatever the targetlist says.
-	 * Otherwise, if it's INSERT/UPDATE/DELETE with RETURNING, it returns
-	 * that. Otherwise, the function return type must be VOID.
+	 * Otherwise, if it's INSERT/UPDATE/DELETE/MERGE with RETURNING, it
+	 * returns that. Otherwise, the function return type must be VOID.
 	 *
 	 * Note: eventually replace this test with QueryReturnsTuples?	We'd need
 	 * a more general method of determining the output type, though.  Also, it
@@ -1679,7 +1679,8 @@ check_sql_fn_retval(List *queryTreeLists
 	else if (parse &&
 			 (parse->commandType == CMD_INSERT ||
 			  parse->commandType == CMD_UPDATE ||
-			  parse->commandType == CMD_DELETE) &&
+			  parse->commandType == CMD_DELETE ||
+			  parse->commandType == CMD_MERGE) &&
 			 parse->returningList)
 	{
 		tlist = parse->returningList;
@@ -1693,7 +1694,7 @@ check_sql_fn_retval(List *queryTreeLists
 				(errcode(ERRCODE_INVALID_FUNCTION_DEFINITION),
 				 errmsg("return type mismatch in function declared to return %s",
 						format_type_be(rettype)),
-				 errdetail("Function's final statement must be SELECT or INSERT/UPDATE/DELETE RETURNING.")));
+				 errdetail("Function's final statement must be SELECT or INSERT/UPDATE/DELETE/MERGE RETURNING.")));
 		return false;			/* keep compiler quiet */
 	}
 
diff --git a/src/backend/executor/nodeModifyTable.c b/src/backend/executor/nodeModifyTable.c
new file mode 100644
index 299c2c7..7745d86
--- a/src/backend/executor/nodeModifyTable.c
+++ b/src/backend/executor/nodeModifyTable.c
@@ -36,8 +36,7 @@
  *		RETURNING tuple after completing each row insert, update, or delete.
  *		It must be called again to continue the operation.  Without RETURNING,
  *		we just loop within the node until all the work is done, then
- *		return NULL.  This avoids useless call/return overhead.  (MERGE does
- *		not support RETURNING.)
+ *		return NULL.  This avoids useless call/return overhead.
  */
 
 #include "postgres.h"
@@ -88,9 +87,6 @@ typedef struct ModifyTableContext
 	 */
 	TupleTableSlot *planSlot;
 
-	/* MERGE specific */
-	MergeActionState *relaction;	/* MERGE action in progress */
-
 	/*
 	 * Information about the changes that were made concurrently to a tuple
 	 * being updated or deleted
@@ -153,13 +149,14 @@ static TupleTableSlot *ExecMerge(ModifyT
 								 ItemPointer tupleid,
 								 bool canSetTag);
 static void ExecInitMerge(ModifyTableState *mtstate, EState *estate);
-static bool ExecMergeMatched(ModifyTableContext *context,
-							 ResultRelInfo *resultRelInfo,
-							 ItemPointer tupleid,
-							 bool canSetTag);
-static void ExecMergeNotMatched(ModifyTableContext *context,
-								ResultRelInfo *resultRelInfo,
-								bool canSetTag);
+static TupleTableSlot *ExecMergeMatched(ModifyTableContext *context,
+										ResultRelInfo *resultRelInfo,
+										ItemPointer tupleid,
+										bool canSetTag,
+										bool *matched);
+static TupleTableSlot *ExecMergeNotMatched(ModifyTableContext *context,
+										   ResultRelInfo *resultRelInfo,
+										   bool canSetTag);
 
 
 /*
@@ -979,7 +976,7 @@ ExecInsert(ModifyTableContext *context,
 		if (mtstate->operation == CMD_UPDATE)
 			wco_kind = WCO_RLS_UPDATE_CHECK;
 		else if (mtstate->operation == CMD_MERGE)
-			wco_kind = (context->relaction->mas_action->commandType == CMD_UPDATE) ?
+			wco_kind = (mtstate->mt_merge_action->mas_action->commandType == CMD_UPDATE) ?
 				WCO_RLS_UPDATE_CHECK : WCO_RLS_INSERT_CHECK;
 		else
 			wco_kind = WCO_RLS_INSERT_CHECK;
@@ -1828,7 +1825,7 @@ ExecCrossPartitionUpdate(ModifyTableCont
 		 * additional rechecking, and might end up executing a different
 		 * action entirely).
 		 */
-		if (context->relaction != NULL)
+		if (mtstate->mt_merge_action != NULL)
 			return false;
 		else if (TupIsNull(epqslot))
 			return true;
@@ -2069,7 +2066,7 @@ lreplace:
 		 * No luck, a retry is needed.  If running MERGE, we do not do so
 		 * here; instead let it handle that on its own rules.
 		 */
-		if (context->relaction != NULL)
+		if (context->mtstate->mt_merge_action != NULL)
 			return TM_Updated;
 
 		/*
@@ -2712,6 +2709,7 @@ static TupleTableSlot *
 ExecMerge(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 		  ItemPointer tupleid, bool canSetTag)
 {
+	TupleTableSlot *rslot = NULL;
 	bool		matched;
 
 	/*-----
@@ -2759,18 +2757,18 @@ ExecMerge(ModifyTableContext *context, R
 	 */
 	matched = tupleid != NULL;
 	if (matched)
-		matched = ExecMergeMatched(context, resultRelInfo, tupleid, canSetTag);
+		rslot = ExecMergeMatched(context, resultRelInfo, tupleid, canSetTag,
+								 &matched);
 
 	/*
-	 * Either we were dealing with a NOT MATCHED tuple or ExecMergeMatched()
-	 * returned "false", indicating the previously MATCHED tuple no longer
-	 * matches.
+	 * Deal with the NOT MATCHED case (either a NOT MATCHED tuple from the
+	 * join, or a previously MATCHED tuple for which ExecMergeMatched() set
+	 * "matched" to false, indicating that it no longer matches).
 	 */
 	if (!matched)
-		ExecMergeNotMatched(context, resultRelInfo, canSetTag);
+		rslot = ExecMergeNotMatched(context, resultRelInfo, canSetTag);
 
-	/* No RETURNING support yet */
-	return NULL;
+	return rslot;
 }
 
 /*
@@ -2780,8 +2778,8 @@ ExecMerge(ModifyTableContext *context, R
  * We start from the first WHEN MATCHED action and check if the WHEN quals
  * pass, if any. If the WHEN quals for the first action do not pass, we
  * check the second, then the third and so on. If we reach to the end, no
- * action is taken and we return true, indicating that no further action is
- * required for this tuple.
+ * action is taken and "matched" is set to true, indicating that no further
+ * action is required for this tuple.
  *
  * If we do find a qualifying action, then we attempt to execute the action.
  *
@@ -2790,16 +2788,17 @@ ExecMerge(ModifyTableContext *context, R
  * with individual actions are evaluated by this routine via ExecQual, while
  * EvalPlanQual checks for the join quals. If EvalPlanQual tells us that the
  * updated tuple still passes the join quals, then we restart from the first
- * action to look for a qualifying action. Otherwise, we return false --
- * meaning that a NOT MATCHED action must now be executed for the current
- * source tuple.
+ * action to look for a qualifying action. Otherwise, "matched" is set to
+ * false -- meaning that a NOT MATCHED action must now be executed for the
+ * current source tuple.
  */
-static bool
+static TupleTableSlot *
 ExecMergeMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
-				 ItemPointer tupleid, bool canSetTag)
+				 ItemPointer tupleid, bool canSetTag, bool *matched)
 {
 	ModifyTableState *mtstate = context->mtstate;
-	TupleTableSlot *newslot;
+	TupleTableSlot *rslot = NULL;
+	TupleTableSlot *newslot = NULL;
 	EState	   *estate = context->estate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
 	bool		isNull;
@@ -2810,7 +2809,10 @@ ExecMergeMatched(ModifyTableContext *con
 	 * If there are no WHEN MATCHED actions, we are done.
 	 */
 	if (resultRelInfo->ri_matchedMergeAction == NIL)
-		return true;
+	{
+		*matched = true;
+		return NULL;
+	}
 
 	/*
 	 * Make tuple and any needed join variables available to ExecQual and
@@ -2889,12 +2891,15 @@ lmerge_matched:
 				 */
 				newslot = ExecProject(relaction->mas_proj);
 
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				if (!ExecUpdatePrologue(context, resultRelInfo,
 										tupleid, NULL, newslot, &result))
 				{
 					if (result == TM_Ok)
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 					break;		/* concurrent update/delete */
 				}
 				result = ExecUpdateAct(context, resultRelInfo, tupleid, NULL,
@@ -2908,12 +2913,15 @@ lmerge_matched:
 				break;
 
 			case CMD_DELETE:
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				if (!ExecDeletePrologue(context, resultRelInfo, tupleid,
 										NULL, NULL, &result))
 				{
 					if (result == TM_Ok)
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 					break;		/* concurrent update/delete */
 				}
 				result = ExecDeleteAct(context, resultRelInfo, tupleid, false);
@@ -2969,7 +2977,8 @@ lmerge_matched:
 				 * If the tuple was already deleted, return to let caller
 				 * handle it under NOT MATCHED clauses.
 				 */
-				return false;
+				*matched = false;
+				return NULL;
 
 			case TM_Updated:
 				{
@@ -3015,13 +3024,19 @@ lmerge_matched:
 							 * NOT MATCHED actions.
 							 */
 							if (TupIsNull(epqslot))
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							(void) ExecGetJunkAttribute(epqslot,
 														resultRelInfo->ri_RowIdAttNo,
 														&isNull);
 							if (isNull)
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							/*
 							 * When a tuple was updated and migrated to
@@ -3056,7 +3071,8 @@ lmerge_matched:
 							 * tuple already deleted; tell caller to run NOT
 							 * MATCHED actions
 							 */
-							return false;
+							*matched = false;
+							return NULL;
 
 						case TM_SelfModified:
 
@@ -3076,13 +3092,14 @@ lmerge_matched:
 										(errcode(ERRCODE_TRIGGERED_DATA_CHANGE_VIOLATION),
 										 errmsg("tuple to be updated or deleted was already modified by an operation triggered by the current command"),
 										 errhint("Consider using an AFTER trigger instead of a BEFORE trigger to propagate changes to other rows.")));
-							return false;
+							*matched = false;
+							return NULL;
 
 						default:
 							/* see table_tuple_lock call in ExecDelete() */
 							elog(ERROR, "unexpected table_tuple_lock status: %u",
 								 result);
-							return false;
+							return NULL;
 					}
 				}
 
@@ -3094,6 +3111,31 @@ lmerge_matched:
 				break;
 		}
 
+		/* Process RETURNING if present */
+		if (resultRelInfo->ri_projectReturning)
+		{
+			switch (commandType)
+			{
+				case CMD_UPDATE:
+					rslot = ExecProcessReturning(resultRelInfo, newslot,
+												 context->planSlot);
+					break;
+
+				case CMD_DELETE:
+					rslot = ExecProcessReturning(resultRelInfo,
+												 resultRelInfo->ri_oldTupleSlot,
+												 context->planSlot);
+					break;
+
+				case CMD_NOTHING:
+					break;
+
+				default:
+					elog(ERROR, "unrecognized commandType: %d",
+						 (int) commandType);
+			}
+		}
+
 		/*
 		 * We've activated one of the WHEN clauses, so we don't search
 		 * further. This is required behaviour, not an optimization.
@@ -3104,19 +3146,22 @@ lmerge_matched:
 	/*
 	 * Successfully executed an action or no qualifying action was found.
 	 */
-	return true;
+	*matched = true;
+
+	return rslot;
 }
 
 /*
  * Execute the first qualifying NOT MATCHED action.
  */
-static void
+static TupleTableSlot *
 ExecMergeNotMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 					bool canSetTag)
 {
 	ModifyTableState *mtstate = context->mtstate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
 	List	   *actionStates = NIL;
+	TupleTableSlot *rslot = NULL;
 	ListCell   *l;
 
 	/*
@@ -3166,10 +3211,10 @@ ExecMergeNotMatched(ModifyTableContext *
 				 * so we don't need to map the tuple here.
 				 */
 				newslot = ExecProject(action->mas_proj);
-				context->relaction = action;
+				mtstate->mt_merge_action = action;
 
-				(void) ExecInsert(context, mtstate->rootResultRelInfo, newslot,
-								  canSetTag, NULL, NULL);
+				rslot = ExecInsert(context, mtstate->rootResultRelInfo,
+								   newslot, canSetTag, NULL, NULL);
 				mtstate->mt_merge_inserted += 1;
 				break;
 			case CMD_NOTHING:
@@ -3185,6 +3230,8 @@ ExecMergeNotMatched(ModifyTableContext *
 		 */
 		break;
 	}
+
+	return rslot;
 }
 
 /*
@@ -3647,8 +3694,17 @@ ExecModifyTable(PlanState *pstate)
 				{
 					EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-					ExecMerge(&context, node->resultRelInfo, NULL, node->canSetTag);
-					continue;	/* no RETURNING support yet */
+					slot = ExecMerge(&context, node->resultRelInfo, NULL,
+									 node->canSetTag);
+
+					/*
+					 * If we got a RETURNING result, return it to the caller.
+					 * We'll continue the work on next call.
+					 */
+					if (slot)
+						return slot;
+
+					continue;	/* continue with the next tuple */
 				}
 
 				elog(ERROR, "tableoid is NULL");
@@ -3725,8 +3781,17 @@ ExecModifyTable(PlanState *pstate)
 					{
 						EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-						ExecMerge(&context, node->resultRelInfo, NULL, node->canSetTag);
-						continue;	/* no RETURNING support yet */
+						slot = ExecMerge(&context, node->resultRelInfo, NULL,
+										 node->canSetTag);
+
+						/*
+						 * If we got a RETURNING result, return it to the
+						 * caller.  We'll continue the work on next call.
+						 */
+						if (slot)
+							return slot;
+
+						continue;	/* continue with the next tuple */
 					}
 
 					elog(ERROR, "ctid is NULL");
@@ -3818,7 +3883,7 @@ ExecModifyTable(PlanState *pstate)
 				}
 				slot = ExecGetUpdateNewTuple(resultRelInfo, context.planSlot,
 											 oldSlot);
-				context.relaction = NULL;
+				node->mt_merge_action = NULL;
 
 				/* Now apply the update. */
 				slot = ExecUpdate(&context, resultRelInfo, tupleid, oldtuple,
diff --git a/src/backend/executor/spi.c b/src/backend/executor/spi.c
new file mode 100644
index 3397568..b9f443c
--- a/src/backend/executor/spi.c
+++ b/src/backend/executor/spi.c
@@ -2033,6 +2033,8 @@ SPI_result_code_string(int code)
 			return "SPI_OK_TD_REGISTER";
 		case SPI_OK_MERGE:
 			return "SPI_OK_MERGE";
+		case SPI_OK_MERGE_RETURNING:
+			return "SPI_OK_MERGE_RETURNING";
 	}
 	/* Unrecognized code ... return something useful ... */
 	sprintf(buf, "Unrecognized SPI code %d", code);
@@ -2886,7 +2888,10 @@ _SPI_pquery(QueryDesc *queryDesc, bool f
 				res = SPI_OK_UPDATE;
 			break;
 		case CMD_MERGE:
-			res = SPI_OK_MERGE;
+			if (queryDesc->plannedstmt->hasReturning)
+				res = SPI_OK_MERGE_RETURNING;
+			else
+				res = SPI_OK_MERGE;
 			break;
 		default:
 			return SPI_ERROR_OPUNKNOWN;
diff --git a/src/backend/jit/llvm/llvmjit_expr.c b/src/backend/jit/llvm/llvmjit_expr.c
new file mode 100644
index a3a0876..e966037
--- a/src/backend/jit/llvm/llvmjit_expr.c
+++ b/src/backend/jit/llvm/llvmjit_expr.c
@@ -1980,6 +1980,12 @@ llvm_compile_expr(ExprState *state)
 					break;
 				}
 
+			case EEOP_MERGING_FUNC:
+				build_EvalXFunc(b, mod, "ExecEvalMergingFunc",
+								v_state, op, v_econtext);
+				LLVMBuildBr(b, opblocks[opno + 1]);
+				break;
+
 			case EEOP_SUBPLAN:
 				build_EvalXFunc(b, mod, "ExecEvalSubPlan",
 								v_state, op, v_econtext);
diff --git a/src/backend/jit/llvm/llvmjit_types.c b/src/backend/jit/llvm/llvmjit_types.c
new file mode 100644
index 791902f..65db8f2
--- a/src/backend/jit/llvm/llvmjit_types.c
+++ b/src/backend/jit/llvm/llvmjit_types.c
@@ -155,6 +155,7 @@ void	   *referenced_functions[] =
 	ExecEvalFuncExprFusage,
 	ExecEvalFuncExprStrictFusage,
 	ExecEvalGroupingFunc,
+	ExecEvalMergingFunc,
 	ExecEvalMinMax,
 	ExecEvalNextValueExpr,
 	ExecEvalParamExec,
diff --git a/src/backend/nodes/nodeFuncs.c b/src/backend/nodes/nodeFuncs.c
new file mode 100644
index c03f4f2..6cf7895
--- a/src/backend/nodes/nodeFuncs.c
+++ b/src/backend/nodes/nodeFuncs.c
@@ -67,6 +67,9 @@ exprType(const Node *expr)
 		case T_WindowFunc:
 			type = ((const WindowFunc *) expr)->wintype;
 			break;
+		case T_MergingFunc:
+			type = ((const MergingFunc *) expr)->mftype;
+			break;
 		case T_SubscriptingRef:
 			type = ((const SubscriptingRef *) expr)->refrestype;
 			break;
@@ -810,6 +813,9 @@ exprCollation(const Node *expr)
 		case T_WindowFunc:
 			coll = ((const WindowFunc *) expr)->wincollid;
 			break;
+		case T_MergingFunc:
+			coll = ((const MergingFunc *) expr)->mfcollid;
+			break;
 		case T_SubscriptingRef:
 			coll = ((const SubscriptingRef *) expr)->refcollid;
 			break;
@@ -1085,6 +1091,9 @@ exprSetCollation(Node *expr, Oid collati
 		case T_WindowFunc:
 			((WindowFunc *) expr)->wincollid = collation;
 			break;
+		case T_MergingFunc:
+			((MergingFunc *) expr)->mfcollid = collation;
+			break;
 		case T_SubscriptingRef:
 			((SubscriptingRef *) expr)->refcollid = collation;
 			break;
@@ -1343,6 +1352,9 @@ exprLocation(const Node *expr)
 			/* function name should always be the first thing */
 			loc = ((const WindowFunc *) expr)->location;
 			break;
+		case T_MergingFunc:
+			loc = ((const MergingFunc *) expr)->location;
+			break;
 		case T_SubscriptingRef:
 			/* just use container argument's location */
 			loc = exprLocation((Node *) ((const SubscriptingRef *) expr)->refexpr);
@@ -2035,6 +2047,7 @@ expression_tree_walker_impl(Node *node,
 		case T_RangeTblRef:
 		case T_SortGroupClause:
 		case T_CTESearchClause:
+		case T_MergingFunc:
 			/* primitive node types with no expression subnodes */
 			break;
 		case T_WithCheckOption:
@@ -2865,6 +2878,7 @@ expression_tree_mutator_impl(Node *node,
 		case T_RangeTblRef:
 		case T_SortGroupClause:
 		case T_CTESearchClause:
+		case T_MergingFunc:
 			return (Node *) copyObject(node);
 		case T_WithCheckOption:
 			{
@@ -3827,6 +3841,7 @@ raw_expression_tree_walker_impl(Node *no
 		case T_ParamRef:
 		case T_A_Const:
 		case T_A_Star:
+		case T_MergingFunc:
 			/* primitive node types with no subnodes */
 			break;
 		case T_Alias:
diff --git a/src/backend/optimizer/plan/subselect.c b/src/backend/optimizer/plan/subselect.c
new file mode 100644
index 7a9fe88..262adf5
--- a/src/backend/optimizer/plan/subselect.c
+++ b/src/backend/optimizer/plan/subselect.c
@@ -1846,7 +1846,8 @@ convert_EXISTS_to_ANY(PlannerInfo *root,
 /*
  * Replace correlation vars (uplevel vars) with Params.
  *
- * Uplevel PlaceHolderVars and aggregates are replaced, too.
+ * Uplevel PlaceHolderVars, aggregates, GROUPING() expressions, and MERGING()
+ * expressions are replaced, too.
  *
  * Note: it is critical that this runs immediately after SS_process_sublinks.
  * Since we do not recurse into the arguments of uplevel PHVs and aggregates,
@@ -1900,6 +1901,11 @@ replace_correlation_vars_mutator(Node *n
 		if (((GroupingFunc *) node)->agglevelsup > 0)
 			return (Node *) replace_outer_grouping(root, (GroupingFunc *) node);
 	}
+	if (IsA(node, MergingFunc))
+	{
+		if (root->parse->commandType != CMD_MERGE)
+			return (Node *) replace_outer_merging(root, (MergingFunc *) node);
+	}
 	return expression_tree_mutator(node,
 								   replace_correlation_vars_mutator,
 								   (void *) root);
diff --git a/src/backend/optimizer/util/paramassign.c b/src/backend/optimizer/util/paramassign.c
new file mode 100644
index d6a923b..b3b4da6
--- a/src/backend/optimizer/util/paramassign.c
+++ b/src/backend/optimizer/util/paramassign.c
@@ -307,6 +307,56 @@ replace_outer_grouping(PlannerInfo *root
 }
 
 /*
+ * Generate a Param node to replace the given MergingFunc expression which is
+ * expected to be in the RETURNING list of an upper-level MERGE query.  Record
+ * the need for the MergingFunc in the proper upper-level root->plan_params.
+ */
+Param *
+replace_outer_merging(PlannerInfo *root, MergingFunc *mf)
+{
+	Param	   *retval;
+	PlannerParamItem *pitem;
+	Oid			ptype = exprType((Node *) mf);
+
+	Assert(root->parse->commandType != CMD_MERGE);
+
+	/*
+	 * The parser should have ensured that the MergingFunc is in the RETURNING
+	 * list of an upper-level MERGE query, so find that query.
+	 */
+	do
+	{
+		root = root->parent_root;
+		if (root == NULL)
+			elog(ERROR, "MergingFunc found outside MERGE");
+	} while (root->parse->commandType != CMD_MERGE);
+
+	/*
+	 * It does not seem worthwhile to try to de-duplicate references to outer
+	 * MergingFunc expressions.  Just make a new slot every time.
+	 */
+	mf = copyObject(mf);
+
+	pitem = makeNode(PlannerParamItem);
+	pitem->item = (Node *) mf;
+	pitem->paramId = list_length(root->glob->paramExecTypes);
+	root->glob->paramExecTypes = lappend_oid(root->glob->paramExecTypes,
+											 ptype);
+
+	root->plan_params = lappend(root->plan_params, pitem);
+
+	retval = makeNode(Param);
+	retval->paramkind = PARAM_EXEC;
+	retval->paramid = pitem->paramId;
+	retval->paramtype = ptype;
+	retval->paramtypmod = -1;
+	retval->paramcollid = InvalidOid;
+	retval->location = mf->location;
+
+	return retval;
+}
+
+/*
  * Generate a Param node to replace the given Var,
  * which is expected to come from some upper NestLoop plan node.
  * Record the need for the Var in root->curOuterParams.
diff --git a/src/backend/parser/analyze.c b/src/backend/parser/analyze.c
new file mode 100644
index 7a1dfb6..5f882b4
--- a/src/backend/parser/analyze.c
+++ b/src/backend/parser/analyze.c
@@ -74,7 +74,6 @@ static void determineRecursiveColTypes(P
 									   Node *larg, List *nrtargetlist);
 static Query *transformReturnStmt(ParseState *pstate, ReturnStmt *stmt);
 static Query *transformUpdateStmt(ParseState *pstate, UpdateStmt *stmt);
-static List *transformReturningList(ParseState *pstate, List *returningList);
 static Query *transformPLAssignStmt(ParseState *pstate,
 									PLAssignStmt *stmt);
 static Query *transformDeclareCursorStmt(ParseState *pstate,
@@ -553,7 +552,8 @@ transformDeleteStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -980,7 +980,8 @@ transformInsertStmt(ParseState *pstate,
 	/* Process RETURNING, if any. */
 	if (stmt->returningList)
 		qry->returningList = transformReturningList(pstate,
-													stmt->returningList);
+													stmt->returningList,
+													EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -2445,7 +2446,8 @@ transformUpdateStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/*
 	 * Now we are done with SELECT-like processing, and can get on with
@@ -2539,10 +2541,11 @@ transformUpdateTargetList(ParseState *ps
 
 /*
  * transformReturningList -
- *	handle a RETURNING clause in INSERT/UPDATE/DELETE
+ *	handle a RETURNING clause in INSERT/UPDATE/DELETE/MERGE
  */
-static List *
-transformReturningList(ParseState *pstate, List *returningList)
+List *
+transformReturningList(ParseState *pstate, List *returningList,
+					   ParseExprKind exprKind)
 {
 	List	   *rlist;
 	int			save_next_resno;
@@ -2559,7 +2562,7 @@ transformReturningList(ParseState *pstat
 	pstate->p_next_resno = 1;
 
 	/* transform RETURNING identically to a SELECT targetlist */
-	rlist = transformTargetList(pstate, returningList, EXPR_KIND_RETURNING);
+	rlist = transformTargetList(pstate, returningList, exprKind);
 
 	/*
 	 * Complain if the nonempty tlist expanded to nothing (which is possible
diff --git a/src/backend/parser/gram.y b/src/backend/parser/gram.y
new file mode 100644
index c224df4..c366c5b
--- a/src/backend/parser/gram.y
+++ b/src/backend/parser/gram.y
@@ -693,7 +693,7 @@ static Node *makeRecursiveViewSelect(cha
 	BOOLEAN_P BOTH BREADTH BY
 
 	CACHE CALL CALLED CASCADE CASCADED CASE CAST CATALOG_P CHAIN CHAR_P
-	CHARACTER CHARACTERISTICS CHECK CHECKPOINT CLASS CLOSE
+	CHARACTER CHARACTERISTICS CHECK CHECKPOINT CLASS CLAUSE_NUMBER CLOSE
 	CLUSTER COALESCE COLLATE COLLATION COLUMN COLUMNS COMMENT COMMENTS COMMIT
 	COMMITTED COMPRESSION CONCURRENTLY CONFIGURATION CONFLICT
 	CONNECTION CONSTRAINT CONSTRAINTS CONTENT_P CONTINUE_P CONVERSION_P COPY
@@ -731,7 +731,7 @@ static Node *makeRecursiveViewSelect(cha
 	LEADING LEAKPROOF LEAST LEFT LEVEL LIKE LIMIT LISTEN LOAD LOCAL
 	LOCALTIME LOCALTIMESTAMP LOCATION LOCK_P LOCKED LOGGED
 
-	MAPPING MATCH MATCHED MATERIALIZED MAXVALUE MERGE METHOD
+	MAPPING MATCH MATCHED MATERIALIZED MAXVALUE MERGE MERGING METHOD
 	MINUTE_P MINVALUE MODE MONTH_P MOVE
 
 	NAME_P NAMES NATIONAL NATURAL NCHAR NEW NEXT NFC NFD NFKC NFKD NO NONE
@@ -12332,6 +12332,7 @@ MergeStmt:
 			USING table_ref
 			ON a_expr
 			merge_when_list
+			returning_clause
 				{
 					MergeStmt  *m = makeNode(MergeStmt);
 
@@ -12340,6 +12341,7 @@ MergeStmt:
 					m->sourceRelation = $6;
 					m->joinCondition = $8;
 					m->mergeWhenClauses = $9;
+					m->returningList = $10;
 
 					$$ = (Node *) m;
 				}
@@ -15755,6 +15757,24 @@ func_expr_common_subexpr:
 					n->location = @1;
 					$$ = (Node *) n;
 				}
+			| MERGING '(' ACTION ')'
+				{
+					MergingFunc *m = makeNode(MergingFunc);
+
+					m->mfop = MERGING_ACTION;
+					m->mftype = TEXTOID;
+					m->location = @1;
+					$$ = (Node *) m;
+				}
+			| MERGING '(' CLAUSE_NUMBER ')'
+				{
+					MergingFunc *m = makeNode(MergingFunc);
+
+					m->mfop = MERGING_CLAUSE_NUMBER;
+					m->mftype = INT4OID;
+					m->location = @1;
+					$$ = (Node *) m;
+				}
 			;
 
 
@@ -17100,6 +17120,7 @@ unreserved_keyword:
 			| CHARACTERISTICS
 			| CHECKPOINT
 			| CLASS
+			| CLAUSE_NUMBER
 			| CLOSE
 			| CLUSTER
 			| COLUMNS
@@ -17419,6 +17440,7 @@ col_name_keyword:
 			| JSON_SCALAR
 			| JSON_SERIALIZE
 			| LEAST
+			| MERGING
 			| NATIONAL
 			| NCHAR
 			| NONE
@@ -17637,6 +17659,7 @@ bare_label_keyword:
 			| CHECK
 			| CHECKPOINT
 			| CLASS
+			| CLAUSE_NUMBER
 			| CLOSE
 			| CLUSTER
 			| COALESCE
@@ -17808,6 +17831,7 @@ bare_label_keyword:
 			| MATERIALIZED
 			| MAXVALUE
 			| MERGE
+			| MERGING
 			| METHOD
 			| MINVALUE
 			| MODE
diff --git a/src/backend/parser/parse_agg.c b/src/backend/parser/parse_agg.c
new file mode 100644
index 85cd47b..e72c37b
--- a/src/backend/parser/parse_agg.c
+++ b/src/backend/parser/parse_agg.c
@@ -456,6 +456,7 @@ check_agglevels_and_constraints(ParseSta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
@@ -903,6 +904,7 @@ transformWindowFuncCall(ParseState *psta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_cte.c b/src/backend/parser/parse_cte.c
new file mode 100644
index 6992a78..5cad6d8
--- a/src/backend/parser/parse_cte.c
+++ b/src/backend/parser/parse_cte.c
@@ -126,13 +126,6 @@ transformWithClause(ParseState *pstate,
 		CommonTableExpr *cte = (CommonTableExpr *) lfirst(lc);
 		ListCell   *rest;
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(cte->ctequery, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in WITH query"),
-					parser_errposition(pstate, cte->location));
-
 		for_each_cell(rest, withClause->ctes, lnext(withClause->ctes, lc))
 		{
 			CommonTableExpr *cte2 = (CommonTableExpr *) lfirst(rest);
@@ -153,7 +146,8 @@ transformWithClause(ParseState *pstate,
 			/* must be a data-modifying statement */
 			Assert(IsA(cte->ctequery, InsertStmt) ||
 				   IsA(cte->ctequery, UpdateStmt) ||
-				   IsA(cte->ctequery, DeleteStmt));
+				   IsA(cte->ctequery, DeleteStmt) ||
+				   IsA(cte->ctequery, MergeStmt));
 
 			pstate->p_hasModifyingCTE = true;
 		}
diff --git a/src/backend/parser/parse_expr.c b/src/backend/parser/parse_expr.c
new file mode 100644
index 64c582c..ad0717e
--- a/src/backend/parser/parse_expr.c
+++ b/src/backend/parser/parse_expr.c
@@ -54,6 +54,7 @@ static Node *transformAExprDistinct(Pars
 static Node *transformAExprNullIf(ParseState *pstate, A_Expr *a);
 static Node *transformAExprIn(ParseState *pstate, A_Expr *a);
 static Node *transformAExprBetween(ParseState *pstate, A_Expr *a);
+static Node *transformMergingFunc(ParseState *pstate, MergingFunc *f);
 static Node *transformBoolExpr(ParseState *pstate, BoolExpr *a);
 static Node *transformFuncCall(ParseState *pstate, FuncCall *fn);
 static Node *transformMultiAssignRef(ParseState *pstate, MultiAssignRef *maref);
@@ -227,6 +228,10 @@ transformExprRecurse(ParseState *pstate,
 			result = transformGroupingFunc(pstate, (GroupingFunc *) expr);
 			break;
 
+		case T_MergingFunc:
+			result = transformMergingFunc(pstate, (MergingFunc *) expr);
+			break;
+
 		case T_NamedArgExpr:
 			{
 				NamedArgExpr *na = (NamedArgExpr *) expr;
@@ -541,6 +546,7 @@ transformColumnRef(ParseState *pstate, C
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CHECK_CONSTRAINT:
@@ -1354,6 +1360,31 @@ transformAExprBetween(ParseState *pstate
 }
 
 static Node *
+transformMergingFunc(ParseState *pstate, MergingFunc *f)
+{
+	/*
+	 * Check that we're in the RETURNING list of a MERGE command.
+	 */
+	if (pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+	{
+		ParseState *parent_pstate = pstate->parentParseState;
+
+		while (parent_pstate &&
+			   parent_pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+			parent_pstate = parent_pstate->parentParseState;
+
+		if (!parent_pstate ||
+			parent_pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+			ereport(ERROR,
+					errcode(ERRCODE_WRONG_OBJECT_TYPE),
+					errmsg("MERGING() can only be used in the RETURNING list of a MERGE command"),
+					parser_errposition(pstate, f->location));
+	}
+
+	return (Node *) f;
+}
+
+static Node *
 transformBoolExpr(ParseState *pstate, BoolExpr *a)
 {
 	List	   *args = NIL;
@@ -1767,6 +1798,7 @@ transformSubLink(ParseState *pstate, Sub
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CYCLE_MARK:
@@ -3115,6 +3147,7 @@ ParseExprKindName(ParseExprKind exprKind
 		case EXPR_KIND_OFFSET:
 			return "OFFSET";
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			return "RETURNING";
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
diff --git a/src/backend/parser/parse_func.c b/src/backend/parser/parse_func.c
new file mode 100644
index 6c29471..1a0f983
--- a/src/backend/parser/parse_func.c
+++ b/src/backend/parser/parse_func.c
@@ -2599,6 +2599,7 @@ check_srf_call_placement(ParseState *pst
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_merge.c b/src/backend/parser/parse_merge.c
new file mode 100644
index 91b1156..33852e9
--- a/src/backend/parser/parse_merge.c
+++ b/src/backend/parser/parse_merge.c
@@ -233,6 +233,10 @@ transformMergeStmt(ParseState *pstate, M
 	 */
 	qry->jointree = makeFromExpr(pstate->p_joinlist, joinExpr);
 
+	/* Transform the RETURNING list, if any */
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_MERGE_RETURNING);
+
 	/*
 	 * We now have a good query shape, so now look at the WHEN conditions and
 	 * action targetlists.
@@ -254,6 +258,7 @@ transformMergeStmt(ParseState *pstate, M
 		MergeAction *action;
 
 		action = makeNode(MergeAction);
+		action->index = foreach_current_index(l) + 1;
 		action->commandType = mergeWhenClause->commandType;
 		action->matched = mergeWhenClause->matched;
 
@@ -390,9 +395,6 @@ transformMergeStmt(ParseState *pstate, M
 
 	qry->mergeActionList = mergeActionList;
 
-	/* RETURNING could potentially be added in the future, but not in SQL std */
-	qry->returningList = NULL;
-
 	qry->hasTargetSRFs = false;
 	qry->hasSubLinks = pstate->p_hasSubLinks;
 
diff --git a/src/backend/parser/parse_relation.c b/src/backend/parser/parse_relation.c
new file mode 100644
index 864ea9b..cc19041
--- a/src/backend/parser/parse_relation.c
+++ b/src/backend/parser/parse_relation.c
@@ -2346,9 +2346,10 @@ addRangeTableEntryForCTE(ParseState *pst
 		cte->cterefcount++;
 
 	/*
-	 * We throw error if the CTE is INSERT/UPDATE/DELETE without RETURNING.
-	 * This won't get checked in case of a self-reference, but that's OK
-	 * because data-modifying CTEs aren't allowed to be recursive anyhow.
+	 * We throw error if the CTE is INSERT/UPDATE/DELETE/MERGE without
+	 * RETURNING.  This won't get checked in case of a self-reference, but
+	 * that's OK because data-modifying CTEs aren't allowed to be recursive
+	 * anyhow.
 	 */
 	if (IsA(cte->ctequery, Query))
 	{
diff --git a/src/backend/parser/parse_target.c b/src/backend/parser/parse_target.c
new file mode 100644
index 3bc62ac..eaa2a2b
--- a/src/backend/parser/parse_target.c
+++ b/src/backend/parser/parse_target.c
@@ -1807,6 +1807,20 @@ FigureColnameInternal(Node *node, char *
 			/* make GROUPING() act like a regular function */
 			*name = "grouping";
 			return 2;
+		case T_MergingFunc:
+			switch (((MergingFunc *) node)->mfop)
+			{
+				case MERGING_ACTION:
+					*name = "action";
+					return 2;
+				case MERGING_CLAUSE_NUMBER:
+					*name = "clause_number";
+					return 2;
+				default:
+					elog(ERROR, "unrecognized MergingFuncOp: %d",
+						 (int) ((MergingFunc *) node)->mfop);
+			}
+			break;
 		case T_SubLink:
 			switch (((SubLink *) node)->subLinkType)
 			{
diff --git a/src/backend/rewrite/rewriteHandler.c b/src/backend/rewrite/rewriteHandler.c
new file mode 100644
index 41a3623..3786a7c
--- a/src/backend/rewrite/rewriteHandler.c
+++ b/src/backend/rewrite/rewriteHandler.c
@@ -3615,9 +3615,9 @@ RewriteQuery(Query *parsetree, List *rew
 	ListCell   *lc1;
 
 	/*
-	 * First, recursively process any insert/update/delete statements in WITH
-	 * clauses.  (We have to do this first because the WITH clauses may get
-	 * copied into rule actions below.)
+	 * First, recursively process any insert/update/delete/merge statements in
+	 * WITH clauses.  (We have to do this first because the WITH clauses may
+	 * get copied into rule actions below.)
 	 */
 	foreach(lc1, parsetree->cteList)
 	{
@@ -3642,7 +3642,8 @@ RewriteQuery(Query *parsetree, List *rew
 			if (!(ctequery->commandType == CMD_SELECT ||
 				  ctequery->commandType == CMD_UPDATE ||
 				  ctequery->commandType == CMD_INSERT ||
-				  ctequery->commandType == CMD_DELETE))
+				  ctequery->commandType == CMD_DELETE ||
+				  ctequery->commandType == CMD_MERGE))
 			{
 				/*
 				 * Currently it could only be NOTIFY; this error message will
diff --git a/src/backend/rewrite/rowsecurity.c b/src/backend/rewrite/rowsecurity.c
new file mode 100644
index b1620e4..966dcf0
--- a/src/backend/rewrite/rowsecurity.c
+++ b/src/backend/rewrite/rowsecurity.c
@@ -395,10 +395,10 @@ get_row_security_policies(Query *root, R
 	 * on the final action we take.
 	 *
 	 * We already fetched the SELECT policies above, to check existing rows,
-	 * but we must also check that new rows created by UPDATE actions are
-	 * visible, if SELECT rights are required for this relation. We don't do
-	 * this for INSERT actions, since an INSERT command would only do this
-	 * check if it had a RETURNING list, and MERGE does not support RETURNING.
+	 * but we must also check that new rows created by INSERT/UPDATE actions
+	 * are visible, if SELECT rights are required. For INSERT actions, we only
+	 * do this if RETURNING is specified, to be consistent with a plain INSERT
+	 * command, which can only require SELECT rights when RETURNING is used.
 	 *
 	 * We don't push the UPDATE/DELETE USING quals to the RTE because we don't
 	 * really want to apply them while scanning the relation since we don't
@@ -420,6 +420,8 @@ get_row_security_policies(Query *root, R
 		List	   *merge_delete_restrictive_policies;
 		List	   *merge_insert_permissive_policies;
 		List	   *merge_insert_restrictive_policies;
+		List	   *merge_select_permissive_policies = NIL;
+		List	   *merge_select_restrictive_policies = NIL;
 
 		/*
 		 * Fetch the UPDATE policies and set them up to execute on the
@@ -457,9 +459,6 @@ get_row_security_policies(Query *root, R
 		 */
 		if (perminfo->requiredPerms & ACL_SELECT)
 		{
-			List	   *merge_select_permissive_policies;
-			List	   *merge_select_restrictive_policies;
-
 			get_policies_for_relation(rel, CMD_SELECT, user_id,
 									  &merge_select_permissive_policies,
 									  &merge_select_restrictive_policies);
@@ -508,6 +507,21 @@ get_row_security_policies(Query *root, R
 							   withCheckOptions,
 							   hasSubLinks,
 							   false);
+
+		/*
+		 * Add ALL/SELECT policies as WCO_RLS_INSERT_CHECK WCOs, to ensure
+		 * that the inserted row is visible when executing an INSERT action,
+		 * if RETURNING is specified and SELECT rights are required for this
+		 * relation.
+		 */
+		if (perminfo->requiredPerms & ACL_SELECT && root->returningList)
+			add_with_check_options(rel, rt_index,
+								   WCO_RLS_INSERT_CHECK,
+								   merge_select_permissive_policies,
+								   merge_select_restrictive_policies,
+								   withCheckOptions,
+								   hasSubLinks,
+								   true);
 	}
 
 	table_close(rel, NoLock);
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
new file mode 100644
index e3ccf6c..6f3dc89
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -2142,11 +2142,10 @@ QueryReturnsTuples(Query *parsetree)
 		case CMD_SELECT:
 			/* returns tuples */
 			return true;
-		case CMD_MERGE:
-			return false;
 		case CMD_INSERT:
 		case CMD_UPDATE:
 		case CMD_DELETE:
+		case CMD_MERGE:
 			/* the forms with RETURNING return tuples */
 			if (parsetree->returningList)
 				return true;
diff --git a/src/backend/utils/adt/ruleutils.c b/src/backend/utils/adt/ruleutils.c
new file mode 100644
index ed7f40f..aca2217
--- a/src/backend/utils/adt/ruleutils.c
+++ b/src/backend/utils/adt/ruleutils.c
@@ -7197,8 +7197,13 @@ get_merge_query_def(Query *query, depars
 			appendStringInfoString(buf, "DO NOTHING");
 	}
 
-	/* No RETURNING support in MERGE yet */
-	Assert(query->returningList == NIL);
+	/* Add RETURNING if present */
+	if (query->returningList)
+	{
+		appendContextKeyword(context, " RETURNING",
+							 -PRETTYINDENT_STD, PRETTYINDENT_STD, 1);
+		get_target_list(query->returningList, context, NULL, colNamesVisible);
+	}
 }
 
 
@@ -8298,6 +8303,7 @@ isSimpleNode(Node *node, Node *parentNod
 		case T_Aggref:
 		case T_GroupingFunc:
 		case T_WindowFunc:
+		case T_MergingFunc:
 		case T_FuncExpr:
 		case T_JsonConstructorExpr:
 			/* function-like: name(..) or name[..] */
@@ -8652,6 +8658,27 @@ get_rule_expr(Node *node, deparse_contex
 			get_windowfunc_expr((WindowFunc *) node, context);
 			break;
 
+		case T_MergingFunc:
+			{
+				MergingFunc *mexpr = (MergingFunc *) node;
+
+				appendStringInfoString(buf, "MERGING(");
+				switch (mexpr->mfop)
+				{
+					case MERGING_ACTION:
+						appendStringInfoString(buf, "ACTION");
+						break;
+					case MERGING_CLAUSE_NUMBER:
+						appendStringInfoString(buf, "CLAUSE_NUMBER");
+						break;
+					default:
+						elog(ERROR, "unrecognized MergingFuncOp: %d",
+							 (int) mexpr->mfop);
+				}
+				appendStringInfoChar(buf, ')');
+			}
+			break;
+
 		case T_SubscriptingRef:
 			{
 				SubscriptingRef *sbsref = (SubscriptingRef *) node;
diff --git a/src/bin/psql/common.c b/src/bin/psql/common.c
new file mode 100644
index daabf6f..6b63214
--- a/src/bin/psql/common.c
+++ b/src/bin/psql/common.c
@@ -983,13 +983,17 @@ PrintQueryResult(PGresult *result, bool
 			else
 				success = true;
 
-			/* if it's INSERT/UPDATE/DELETE RETURNING, also print status */
+			/*
+			 * If it's INSERT/UPDATE/DELETE/MERGE RETURNING, also print
+			 * status.
+			 */
 			if (last || pset.show_all_results)
 			{
 				cmdstatus = PQcmdStatus(result);
 				if (strncmp(cmdstatus, "INSERT", 6) == 0 ||
 					strncmp(cmdstatus, "UPDATE", 6) == 0 ||
-					strncmp(cmdstatus, "DELETE", 6) == 0)
+					strncmp(cmdstatus, "DELETE", 6) == 0 ||
+					strncmp(cmdstatus, "MERGE", 5) == 0)
 					PrintQueryStatus(result, printStatusFout);
 			}
 
diff --git a/src/include/executor/execExpr.h b/src/include/executor/execExpr.h
new file mode 100644
index 048573c..7580ab9
--- a/src/include/executor/execExpr.h
+++ b/src/include/executor/execExpr.h
@@ -241,6 +241,7 @@ typedef enum ExprEvalOp
 	EEOP_AGGREF,
 	EEOP_GROUPING_FUNC,
 	EEOP_WINDOW_FUNC,
+	EEOP_MERGING_FUNC,
 	EEOP_SUBPLAN,
 
 	/* aggregation related nodes */
@@ -623,6 +624,12 @@ typedef struct ExprEvalStep
 			WindowFuncExprState *wfstate;
 		}			window_func;
 
+		/* for EEOP_MERGING_FUNC */
+		struct
+		{
+			MergingFuncOp mfop; /* the MERGING() operation to perform */
+		}			merging_func;
+
 		/* for EEOP_SUBPLAN */
 		struct
 		{
@@ -806,6 +813,8 @@ extern void ExecEvalJsonConstructor(Expr
 									ExprContext *econtext);
 extern void ExecEvalJsonIsPredicate(ExprState *state, ExprEvalStep *op);
 extern void ExecEvalGroupingFunc(ExprState *state, ExprEvalStep *op);
+extern void ExecEvalMergingFunc(ExprState *state, ExprEvalStep *op,
+								ExprContext *econtext);
 extern void ExecEvalSubPlan(ExprState *state, ExprEvalStep *op,
 							ExprContext *econtext);
 extern void ExecEvalWholeRowVar(ExprState *state, ExprEvalStep *op,
diff --git a/src/include/executor/spi.h b/src/include/executor/spi.h
new file mode 100644
index d1de139..ed8787d
--- a/src/include/executor/spi.h
+++ b/src/include/executor/spi.h
@@ -97,6 +97,7 @@ typedef struct _SPI_plan *SPIPlanPtr;
 #define SPI_OK_REL_UNREGISTER	16
 #define SPI_OK_TD_REGISTER		17
 #define SPI_OK_MERGE			18
+#define SPI_OK_MERGE_RETURNING	19
 
 #define SPI_OPT_NONATOMIC		(1 << 0)
 
diff --git a/src/include/nodes/execnodes.h b/src/include/nodes/execnodes.h
new file mode 100644
index 5d7f17d..08bc9ee
--- a/src/include/nodes/execnodes.h
+++ b/src/include/nodes/execnodes.h
@@ -1318,6 +1318,9 @@ typedef struct ModifyTableState
 	/* Flags showing which subcommands are present INS/UPD/DEL/DO NOTHING */
 	int			mt_merge_subcommands;
 
+	/* For MERGE, the action currently being executed */
+	MergeActionState *mt_merge_action;
+
 	/* tuple counters for MERGE */
 	double		mt_merge_inserted;
 	double		mt_merge_updated;
diff --git a/src/include/nodes/parsenodes.h b/src/include/nodes/parsenodes.h
new file mode 100644
index cf7e790..0dd872a
--- a/src/include/nodes/parsenodes.h
+++ b/src/include/nodes/parsenodes.h
@@ -1688,6 +1688,7 @@ typedef struct MergeWhenClause
 typedef struct MergeAction
 {
 	NodeTag		type;
+	int			index;			/* 1-based index of the clause */
 	bool		matched;		/* true=MATCHED, false=NOT MATCHED */
 	CmdType		commandType;	/* INSERT/UPDATE/DELETE/DO NOTHING */
 	/* OVERRIDING clause */
@@ -1951,6 +1952,7 @@ typedef struct MergeStmt
 	Node	   *sourceRelation; /* source relation */
 	Node	   *joinCondition;	/* join condition between source and target */
 	List	   *mergeWhenClauses;	/* list of MergeWhenClause(es) */
+	List	   *returningList;	/* list of expressions to return */
 	WithClause *withClause;		/* WITH clause */
 } MergeStmt;
 
diff --git a/src/include/nodes/primnodes.h b/src/include/nodes/primnodes.h
new file mode 100644
index ab6d7fd..7a2fe8e
--- a/src/include/nodes/primnodes.h
+++ b/src/include/nodes/primnodes.h
@@ -564,6 +564,39 @@ typedef struct WindowFunc
 } WindowFunc;
 
 /*
+ * MergingFunc
+ *
+ * A MergingFunc is a MERGING(...) expression that can only appear in the
+ * RETURNING list of a MERGE command.  It returns information about the
+ * currently executing merge action.  Possible operations are:
+ *
+ *	MERGING_ACTION:
+ *		Return the command string of the current merge action ("INSERT",
+ *		"UPDATE" or "DELETE").
+ *
+ *	MERGING_CLAUSE_NUMBER:
+ *		Return the 1-based index of the current merge WHEN clause.
+ */
+typedef enum MergingFuncOp
+{
+	MERGING_ACTION,
+	MERGING_CLAUSE_NUMBER,
+} MergingFuncOp;
+
+typedef struct MergingFunc
+{
+	Expr		xpr;
+	/* operation to perform */
+	MergingFuncOp mfop;
+	/* type Oid of result */
+	Oid			mftype pg_node_attr(query_jumble_ignore);
+	/* OID of collation, or InvalidOid if none */
+	Oid			mfcollid pg_node_attr(query_jumble_ignore);
+	/* token location, or -1 if unknown */
+	int			location;
+} MergingFunc;
+
+/*
  * SubscriptingRef: describes a subscripting operation over a container
  * (array, etc).
  *
diff --git a/src/include/optimizer/paramassign.h b/src/include/optimizer/paramassign.h
new file mode 100644
index 55c27a6..cd0e629
--- a/src/include/optimizer/paramassign.h
+++ b/src/include/optimizer/paramassign.h
@@ -20,6 +20,7 @@ extern Param *replace_outer_placeholderv
 										   PlaceHolderVar *phv);
 extern Param *replace_outer_agg(PlannerInfo *root, Aggref *agg);
 extern Param *replace_outer_grouping(PlannerInfo *root, GroupingFunc *grp);
+extern Param *replace_outer_merging(PlannerInfo *root, MergingFunc *mf);
 extern Param *replace_nestloop_param_var(PlannerInfo *root, Var *var);
 extern Param *replace_nestloop_param_placeholdervar(PlannerInfo *root,
 													PlaceHolderVar *phv);
diff --git a/src/include/parser/analyze.h b/src/include/parser/analyze.h
new file mode 100644
index c96483a..47d988b
--- a/src/include/parser/analyze.h
+++ b/src/include/parser/analyze.h
@@ -44,6 +44,8 @@ extern List *transformInsertRow(ParseSta
 								bool strip_indirection);
 extern List *transformUpdateTargetList(ParseState *pstate,
 									   List *origTlist);
+extern List *transformReturningList(ParseState *pstate, List *returningList,
+									ParseExprKind exprKind);
 extern Query *transformTopLevelStmt(ParseState *pstate, RawStmt *parseTree);
 extern Query *transformStmt(ParseState *pstate, Node *parseTree);
 
diff --git a/src/include/parser/kwlist.h b/src/include/parser/kwlist.h
new file mode 100644
index 5984dcf..c283a3f
--- a/src/include/parser/kwlist.h
+++ b/src/include/parser/kwlist.h
@@ -80,6 +80,7 @@ PG_KEYWORD("characteristics", CHARACTERI
 PG_KEYWORD("check", CHECK, RESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("checkpoint", CHECKPOINT, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("class", CLASS, UNRESERVED_KEYWORD, BARE_LABEL)
+PG_KEYWORD("clause_number", CLAUSE_NUMBER, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("close", CLOSE, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("cluster", CLUSTER, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("coalesce", COALESCE, COL_NAME_KEYWORD, BARE_LABEL)
@@ -266,6 +267,7 @@ PG_KEYWORD("matched", MATCHED, UNRESERVE
 PG_KEYWORD("materialized", MATERIALIZED, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("maxvalue", MAXVALUE, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("merge", MERGE, UNRESERVED_KEYWORD, BARE_LABEL)
+PG_KEYWORD("merging", MERGING, COL_NAME_KEYWORD, BARE_LABEL)
 PG_KEYWORD("method", METHOD, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("minute", MINUTE_P, UNRESERVED_KEYWORD, AS_LABEL)
 PG_KEYWORD("minvalue", MINVALUE, UNRESERVED_KEYWORD, BARE_LABEL)
diff --git a/src/include/parser/parse_node.h b/src/include/parser/parse_node.h
new file mode 100644
index f589112..836819d
--- a/src/include/parser/parse_node.h
+++ b/src/include/parser/parse_node.h
@@ -61,7 +61,8 @@ typedef enum ParseExprKind
 	EXPR_KIND_DISTINCT_ON,		/* DISTINCT ON */
 	EXPR_KIND_LIMIT,			/* LIMIT */
 	EXPR_KIND_OFFSET,			/* OFFSET */
-	EXPR_KIND_RETURNING,		/* RETURNING */
+	EXPR_KIND_RETURNING,		/* RETURNING in INSERT/UPDATE/DELETE */
+	EXPR_KIND_MERGE_RETURNING,	/* RETURNING in MERGE */
 	EXPR_KIND_VALUES,			/* VALUES */
 	EXPR_KIND_VALUES_SINGLE,	/* single-row VALUES (in INSERT only) */
 	EXPR_KIND_CHECK_CONSTRAINT, /* CHECK constraint for a table */
diff --git a/src/pl/plpgsql/src/pl_exec.c b/src/pl/plpgsql/src/pl_exec.c
new file mode 100644
index 4b76f76..fb6289c
--- a/src/pl/plpgsql/src/pl_exec.c
+++ b/src/pl/plpgsql/src/pl_exec.c
@@ -4267,9 +4267,9 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 	/*
 	 * If we have INTO, then we only need one row back ... but if we have INTO
 	 * STRICT or extra check too_many_rows, ask for two rows, so that we can
-	 * verify the statement returns only one.  INSERT/UPDATE/DELETE are always
-	 * treated strictly. Without INTO, just run the statement to completion
-	 * (tcount = 0).
+	 * verify the statement returns only one.  INSERT/UPDATE/DELETE/MERGE are
+	 * always treated strictly. Without INTO, just run the statement to
+	 * completion (tcount = 0).
 	 *
 	 * We could just ask for two rows always when using INTO, but there are
 	 * some cases where demanding the extra row costs significant time, eg by
@@ -4307,10 +4307,11 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 			Assert(stmt->mod_stmt);
 			exec_set_found(estate, (SPI_processed != 0));
 			break;
@@ -4489,10 +4490,11 @@ exec_stmt_dynexecute(PLpgSQL_execstate *
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 		case SPI_OK_UTILITY:
 		case SPI_OK_REWRITTEN:
 			break;
diff --git a/src/pl/tcl/pltcl.c b/src/pl/tcl/pltcl.c
new file mode 100644
index e8f9d7b..0c9f81e
--- a/src/pl/tcl/pltcl.c
+++ b/src/pl/tcl/pltcl.c
@@ -2456,6 +2456,7 @@ pltcl_process_SPI_result(Tcl_Interp *int
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
+		case SPI_OK_MERGE_RETURNING:
 
 			/*
 			 * Process the tuples we got
diff --git a/src/test/regress/expected/merge.out b/src/test/regress/expected/merge.out
new file mode 100644
index 28a6d0b..183d160
--- a/src/test/regress/expected/merge.out
+++ b/src/test/regress/expected/merge.out
@@ -123,20 +123,20 @@ ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
 ERROR:  name "target" specified more than once
 DETAIL:  The name is used both as MERGE target table and data source.
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
-ERROR:  MERGE not supported in WITH query
-LINE 1: WITH foo AS (
-             ^
--- used in COPY
+ERROR:  WITH query "foo" does not have a RETURNING clause
+LINE 4: ) SELECT * FROM foo;
+                        ^
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) TO stdout;
-ERROR:  MERGE not supported in COPY
+ERROR:  COPY query must have a RETURNING clause
 -- unsupported relation types
 -- view
 CREATE VIEW tv AS SELECT * FROM target;
@@ -1256,7 +1256,7 @@ BEGIN;
 MERGE INTO sq_target
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1271,7 +1271,7 @@ INSERT INTO sq_source (sid, balance, del
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1295,7 +1295,7 @@ WITH targq AS (
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1304,20 +1304,191 @@ WHEN MATCHED AND tid < 2 THEN
 ROLLBACK;
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
+CREATE TABLE merge_actions(action text, abbrev text);
+INSERT INTO merge_actions VALUES ('INSERT', 'ins'), ('UPDATE', 'upd'), ('DELETE', 'del');
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
-ERROR:  syntax error at or near "RETURNING"
-LINE 10: RETURNING *;
-         ^
+    DELETE
+RETURNING MERGING(CLAUSE_NUMBER) AS clause,
+          (SELECT abbrev FROM merge_actions
+            WHERE action = MERGING(ACTION)) AS action,
+          t.*,
+          CASE MERGING(ACTION)
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ clause | action | tid | balance |     description     
+--------+--------+-----+---------+---------------------
+      3 | del    |   1 |     100 | Removed (1,100)
+      1 | upd    |   2 |     220 | Added 20 to balance
+      2 | ins    |   4 |      40 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- error when using MERGING() outside MERGE
+SELECT MERGING(ACTION) FROM sq_target;
+ERROR:  MERGING() can only be used in the RETURNING list of a MERGE command
+LINE 1: SELECT MERGING(ACTION) FROM sq_target;
+               ^
+UPDATE sq_target SET balance = balance + 1 RETURNING MERGING(CLAUSE_NUMBER);
+ERROR:  MERGING() can only be used in the RETURNING list of a MERGE command
+LINE 1: ...ATE sq_target SET balance = balance + 1 RETURNING MERGING(CL...
+                                                             ^
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(CLAUSE_NUMBER), MERGING(ACTION), t.*,
+              CASE MERGING(ACTION)
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING action, MERGING(ACTION) AS log_action, l.*
+)
+SELECT * FROM m2;
+ action | log_action | tid |     last_change     
+--------+------------+-----+---------------------
+ DELETE | UPDATE     |   1 | Removed (1,100)
+ UPDATE | INSERT     |   2 | Added 20 to balance
+ INSERT | INSERT     |   4 | Inserted (4,40)
+(3 rows)
+
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ tid |     last_change     
+-----+---------------------
+   1 | Removed (1,100)
+   2 | Added 20 to balance
+   4 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.*
+) TO stdout;
+DELETE	1	100
+UPDATE	2	220
+INSERT	4	40
+ROLLBACK;
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ action | tid | new_balance 
+--------+-----+-------------
+ DELETE |   1 |         100
+ UPDATE |   3 |         320
+ INSERT |   4 |         110
+(3 rows)
+
+ROLLBACK;
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ action | tid | balance 
+--------+-----+---------
+ DELETE |   1 |     100
+ UPDATE |   2 |     220
+ INSERT |   4 |      40
+(3 rows)
+
+ROLLBACK;
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ r_action | r_tid | r_balance 
+----------+-------+-----------
+ DELETE   |     1 |       100
+ UPDATE   |     3 |       320
+ INSERT   |     4 |       110
+(3 rows)
+
 ROLLBACK;
 -- EXPLAIN
 CREATE TABLE ex_mtarget (a int, b int)
@@ -1514,7 +1685,7 @@ SELECT * FROM sq_target WHERE tid = 1;
 (1 row)
 
 ROLLBACK;
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 NOTICE:  drop cascades to view v
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1640,6 +1811,32 @@ SELECT * FROM pa_target ORDER BY tid;
 (14 rows)
 
 ROLLBACK;
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING MERGING(ACTION), MERGING(CLAUSE_NUMBER), t.*;
+ action | clause_number | tid | balance |           val            
+--------+---------------+-----+---------+--------------------------
+ UPDATE |             1 |   2 |     110 | initial updated by merge
+(1 row)
+
+SELECT * FROM pa_target ORDER BY tid;
+ tid | balance |           val            
+-----+---------+--------------------------
+   2 |     110 | initial updated by merge
+   3 |     300 | initial
+   5 |     500 | initial
+   7 |     700 | initial
+   9 |     900 | initial
+  11 |    1100 | initial
+  13 |    1300 | initial
+(7 rows)
+
+ROLLBACK;
 DROP TABLE pa_target CASCADE;
 -- The target table is partitioned in the same way, but this time by attaching
 -- partitions which have columns in different order, dropped columns etc.
@@ -1731,7 +1928,26 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (sid, delta, 'inserted by merge');
+    INSERT VALUES (sid, delta, 'inserted by merge')
+  RETURNING MERGING(ACTION), MERGING(CLAUSE_NUMBER), t.*;
+ action | clause_number | tid | balance |           val            
+--------+---------------+-----+---------+--------------------------
+ UPDATE |             1 |   2 |     110 | initial updated by merge
+ INSERT |             2 |   2 |      20 | inserted by merge
+ UPDATE |             1 |   4 |     330 | initial updated by merge
+ INSERT |             2 |   4 |      40 | inserted by merge
+ UPDATE |             1 |   6 |     550 | initial updated by merge
+ INSERT |             2 |   6 |      60 | inserted by merge
+ UPDATE |             1 |   8 |     770 | initial updated by merge
+ INSERT |             2 |   8 |      80 | inserted by merge
+ UPDATE |             1 |  10 |     990 | initial updated by merge
+ INSERT |             2 |  10 |     100 | inserted by merge
+ UPDATE |             1 |  12 |    1210 | initial updated by merge
+ INSERT |             2 |  12 |     120 | inserted by merge
+ UPDATE |             1 |  14 |    1430 | initial updated by merge
+ INSERT |             2 |  14 |     140 | inserted by merge
+(14 rows)
+
 SELECT * FROM pa_target ORDER BY tid;
  tid | balance |           val            
 -----+---------+--------------------------
@@ -1798,7 +2014,21 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING MERGING(ACTION), MERGING(CLAUSE_NUMBER), t.*;
+ action | clause_number |          logts           | tid | balance |           val            
+--------+---------------+--------------------------+-----+---------+--------------------------
+ UPDATE |             1 | Tue Jan 31 00:00:00 2017 |   1 |     110 | initial updated by merge
+ UPDATE |             1 | Tue Feb 28 00:00:00 2017 |   2 |     220 | initial updated by merge
+ INSERT |             2 | Sun Jan 15 00:00:00 2017 |   3 |      30 | inserted by merge
+ UPDATE |             1 | Tue Jan 31 00:00:00 2017 |   4 |     440 | initial updated by merge
+ UPDATE |             1 | Tue Feb 28 00:00:00 2017 |   5 |     550 | initial updated by merge
+ INSERT |             2 | Sun Jan 15 00:00:00 2017 |   6 |      60 | inserted by merge
+ UPDATE |             1 | Tue Jan 31 00:00:00 2017 |   7 |     770 | initial updated by merge
+ UPDATE |             1 | Tue Feb 28 00:00:00 2017 |   8 |     880 | initial updated by merge
+ INSERT |             2 | Sun Jan 15 00:00:00 2017 |   9 |      90 | inserted by merge
+(9 rows)
+
 SELECT * FROM pa_target ORDER BY tid;
           logts           | tid | balance |           val            
 --------------------------+-----+---------+--------------------------
diff --git a/src/test/regress/expected/rowsecurity.out b/src/test/regress/expected/rowsecurity.out
new file mode 100644
index 6988128..3fd4fab
--- a/src/test/regress/expected/rowsecurity.out
+++ b/src/test/regress/expected/rowsecurity.out
@@ -2302,6 +2302,35 @@ WHEN MATCHED THEN
 	UPDATE SET dnotes = dnotes || ' notes added by merge8 '
 WHEN NOT MATCHED THEN
 	INSERT VALUES (13, 44, 1, 'regress_rls_bob', 'new manga');
+SELECT * FROM document WHERE did = 13;
+ did | cid | dlevel | dauthor | dtitle | dnotes 
+-----+-----+--------+---------+--------+--------
+(0 rows)
+
+-- but not OK if RETURNING is used
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge9 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 44, 1, 'regress_rls_bob', 'new manga')
+RETURNING *;
+ERROR:  new row violates row-level security policy for table "document"
+-- but OK if new row is visible
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge10 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 11, 1, 'regress_rls_bob', 'new novel')
+RETURNING *;
+ sdid | did | cid | dlevel |     dauthor     |  dtitle   | dnotes 
+------+-----+-----+--------+-----------------+-----------+--------
+   14 |  14 |  11 |      1 | regress_rls_bob | new novel | 
+(1 row)
+
 RESET SESSION AUTHORIZATION;
 -- drop the restrictive SELECT policy so that we can look at the
 -- final state of the table
@@ -2325,7 +2354,8 @@ SELECT * FROM document;
   12 |  11 |      1 | regress_rls_bob   | another novel                    | 
    1 |  11 |      1 | regress_rls_bob   | my first novel                   |  notes added by merge2  notes added by merge3  notes added by merge4  notes added by merge7 
   13 |  44 |      1 | regress_rls_bob   | new manga                        | 
-(15 rows)
+  14 |  11 |      1 | regress_rls_bob   | new novel                        | 
+(16 rows)
 
 --
 -- ROLE/GROUP
diff --git a/src/test/regress/expected/rules.out b/src/test/regress/expected/rules.out
new file mode 100644
index 1442c43..f187dec
--- a/src/test/regress/expected/rules.out
+++ b/src/test/regress/expected/rules.out
@@ -3596,7 +3596,8 @@ MERGE INTO rule_merge2 t USING (SELECT 1
 -- test deparsing
 CREATE TABLE sf_target(id int, data text, filling int[]);
 CREATE FUNCTION merge_sf_test()
- RETURNS void
+ RETURNS TABLE(action text, clause_number int,
+               a int, b text, id int, data text, filling int[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -3633,11 +3634,13 @@ WHEN NOT MATCHED
    VALUES (s.a, s.b, DEFAULT)
 WHEN NOT MATCHED
    THEN INSERT (filling[1], id)
-   VALUES (s.a, s.a);
+   VALUES (s.a, s.a)
+RETURNING
+   merging(action), merging(clause_number) AS clause_number, *;
 END;
 \sf merge_sf_test
 CREATE OR REPLACE FUNCTION public.merge_sf_test()
- RETURNS void
+ RETURNS TABLE(action text, clause_number integer, a integer, b text, id integer, data text, filling integer[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -3674,7 +3677,14 @@ BEGIN ATOMIC
       VALUES (s.a, s.b, DEFAULT)
     WHEN NOT MATCHED
      THEN INSERT (filling[1], id)
-      VALUES (s.a, s.a);
+      VALUES (s.a, s.a)
+   RETURNING MERGING(ACTION) AS action,
+     MERGING(CLAUSE_NUMBER) AS clause_number,
+     s.a,
+     s.b,
+     t.id,
+     t.data,
+     t.filling;
 END
 DROP FUNCTION merge_sf_test;
 DROP TABLE sf_target;
diff --git a/src/test/regress/sql/merge.sql b/src/test/regress/sql/merge.sql
new file mode 100644
index 4cf6db9..0eb2f97
--- a/src/test/regress/sql/merge.sql
+++ b/src/test/regress/sql/merge.sql
@@ -86,12 +86,12 @@ MERGE INTO target
 USING target
 ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
--- used in COPY
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
@@ -816,7 +816,7 @@ BEGIN;
 MERGE INTO sq_target
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -829,7 +829,7 @@ INSERT INTO sq_source (sid, balance, del
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -847,7 +847,7 @@ WITH targq AS (
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -857,17 +857,146 @@ ROLLBACK;
 
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
+CREATE TABLE merge_actions(action text, abbrev text);
+INSERT INTO merge_actions VALUES ('INSERT', 'ins'), ('UPDATE', 'upd'), ('DELETE', 'del');
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
+    DELETE
+RETURNING MERGING(CLAUSE_NUMBER) AS clause,
+          (SELECT abbrev FROM merge_actions
+            WHERE action = MERGING(ACTION)) AS action,
+          t.*,
+          CASE MERGING(ACTION)
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ROLLBACK;
+
+-- error when using MERGING() outside MERGE
+SELECT MERGING(ACTION) FROM sq_target;
+UPDATE sq_target SET balance = balance + 1 RETURNING MERGING(CLAUSE_NUMBER);
+
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(CLAUSE_NUMBER), MERGING(ACTION), t.*,
+              CASE MERGING(ACTION)
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING action, MERGING(ACTION) AS log_action, l.*
+)
+SELECT * FROM m2;
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ROLLBACK;
+
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.*
+) TO stdout;
+ROLLBACK;
+
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ROLLBACK;
+
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ROLLBACK;
+
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
 ROLLBACK;
 
 -- EXPLAIN
@@ -966,7 +1095,7 @@ WHEN MATCHED THEN
 SELECT * FROM sq_target WHERE tid = 1;
 ROLLBACK;
 
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1033,6 +1162,17 @@ SELECT merge_func();
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING MERGING(ACTION), MERGING(CLAUSE_NUMBER), t.*;
+SELECT * FROM pa_target ORDER BY tid;
+ROLLBACK;
+
 DROP TABLE pa_target CASCADE;
 
 -- The target table is partitioned in the same way, but this time by attaching
@@ -1091,7 +1231,8 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (sid, delta, 'inserted by merge');
+    INSERT VALUES (sid, delta, 'inserted by merge')
+  RETURNING MERGING(ACTION), MERGING(CLAUSE_NUMBER), t.*;
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
@@ -1145,7 +1286,8 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING MERGING(ACTION), MERGING(CLAUSE_NUMBER), t.*;
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
diff --git a/src/test/regress/sql/rowsecurity.sql b/src/test/regress/sql/rowsecurity.sql
new file mode 100644
index dec7340..1d5ed0a
--- a/src/test/regress/sql/rowsecurity.sql
+++ b/src/test/regress/sql/rowsecurity.sql
@@ -992,6 +992,27 @@ WHEN MATCHED THEN
 	UPDATE SET dnotes = dnotes || ' notes added by merge8 '
 WHEN NOT MATCHED THEN
 	INSERT VALUES (13, 44, 1, 'regress_rls_bob', 'new manga');
+SELECT * FROM document WHERE did = 13;
+
+-- but not OK if RETURNING is used
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge9 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 44, 1, 'regress_rls_bob', 'new manga')
+RETURNING *;
+
+-- but OK if new row is visible
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge10 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 11, 1, 'regress_rls_bob', 'new novel')
+RETURNING *;
 
 RESET SESSION AUTHORIZATION;
 -- drop the restrictive SELECT policy so that we can look at the
diff --git a/src/test/regress/sql/rules.sql b/src/test/regress/sql/rules.sql
new file mode 100644
index 8b7e255..d820b9b
--- a/src/test/regress/sql/rules.sql
+++ b/src/test/regress/sql/rules.sql
@@ -1281,7 +1281,8 @@ MERGE INTO rule_merge2 t USING (SELECT 1
 CREATE TABLE sf_target(id int, data text, filling int[]);
 
 CREATE FUNCTION merge_sf_test()
- RETURNS void
+ RETURNS TABLE(action text, clause_number int,
+               a int, b text, id int, data text, filling int[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -1318,7 +1319,9 @@ WHEN NOT MATCHED
    VALUES (s.a, s.b, DEFAULT)
 WHEN NOT MATCHED
    THEN INSERT (filling[1], id)
-   VALUES (s.a, s.a);
+   VALUES (s.a, s.a)
+RETURNING
+   merging(action), merging(clause_number) AS clause_number, *;
 END;
 
 \sf merge_sf_test
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
new file mode 100644
index 87c1aee..31a552e
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -1559,6 +1559,8 @@ MergePath
 MergeScanSelCache
 MergeStmt
 MergeWhenClause
+MergingFunc
+MergingFuncOp
 MetaCommand
 MinMaxAggInfo
 MinMaxAggPath

diff --git a/doc/src/sgml/dml.sgml b/doc/src/sgml/dml.sgml
new file mode 100644
index cbbc5e2..3d95bdb
--- a/doc/src/sgml/dml.sgml
+++ b/doc/src/sgml/dml.sgml
@@ -283,10 +283,15 @@ DELETE FROM products;
    <secondary>RETURNING</secondary>
   </indexterm>
 
+  <indexterm zone="dml-returning">
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
   <para>
    Sometimes it is useful to obtain data from modified rows while they are
    being manipulated.  The <command>INSERT</command>, <command>UPDATE</command>,
-   and <command>DELETE</command> commands all have an
+   <command>DELETE</command>, and <command>MERGE</command> commands all have an
    optional <literal>RETURNING</literal> clause that supports this.  Use
    of <literal>RETURNING</literal> avoids performing an extra database query to
    collect the data, and is especially valuable when it would otherwise be
@@ -339,6 +344,21 @@ DELETE FROM products
 </programlisting>
   </para>
 
+  <para>
+   In a <command>MERGE</command>, the data available to <literal>RETURNING</literal> is
+   the content of the source row plus the content of the inserted, updated, or
+   deleted target row.  Since it is quite common for the source and target to
+   have many of the same columns, specifying <literal>RETURNING *</literal>
+   can lead to a lot of duplicated columns, so it is often more useful to
+   qualify it so as to return just the source or target row.  For example:
+<programlisting>
+MERGE INTO products p USING new_products n ON p.product_no = n.product_no
+  WHEN NOT MATCHED THEN INSERT VALUES (n.product_no, n.name, n.price)
+  WHEN MATCHED THEN UPDATE SET name = n.name, price = n.price
+  RETURNING p.*;
+</programlisting>
+  </para>
+
   <para>
    If there are triggers (<xref linkend="triggers"/>) on the target table,
    the data available to <literal>RETURNING</literal> is the row as modified by
diff --git a/doc/src/sgml/func.sgml b/doc/src/sgml/func.sgml
new file mode 100644
index d963f0a..b25de53
--- a/doc/src/sgml/func.sgml
+++ b/doc/src/sgml/func.sgml
@@ -21990,6 +21990,84 @@ SELECT count(*) FROM sometable;
 
  </sect1>
 
+ <sect1 id="functions-merge-support">
+  <title>Merge Support Functions</title>
+
+  <indexterm>
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
+  <indexterm>
+   <primary>MERGING</primary>
+  </indexterm>
+
+  <para>
+   <productname>PostgreSQL</productname> includes one merge support function
+   that may be used in the <literal>RETURNING</literal> list of a
+   <xref linkend="sql-merge"/> command to return additional information about
+   the action taken for each row:
+<synopsis>
+<function id="function-merging">MERGING</function> ( <replaceable>property</replaceable> )
+</synopsis>
+   The following are valid property values specifying what to return:
+
+   <variablelist>
+    <varlistentry>
+     <term><literal>ACTION</literal></term>
+     <listitem>
+      <para>
+       The merge action command executed for the current row
+       (<literal>'INSERT'</literal>, <literal>'UPDATE'</literal>, or
+       <literal>'DELETE'</literal>).
+      </para>
+     </listitem>
+    </varlistentry>
+
+    <varlistentry>
+     <term><literal>CLAUSE_NUMBER</literal></term>
+     <listitem>
+      <para>
+       The 1-based index of the <literal>WHEN</literal> clause executed for
+       the current row.
+      </para>
+     </listitem>
+    </varlistentry>
+   </variablelist>
+  </para>
+
+  <para>
+   The following example illustrates how this may be used to determine
+   which actions were performed for each row affected by the
+   <command>MERGE</command> command:
+<screen><![CDATA[
+MERGE INTO products p
+  USING stock s ON p.product_id = s.product_id
+  WHEN MATCHED AND s.quantity > 0 THEN
+    UPDATE SET in_stock = true, quantity = s.quantity
+  WHEN MATCHED THEN
+    UPDATE SET in_stock = false, quantity = 0
+  WHEN NOT MATCHED THEN
+    INSERT (product_id, in_stock, quantity)
+      VALUES (s.product_id, true, s.quantity)
+  RETURNING MERGING(ACTION), MERGING(CLAUSE_NUMBER), p.*;
+
+ action | clause_number | product_id | in_stock | quantity 
+--------+---------------+------------+----------+----------
+ UPDATE |             1 |       1001 | t        |       50
+ UPDATE |             2 |       1002 | f        |        0
+ INSERT |             3 |       1003 | t        |       10
+]]></screen>
+  </para>
+
+  <para>
+   Note that this function can only be used in the <literal>RETURNING</literal>
+   list of a <command>MERGE</command> command.  It is an error to use it in any
+   other part of a query.
+  </para>
+
+ </sect1>
+
  <sect1 id="functions-subquery">
   <title>Subquery Expressions</title>
 
diff --git a/doc/src/sgml/glossary.sgml b/doc/src/sgml/glossary.sgml
new file mode 100644
index fe8def4..51a15ca
--- a/doc/src/sgml/glossary.sgml
+++ b/doc/src/sgml/glossary.sgml
@@ -1387,9 +1387,9 @@
      to a <glossterm linkend="glossary-client">client</glossterm> upon the
      completion of an <acronym>SQL</acronym> command, usually a
      <command>SELECT</command> but it can be an
-     <command>INSERT</command>, <command>UPDATE</command>, or
-     <command>DELETE</command> command if the <literal>RETURNING</literal>
-     clause is specified.
+     <command>INSERT</command>, <command>UPDATE</command>,
+     <command>DELETE</command>, or <command>MERGE</command> command if the
+     <literal>RETURNING</literal> clause is specified.
     </para>
     <para>
      The fact that a result set is a relation means that a query can be used
diff --git a/doc/src/sgml/plpgsql.sgml b/doc/src/sgml/plpgsql.sgml
new file mode 100644
index 5977534..4ad9894
--- a/doc/src/sgml/plpgsql.sgml
+++ b/doc/src/sgml/plpgsql.sgml
@@ -1020,8 +1020,8 @@ INSERT INTO mytable VALUES (1,'one'), (2
     </para>
 
     <para>
-     If the command does return rows (for example <command>SELECT</command>,
-     or <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>
+     If the command does return rows (for example <command>SELECT</command>, or
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
      with <literal>RETURNING</literal>), there are two ways to proceed.
      When the command will return at most one row, or you only care about
      the first row of output, write the command as usual but add
@@ -1149,6 +1149,7 @@ SELECT <replaceable>select_expressions</
 INSERT ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 UPDATE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 DELETE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
+MERGE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 </synopsis>
 
      where <replaceable>target</replaceable> can be a record variable, a row
@@ -1159,8 +1160,8 @@ DELETE ... RETURNING <replaceable>expres
      <literal>INTO</literal> clause) just as described above,
      and the plan is cached in the same way.
      This works for <command>SELECT</command>,
-     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
-     <literal>RETURNING</literal>, and certain utility commands
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
+     with <literal>RETURNING</literal>, and certain utility commands
      that return row sets, such as <command>EXPLAIN</command>.
      Except for the <literal>INTO</literal> clause, the SQL command is the same
      as it would be written outside <application>PL/pgSQL</application>.
@@ -1236,7 +1237,7 @@ END;
     </para>
 
     <para>
-     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
+     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command> with
      <literal>RETURNING</literal>, <application>PL/pgSQL</application> reports
      an error for more than one returned row, even when
      <literal>STRICT</literal> is not specified.  This is because there
diff --git a/doc/src/sgml/queries.sgml b/doc/src/sgml/queries.sgml
new file mode 100644
index 8a4674e..ae1f738
--- a/doc/src/sgml/queries.sgml
+++ b/doc/src/sgml/queries.sgml
@@ -2062,9 +2062,10 @@ SELECT <replaceable>select_list</replace
    Table Expressions or <acronym>CTE</acronym>s, can be thought of as defining
    temporary tables that exist just for one query.  Each auxiliary statement
    in a <literal>WITH</literal> clause can be a <command>SELECT</command>,
-   <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>; and the
+   <command>INSERT</command>, <command>UPDATE</command>, <command>DELETE</command>,
+   or <command>MERGE</command>; and the
    <literal>WITH</literal> clause itself is attached to a primary statement that can
-   be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
+   also be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
    <command>DELETE</command>, or <command>MERGE</command>.
   </para>
 
@@ -2598,8 +2599,8 @@ SELECT * FROM w AS w1 JOIN w AS w2 ON w1
    <title>Data-Modifying Statements in <literal>WITH</literal></title>
 
    <para>
-    You can use most data-modifying statements (<command>INSERT</command>,
-    <command>UPDATE</command>, or <command>DELETE</command>, but not
+    You can use data-modifying statements (<command>INSERT</command>,
+    <command>UPDATE</command>, <command>DELETE</command>, or
     <command>MERGE</command>) in <literal>WITH</literal>.  This
     allows you to perform several different operations in the same query.
     An example is:
diff --git a/doc/src/sgml/ref/copy.sgml b/doc/src/sgml/ref/copy.sgml
new file mode 100644
index d12ba96..9b9536b
--- a/doc/src/sgml/ref/copy.sgml
+++ b/doc/src/sgml/ref/copy.sgml
@@ -122,13 +122,16 @@ COPY { <replaceable class="parameter">ta
       A <link linkend="sql-select"><command>SELECT</command></link>,
       <link linkend="sql-values"><command>VALUES</command></link>,
       <link linkend="sql-insert"><command>INSERT</command></link>,
-      <link linkend="sql-update"><command>UPDATE</command></link>, or
-      <link linkend="sql-delete"><command>DELETE</command></link> command whose results are to be
-      copied.  Note that parentheses are required around the query.
+      <link linkend="sql-update"><command>UPDATE</command></link>,
+      <link linkend="sql-delete"><command>DELETE</command></link>, or
+      <link linkend="sql-merge"><command>MERGE</command></link> command
+      whose results are to be copied.  Note that parentheses are required
+      around the query.
      </para>
      <para>
-      For <command>INSERT</command>, <command>UPDATE</command> and
-      <command>DELETE</command> queries a RETURNING clause must be provided,
+      For <command>INSERT</command>, <command>UPDATE</command>,
+      <command>DELETE</command>, and <command>MERGE</command> queries a
+      RETURNING clause must be provided,
       and the target relation must not have a conditional rule, nor
       an <literal>ALSO</literal> rule, nor an <literal>INSTEAD</literal> rule
       that expands to multiple statements.
diff --git a/doc/src/sgml/ref/merge.sgml b/doc/src/sgml/ref/merge.sgml
new file mode 100644
index 0f61d47..1ef29ab
--- a/doc/src/sgml/ref/merge.sgml
+++ b/doc/src/sgml/ref/merge.sgml
@@ -25,6 +25,7 @@ PostgreSQL documentation
 MERGE INTO [ ONLY ] <replaceable class="parameter">target_table_name</replaceable> [ * ] [ [ AS ] <replaceable class="parameter">target_alias</replaceable> ]
 USING <replaceable class="parameter">data_source</replaceable> ON <replaceable class="parameter">join_condition</replaceable>
 <replaceable class="parameter">when_clause</replaceable> [...]
+[ RETURNING * | <replaceable class="parameter">output_expression</replaceable> [ [ AS ] <replaceable class="parameter">output_name</replaceable> ] [, ...] ]
 
 <phrase>where <replaceable class="parameter">data_source</replaceable> is:</phrase>
 
@@ -95,6 +96,18 @@ DELETE
   </para>
 
   <para>
+   The optional <literal>RETURNING</literal> clause causes <command>MERGE</command>
+   to compute and return value(s) based on each row inserted, updated, or
+   deleted.  Any expression using the source or target table's columns, or
+   the <xref linkend="function-merging"/> function can be computed.  When an
+   <command>INSERT</command> or <command>UPDATE</command> action is performed,
+   the new values of the target table's columns are used.  When a
+   <command>DELETE</command> is performed, the old values of the target table's
+   columns are used.  The syntax of the <literal>RETURNING</literal> list is
+   identical to that of the output list of <command>SELECT</command>.
+  </para>
+
+  <para>
    There is no separate <literal>MERGE</literal> privilege.
    If you specify an update action, you must have the
    <literal>UPDATE</literal> privilege on the column(s)
@@ -125,6 +138,18 @@ DELETE
 
   <variablelist>
    <varlistentry>
+    <term><replaceable class="parameter">with_query</replaceable></term>
+    <listitem>
+     <para>
+      The <literal>WITH</literal> clause allows you to specify one or more
+      subqueries that can be referenced by name in the <command>MERGE</command>
+      query. See <xref linkend="queries-with"/> and <xref linkend="sql-select"/>
+      for details.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
     <term><replaceable class="parameter">target_table_name</replaceable></term>
     <listitem>
      <para>
@@ -399,13 +424,30 @@ DELETE
    </varlistentry>
 
    <varlistentry>
-    <term><replaceable class="parameter">with_query</replaceable></term>
+    <term><replaceable class="parameter">output_expression</replaceable></term>
     <listitem>
      <para>
-      The <literal>WITH</literal> clause allows you to specify one or more
-      subqueries that can be referenced by name in the <command>MERGE</command>
-      query. See <xref linkend="queries-with"/> and <xref linkend="sql-select"/>
-      for details.
+      An expression to be computed and returned by the <command>MERGE</command>
+      command after each row is merged.  The expression can use any columns of
+      the source or target tables, or the <xref linkend="function-merging"/>
+      function to return additional information about the action executed.
+     </para>
+     <para>
+      Writing <literal>*</literal> will return all columns from the source
+      table, followed by all columns from the target table.  Often this will
+      lead to a lot of duplication, since it is common for the source and
+      target tables to have a lot of the same columns.  This can be avoided by
+      qualifying the <literal>*</literal> with the name or alias of the source
+      or target table.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
+    <term><replaceable class="parameter">output_name</replaceable></term>
+    <listitem>
+     <para>
+      A name to use for a returned column.
      </para>
     </listitem>
    </varlistentry>
@@ -428,6 +470,13 @@ MERGE <replaceable class="parameter">tot
    were changed in any way.
   </para>
 
+  <para>
+   If the <command>MERGE</command> command contains a <literal>RETURNING</literal>
+   clause, the result will be similar to that of a <command>SELECT</command>
+   statement containing the columns and values defined in the
+   <literal>RETURNING</literal> list, computed over the row(s) inserted, updated,
+   or deleted by the command.
+  </para>
  </refsect1>
 
  <refsect1>
@@ -544,13 +593,6 @@ MERGE <replaceable class="parameter">tot
   </para>
 
   <para>
-   There is no <literal>RETURNING</literal> clause with
-   <command>MERGE</command>.  Actions of <command>INSERT</command>,
-   <command>UPDATE</command> and <command>DELETE</command> cannot contain
-   <literal>RETURNING</literal> or <literal>WITH</literal> clauses.
-  </para>
-
-  <para>
    When <command>MERGE</command> is run concurrently with other commands
    that modify the target table, the usual transaction isolation rules
    apply; see <xref linkend="transaction-iso"/> for an explanation
@@ -602,7 +644,8 @@ WHEN NOT MATCHED THEN
   <para>
    Attempt to insert a new stock item along with the quantity of stock. If
    the item already exists, instead update the stock count of the existing
-   item. Don't allow entries that have zero stock.
+   item. Don't allow entries that have zero stock. Return details of all
+   changes made.
 <programlisting>
 MERGE INTO wines w
 USING wine_stock_changes s
@@ -612,7 +655,8 @@ WHEN NOT MATCHED AND s.stock_delta > 0 T
 WHEN MATCHED AND w.stock + s.stock_delta > 0 THEN
   UPDATE SET stock = w.stock + s.stock_delta
 WHEN MATCHED THEN
-  DELETE;
+  DELETE
+RETURNING MERGING(ACTION), w.*;
 </programlisting>
 
    The <literal>wine_stock_changes</literal> table might be, for example, a
@@ -627,7 +671,8 @@ WHEN MATCHED THEN
     This command conforms to the <acronym>SQL</acronym> standard.
   </para>
    <para>
-    The WITH clause and <literal>DO NOTHING</literal> action are extensions to
+    The WITH clause, <literal>DO NOTHING</literal> action, and
+    <literal>RETURNING</literal> clause are extensions to
     the <acronym>SQL</acronym> standard.
   </para>
  </refsect1>
diff --git a/doc/src/sgml/ref/select.sgml b/doc/src/sgml/ref/select.sgml
new file mode 100644
index 42d7891..8513ca6
--- a/doc/src/sgml/ref/select.sgml
+++ b/doc/src/sgml/ref/select.sgml
@@ -74,7 +74,7 @@ SELECT [ ALL | DISTINCT [ ON ( <replacea
 
 <phrase>and <replaceable class="parameter">with_query</replaceable> is:</phrase>
 
-    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> )
+    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> | <replaceable class="parameter">merge</replaceable> )
         [ SEARCH { BREADTH | DEPTH } FIRST BY <replaceable>column_name</replaceable> [, ...] SET <replaceable>search_seq_col_name</replaceable> ]
         [ CYCLE <replaceable>column_name</replaceable> [, ...] SET <replaceable>cycle_mark_col_name</replaceable> [ TO <replaceable>cycle_mark_value</replaceable> DEFAULT <replaceable>cycle_mark_default</replaceable> ] USING <replaceable>cycle_path_col_name</replaceable> ]
 
@@ -230,10 +230,10 @@ TABLE [ ONLY ] <replaceable class="param
     The subqueries effectively act as temporary tables or views
     for the duration of the primary query.
     Each subquery can be a <command>SELECT</command>, <command>TABLE</command>, <command>VALUES</command>,
-    <command>INSERT</command>, <command>UPDATE</command> or
-    <command>DELETE</command> statement.
+    <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command> statement.
     When writing a data-modifying statement (<command>INSERT</command>,
-    <command>UPDATE</command> or <command>DELETE</command>) in
+    <command>UPDATE</command>, <command>DELETE</command>, or <command>MERGE</command>) in
     <literal>WITH</literal>, it is usual to include a <literal>RETURNING</literal> clause.
     It is the output of <literal>RETURNING</literal>, <emphasis>not</emphasis> the underlying
     table that the statement modifies, that forms the temporary table that is
@@ -2182,7 +2182,8 @@ SELECT 2+2;
 
    <para>
     <productname>PostgreSQL</productname> allows <command>INSERT</command>,
-    <command>UPDATE</command>, and <command>DELETE</command> to be used as <literal>WITH</literal>
+    <command>UPDATE</command>, <command>DELETE</command>, and
+    <command>MERGE</command> to be used as <literal>WITH</literal>
     queries.  This is not found in the SQL standard.
    </para>
   </refsect2>
diff --git a/doc/src/sgml/rowtypes.sgml b/doc/src/sgml/rowtypes.sgml
new file mode 100644
index 4d86f97..bbeac84
--- a/doc/src/sgml/rowtypes.sgml
+++ b/doc/src/sgml/rowtypes.sgml
@@ -348,7 +348,7 @@ SELECT m.* FROM some_table, LATERAL myfu
    column expansion of this kind when it appears at the top level of
    a <link linkend="queries-select-lists"><command>SELECT</command> output
    list</link>, a <link linkend="dml-returning"><literal>RETURNING</literal>
-   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>,
+   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>,
    a <link linkend="queries-values"><literal>VALUES</literal> clause</link>, or
    a <link linkend="sql-syntax-row-constructors">row constructor</link>.
    In all other contexts (including when nested inside one of those
diff --git a/doc/src/sgml/spi.sgml b/doc/src/sgml/spi.sgml
new file mode 100644
index 47f4b5b..bb37786
--- a/doc/src/sgml/spi.sgml
+++ b/doc/src/sgml/spi.sgml
@@ -301,8 +301,9 @@ SPI_execute("INSERT INTO foo SELECT * FR
    is returned in the global variable <varname>SPI_processed</varname>.
    If the return value of the function is <symbol>SPI_OK_SELECT</symbol>,
    <symbol>SPI_OK_INSERT_RETURNING</symbol>,
-   <symbol>SPI_OK_DELETE_RETURNING</symbol>, or
-   <symbol>SPI_OK_UPDATE_RETURNING</symbol>,
+   <symbol>SPI_OK_DELETE_RETURNING</symbol>,
+   <symbol>SPI_OK_UPDATE_RETURNING</symbol>, or
+   <symbol>SPI_OK_MERGE_RETURNING</symbol>,
    then you can use the
    global pointer <literal>SPITupleTable *SPI_tuptable</literal> to
    access the result rows.  Some utility commands (such as
@@ -472,6 +473,15 @@ typedef struct SPITupleTable
       </para>
      </listitem>
     </varlistentry>
+
+    <varlistentry>
+     <term><symbol>SPI_OK_MERGE_RETURNING</symbol></term>
+     <listitem>
+      <para>
+       if a <command>MERGE RETURNING</command> was executed
+      </para>
+     </listitem>
+    </varlistentry>
 
     <varlistentry>
      <term><symbol>SPI_OK_UTILITY</symbol></term>
diff --git a/doc/src/sgml/xfunc.sgml b/doc/src/sgml/xfunc.sgml
new file mode 100644
index 89116ae..cb0aa55
--- a/doc/src/sgml/xfunc.sgml
+++ b/doc/src/sgml/xfunc.sgml
@@ -177,7 +177,8 @@
     statements separated by semicolons.  A semicolon after the last
     statement is optional.  Unless the function is declared to return
     <type>void</type>, the last statement must be a <command>SELECT</command>,
-    or an <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>
+    or an <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command>
     that has a <literal>RETURNING</literal> clause.
    </para>
 
@@ -1236,8 +1237,9 @@ SELECT x, CASE WHEN x &gt; 0 THEN genera
 
     <note>
      <para>
-      If a function's last command is <command>INSERT</command>, <command>UPDATE</command>,
-      or <command>DELETE</command> with <literal>RETURNING</literal>, that command will
+      If a function's last command is <command>INSERT</command>,
+      <command>UPDATE</command>, <command>DELETE</command>, or
+      <command>MERGE</command> with <literal>RETURNING</literal>, that command will
       always be executed to completion, even if the function is not declared
       with <literal>SETOF</literal> or the calling query does not fetch all the
       result rows.  Any extra rows produced by the <literal>RETURNING</literal>
diff --git a/src/backend/commands/copy.c b/src/backend/commands/copy.c
new file mode 100644
index c5d7d78..7977873
--- a/src/backend/commands/copy.c
+++ b/src/backend/commands/copy.c
@@ -283,12 +283,6 @@ DoCopy(ParseState *pstate, const CopyStm
 	{
 		Assert(stmt->query);
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(stmt->query, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in COPY"));
-
 		query = makeNode(RawStmt);
 		query->stmt = stmt->query;
 		query->stmt_location = stmt_location;
diff --git a/src/backend/commands/copyto.c b/src/backend/commands/copyto.c
new file mode 100644
index c66a047..1145aaf
--- a/src/backend/commands/copyto.c
+++ b/src/backend/commands/copyto.c
@@ -510,7 +510,8 @@ BeginCopyTo(ParseState *pstate,
 		{
 			Assert(query->commandType == CMD_INSERT ||
 				   query->commandType == CMD_UPDATE ||
-				   query->commandType == CMD_DELETE);
+				   query->commandType == CMD_DELETE ||
+				   query->commandType == CMD_MERGE);
 
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
diff --git a/src/backend/executor/execExpr.c b/src/backend/executor/execExpr.c
new file mode 100644
index 2c62b0c..6414c0d
--- a/src/backend/executor/execExpr.c
+++ b/src/backend/executor/execExpr.c
@@ -1111,6 +1111,21 @@ ExecInitExprRec(Expr *node, ExprState *s
 				break;
 			}
 
+		case T_MergingFunc:
+			{
+				MergingFunc *mf_node = (MergingFunc *) node;
+
+				if (!state->parent ||
+					!IsA(state->parent, ModifyTableState) ||
+					((ModifyTableState *) state->parent)->operation != CMD_MERGE)
+					elog(ERROR, "MergingFunc found in non-merge plan node");
+
+				scratch.opcode = EEOP_MERGING_FUNC;
+				scratch.d.merging_func.mfop = mf_node->mfop;
+				ExprEvalPushStep(state, &scratch);
+				break;
+			}
+
 		case T_SubscriptingRef:
 			{
 				SubscriptingRef *sbsref = (SubscriptingRef *) node;
diff --git a/src/backend/executor/execExprInterp.c b/src/backend/executor/execExprInterp.c
new file mode 100644
index 24c2b60..b176ab6
--- a/src/backend/executor/execExprInterp.c
+++ b/src/backend/executor/execExprInterp.c
@@ -483,6 +483,7 @@ ExecInterpExpr(ExprState *state, ExprCon
 		&&CASE_EEOP_AGGREF,
 		&&CASE_EEOP_GROUPING_FUNC,
 		&&CASE_EEOP_WINDOW_FUNC,
+		&&CASE_EEOP_MERGING_FUNC,
 		&&CASE_EEOP_SUBPLAN,
 		&&CASE_EEOP_AGG_STRICT_DESERIALIZE,
 		&&CASE_EEOP_AGG_DESERIALIZE,
@@ -1582,6 +1583,14 @@ ExecInterpExpr(ExprState *state, ExprCon
 			EEO_NEXT();
 		}
 
+		EEO_CASE(EEOP_MERGING_FUNC)
+		{
+			/* too complex/uncommon for an inline implementation */
+			ExecEvalMergingFunc(state, op, econtext);
+
+			EEO_NEXT();
+		}
+
 		EEO_CASE(EEOP_SUBPLAN)
 		{
 			/* too complex for an inline implementation */
@@ -4171,6 +4180,59 @@ ExecEvalGroupingFunc(ExprState *state, E
 }
 
 /*
+ * ExecEvalMergingFunc
+ *
+ * Returns information about the current MERGE action for its RETURNING list.
+ */
+void
+ExecEvalMergingFunc(ExprState *state, ExprEvalStep *op, ExprContext *econtext)
+{
+	MergingFuncOp mfop = op->d.merging_func.mfop;
+	ModifyTableState *mtstate = castNode(ModifyTableState, state->parent);
+	MergeActionState *relaction = mtstate->mt_merge_action;
+
+	if (!relaction)
+		elog(ERROR, "no merge action in progress");
+
+	switch (mfop)
+	{
+		case MERGING_ACTION:
+			/* Return the MERGE action ("INSERT", "UPDATE", or "DELETE") */
+			switch (relaction->mas_action->commandType)
+			{
+				case CMD_INSERT:
+					*op->resvalue = PointerGetDatum(cstring_to_text_with_len("INSERT", 6));
+					*op->resnull = false;
+					break;
+				case CMD_UPDATE:
+					*op->resvalue = PointerGetDatum(cstring_to_text_with_len("UPDATE", 6));
+					*op->resnull = false;
+					break;
+				case CMD_DELETE:
+					*op->resvalue = PointerGetDatum(cstring_to_text_with_len("DELETE", 6));
+					*op->resnull = false;
+					break;
+				case CMD_NOTHING:
+					elog(ERROR, "unexpected merge action: DO NOTHING");
+					break;
+				default:
+					elog(ERROR, "unrecognized commandType: %d",
+						 (int) relaction->mas_action->commandType);
+			}
+			break;
+
+		case MERGING_CLAUSE_NUMBER:
+			/* Return the 1-based index of the MERGE action */
+			*op->resvalue = Int32GetDatum(relaction->mas_action->index);
+			*op->resnull = false;
+			break;
+
+		default:
+			elog(ERROR, "unrecognized MergingFuncOp: %d", (int) mfop);
+	}
+}
+
+/*
  * Hand off evaluation of a subplan to nodeSubplan.c
  */
 void
diff --git a/src/backend/executor/execPartition.c b/src/backend/executor/execPartition.c
new file mode 100644
index f6c3432..34d8231
--- a/src/backend/executor/execPartition.c
+++ b/src/backend/executor/execPartition.c
@@ -612,8 +612,8 @@ ExecInitPartitionInfo(ModifyTableState *
 	 * Build the RETURNING projection for the partition.  Note that we didn't
 	 * build the returningList for partitions within the planner, but simple
 	 * translation of varattnos will suffice.  This only occurs for the INSERT
-	 * case or in the case of UPDATE tuple routing where we didn't find a
-	 * result rel to reuse.
+	 * case or in the case of UPDATE/MERGE tuple routing where we didn't find
+	 * a result rel to reuse.
 	 */
 	if (node && node->returningLists != NIL)
 	{
@@ -622,12 +622,14 @@ ExecInitPartitionInfo(ModifyTableState *
 		List	   *returningList;
 
 		/* See the comment above for WCO lists. */
-		/* (except no RETURNING support for MERGE yet) */
 		Assert((node->operation == CMD_INSERT &&
 				list_length(node->returningLists) == 1 &&
 				list_length(node->resultRelations) == 1) ||
 			   (node->operation == CMD_UPDATE &&
 				list_length(node->returningLists) ==
+				list_length(node->resultRelations)) ||
+			   (node->operation == CMD_MERGE &&
+				list_length(node->returningLists) ==
 				list_length(node->resultRelations)));
 
 		/*
diff --git a/src/backend/executor/functions.c b/src/backend/executor/functions.c
new file mode 100644
index bace252..68cb4fa
--- a/src/backend/executor/functions.c
+++ b/src/backend/executor/functions.c
@@ -1660,8 +1660,8 @@ check_sql_fn_retval(List *queryTreeLists
 
 	/*
 	 * If it's a plain SELECT, it returns whatever the targetlist says.
-	 * Otherwise, if it's INSERT/UPDATE/DELETE with RETURNING, it returns
-	 * that. Otherwise, the function return type must be VOID.
+	 * Otherwise, if it's INSERT/UPDATE/DELETE/MERGE with RETURNING, it
+	 * returns that. Otherwise, the function return type must be VOID.
 	 *
 	 * Note: eventually replace this test with QueryReturnsTuples?	We'd need
 	 * a more general method of determining the output type, though.  Also, it
@@ -1679,7 +1679,8 @@ check_sql_fn_retval(List *queryTreeLists
 	else if (parse &&
 			 (parse->commandType == CMD_INSERT ||
 			  parse->commandType == CMD_UPDATE ||
-			  parse->commandType == CMD_DELETE) &&
+			  parse->commandType == CMD_DELETE ||
+			  parse->commandType == CMD_MERGE) &&
 			 parse->returningList)
 	{
 		tlist = parse->returningList;
@@ -1693,7 +1694,7 @@ check_sql_fn_retval(List *queryTreeLists
 				(errcode(ERRCODE_INVALID_FUNCTION_DEFINITION),
 				 errmsg("return type mismatch in function declared to return %s",
 						format_type_be(rettype)),
-				 errdetail("Function's final statement must be SELECT or INSERT/UPDATE/DELETE RETURNING.")));
+				 errdetail("Function's final statement must be SELECT or INSERT/UPDATE/DELETE/MERGE RETURNING.")));
 		return false;			/* keep compiler quiet */
 	}
 
diff --git a/src/backend/executor/nodeModifyTable.c b/src/backend/executor/nodeModifyTable.c
new file mode 100644
index b16fbe9..6f9c5c9
--- a/src/backend/executor/nodeModifyTable.c
+++ b/src/backend/executor/nodeModifyTable.c
@@ -36,8 +36,7 @@
  *		RETURNING tuple after completing each row insert, update, or delete.
  *		It must be called again to continue the operation.  Without RETURNING,
  *		we just loop within the node until all the work is done, then
- *		return NULL.  This avoids useless call/return overhead.  (MERGE does
- *		not support RETURNING.)
+ *		return NULL.  This avoids useless call/return overhead.
  */
 
 #include "postgres.h"
@@ -88,9 +87,6 @@ typedef struct ModifyTableContext
 	 */
 	TupleTableSlot *planSlot;
 
-	/* MERGE specific */
-	MergeActionState *relaction;	/* MERGE action in progress */
-
 	/*
 	 * Information about the changes that were made concurrently to a tuple
 	 * being updated or deleted
@@ -153,13 +149,14 @@ static TupleTableSlot *ExecMerge(ModifyT
 								 ItemPointer tupleid,
 								 bool canSetTag);
 static void ExecInitMerge(ModifyTableState *mtstate, EState *estate);
-static bool ExecMergeMatched(ModifyTableContext *context,
-							 ResultRelInfo *resultRelInfo,
-							 ItemPointer tupleid,
-							 bool canSetTag);
-static void ExecMergeNotMatched(ModifyTableContext *context,
-								ResultRelInfo *resultRelInfo,
-								bool canSetTag);
+static TupleTableSlot *ExecMergeMatched(ModifyTableContext *context,
+										ResultRelInfo *resultRelInfo,
+										ItemPointer tupleid,
+										bool canSetTag,
+										bool *matched);
+static TupleTableSlot *ExecMergeNotMatched(ModifyTableContext *context,
+										   ResultRelInfo *resultRelInfo,
+										   bool canSetTag);
 
 
 /*
@@ -979,7 +976,7 @@ ExecInsert(ModifyTableContext *context,
 		if (mtstate->operation == CMD_UPDATE)
 			wco_kind = WCO_RLS_UPDATE_CHECK;
 		else if (mtstate->operation == CMD_MERGE)
-			wco_kind = (context->relaction->mas_action->commandType == CMD_UPDATE) ?
+			wco_kind = (mtstate->mt_merge_action->mas_action->commandType == CMD_UPDATE) ?
 				WCO_RLS_UPDATE_CHECK : WCO_RLS_INSERT_CHECK;
 		else
 			wco_kind = WCO_RLS_INSERT_CHECK;
@@ -1828,7 +1825,7 @@ ExecCrossPartitionUpdate(ModifyTableCont
 		 * additional rechecking, and might end up executing a different
 		 * action entirely).
 		 */
-		if (context->relaction != NULL)
+		if (mtstate->mt_merge_action != NULL)
 			return false;
 		else if (TupIsNull(epqslot))
 			return true;
@@ -2069,7 +2066,7 @@ lreplace:
 		 * No luck, a retry is needed.  If running MERGE, we do not do so
 		 * here; instead let it handle that on its own rules.
 		 */
-		if (context->relaction != NULL)
+		if (context->mtstate->mt_merge_action != NULL)
 			return TM_Updated;
 
 		/*
@@ -2712,6 +2709,7 @@ static TupleTableSlot *
 ExecMerge(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 		  ItemPointer tupleid, bool canSetTag)
 {
+	TupleTableSlot *rslot = NULL;
 	bool		matched;
 
 	/*-----
@@ -2759,18 +2757,18 @@ ExecMerge(ModifyTableContext *context, R
 	 */
 	matched = tupleid != NULL;
 	if (matched)
-		matched = ExecMergeMatched(context, resultRelInfo, tupleid, canSetTag);
+		rslot = ExecMergeMatched(context, resultRelInfo, tupleid, canSetTag,
+								 &matched);
 
 	/*
-	 * Either we were dealing with a NOT MATCHED tuple or ExecMergeMatched()
-	 * returned "false", indicating the previously MATCHED tuple no longer
-	 * matches.
+	 * Deal with the NOT MATCHED case (either a NOT MATCHED tuple from the
+	 * join, or a previously MATCHED tuple for which ExecMergeMatched() set
+	 * "matched" to false, indicating that it no longer matches).
 	 */
 	if (!matched)
-		ExecMergeNotMatched(context, resultRelInfo, canSetTag);
+		rslot = ExecMergeNotMatched(context, resultRelInfo, canSetTag);
 
-	/* No RETURNING support yet */
-	return NULL;
+	return rslot;
 }
 
 /*
@@ -2780,8 +2778,8 @@ ExecMerge(ModifyTableContext *context, R
  * We start from the first WHEN MATCHED action and check if the WHEN quals
  * pass, if any. If the WHEN quals for the first action do not pass, we
  * check the second, then the third and so on. If we reach to the end, no
- * action is taken and we return true, indicating that no further action is
- * required for this tuple.
+ * action is taken and "matched" is set to true, indicating that no further
+ * action is required for this tuple.
  *
  * If we do find a qualifying action, then we attempt to execute the action.
  *
@@ -2790,16 +2788,17 @@ ExecMerge(ModifyTableContext *context, R
  * with individual actions are evaluated by this routine via ExecQual, while
  * EvalPlanQual checks for the join quals. If EvalPlanQual tells us that the
  * updated tuple still passes the join quals, then we restart from the first
- * action to look for a qualifying action. Otherwise, we return false --
- * meaning that a NOT MATCHED action must now be executed for the current
- * source tuple.
+ * action to look for a qualifying action. Otherwise, "matched" is set to
+ * false -- meaning that a NOT MATCHED action must now be executed for the
+ * current source tuple.
  */
-static bool
+static TupleTableSlot *
 ExecMergeMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
-				 ItemPointer tupleid, bool canSetTag)
+				 ItemPointer tupleid, bool canSetTag, bool *matched)
 {
 	ModifyTableState *mtstate = context->mtstate;
-	TupleTableSlot *newslot;
+	TupleTableSlot *rslot = NULL;
+	TupleTableSlot *newslot = NULL;
 	EState	   *estate = context->estate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
 	bool		isNull;
@@ -2810,7 +2809,10 @@ ExecMergeMatched(ModifyTableContext *con
 	 * If there are no WHEN MATCHED actions, we are done.
 	 */
 	if (resultRelInfo->ri_matchedMergeAction == NIL)
-		return true;
+	{
+		*matched = true;
+		return NULL;
+	}
 
 	/*
 	 * Make tuple and any needed join variables available to ExecQual and
@@ -2889,12 +2891,15 @@ lmerge_matched:
 				 */
 				newslot = ExecProject(relaction->mas_proj);
 
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				if (!ExecUpdatePrologue(context, resultRelInfo,
 										tupleid, NULL, newslot, &result))
 				{
 					if (result == TM_Ok)
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 					break;		/* concurrent update/delete */
 				}
 				result = ExecUpdateAct(context, resultRelInfo, tupleid, NULL,
@@ -2905,14 +2910,15 @@ lmerge_matched:
 				 * cross-partition update was done, then there's nothing else
 				 * for us to do --- the UPDATE has been turned into a DELETE
 				 * and an INSERT, and we must not perform any of the usual
-				 * post-update tasks.
+				 * post-update tasks.  Also, the RETURNING tuple (if any) has
+				 * been projected, so we can just return that.
 				 */
 				if (updateCxt.crossPartUpdate)
 				{
 					mtstate->mt_merge_updated += 1;
 					if (canSetTag)
 						(estate->es_processed)++;
-					return true;
+					return context->cpUpdateReturningSlot;
 				}
 
 				if (result == TM_Ok && updateCxt.updated)
@@ -2924,12 +2930,15 @@ lmerge_matched:
 				break;
 
 			case CMD_DELETE:
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				if (!ExecDeletePrologue(context, resultRelInfo, tupleid,
 										NULL, NULL, &result))
 				{
 					if (result == TM_Ok)
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 					break;		/* concurrent update/delete */
 				}
 				result = ExecDeleteAct(context, resultRelInfo, tupleid, false);
@@ -2985,7 +2994,8 @@ lmerge_matched:
 				 * If the tuple was already deleted, return to let caller
 				 * handle it under NOT MATCHED clauses.
 				 */
-				return false;
+				*matched = false;
+				return NULL;
 
 			case TM_Updated:
 				{
@@ -3031,13 +3041,19 @@ lmerge_matched:
 							 * NOT MATCHED actions.
 							 */
 							if (TupIsNull(epqslot))
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							(void) ExecGetJunkAttribute(epqslot,
 														resultRelInfo->ri_RowIdAttNo,
 														&isNull);
 							if (isNull)
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							/*
 							 * When a tuple was updated and migrated to
@@ -3072,7 +3088,8 @@ lmerge_matched:
 							 * tuple already deleted; tell caller to run NOT
 							 * MATCHED actions
 							 */
-							return false;
+							*matched = false;
+							return NULL;
 
 						case TM_SelfModified:
 
@@ -3092,13 +3109,14 @@ lmerge_matched:
 										(errcode(ERRCODE_TRIGGERED_DATA_CHANGE_VIOLATION),
 										 errmsg("tuple to be updated or deleted was already modified by an operation triggered by the current command"),
 										 errhint("Consider using an AFTER trigger instead of a BEFORE trigger to propagate changes to other rows.")));
-							return false;
+							*matched = false;
+							return NULL;
 
 						default:
 							/* see table_tuple_lock call in ExecDelete() */
 							elog(ERROR, "unexpected table_tuple_lock status: %u",
 								 result);
-							return false;
+							return NULL;
 					}
 				}
 
@@ -3110,6 +3128,31 @@ lmerge_matched:
 				break;
 		}
 
+		/* Process RETURNING if present */
+		if (resultRelInfo->ri_projectReturning)
+		{
+			switch (commandType)
+			{
+				case CMD_UPDATE:
+					rslot = ExecProcessReturning(resultRelInfo, newslot,
+												 context->planSlot);
+					break;
+
+				case CMD_DELETE:
+					rslot = ExecProcessReturning(resultRelInfo,
+												 resultRelInfo->ri_oldTupleSlot,
+												 context->planSlot);
+					break;
+
+				case CMD_NOTHING:
+					break;
+
+				default:
+					elog(ERROR, "unrecognized commandType: %d",
+						 (int) commandType);
+			}
+		}
+
 		/*
 		 * We've activated one of the WHEN clauses, so we don't search
 		 * further. This is required behaviour, not an optimization.
@@ -3120,19 +3163,22 @@ lmerge_matched:
 	/*
 	 * Successfully executed an action or no qualifying action was found.
 	 */
-	return true;
+	*matched = true;
+
+	return rslot;
 }
 
 /*
  * Execute the first qualifying NOT MATCHED action.
  */
-static void
+static TupleTableSlot *
 ExecMergeNotMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 					bool canSetTag)
 {
 	ModifyTableState *mtstate = context->mtstate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
 	List	   *actionStates = NIL;
+	TupleTableSlot *rslot = NULL;
 	ListCell   *l;
 
 	/*
@@ -3182,10 +3228,10 @@ ExecMergeNotMatched(ModifyTableContext *
 				 * so we don't need to map the tuple here.
 				 */
 				newslot = ExecProject(action->mas_proj);
-				context->relaction = action;
+				mtstate->mt_merge_action = action;
 
-				(void) ExecInsert(context, mtstate->rootResultRelInfo, newslot,
-								  canSetTag, NULL, NULL);
+				rslot = ExecInsert(context, mtstate->rootResultRelInfo,
+								   newslot, canSetTag, NULL, NULL);
 				mtstate->mt_merge_inserted += 1;
 				break;
 			case CMD_NOTHING:
@@ -3201,6 +3247,8 @@ ExecMergeNotMatched(ModifyTableContext *
 		 */
 		break;
 	}
+
+	return rslot;
 }
 
 /*
@@ -3663,8 +3711,17 @@ ExecModifyTable(PlanState *pstate)
 				{
 					EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-					ExecMerge(&context, node->resultRelInfo, NULL, node->canSetTag);
-					continue;	/* no RETURNING support yet */
+					slot = ExecMerge(&context, node->resultRelInfo, NULL,
+									 node->canSetTag);
+
+					/*
+					 * If we got a RETURNING result, return it to the caller.
+					 * We'll continue the work on next call.
+					 */
+					if (slot)
+						return slot;
+
+					continue;	/* continue with the next tuple */
 				}
 
 				elog(ERROR, "tableoid is NULL");
@@ -3741,8 +3798,17 @@ ExecModifyTable(PlanState *pstate)
 					{
 						EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-						ExecMerge(&context, node->resultRelInfo, NULL, node->canSetTag);
-						continue;	/* no RETURNING support yet */
+						slot = ExecMerge(&context, node->resultRelInfo, NULL,
+										 node->canSetTag);
+
+						/*
+						 * If we got a RETURNING result, return it to the
+						 * caller.  We'll continue the work on next call.
+						 */
+						if (slot)
+							return slot;
+
+						continue;	/* continue with the next tuple */
 					}
 
 					elog(ERROR, "ctid is NULL");
@@ -3834,7 +3900,7 @@ ExecModifyTable(PlanState *pstate)
 				}
 				slot = ExecGetUpdateNewTuple(resultRelInfo, context.planSlot,
 											 oldSlot);
-				context.relaction = NULL;
+				node->mt_merge_action = NULL;
 
 				/* Now apply the update. */
 				slot = ExecUpdate(&context, resultRelInfo, tupleid, oldtuple,
diff --git a/src/backend/executor/spi.c b/src/backend/executor/spi.c
new file mode 100644
index 3397568..b9f443c
--- a/src/backend/executor/spi.c
+++ b/src/backend/executor/spi.c
@@ -2033,6 +2033,8 @@ SPI_result_code_string(int code)
 			return "SPI_OK_TD_REGISTER";
 		case SPI_OK_MERGE:
 			return "SPI_OK_MERGE";
+		case SPI_OK_MERGE_RETURNING:
+			return "SPI_OK_MERGE_RETURNING";
 	}
 	/* Unrecognized code ... return something useful ... */
 	sprintf(buf, "Unrecognized SPI code %d", code);
@@ -2886,7 +2888,10 @@ _SPI_pquery(QueryDesc *queryDesc, bool f
 				res = SPI_OK_UPDATE;
 			break;
 		case CMD_MERGE:
-			res = SPI_OK_MERGE;
+			if (queryDesc->plannedstmt->hasReturning)
+				res = SPI_OK_MERGE_RETURNING;
+			else
+				res = SPI_OK_MERGE;
 			break;
 		default:
 			return SPI_ERROR_OPUNKNOWN;
diff --git a/src/backend/jit/llvm/llvmjit_expr.c b/src/backend/jit/llvm/llvmjit_expr.c
new file mode 100644
index a3a0876..e966037
--- a/src/backend/jit/llvm/llvmjit_expr.c
+++ b/src/backend/jit/llvm/llvmjit_expr.c
@@ -1980,6 +1980,12 @@ llvm_compile_expr(ExprState *state)
 					break;
 				}
 
+			case EEOP_MERGING_FUNC:
+				build_EvalXFunc(b, mod, "ExecEvalMergingFunc",
+								v_state, op, v_econtext);
+				LLVMBuildBr(b, opblocks[opno + 1]);
+				break;
+
 			case EEOP_SUBPLAN:
 				build_EvalXFunc(b, mod, "ExecEvalSubPlan",
 								v_state, op, v_econtext);
diff --git a/src/backend/jit/llvm/llvmjit_types.c b/src/backend/jit/llvm/llvmjit_types.c
new file mode 100644
index 791902f..65db8f2
--- a/src/backend/jit/llvm/llvmjit_types.c
+++ b/src/backend/jit/llvm/llvmjit_types.c
@@ -155,6 +155,7 @@ void	   *referenced_functions[] =
 	ExecEvalFuncExprFusage,
 	ExecEvalFuncExprStrictFusage,
 	ExecEvalGroupingFunc,
+	ExecEvalMergingFunc,
 	ExecEvalMinMax,
 	ExecEvalNextValueExpr,
 	ExecEvalParamExec,
diff --git a/src/backend/nodes/nodeFuncs.c b/src/backend/nodes/nodeFuncs.c
new file mode 100644
index c03f4f2..6cf7895
--- a/src/backend/nodes/nodeFuncs.c
+++ b/src/backend/nodes/nodeFuncs.c
@@ -67,6 +67,9 @@ exprType(const Node *expr)
 		case T_WindowFunc:
 			type = ((const WindowFunc *) expr)->wintype;
 			break;
+		case T_MergingFunc:
+			type = ((const MergingFunc *) expr)->mftype;
+			break;
 		case T_SubscriptingRef:
 			type = ((const SubscriptingRef *) expr)->refrestype;
 			break;
@@ -810,6 +813,9 @@ exprCollation(const Node *expr)
 		case T_WindowFunc:
 			coll = ((const WindowFunc *) expr)->wincollid;
 			break;
+		case T_MergingFunc:
+			coll = ((const MergingFunc *) expr)->mfcollid;
+			break;
 		case T_SubscriptingRef:
 			coll = ((const SubscriptingRef *) expr)->refcollid;
 			break;
@@ -1085,6 +1091,9 @@ exprSetCollation(Node *expr, Oid collati
 		case T_WindowFunc:
 			((WindowFunc *) expr)->wincollid = collation;
 			break;
+		case T_MergingFunc:
+			((MergingFunc *) expr)->mfcollid = collation;
+			break;
 		case T_SubscriptingRef:
 			((SubscriptingRef *) expr)->refcollid = collation;
 			break;
@@ -1343,6 +1352,9 @@ exprLocation(const Node *expr)
 			/* function name should always be the first thing */
 			loc = ((const WindowFunc *) expr)->location;
 			break;
+		case T_MergingFunc:
+			loc = ((const MergingFunc *) expr)->location;
+			break;
 		case T_SubscriptingRef:
 			/* just use container argument's location */
 			loc = exprLocation((Node *) ((const SubscriptingRef *) expr)->refexpr);
@@ -2035,6 +2047,7 @@ expression_tree_walker_impl(Node *node,
 		case T_RangeTblRef:
 		case T_SortGroupClause:
 		case T_CTESearchClause:
+		case T_MergingFunc:
 			/* primitive node types with no expression subnodes */
 			break;
 		case T_WithCheckOption:
@@ -2865,6 +2878,7 @@ expression_tree_mutator_impl(Node *node,
 		case T_RangeTblRef:
 		case T_SortGroupClause:
 		case T_CTESearchClause:
+		case T_MergingFunc:
 			return (Node *) copyObject(node);
 		case T_WithCheckOption:
 			{
@@ -3827,6 +3841,7 @@ raw_expression_tree_walker_impl(Node *no
 		case T_ParamRef:
 		case T_A_Const:
 		case T_A_Star:
+		case T_MergingFunc:
 			/* primitive node types with no subnodes */
 			break;
 		case T_Alias:
diff --git a/src/backend/optimizer/plan/subselect.c b/src/backend/optimizer/plan/subselect.c
new file mode 100644
index 7a9fe88..262adf5
--- a/src/backend/optimizer/plan/subselect.c
+++ b/src/backend/optimizer/plan/subselect.c
@@ -1846,7 +1846,8 @@ convert_EXISTS_to_ANY(PlannerInfo *root,
 /*
  * Replace correlation vars (uplevel vars) with Params.
  *
- * Uplevel PlaceHolderVars and aggregates are replaced, too.
+ * Uplevel PlaceHolderVars, aggregates, GROUPING() expressions, and MERGING()
+ * expressions are replaced, too.
  *
  * Note: it is critical that this runs immediately after SS_process_sublinks.
  * Since we do not recurse into the arguments of uplevel PHVs and aggregates,
@@ -1900,6 +1901,11 @@ replace_correlation_vars_mutator(Node *n
 		if (((GroupingFunc *) node)->agglevelsup > 0)
 			return (Node *) replace_outer_grouping(root, (GroupingFunc *) node);
 	}
+	if (IsA(node, MergingFunc))
+	{
+		if (root->parse->commandType != CMD_MERGE)
+			return (Node *) replace_outer_merging(root, (MergingFunc *) node);
+	}
 	return expression_tree_mutator(node,
 								   replace_correlation_vars_mutator,
 								   (void *) root);
diff --git a/src/backend/optimizer/util/paramassign.c b/src/backend/optimizer/util/paramassign.c
new file mode 100644
index d6a923b..b3b4da6
--- a/src/backend/optimizer/util/paramassign.c
+++ b/src/backend/optimizer/util/paramassign.c
@@ -307,6 +307,56 @@ replace_outer_grouping(PlannerInfo *root
 }
 
 /*
+ * Generate a Param node to replace the given MergingFunc expression which is
+ * expected to be in the RETURNING list of an upper-level MERGE query.  Record
+ * the need for the MergingFunc in the proper upper-level root->plan_params.
+ */
+Param *
+replace_outer_merging(PlannerInfo *root, MergingFunc *mf)
+{
+	Param	   *retval;
+	PlannerParamItem *pitem;
+	Oid			ptype = exprType((Node *) mf);
+
+	Assert(root->parse->commandType != CMD_MERGE);
+
+	/*
+	 * The parser should have ensured that the MergingFunc is in the RETURNING
+	 * list of an upper-level MERGE query, so find that query.
+	 */
+	do
+	{
+		root = root->parent_root;
+		if (root == NULL)
+			elog(ERROR, "MergingFunc found outside MERGE");
+	} while (root->parse->commandType != CMD_MERGE);
+
+	/*
+	 * It does not seem worthwhile to try to de-duplicate references to outer
+	 * MergingFunc expressions.  Just make a new slot every time.
+	 */
+	mf = copyObject(mf);
+
+	pitem = makeNode(PlannerParamItem);
+	pitem->item = (Node *) mf;
+	pitem->paramId = list_length(root->glob->paramExecTypes);
+	root->glob->paramExecTypes = lappend_oid(root->glob->paramExecTypes,
+											 ptype);
+
+	root->plan_params = lappend(root->plan_params, pitem);
+
+	retval = makeNode(Param);
+	retval->paramkind = PARAM_EXEC;
+	retval->paramid = pitem->paramId;
+	retval->paramtype = ptype;
+	retval->paramtypmod = -1;
+	retval->paramcollid = InvalidOid;
+	retval->location = mf->location;
+
+	return retval;
+}
+
+/*
  * Generate a Param node to replace the given Var,
  * which is expected to come from some upper NestLoop plan node.
  * Record the need for the Var in root->curOuterParams.
diff --git a/src/backend/parser/analyze.c b/src/backend/parser/analyze.c
new file mode 100644
index 7a1dfb6..5f882b4
--- a/src/backend/parser/analyze.c
+++ b/src/backend/parser/analyze.c
@@ -74,7 +74,6 @@ static void determineRecursiveColTypes(P
 									   Node *larg, List *nrtargetlist);
 static Query *transformReturnStmt(ParseState *pstate, ReturnStmt *stmt);
 static Query *transformUpdateStmt(ParseState *pstate, UpdateStmt *stmt);
-static List *transformReturningList(ParseState *pstate, List *returningList);
 static Query *transformPLAssignStmt(ParseState *pstate,
 									PLAssignStmt *stmt);
 static Query *transformDeclareCursorStmt(ParseState *pstate,
@@ -553,7 +552,8 @@ transformDeleteStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -980,7 +980,8 @@ transformInsertStmt(ParseState *pstate,
 	/* Process RETURNING, if any. */
 	if (stmt->returningList)
 		qry->returningList = transformReturningList(pstate,
-													stmt->returningList);
+													stmt->returningList,
+													EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -2445,7 +2446,8 @@ transformUpdateStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/*
 	 * Now we are done with SELECT-like processing, and can get on with
@@ -2539,10 +2541,11 @@ transformUpdateTargetList(ParseState *ps
 
 /*
  * transformReturningList -
- *	handle a RETURNING clause in INSERT/UPDATE/DELETE
+ *	handle a RETURNING clause in INSERT/UPDATE/DELETE/MERGE
  */
-static List *
-transformReturningList(ParseState *pstate, List *returningList)
+List *
+transformReturningList(ParseState *pstate, List *returningList,
+					   ParseExprKind exprKind)
 {
 	List	   *rlist;
 	int			save_next_resno;
@@ -2559,7 +2562,7 @@ transformReturningList(ParseState *pstat
 	pstate->p_next_resno = 1;
 
 	/* transform RETURNING identically to a SELECT targetlist */
-	rlist = transformTargetList(pstate, returningList, EXPR_KIND_RETURNING);
+	rlist = transformTargetList(pstate, returningList, exprKind);
 
 	/*
 	 * Complain if the nonempty tlist expanded to nothing (which is possible
diff --git a/src/backend/parser/gram.y b/src/backend/parser/gram.y
new file mode 100644
index c224df4..c366c5b
--- a/src/backend/parser/gram.y
+++ b/src/backend/parser/gram.y
@@ -693,7 +693,7 @@ static Node *makeRecursiveViewSelect(cha
 	BOOLEAN_P BOTH BREADTH BY
 
 	CACHE CALL CALLED CASCADE CASCADED CASE CAST CATALOG_P CHAIN CHAR_P
-	CHARACTER CHARACTERISTICS CHECK CHECKPOINT CLASS CLOSE
+	CHARACTER CHARACTERISTICS CHECK CHECKPOINT CLASS CLAUSE_NUMBER CLOSE
 	CLUSTER COALESCE COLLATE COLLATION COLUMN COLUMNS COMMENT COMMENTS COMMIT
 	COMMITTED COMPRESSION CONCURRENTLY CONFIGURATION CONFLICT
 	CONNECTION CONSTRAINT CONSTRAINTS CONTENT_P CONTINUE_P CONVERSION_P COPY
@@ -731,7 +731,7 @@ static Node *makeRecursiveViewSelect(cha
 	LEADING LEAKPROOF LEAST LEFT LEVEL LIKE LIMIT LISTEN LOAD LOCAL
 	LOCALTIME LOCALTIMESTAMP LOCATION LOCK_P LOCKED LOGGED
 
-	MAPPING MATCH MATCHED MATERIALIZED MAXVALUE MERGE METHOD
+	MAPPING MATCH MATCHED MATERIALIZED MAXVALUE MERGE MERGING METHOD
 	MINUTE_P MINVALUE MODE MONTH_P MOVE
 
 	NAME_P NAMES NATIONAL NATURAL NCHAR NEW NEXT NFC NFD NFKC NFKD NO NONE
@@ -12332,6 +12332,7 @@ MergeStmt:
 			USING table_ref
 			ON a_expr
 			merge_when_list
+			returning_clause
 				{
 					MergeStmt  *m = makeNode(MergeStmt);
 
@@ -12340,6 +12341,7 @@ MergeStmt:
 					m->sourceRelation = $6;
 					m->joinCondition = $8;
 					m->mergeWhenClauses = $9;
+					m->returningList = $10;
 
 					$$ = (Node *) m;
 				}
@@ -15755,6 +15757,24 @@ func_expr_common_subexpr:
 					n->location = @1;
 					$$ = (Node *) n;
 				}
+			| MERGING '(' ACTION ')'
+				{
+					MergingFunc *m = makeNode(MergingFunc);
+
+					m->mfop = MERGING_ACTION;
+					m->mftype = TEXTOID;
+					m->location = @1;
+					$$ = (Node *) m;
+				}
+			| MERGING '(' CLAUSE_NUMBER ')'
+				{
+					MergingFunc *m = makeNode(MergingFunc);
+
+					m->mfop = MERGING_CLAUSE_NUMBER;
+					m->mftype = INT4OID;
+					m->location = @1;
+					$$ = (Node *) m;
+				}
 			;
 
 
@@ -17100,6 +17120,7 @@ unreserved_keyword:
 			| CHARACTERISTICS
 			| CHECKPOINT
 			| CLASS
+			| CLAUSE_NUMBER
 			| CLOSE
 			| CLUSTER
 			| COLUMNS
@@ -17419,6 +17440,7 @@ col_name_keyword:
 			| JSON_SCALAR
 			| JSON_SERIALIZE
 			| LEAST
+			| MERGING
 			| NATIONAL
 			| NCHAR
 			| NONE
@@ -17637,6 +17659,7 @@ bare_label_keyword:
 			| CHECK
 			| CHECKPOINT
 			| CLASS
+			| CLAUSE_NUMBER
 			| CLOSE
 			| CLUSTER
 			| COALESCE
@@ -17808,6 +17831,7 @@ bare_label_keyword:
 			| MATERIALIZED
 			| MAXVALUE
 			| MERGE
+			| MERGING
 			| METHOD
 			| MINVALUE
 			| MODE
diff --git a/src/backend/parser/parse_agg.c b/src/backend/parser/parse_agg.c
new file mode 100644
index 9bbad33..f57098f
--- a/src/backend/parser/parse_agg.c
+++ b/src/backend/parser/parse_agg.c
@@ -467,6 +467,7 @@ check_agglevels_and_constraints(ParseSta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
@@ -914,6 +915,7 @@ transformWindowFuncCall(ParseState *psta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_cte.c b/src/backend/parser/parse_cte.c
new file mode 100644
index 6992a78..5cad6d8
--- a/src/backend/parser/parse_cte.c
+++ b/src/backend/parser/parse_cte.c
@@ -126,13 +126,6 @@ transformWithClause(ParseState *pstate,
 		CommonTableExpr *cte = (CommonTableExpr *) lfirst(lc);
 		ListCell   *rest;
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(cte->ctequery, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in WITH query"),
-					parser_errposition(pstate, cte->location));
-
 		for_each_cell(rest, withClause->ctes, lnext(withClause->ctes, lc))
 		{
 			CommonTableExpr *cte2 = (CommonTableExpr *) lfirst(rest);
@@ -153,7 +146,8 @@ transformWithClause(ParseState *pstate,
 			/* must be a data-modifying statement */
 			Assert(IsA(cte->ctequery, InsertStmt) ||
 				   IsA(cte->ctequery, UpdateStmt) ||
-				   IsA(cte->ctequery, DeleteStmt));
+				   IsA(cte->ctequery, DeleteStmt) ||
+				   IsA(cte->ctequery, MergeStmt));
 
 			pstate->p_hasModifyingCTE = true;
 		}
diff --git a/src/backend/parser/parse_expr.c b/src/backend/parser/parse_expr.c
new file mode 100644
index 64c582c..ad0717e
--- a/src/backend/parser/parse_expr.c
+++ b/src/backend/parser/parse_expr.c
@@ -54,6 +54,7 @@ static Node *transformAExprDistinct(Pars
 static Node *transformAExprNullIf(ParseState *pstate, A_Expr *a);
 static Node *transformAExprIn(ParseState *pstate, A_Expr *a);
 static Node *transformAExprBetween(ParseState *pstate, A_Expr *a);
+static Node *transformMergingFunc(ParseState *pstate, MergingFunc *f);
 static Node *transformBoolExpr(ParseState *pstate, BoolExpr *a);
 static Node *transformFuncCall(ParseState *pstate, FuncCall *fn);
 static Node *transformMultiAssignRef(ParseState *pstate, MultiAssignRef *maref);
@@ -227,6 +228,10 @@ transformExprRecurse(ParseState *pstate,
 			result = transformGroupingFunc(pstate, (GroupingFunc *) expr);
 			break;
 
+		case T_MergingFunc:
+			result = transformMergingFunc(pstate, (MergingFunc *) expr);
+			break;
+
 		case T_NamedArgExpr:
 			{
 				NamedArgExpr *na = (NamedArgExpr *) expr;
@@ -541,6 +546,7 @@ transformColumnRef(ParseState *pstate, C
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CHECK_CONSTRAINT:
@@ -1354,6 +1360,31 @@ transformAExprBetween(ParseState *pstate
 }
 
 static Node *
+transformMergingFunc(ParseState *pstate, MergingFunc *f)
+{
+	/*
+	 * Check that we're in the RETURNING list of a MERGE command.
+	 */
+	if (pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+	{
+		ParseState *parent_pstate = pstate->parentParseState;
+
+		while (parent_pstate &&
+			   parent_pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+			parent_pstate = parent_pstate->parentParseState;
+
+		if (!parent_pstate ||
+			parent_pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+			ereport(ERROR,
+					errcode(ERRCODE_WRONG_OBJECT_TYPE),
+					errmsg("MERGING() can only be used in the RETURNING list of a MERGE command"),
+					parser_errposition(pstate, f->location));
+	}
+
+	return (Node *) f;
+}
+
+static Node *
 transformBoolExpr(ParseState *pstate, BoolExpr *a)
 {
 	List	   *args = NIL;
@@ -1767,6 +1798,7 @@ transformSubLink(ParseState *pstate, Sub
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CYCLE_MARK:
@@ -3115,6 +3147,7 @@ ParseExprKindName(ParseExprKind exprKind
 		case EXPR_KIND_OFFSET:
 			return "OFFSET";
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			return "RETURNING";
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
diff --git a/src/backend/parser/parse_func.c b/src/backend/parser/parse_func.c
new file mode 100644
index 6c29471..1a0f983
--- a/src/backend/parser/parse_func.c
+++ b/src/backend/parser/parse_func.c
@@ -2599,6 +2599,7 @@ check_srf_call_placement(ParseState *pst
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_merge.c b/src/backend/parser/parse_merge.c
new file mode 100644
index 91b1156..33852e9
--- a/src/backend/parser/parse_merge.c
+++ b/src/backend/parser/parse_merge.c
@@ -233,6 +233,10 @@ transformMergeStmt(ParseState *pstate, M
 	 */
 	qry->jointree = makeFromExpr(pstate->p_joinlist, joinExpr);
 
+	/* Transform the RETURNING list, if any */
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_MERGE_RETURNING);
+
 	/*
 	 * We now have a good query shape, so now look at the WHEN conditions and
 	 * action targetlists.
@@ -254,6 +258,7 @@ transformMergeStmt(ParseState *pstate, M
 		MergeAction *action;
 
 		action = makeNode(MergeAction);
+		action->index = foreach_current_index(l) + 1;
 		action->commandType = mergeWhenClause->commandType;
 		action->matched = mergeWhenClause->matched;
 
@@ -390,9 +395,6 @@ transformMergeStmt(ParseState *pstate, M
 
 	qry->mergeActionList = mergeActionList;
 
-	/* RETURNING could potentially be added in the future, but not in SQL std */
-	qry->returningList = NULL;
-
 	qry->hasTargetSRFs = false;
 	qry->hasSubLinks = pstate->p_hasSubLinks;
 
diff --git a/src/backend/parser/parse_relation.c b/src/backend/parser/parse_relation.c
new file mode 100644
index 864ea9b..cc19041
--- a/src/backend/parser/parse_relation.c
+++ b/src/backend/parser/parse_relation.c
@@ -2346,9 +2346,10 @@ addRangeTableEntryForCTE(ParseState *pst
 		cte->cterefcount++;
 
 	/*
-	 * We throw error if the CTE is INSERT/UPDATE/DELETE without RETURNING.
-	 * This won't get checked in case of a self-reference, but that's OK
-	 * because data-modifying CTEs aren't allowed to be recursive anyhow.
+	 * We throw error if the CTE is INSERT/UPDATE/DELETE/MERGE without
+	 * RETURNING.  This won't get checked in case of a self-reference, but
+	 * that's OK because data-modifying CTEs aren't allowed to be recursive
+	 * anyhow.
 	 */
 	if (IsA(cte->ctequery, Query))
 	{
diff --git a/src/backend/parser/parse_target.c b/src/backend/parser/parse_target.c
new file mode 100644
index 3bc62ac..eaa2a2b
--- a/src/backend/parser/parse_target.c
+++ b/src/backend/parser/parse_target.c
@@ -1807,6 +1807,20 @@ FigureColnameInternal(Node *node, char *
 			/* make GROUPING() act like a regular function */
 			*name = "grouping";
 			return 2;
+		case T_MergingFunc:
+			switch (((MergingFunc *) node)->mfop)
+			{
+				case MERGING_ACTION:
+					*name = "action";
+					return 2;
+				case MERGING_CLAUSE_NUMBER:
+					*name = "clause_number";
+					return 2;
+				default:
+					elog(ERROR, "unrecognized MergingFuncOp: %d",
+						 (int) ((MergingFunc *) node)->mfop);
+			}
+			break;
 		case T_SubLink:
 			switch (((SubLink *) node)->subLinkType)
 			{
diff --git a/src/backend/rewrite/rewriteHandler.c b/src/backend/rewrite/rewriteHandler.c
new file mode 100644
index 41a3623..3786a7c
--- a/src/backend/rewrite/rewriteHandler.c
+++ b/src/backend/rewrite/rewriteHandler.c
@@ -3615,9 +3615,9 @@ RewriteQuery(Query *parsetree, List *rew
 	ListCell   *lc1;
 
 	/*
-	 * First, recursively process any insert/update/delete statements in WITH
-	 * clauses.  (We have to do this first because the WITH clauses may get
-	 * copied into rule actions below.)
+	 * First, recursively process any insert/update/delete/merge statements in
+	 * WITH clauses.  (We have to do this first because the WITH clauses may
+	 * get copied into rule actions below.)
 	 */
 	foreach(lc1, parsetree->cteList)
 	{
@@ -3642,7 +3642,8 @@ RewriteQuery(Query *parsetree, List *rew
 			if (!(ctequery->commandType == CMD_SELECT ||
 				  ctequery->commandType == CMD_UPDATE ||
 				  ctequery->commandType == CMD_INSERT ||
-				  ctequery->commandType == CMD_DELETE))
+				  ctequery->commandType == CMD_DELETE ||
+				  ctequery->commandType == CMD_MERGE))
 			{
 				/*
 				 * Currently it could only be NOTIFY; this error message will
diff --git a/src/backend/rewrite/rowsecurity.c b/src/backend/rewrite/rowsecurity.c
new file mode 100644
index b1620e4..966dcf0
--- a/src/backend/rewrite/rowsecurity.c
+++ b/src/backend/rewrite/rowsecurity.c
@@ -395,10 +395,10 @@ get_row_security_policies(Query *root, R
 	 * on the final action we take.
 	 *
 	 * We already fetched the SELECT policies above, to check existing rows,
-	 * but we must also check that new rows created by UPDATE actions are
-	 * visible, if SELECT rights are required for this relation. We don't do
-	 * this for INSERT actions, since an INSERT command would only do this
-	 * check if it had a RETURNING list, and MERGE does not support RETURNING.
+	 * but we must also check that new rows created by INSERT/UPDATE actions
+	 * are visible, if SELECT rights are required. For INSERT actions, we only
+	 * do this if RETURNING is specified, to be consistent with a plain INSERT
+	 * command, which can only require SELECT rights when RETURNING is used.
 	 *
 	 * We don't push the UPDATE/DELETE USING quals to the RTE because we don't
 	 * really want to apply them while scanning the relation since we don't
@@ -420,6 +420,8 @@ get_row_security_policies(Query *root, R
 		List	   *merge_delete_restrictive_policies;
 		List	   *merge_insert_permissive_policies;
 		List	   *merge_insert_restrictive_policies;
+		List	   *merge_select_permissive_policies = NIL;
+		List	   *merge_select_restrictive_policies = NIL;
 
 		/*
 		 * Fetch the UPDATE policies and set them up to execute on the
@@ -457,9 +459,6 @@ get_row_security_policies(Query *root, R
 		 */
 		if (perminfo->requiredPerms & ACL_SELECT)
 		{
-			List	   *merge_select_permissive_policies;
-			List	   *merge_select_restrictive_policies;
-
 			get_policies_for_relation(rel, CMD_SELECT, user_id,
 									  &merge_select_permissive_policies,
 									  &merge_select_restrictive_policies);
@@ -508,6 +507,21 @@ get_row_security_policies(Query *root, R
 							   withCheckOptions,
 							   hasSubLinks,
 							   false);
+
+		/*
+		 * Add ALL/SELECT policies as WCO_RLS_INSERT_CHECK WCOs, to ensure
+		 * that the inserted row is visible when executing an INSERT action,
+		 * if RETURNING is specified and SELECT rights are required for this
+		 * relation.
+		 */
+		if (perminfo->requiredPerms & ACL_SELECT && root->returningList)
+			add_with_check_options(rel, rt_index,
+								   WCO_RLS_INSERT_CHECK,
+								   merge_select_permissive_policies,
+								   merge_select_restrictive_policies,
+								   withCheckOptions,
+								   hasSubLinks,
+								   true);
 	}
 
 	table_close(rel, NoLock);
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
new file mode 100644
index e3ccf6c..6f3dc89
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -2142,11 +2142,10 @@ QueryReturnsTuples(Query *parsetree)
 		case CMD_SELECT:
 			/* returns tuples */
 			return true;
-		case CMD_MERGE:
-			return false;
 		case CMD_INSERT:
 		case CMD_UPDATE:
 		case CMD_DELETE:
+		case CMD_MERGE:
 			/* the forms with RETURNING return tuples */
 			if (parsetree->returningList)
 				return true;
diff --git a/src/backend/utils/adt/ruleutils.c b/src/backend/utils/adt/ruleutils.c
new file mode 100644
index ed7f40f..aca2217
--- a/src/backend/utils/adt/ruleutils.c
+++ b/src/backend/utils/adt/ruleutils.c
@@ -7197,8 +7197,13 @@ get_merge_query_def(Query *query, depars
 			appendStringInfoString(buf, "DO NOTHING");
 	}
 
-	/* No RETURNING support in MERGE yet */
-	Assert(query->returningList == NIL);
+	/* Add RETURNING if present */
+	if (query->returningList)
+	{
+		appendContextKeyword(context, " RETURNING",
+							 -PRETTYINDENT_STD, PRETTYINDENT_STD, 1);
+		get_target_list(query->returningList, context, NULL, colNamesVisible);
+	}
 }
 
 
@@ -8298,6 +8303,7 @@ isSimpleNode(Node *node, Node *parentNod
 		case T_Aggref:
 		case T_GroupingFunc:
 		case T_WindowFunc:
+		case T_MergingFunc:
 		case T_FuncExpr:
 		case T_JsonConstructorExpr:
 			/* function-like: name(..) or name[..] */
@@ -8652,6 +8658,27 @@ get_rule_expr(Node *node, deparse_contex
 			get_windowfunc_expr((WindowFunc *) node, context);
 			break;
 
+		case T_MergingFunc:
+			{
+				MergingFunc *mexpr = (MergingFunc *) node;
+
+				appendStringInfoString(buf, "MERGING(");
+				switch (mexpr->mfop)
+				{
+					case MERGING_ACTION:
+						appendStringInfoString(buf, "ACTION");
+						break;
+					case MERGING_CLAUSE_NUMBER:
+						appendStringInfoString(buf, "CLAUSE_NUMBER");
+						break;
+					default:
+						elog(ERROR, "unrecognized MergingFuncOp: %d",
+							 (int) mexpr->mfop);
+				}
+				appendStringInfoChar(buf, ')');
+			}
+			break;
+
 		case T_SubscriptingRef:
 			{
 				SubscriptingRef *sbsref = (SubscriptingRef *) node;
diff --git a/src/bin/psql/common.c b/src/bin/psql/common.c
new file mode 100644
index daabf6f..6b63214
--- a/src/bin/psql/common.c
+++ b/src/bin/psql/common.c
@@ -983,13 +983,17 @@ PrintQueryResult(PGresult *result, bool
 			else
 				success = true;
 
-			/* if it's INSERT/UPDATE/DELETE RETURNING, also print status */
+			/*
+			 * If it's INSERT/UPDATE/DELETE/MERGE RETURNING, also print
+			 * status.
+			 */
 			if (last || pset.show_all_results)
 			{
 				cmdstatus = PQcmdStatus(result);
 				if (strncmp(cmdstatus, "INSERT", 6) == 0 ||
 					strncmp(cmdstatus, "UPDATE", 6) == 0 ||
-					strncmp(cmdstatus, "DELETE", 6) == 0)
+					strncmp(cmdstatus, "DELETE", 6) == 0 ||
+					strncmp(cmdstatus, "MERGE", 5) == 0)
 					PrintQueryStatus(result, printStatusFout);
 			}
 
diff --git a/src/include/executor/execExpr.h b/src/include/executor/execExpr.h
new file mode 100644
index 048573c..7580ab9
--- a/src/include/executor/execExpr.h
+++ b/src/include/executor/execExpr.h
@@ -241,6 +241,7 @@ typedef enum ExprEvalOp
 	EEOP_AGGREF,
 	EEOP_GROUPING_FUNC,
 	EEOP_WINDOW_FUNC,
+	EEOP_MERGING_FUNC,
 	EEOP_SUBPLAN,
 
 	/* aggregation related nodes */
@@ -623,6 +624,12 @@ typedef struct ExprEvalStep
 			WindowFuncExprState *wfstate;
 		}			window_func;
 
+		/* for EEOP_MERGING_FUNC */
+		struct
+		{
+			MergingFuncOp mfop; /* the MERGING() operation to perform */
+		}			merging_func;
+
 		/* for EEOP_SUBPLAN */
 		struct
 		{
@@ -806,6 +813,8 @@ extern void ExecEvalJsonConstructor(Expr
 									ExprContext *econtext);
 extern void ExecEvalJsonIsPredicate(ExprState *state, ExprEvalStep *op);
 extern void ExecEvalGroupingFunc(ExprState *state, ExprEvalStep *op);
+extern void ExecEvalMergingFunc(ExprState *state, ExprEvalStep *op,
+								ExprContext *econtext);
 extern void ExecEvalSubPlan(ExprState *state, ExprEvalStep *op,
 							ExprContext *econtext);
 extern void ExecEvalWholeRowVar(ExprState *state, ExprEvalStep *op,
diff --git a/src/include/executor/spi.h b/src/include/executor/spi.h
new file mode 100644
index d1de139..ed8787d
--- a/src/include/executor/spi.h
+++ b/src/include/executor/spi.h
@@ -97,6 +97,7 @@ typedef struct _SPI_plan *SPIPlanPtr;
 #define SPI_OK_REL_UNREGISTER	16
 #define SPI_OK_TD_REGISTER		17
 #define SPI_OK_MERGE			18
+#define SPI_OK_MERGE_RETURNING	19
 
 #define SPI_OPT_NONATOMIC		(1 << 0)
 
diff --git a/src/include/nodes/execnodes.h b/src/include/nodes/execnodes.h
new file mode 100644
index 5d7f17d..08bc9ee
--- a/src/include/nodes/execnodes.h
+++ b/src/include/nodes/execnodes.h
@@ -1318,6 +1318,9 @@ typedef struct ModifyTableState
 	/* Flags showing which subcommands are present INS/UPD/DEL/DO NOTHING */
 	int			mt_merge_subcommands;
 
+	/* For MERGE, the action currently being executed */
+	MergeActionState *mt_merge_action;
+
 	/* tuple counters for MERGE */
 	double		mt_merge_inserted;
 	double		mt_merge_updated;
diff --git a/src/include/nodes/parsenodes.h b/src/include/nodes/parsenodes.h
new file mode 100644
index e494309..67ec626
--- a/src/include/nodes/parsenodes.h
+++ b/src/include/nodes/parsenodes.h
@@ -1927,6 +1927,7 @@ typedef struct MergeStmt
 	Node	   *sourceRelation; /* source relation */
 	Node	   *joinCondition;	/* join condition between source and target */
 	List	   *mergeWhenClauses;	/* list of MergeWhenClause(es) */
+	List	   *returningList;	/* list of expressions to return */
 	WithClause *withClause;		/* WITH clause */
 } MergeStmt;
 
diff --git a/src/include/nodes/primnodes.h b/src/include/nodes/primnodes.h
new file mode 100644
index bb930af..00e0992
--- a/src/include/nodes/primnodes.h
+++ b/src/include/nodes/primnodes.h
@@ -572,6 +572,39 @@ typedef struct WindowFunc
 } WindowFunc;
 
 /*
+ * MergingFunc
+ *
+ * A MergingFunc is a MERGING(...) expression that can only appear in the
+ * RETURNING list of a MERGE command.  It returns information about the
+ * currently executing merge action.  Possible operations are:
+ *
+ *	MERGING_ACTION:
+ *		Return the command string of the current merge action ("INSERT",
+ *		"UPDATE" or "DELETE").
+ *
+ *	MERGING_CLAUSE_NUMBER:
+ *		Return the 1-based index of the current merge WHEN clause.
+ */
+typedef enum MergingFuncOp
+{
+	MERGING_ACTION,
+	MERGING_CLAUSE_NUMBER,
+} MergingFuncOp;
+
+typedef struct MergingFunc
+{
+	Expr		xpr;
+	/* operation to perform */
+	MergingFuncOp mfop;
+	/* type Oid of result */
+	Oid			mftype pg_node_attr(query_jumble_ignore);
+	/* OID of collation, or InvalidOid if none */
+	Oid			mfcollid pg_node_attr(query_jumble_ignore);
+	/* token location, or -1 if unknown */
+	int			location;
+} MergingFunc;
+
+/*
  * SubscriptingRef: describes a subscripting operation over a container
  * (array, etc).
  *
@@ -1735,6 +1768,7 @@ typedef struct BooleanTest
 typedef struct MergeAction
 {
 	NodeTag		type;
+	int			index;			/* 1-based index of the clause */
 	bool		matched;		/* true=MATCHED, false=NOT MATCHED */
 	CmdType		commandType;	/* INSERT/UPDATE/DELETE/DO NOTHING */
 	/* OVERRIDING clause */
diff --git a/src/include/optimizer/paramassign.h b/src/include/optimizer/paramassign.h
new file mode 100644
index 55c27a6..cd0e629
--- a/src/include/optimizer/paramassign.h
+++ b/src/include/optimizer/paramassign.h
@@ -20,6 +20,7 @@ extern Param *replace_outer_placeholderv
 										   PlaceHolderVar *phv);
 extern Param *replace_outer_agg(PlannerInfo *root, Aggref *agg);
 extern Param *replace_outer_grouping(PlannerInfo *root, GroupingFunc *grp);
+extern Param *replace_outer_merging(PlannerInfo *root, MergingFunc *mf);
 extern Param *replace_nestloop_param_var(PlannerInfo *root, Var *var);
 extern Param *replace_nestloop_param_placeholdervar(PlannerInfo *root,
 													PlaceHolderVar *phv);
diff --git a/src/include/parser/analyze.h b/src/include/parser/analyze.h
new file mode 100644
index c96483a..47d988b
--- a/src/include/parser/analyze.h
+++ b/src/include/parser/analyze.h
@@ -44,6 +44,8 @@ extern List *transformInsertRow(ParseSta
 								bool strip_indirection);
 extern List *transformUpdateTargetList(ParseState *pstate,
 									   List *origTlist);
+extern List *transformReturningList(ParseState *pstate, List *returningList,
+									ParseExprKind exprKind);
 extern Query *transformTopLevelStmt(ParseState *pstate, RawStmt *parseTree);
 extern Query *transformStmt(ParseState *pstate, Node *parseTree);
 
diff --git a/src/include/parser/kwlist.h b/src/include/parser/kwlist.h
new file mode 100644
index 5984dcf..c283a3f
--- a/src/include/parser/kwlist.h
+++ b/src/include/parser/kwlist.h
@@ -80,6 +80,7 @@ PG_KEYWORD("characteristics", CHARACTERI
 PG_KEYWORD("check", CHECK, RESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("checkpoint", CHECKPOINT, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("class", CLASS, UNRESERVED_KEYWORD, BARE_LABEL)
+PG_KEYWORD("clause_number", CLAUSE_NUMBER, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("close", CLOSE, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("cluster", CLUSTER, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("coalesce", COALESCE, COL_NAME_KEYWORD, BARE_LABEL)
@@ -266,6 +267,7 @@ PG_KEYWORD("matched", MATCHED, UNRESERVE
 PG_KEYWORD("materialized", MATERIALIZED, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("maxvalue", MAXVALUE, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("merge", MERGE, UNRESERVED_KEYWORD, BARE_LABEL)
+PG_KEYWORD("merging", MERGING, COL_NAME_KEYWORD, BARE_LABEL)
 PG_KEYWORD("method", METHOD, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("minute", MINUTE_P, UNRESERVED_KEYWORD, AS_LABEL)
 PG_KEYWORD("minvalue", MINVALUE, UNRESERVED_KEYWORD, BARE_LABEL)
diff --git a/src/include/parser/parse_node.h b/src/include/parser/parse_node.h
new file mode 100644
index f589112..836819d
--- a/src/include/parser/parse_node.h
+++ b/src/include/parser/parse_node.h
@@ -61,7 +61,8 @@ typedef enum ParseExprKind
 	EXPR_KIND_DISTINCT_ON,		/* DISTINCT ON */
 	EXPR_KIND_LIMIT,			/* LIMIT */
 	EXPR_KIND_OFFSET,			/* OFFSET */
-	EXPR_KIND_RETURNING,		/* RETURNING */
+	EXPR_KIND_RETURNING,		/* RETURNING in INSERT/UPDATE/DELETE */
+	EXPR_KIND_MERGE_RETURNING,	/* RETURNING in MERGE */
 	EXPR_KIND_VALUES,			/* VALUES */
 	EXPR_KIND_VALUES_SINGLE,	/* single-row VALUES (in INSERT only) */
 	EXPR_KIND_CHECK_CONSTRAINT, /* CHECK constraint for a table */
diff --git a/src/pl/plpgsql/src/pl_exec.c b/src/pl/plpgsql/src/pl_exec.c
new file mode 100644
index f8c7f48..e5d92f4
--- a/src/pl/plpgsql/src/pl_exec.c
+++ b/src/pl/plpgsql/src/pl_exec.c
@@ -4267,9 +4267,9 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 	/*
 	 * If we have INTO, then we only need one row back ... but if we have INTO
 	 * STRICT or extra check too_many_rows, ask for two rows, so that we can
-	 * verify the statement returns only one.  INSERT/UPDATE/DELETE are always
-	 * treated strictly. Without INTO, just run the statement to completion
-	 * (tcount = 0).
+	 * verify the statement returns only one.  INSERT/UPDATE/DELETE/MERGE are
+	 * always treated strictly. Without INTO, just run the statement to
+	 * completion (tcount = 0).
 	 *
 	 * We could just ask for two rows always when using INTO, but there are
 	 * some cases where demanding the extra row costs significant time, eg by
@@ -4307,10 +4307,11 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 			Assert(stmt->mod_stmt);
 			exec_set_found(estate, (SPI_processed != 0));
 			break;
@@ -4489,10 +4490,11 @@ exec_stmt_dynexecute(PLpgSQL_execstate *
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 		case SPI_OK_UTILITY:
 		case SPI_OK_REWRITTEN:
 			break;
diff --git a/src/pl/tcl/pltcl.c b/src/pl/tcl/pltcl.c
new file mode 100644
index e8f9d7b..0c9f81e
--- a/src/pl/tcl/pltcl.c
+++ b/src/pl/tcl/pltcl.c
@@ -2456,6 +2456,7 @@ pltcl_process_SPI_result(Tcl_Interp *int
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
+		case SPI_OK_MERGE_RETURNING:
 
 			/*
 			 * Process the tuples we got
diff --git a/src/test/regress/expected/merge.out b/src/test/regress/expected/merge.out
new file mode 100644
index 28a6d0b..183d160
--- a/src/test/regress/expected/merge.out
+++ b/src/test/regress/expected/merge.out
@@ -123,20 +123,20 @@ ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
 ERROR:  name "target" specified more than once
 DETAIL:  The name is used both as MERGE target table and data source.
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
-ERROR:  MERGE not supported in WITH query
-LINE 1: WITH foo AS (
-             ^
--- used in COPY
+ERROR:  WITH query "foo" does not have a RETURNING clause
+LINE 4: ) SELECT * FROM foo;
+                        ^
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) TO stdout;
-ERROR:  MERGE not supported in COPY
+ERROR:  COPY query must have a RETURNING clause
 -- unsupported relation types
 -- view
 CREATE VIEW tv AS SELECT * FROM target;
@@ -1256,7 +1256,7 @@ BEGIN;
 MERGE INTO sq_target
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1271,7 +1271,7 @@ INSERT INTO sq_source (sid, balance, del
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1295,7 +1295,7 @@ WITH targq AS (
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1304,20 +1304,191 @@ WHEN MATCHED AND tid < 2 THEN
 ROLLBACK;
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
+CREATE TABLE merge_actions(action text, abbrev text);
+INSERT INTO merge_actions VALUES ('INSERT', 'ins'), ('UPDATE', 'upd'), ('DELETE', 'del');
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
-ERROR:  syntax error at or near "RETURNING"
-LINE 10: RETURNING *;
-         ^
+    DELETE
+RETURNING MERGING(CLAUSE_NUMBER) AS clause,
+          (SELECT abbrev FROM merge_actions
+            WHERE action = MERGING(ACTION)) AS action,
+          t.*,
+          CASE MERGING(ACTION)
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ clause | action | tid | balance |     description     
+--------+--------+-----+---------+---------------------
+      3 | del    |   1 |     100 | Removed (1,100)
+      1 | upd    |   2 |     220 | Added 20 to balance
+      2 | ins    |   4 |      40 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- error when using MERGING() outside MERGE
+SELECT MERGING(ACTION) FROM sq_target;
+ERROR:  MERGING() can only be used in the RETURNING list of a MERGE command
+LINE 1: SELECT MERGING(ACTION) FROM sq_target;
+               ^
+UPDATE sq_target SET balance = balance + 1 RETURNING MERGING(CLAUSE_NUMBER);
+ERROR:  MERGING() can only be used in the RETURNING list of a MERGE command
+LINE 1: ...ATE sq_target SET balance = balance + 1 RETURNING MERGING(CL...
+                                                             ^
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(CLAUSE_NUMBER), MERGING(ACTION), t.*,
+              CASE MERGING(ACTION)
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING action, MERGING(ACTION) AS log_action, l.*
+)
+SELECT * FROM m2;
+ action | log_action | tid |     last_change     
+--------+------------+-----+---------------------
+ DELETE | UPDATE     |   1 | Removed (1,100)
+ UPDATE | INSERT     |   2 | Added 20 to balance
+ INSERT | INSERT     |   4 | Inserted (4,40)
+(3 rows)
+
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ tid |     last_change     
+-----+---------------------
+   1 | Removed (1,100)
+   2 | Added 20 to balance
+   4 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.*
+) TO stdout;
+DELETE	1	100
+UPDATE	2	220
+INSERT	4	40
+ROLLBACK;
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ action | tid | new_balance 
+--------+-----+-------------
+ DELETE |   1 |         100
+ UPDATE |   3 |         320
+ INSERT |   4 |         110
+(3 rows)
+
+ROLLBACK;
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ action | tid | balance 
+--------+-----+---------
+ DELETE |   1 |     100
+ UPDATE |   2 |     220
+ INSERT |   4 |      40
+(3 rows)
+
+ROLLBACK;
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ r_action | r_tid | r_balance 
+----------+-------+-----------
+ DELETE   |     1 |       100
+ UPDATE   |     3 |       320
+ INSERT   |     4 |       110
+(3 rows)
+
 ROLLBACK;
 -- EXPLAIN
 CREATE TABLE ex_mtarget (a int, b int)
@@ -1514,7 +1685,7 @@ SELECT * FROM sq_target WHERE tid = 1;
 (1 row)
 
 ROLLBACK;
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 NOTICE:  drop cascades to view v
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1640,6 +1811,32 @@ SELECT * FROM pa_target ORDER BY tid;
 (14 rows)
 
 ROLLBACK;
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING MERGING(ACTION), MERGING(CLAUSE_NUMBER), t.*;
+ action | clause_number | tid | balance |           val            
+--------+---------------+-----+---------+--------------------------
+ UPDATE |             1 |   2 |     110 | initial updated by merge
+(1 row)
+
+SELECT * FROM pa_target ORDER BY tid;
+ tid | balance |           val            
+-----+---------+--------------------------
+   2 |     110 | initial updated by merge
+   3 |     300 | initial
+   5 |     500 | initial
+   7 |     700 | initial
+   9 |     900 | initial
+  11 |    1100 | initial
+  13 |    1300 | initial
+(7 rows)
+
+ROLLBACK;
 DROP TABLE pa_target CASCADE;
 -- The target table is partitioned in the same way, but this time by attaching
 -- partitions which have columns in different order, dropped columns etc.
@@ -1731,7 +1928,26 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (sid, delta, 'inserted by merge');
+    INSERT VALUES (sid, delta, 'inserted by merge')
+  RETURNING MERGING(ACTION), MERGING(CLAUSE_NUMBER), t.*;
+ action | clause_number | tid | balance |           val            
+--------+---------------+-----+---------+--------------------------
+ UPDATE |             1 |   2 |     110 | initial updated by merge
+ INSERT |             2 |   2 |      20 | inserted by merge
+ UPDATE |             1 |   4 |     330 | initial updated by merge
+ INSERT |             2 |   4 |      40 | inserted by merge
+ UPDATE |             1 |   6 |     550 | initial updated by merge
+ INSERT |             2 |   6 |      60 | inserted by merge
+ UPDATE |             1 |   8 |     770 | initial updated by merge
+ INSERT |             2 |   8 |      80 | inserted by merge
+ UPDATE |             1 |  10 |     990 | initial updated by merge
+ INSERT |             2 |  10 |     100 | inserted by merge
+ UPDATE |             1 |  12 |    1210 | initial updated by merge
+ INSERT |             2 |  12 |     120 | inserted by merge
+ UPDATE |             1 |  14 |    1430 | initial updated by merge
+ INSERT |             2 |  14 |     140 | inserted by merge
+(14 rows)
+
 SELECT * FROM pa_target ORDER BY tid;
  tid | balance |           val            
 -----+---------+--------------------------
@@ -1798,7 +2014,21 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING MERGING(ACTION), MERGING(CLAUSE_NUMBER), t.*;
+ action | clause_number |          logts           | tid | balance |           val            
+--------+---------------+--------------------------+-----+---------+--------------------------
+ UPDATE |             1 | Tue Jan 31 00:00:00 2017 |   1 |     110 | initial updated by merge
+ UPDATE |             1 | Tue Feb 28 00:00:00 2017 |   2 |     220 | initial updated by merge
+ INSERT |             2 | Sun Jan 15 00:00:00 2017 |   3 |      30 | inserted by merge
+ UPDATE |             1 | Tue Jan 31 00:00:00 2017 |   4 |     440 | initial updated by merge
+ UPDATE |             1 | Tue Feb 28 00:00:00 2017 |   5 |     550 | initial updated by merge
+ INSERT |             2 | Sun Jan 15 00:00:00 2017 |   6 |      60 | inserted by merge
+ UPDATE |             1 | Tue Jan 31 00:00:00 2017 |   7 |     770 | initial updated by merge
+ UPDATE |             1 | Tue Feb 28 00:00:00 2017 |   8 |     880 | initial updated by merge
+ INSERT |             2 | Sun Jan 15 00:00:00 2017 |   9 |      90 | inserted by merge
+(9 rows)
+
 SELECT * FROM pa_target ORDER BY tid;
           logts           | tid | balance |           val            
 --------------------------+-----+---------+--------------------------
diff --git a/src/test/regress/expected/rowsecurity.out b/src/test/regress/expected/rowsecurity.out
new file mode 100644
index 6988128..3fd4fab
--- a/src/test/regress/expected/rowsecurity.out
+++ b/src/test/regress/expected/rowsecurity.out
@@ -2302,6 +2302,35 @@ WHEN MATCHED THEN
 	UPDATE SET dnotes = dnotes || ' notes added by merge8 '
 WHEN NOT MATCHED THEN
 	INSERT VALUES (13, 44, 1, 'regress_rls_bob', 'new manga');
+SELECT * FROM document WHERE did = 13;
+ did | cid | dlevel | dauthor | dtitle | dnotes 
+-----+-----+--------+---------+--------+--------
+(0 rows)
+
+-- but not OK if RETURNING is used
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge9 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 44, 1, 'regress_rls_bob', 'new manga')
+RETURNING *;
+ERROR:  new row violates row-level security policy for table "document"
+-- but OK if new row is visible
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge10 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 11, 1, 'regress_rls_bob', 'new novel')
+RETURNING *;
+ sdid | did | cid | dlevel |     dauthor     |  dtitle   | dnotes 
+------+-----+-----+--------+-----------------+-----------+--------
+   14 |  14 |  11 |      1 | regress_rls_bob | new novel | 
+(1 row)
+
 RESET SESSION AUTHORIZATION;
 -- drop the restrictive SELECT policy so that we can look at the
 -- final state of the table
@@ -2325,7 +2354,8 @@ SELECT * FROM document;
   12 |  11 |      1 | regress_rls_bob   | another novel                    | 
    1 |  11 |      1 | regress_rls_bob   | my first novel                   |  notes added by merge2  notes added by merge3  notes added by merge4  notes added by merge7 
   13 |  44 |      1 | regress_rls_bob   | new manga                        | 
-(15 rows)
+  14 |  11 |      1 | regress_rls_bob   | new novel                        | 
+(16 rows)
 
 --
 -- ROLE/GROUP
diff --git a/src/test/regress/expected/rules.out b/src/test/regress/expected/rules.out
new file mode 100644
index 1442c43..f187dec
--- a/src/test/regress/expected/rules.out
+++ b/src/test/regress/expected/rules.out
@@ -3596,7 +3596,8 @@ MERGE INTO rule_merge2 t USING (SELECT 1
 -- test deparsing
 CREATE TABLE sf_target(id int, data text, filling int[]);
 CREATE FUNCTION merge_sf_test()
- RETURNS void
+ RETURNS TABLE(action text, clause_number int,
+               a int, b text, id int, data text, filling int[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -3633,11 +3634,13 @@ WHEN NOT MATCHED
    VALUES (s.a, s.b, DEFAULT)
 WHEN NOT MATCHED
    THEN INSERT (filling[1], id)
-   VALUES (s.a, s.a);
+   VALUES (s.a, s.a)
+RETURNING
+   merging(action), merging(clause_number) AS clause_number, *;
 END;
 \sf merge_sf_test
 CREATE OR REPLACE FUNCTION public.merge_sf_test()
- RETURNS void
+ RETURNS TABLE(action text, clause_number integer, a integer, b text, id integer, data text, filling integer[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -3674,7 +3677,14 @@ BEGIN ATOMIC
       VALUES (s.a, s.b, DEFAULT)
     WHEN NOT MATCHED
      THEN INSERT (filling[1], id)
-      VALUES (s.a, s.a);
+      VALUES (s.a, s.a)
+   RETURNING MERGING(ACTION) AS action,
+     MERGING(CLAUSE_NUMBER) AS clause_number,
+     s.a,
+     s.b,
+     t.id,
+     t.data,
+     t.filling;
 END
 DROP FUNCTION merge_sf_test;
 DROP TABLE sf_target;
diff --git a/src/test/regress/sql/merge.sql b/src/test/regress/sql/merge.sql
new file mode 100644
index 4cf6db9..0eb2f97
--- a/src/test/regress/sql/merge.sql
+++ b/src/test/regress/sql/merge.sql
@@ -86,12 +86,12 @@ MERGE INTO target
 USING target
 ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
--- used in COPY
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
@@ -816,7 +816,7 @@ BEGIN;
 MERGE INTO sq_target
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -829,7 +829,7 @@ INSERT INTO sq_source (sid, balance, del
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -847,7 +847,7 @@ WITH targq AS (
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -857,17 +857,146 @@ ROLLBACK;
 
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
+CREATE TABLE merge_actions(action text, abbrev text);
+INSERT INTO merge_actions VALUES ('INSERT', 'ins'), ('UPDATE', 'upd'), ('DELETE', 'del');
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
+    DELETE
+RETURNING MERGING(CLAUSE_NUMBER) AS clause,
+          (SELECT abbrev FROM merge_actions
+            WHERE action = MERGING(ACTION)) AS action,
+          t.*,
+          CASE MERGING(ACTION)
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ROLLBACK;
+
+-- error when using MERGING() outside MERGE
+SELECT MERGING(ACTION) FROM sq_target;
+UPDATE sq_target SET balance = balance + 1 RETURNING MERGING(CLAUSE_NUMBER);
+
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(CLAUSE_NUMBER), MERGING(ACTION), t.*,
+              CASE MERGING(ACTION)
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING action, MERGING(ACTION) AS log_action, l.*
+)
+SELECT * FROM m2;
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ROLLBACK;
+
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.*
+) TO stdout;
+ROLLBACK;
+
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ROLLBACK;
+
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ROLLBACK;
+
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
 ROLLBACK;
 
 -- EXPLAIN
@@ -966,7 +1095,7 @@ WHEN MATCHED THEN
 SELECT * FROM sq_target WHERE tid = 1;
 ROLLBACK;
 
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1033,6 +1162,17 @@ SELECT merge_func();
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING MERGING(ACTION), MERGING(CLAUSE_NUMBER), t.*;
+SELECT * FROM pa_target ORDER BY tid;
+ROLLBACK;
+
 DROP TABLE pa_target CASCADE;
 
 -- The target table is partitioned in the same way, but this time by attaching
@@ -1091,7 +1231,8 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (sid, delta, 'inserted by merge');
+    INSERT VALUES (sid, delta, 'inserted by merge')
+  RETURNING MERGING(ACTION), MERGING(CLAUSE_NUMBER), t.*;
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
@@ -1145,7 +1286,8 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING MERGING(ACTION), MERGING(CLAUSE_NUMBER), t.*;
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
diff --git a/src/test/regress/sql/rowsecurity.sql b/src/test/regress/sql/rowsecurity.sql
new file mode 100644
index dec7340..1d5ed0a
--- a/src/test/regress/sql/rowsecurity.sql
+++ b/src/test/regress/sql/rowsecurity.sql
@@ -992,6 +992,27 @@ WHEN MATCHED THEN
 	UPDATE SET dnotes = dnotes || ' notes added by merge8 '
 WHEN NOT MATCHED THEN
 	INSERT VALUES (13, 44, 1, 'regress_rls_bob', 'new manga');
+SELECT * FROM document WHERE did = 13;
+
+-- but not OK if RETURNING is used
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge9 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 44, 1, 'regress_rls_bob', 'new manga')
+RETURNING *;
+
+-- but OK if new row is visible
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge10 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 11, 1, 'regress_rls_bob', 'new novel')
+RETURNING *;
 
 RESET SESSION AUTHORIZATION;
 -- drop the restrictive SELECT policy so that we can look at the
diff --git a/src/test/regress/sql/rules.sql b/src/test/regress/sql/rules.sql
new file mode 100644
index 8b7e255..d820b9b
--- a/src/test/regress/sql/rules.sql
+++ b/src/test/regress/sql/rules.sql
@@ -1281,7 +1281,8 @@ MERGE INTO rule_merge2 t USING (SELECT 1
 CREATE TABLE sf_target(id int, data text, filling int[]);
 
 CREATE FUNCTION merge_sf_test()
- RETURNS void
+ RETURNS TABLE(action text, clause_number int,
+               a int, b text, id int, data text, filling int[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -1318,7 +1319,9 @@ WHEN NOT MATCHED
    VALUES (s.a, s.b, DEFAULT)
 WHEN NOT MATCHED
    THEN INSERT (filling[1], id)
-   VALUES (s.a, s.a);
+   VALUES (s.a, s.a)
+RETURNING
+   merging(action), merging(clause_number) AS clause_number, *;
 END;
 
 \sf merge_sf_test
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
new file mode 100644
index bf50a32..0ca25c6
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -1561,6 +1561,8 @@ MergePath
 MergeScanSelCache
 MergeStmt
 MergeWhenClause
+MergingFunc
+MergingFuncOp
 MetaCommand
 MinMaxAggInfo
 MinMaxAggPath

diff --git a/doc/src/sgml/glossary.sgml b/doc/src/sgml/glossary.sgml
new file mode 100644
index fe8def4..5815fa4
--- a/doc/src/sgml/glossary.sgml
+++ b/doc/src/sgml/glossary.sgml
@@ -1097,8 +1097,8 @@
      (just like a <glossterm linkend="glossary-view">view</glossterm>),
      but stores data in the same way that a
      <glossterm linkend="glossary-table">table</glossterm> does. It cannot be
-     modified via <command>INSERT</command>, <command>UPDATE</command>, or
-     <command>DELETE</command> operations.
+     modified via <command>INSERT</command>, <command>UPDATE</command>,
+     <command>DELETE</command>, or <command>MERGE</command> operations.
     </para>
     <para>
      For more information, see
@@ -1106,6 +1106,23 @@
     </para>
    </glossdef>
   </glossentry>
+
+  <glossentry id="glossary-merge">
+   <glossterm>Merge</glossterm>
+   <glossdef>
+    <para>
+     An <acronym>SQL</acronym> command used to conditionally add, modify,
+     or remove <glossterm linkend="glossary-tuple">rows</glossterm>
+     in a given <glossterm linkend="glossary-table">table</glossterm>,
+     using data from a source
+     <glossterm linkend="glossary-relation">relation</glossterm>.
+    </para>
+    <para>
+     For more information, see
+     <xref linkend="sql-merge"/>.
+    </para>
+   </glossdef>
+  </glossentry>
 
   <glossentry id="glossary-mvcc">
    <glossterm>Multi-version concurrency control (MVCC)</glossterm>
diff --git a/doc/src/sgml/ref/merge.sgml b/doc/src/sgml/ref/merge.sgml
new file mode 100644
index 0f61d47..ab78510
--- a/doc/src/sgml/ref/merge.sgml
+++ b/doc/src/sgml/ref/merge.sgml
@@ -102,14 +102,14 @@ DELETE
    that are referred to in the <literal>SET</literal> clause.
    If you specify an insert action, you must have the <literal>INSERT</literal>
    privilege on the <replaceable class="parameter">target_table_name</replaceable>.
-   If you specify an delete action, you must have the <literal>DELETE</literal>
+   If you specify a delete action, you must have the <literal>DELETE</literal>
    privilege on the <replaceable class="parameter">target_table_name</replaceable>.
    Privileges are tested once at statement start and are checked
    whether or not particular <literal>WHEN</literal> clauses are executed.
-   You will require the <literal>SELECT</literal> privilege on the
-   <replaceable class="parameter">data_source</replaceable> and any column(s)
-   of the <replaceable class="parameter">target_table_name</replaceable>
-   referred to in a <literal>condition</literal>.
+   You will require the <literal>SELECT</literal> privilege and any column(s)
+   of the <replaceable class="parameter">data_source</replaceable> and
+   <replaceable class="parameter">target_table_name</replaceable> referred to
+   in any <literal>condition</literal> or <literal>expression</literal>.
   </para>
 
   <para>

diff --git a/doc/src/sgml/dml.sgml b/doc/src/sgml/dml.sgml
new file mode 100644
index cbbc5e2..3d95bdb
--- a/doc/src/sgml/dml.sgml
+++ b/doc/src/sgml/dml.sgml
@@ -283,10 +283,15 @@ DELETE FROM products;
    <secondary>RETURNING</secondary>
   </indexterm>
 
+  <indexterm zone="dml-returning">
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
   <para>
    Sometimes it is useful to obtain data from modified rows while they are
    being manipulated.  The <command>INSERT</command>, <command>UPDATE</command>,
-   and <command>DELETE</command> commands all have an
+   <command>DELETE</command>, and <command>MERGE</command> commands all have an
    optional <literal>RETURNING</literal> clause that supports this.  Use
    of <literal>RETURNING</literal> avoids performing an extra database query to
    collect the data, and is especially valuable when it would otherwise be
@@ -339,6 +344,21 @@ DELETE FROM products
 </programlisting>
   </para>
 
+  <para>
+   In a <command>MERGE</command>, the data available to <literal>RETURNING</literal> is
+   the content of the source row plus the content of the inserted, updated, or
+   deleted target row.  Since it is quite common for the source and target to
+   have many of the same columns, specifying <literal>RETURNING *</literal>
+   can lead to a lot of duplicated columns, so it is often more useful to
+   qualify it so as to return just the source or target row.  For example:
+<programlisting>
+MERGE INTO products p USING new_products n ON p.product_no = n.product_no
+  WHEN NOT MATCHED THEN INSERT VALUES (n.product_no, n.name, n.price)
+  WHEN MATCHED THEN UPDATE SET name = n.name, price = n.price
+  RETURNING p.*;
+</programlisting>
+  </para>
+
   <para>
    If there are triggers (<xref linkend="triggers"/>) on the target table,
    the data available to <literal>RETURNING</literal> is the row as modified by
diff --git a/doc/src/sgml/func.sgml b/doc/src/sgml/func.sgml
new file mode 100644
index 210c7c0..51db66f
--- a/doc/src/sgml/func.sgml
+++ b/doc/src/sgml/func.sgml
@@ -21998,6 +21998,84 @@ SELECT count(*) FROM sometable;
 
  </sect1>
 
+ <sect1 id="functions-merge-support">
+  <title>Merge Support Functions</title>
+
+  <indexterm>
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
+  <indexterm>
+   <primary>MERGING</primary>
+  </indexterm>
+
+  <para>
+   <productname>PostgreSQL</productname> includes one merge support function
+   that may be used in the <literal>RETURNING</literal> list of a
+   <xref linkend="sql-merge"/> command to return additional information about
+   the action taken for each row:
+<synopsis>
+<function id="function-merging">MERGING</function> ( <parameter>property</parameter> )
+</synopsis>
+   The <parameter>property</parameter> key word specifies what to return:
+
+   <variablelist>
+    <varlistentry>
+     <term><literal>ACTION</literal></term>
+     <listitem>
+      <para>
+       The merge action command executed for the current row.  This will be
+       <literal>'INSERT'</literal>, <literal>'UPDATE'</literal>, or
+       <literal>'DELETE'</literal>.
+      </para>
+     </listitem>
+    </varlistentry>
+
+    <varlistentry>
+     <term><literal>CLAUSE_NUMBER</literal></term>
+     <listitem>
+      <para>
+       The 1-based index of the <literal>WHEN</literal> clause executed for
+       the current row.
+      </para>
+     </listitem>
+    </varlistentry>
+   </variablelist>
+  </para>
+
+  <para>
+   The following example illustrates how this may be used to determine
+   which actions were performed for each row affected by the
+   <command>MERGE</command> command:
+<screen><![CDATA[
+MERGE INTO products p
+  USING stock s ON p.product_id = s.product_id
+  WHEN MATCHED AND s.quantity > 0 THEN
+    UPDATE SET in_stock = true, quantity = s.quantity
+  WHEN MATCHED THEN
+    UPDATE SET in_stock = false, quantity = 0
+  WHEN NOT MATCHED THEN
+    INSERT (product_id, in_stock, quantity)
+      VALUES (s.product_id, true, s.quantity)
+  RETURNING MERGING(ACTION), MERGING(CLAUSE_NUMBER), p.*;
+
+ action | clause_number | product_id | in_stock | quantity 
+--------+---------------+------------+----------+----------
+ UPDATE |             1 |       1001 | t        |       50
+ UPDATE |             2 |       1002 | f        |        0
+ INSERT |             3 |       1003 | t        |       10
+]]></screen>
+  </para>
+
+  <para>
+   Note that this function can only be used in the <literal>RETURNING</literal>
+   list of a <command>MERGE</command> command.  It is an error to use it in any
+   other part of a query.
+  </para>
+
+ </sect1>
+
  <sect1 id="functions-subquery">
   <title>Subquery Expressions</title>
 
diff --git a/doc/src/sgml/glossary.sgml b/doc/src/sgml/glossary.sgml
new file mode 100644
index 881026d..b609bd3
--- a/doc/src/sgml/glossary.sgml
+++ b/doc/src/sgml/glossary.sgml
@@ -1426,9 +1426,9 @@
      to a <glossterm linkend="glossary-client">client</glossterm> upon the
      completion of an <acronym>SQL</acronym> command, usually a
      <command>SELECT</command> but it can be an
-     <command>INSERT</command>, <command>UPDATE</command>, or
-     <command>DELETE</command> command if the <literal>RETURNING</literal>
-     clause is specified.
+     <command>INSERT</command>, <command>UPDATE</command>,
+     <command>DELETE</command>, or <command>MERGE</command> command if the
+     <literal>RETURNING</literal> clause is specified.
     </para>
     <para>
      The fact that a result set is a relation means that a query can be used
diff --git a/doc/src/sgml/plpgsql.sgml b/doc/src/sgml/plpgsql.sgml
new file mode 100644
index c2b9c6a..406834e
--- a/doc/src/sgml/plpgsql.sgml
+++ b/doc/src/sgml/plpgsql.sgml
@@ -1043,8 +1043,8 @@ INSERT INTO mytable VALUES (1,'one'), (2
     </para>
 
     <para>
-     If the command does return rows (for example <command>SELECT</command>,
-     or <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>
+     If the command does return rows (for example <command>SELECT</command>, or
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
      with <literal>RETURNING</literal>), there are two ways to proceed.
      When the command will return at most one row, or you only care about
      the first row of output, write the command as usual but add
@@ -1172,6 +1172,7 @@ SELECT <replaceable>select_expressions</
 INSERT ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 UPDATE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 DELETE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
+MERGE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 </synopsis>
 
      where <replaceable>target</replaceable> can be a record variable, a row
@@ -1182,8 +1183,8 @@ DELETE ... RETURNING <replaceable>expres
      <literal>INTO</literal> clause) just as described above,
      and the plan is cached in the same way.
      This works for <command>SELECT</command>,
-     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
-     <literal>RETURNING</literal>, and certain utility commands
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
+     with <literal>RETURNING</literal>, and certain utility commands
      that return row sets, such as <command>EXPLAIN</command>.
      Except for the <literal>INTO</literal> clause, the SQL command is the same
      as it would be written outside <application>PL/pgSQL</application>.
@@ -1259,7 +1260,7 @@ END;
     </para>
 
     <para>
-     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
+     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command> with
      <literal>RETURNING</literal>, <application>PL/pgSQL</application> reports
      an error for more than one returned row, even when
      <literal>STRICT</literal> is not specified.  This is because there
diff --git a/doc/src/sgml/queries.sgml b/doc/src/sgml/queries.sgml
new file mode 100644
index 648b283..372cce1
--- a/doc/src/sgml/queries.sgml
+++ b/doc/src/sgml/queries.sgml
@@ -2063,9 +2063,10 @@ SELECT <replaceable>select_list</replace
    Table Expressions or <acronym>CTE</acronym>s, can be thought of as defining
    temporary tables that exist just for one query.  Each auxiliary statement
    in a <literal>WITH</literal> clause can be a <command>SELECT</command>,
-   <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>; and the
+   <command>INSERT</command>, <command>UPDATE</command>, <command>DELETE</command>,
+   or <command>MERGE</command>; and the
    <literal>WITH</literal> clause itself is attached to a primary statement that can
-   be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
+   also be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
    <command>DELETE</command>, or <command>MERGE</command>.
   </para>
 
@@ -2599,8 +2600,8 @@ SELECT * FROM w AS w1 JOIN w AS w2 ON w1
    <title>Data-Modifying Statements in <literal>WITH</literal></title>
 
    <para>
-    You can use most data-modifying statements (<command>INSERT</command>,
-    <command>UPDATE</command>, or <command>DELETE</command>, but not
+    You can use data-modifying statements (<command>INSERT</command>,
+    <command>UPDATE</command>, <command>DELETE</command>, or
     <command>MERGE</command>) in <literal>WITH</literal>.  This
     allows you to perform several different operations in the same query.
     An example is:
diff --git a/doc/src/sgml/ref/copy.sgml b/doc/src/sgml/ref/copy.sgml
new file mode 100644
index 85881ca..49ce67c
--- a/doc/src/sgml/ref/copy.sgml
+++ b/doc/src/sgml/ref/copy.sgml
@@ -123,16 +123,19 @@ COPY { <replaceable class="parameter">ta
       A <link linkend="sql-select"><command>SELECT</command></link>,
       <link linkend="sql-values"><command>VALUES</command></link>,
       <link linkend="sql-insert"><command>INSERT</command></link>,
-      <link linkend="sql-update"><command>UPDATE</command></link>, or
-      <link linkend="sql-delete"><command>DELETE</command></link> command whose results are to be
-      copied.  Note that parentheses are required around the query.
+      <link linkend="sql-update"><command>UPDATE</command></link>,
+      <link linkend="sql-delete"><command>DELETE</command></link>, or
+      <link linkend="sql-merge"><command>MERGE</command></link> command
+      whose results are to be copied.  Note that parentheses are required
+      around the query.
      </para>
      <para>
-      For <command>INSERT</command>, <command>UPDATE</command> and
-      <command>DELETE</command> queries a <literal>RETURNING</literal> clause
-      must be provided, and the target relation must not have a conditional
-      rule, nor an <literal>ALSO</literal> rule, nor an
-      <literal>INSTEAD</literal> rule that expands to multiple statements.
+      For <command>INSERT</command>, <command>UPDATE</command>,
+      <command>DELETE</command>, and <command>MERGE</command> queries a
+      <literal>RETURNING</literal> clause must be provided, and the target
+      relation must not have a conditional rule, nor an
+      <literal>ALSO</literal> rule, nor an <literal>INSTEAD</literal> rule
+      that expands to multiple statements.
      </para>
     </listitem>
    </varlistentry>
diff --git a/doc/src/sgml/ref/merge.sgml b/doc/src/sgml/ref/merge.sgml
new file mode 100644
index 655f7dc..44b3fa4
--- a/doc/src/sgml/ref/merge.sgml
+++ b/doc/src/sgml/ref/merge.sgml
@@ -25,6 +25,7 @@ PostgreSQL documentation
 MERGE INTO [ ONLY ] <replaceable class="parameter">target_table_name</replaceable> [ * ] [ [ AS ] <replaceable class="parameter">target_alias</replaceable> ]
 USING <replaceable class="parameter">data_source</replaceable> ON <replaceable class="parameter">join_condition</replaceable>
 <replaceable class="parameter">when_clause</replaceable> [...]
+[ RETURNING * | <replaceable class="parameter">output_expression</replaceable> [ [ AS ] <replaceable class="parameter">output_name</replaceable> ] [, ...] ]
 
 <phrase>where <replaceable class="parameter">data_source</replaceable> is:</phrase>
 
@@ -95,6 +96,18 @@ DELETE
   </para>
 
   <para>
+   The optional <literal>RETURNING</literal> clause causes <command>MERGE</command>
+   to compute and return value(s) based on each row inserted, updated, or
+   deleted.  Any expression using the source or target table's columns, or
+   the <xref linkend="function-merging"/> function can be computed.  When an
+   <command>INSERT</command> or <command>UPDATE</command> action is performed,
+   the new values of the target table's columns are used.  When a
+   <command>DELETE</command> is performed, the old values of the target table's
+   columns are used.  The syntax of the <literal>RETURNING</literal> list is
+   identical to that of the output list of <command>SELECT</command>.
+  </para>
+
+  <para>
    There is no separate <literal>MERGE</literal> privilege.
    If you specify an update action, you must have the
    <literal>UPDATE</literal> privilege on the column(s)
@@ -125,6 +138,18 @@ DELETE
 
   <variablelist>
    <varlistentry>
+    <term><replaceable class="parameter">with_query</replaceable></term>
+    <listitem>
+     <para>
+      The <literal>WITH</literal> clause allows you to specify one or more
+      subqueries that can be referenced by name in the <command>MERGE</command>
+      query. See <xref linkend="queries-with"/> and <xref linkend="sql-select"/>
+      for details.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
     <term><replaceable class="parameter">target_table_name</replaceable></term>
     <listitem>
      <para>
@@ -399,13 +424,30 @@ DELETE
    </varlistentry>
 
    <varlistentry>
-    <term><replaceable class="parameter">with_query</replaceable></term>
+    <term><replaceable class="parameter">output_expression</replaceable></term>
     <listitem>
      <para>
-      The <literal>WITH</literal> clause allows you to specify one or more
-      subqueries that can be referenced by name in the <command>MERGE</command>
-      query. See <xref linkend="queries-with"/> and <xref linkend="sql-select"/>
-      for details.
+      An expression to be computed and returned by the <command>MERGE</command>
+      command after each row is merged.  The expression can use any columns of
+      the source or target tables, or the <xref linkend="function-merging"/>
+      function to return additional information about the action executed.
+     </para>
+     <para>
+      Writing <literal>*</literal> will return all columns from the source
+      table, followed by all columns from the target table.  Often this will
+      lead to a lot of duplication, since it is common for the source and
+      target tables to have a lot of the same columns.  This can be avoided by
+      qualifying the <literal>*</literal> with the name or alias of the source
+      or target table.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
+    <term><replaceable class="parameter">output_name</replaceable></term>
+    <listitem>
+     <para>
+      A name to use for a returned column.
      </para>
     </listitem>
    </varlistentry>
@@ -428,6 +470,13 @@ MERGE <replaceable class="parameter">tot
    were changed in any way.
   </para>
 
+  <para>
+   If the <command>MERGE</command> command contains a <literal>RETURNING</literal>
+   clause, the result will be similar to that of a <command>SELECT</command>
+   statement containing the columns and values defined in the
+   <literal>RETURNING</literal> list, computed over the row(s) inserted, updated,
+   or deleted by the command.
+  </para>
  </refsect1>
 
  <refsect1>
@@ -544,13 +593,6 @@ MERGE <replaceable class="parameter">tot
   </para>
 
   <para>
-   There is no <literal>RETURNING</literal> clause with
-   <command>MERGE</command>.  Actions of <command>INSERT</command>,
-   <command>UPDATE</command> and <command>DELETE</command> cannot contain
-   <literal>RETURNING</literal> or <literal>WITH</literal> clauses.
-  </para>
-
-  <para>
    When <command>MERGE</command> is run concurrently with other commands
    that modify the target table, the usual transaction isolation rules
    apply; see <xref linkend="transaction-iso"/> for an explanation
@@ -602,7 +644,8 @@ WHEN NOT MATCHED THEN
   <para>
    Attempt to insert a new stock item along with the quantity of stock. If
    the item already exists, instead update the stock count of the existing
-   item. Don't allow entries that have zero stock.
+   item. Don't allow entries that have zero stock. Return details of all
+   changes made.
 <programlisting>
 MERGE INTO wines w
 USING wine_stock_changes s
@@ -612,7 +655,8 @@ WHEN NOT MATCHED AND s.stock_delta > 0 T
 WHEN MATCHED AND w.stock + s.stock_delta > 0 THEN
   UPDATE SET stock = w.stock + s.stock_delta
 WHEN MATCHED THEN
-  DELETE;
+  DELETE
+RETURNING MERGING(ACTION), w.*;
 </programlisting>
 
    The <literal>wine_stock_changes</literal> table might be, for example, a
@@ -627,8 +671,9 @@ WHEN MATCHED THEN
     This command conforms to the <acronym>SQL</acronym> standard.
   </para>
    <para>
-    The <literal>WITH</literal> clause and <literal>DO NOTHING</literal>
-    action are extensions to the <acronym>SQL</acronym> standard.
+    The <literal>WITH</literal> clause, <literal>DO NOTHING</literal> action,
+    and <literal>RETURNING</literal> clause are extensions to the
+    <acronym>SQL</acronym> standard.
   </para>
  </refsect1>
 </refentry>
diff --git a/doc/src/sgml/ref/select.sgml b/doc/src/sgml/ref/select.sgml
new file mode 100644
index 9917df7..066aed4
--- a/doc/src/sgml/ref/select.sgml
+++ b/doc/src/sgml/ref/select.sgml
@@ -74,7 +74,7 @@ SELECT [ ALL | DISTINCT [ ON ( <replacea
 
 <phrase>and <replaceable class="parameter">with_query</replaceable> is:</phrase>
 
-    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> )
+    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> | <replaceable class="parameter">merge</replaceable> )
         [ SEARCH { BREADTH | DEPTH } FIRST BY <replaceable>column_name</replaceable> [, ...] SET <replaceable>search_seq_col_name</replaceable> ]
         [ CYCLE <replaceable>column_name</replaceable> [, ...] SET <replaceable>cycle_mark_col_name</replaceable> [ TO <replaceable>cycle_mark_value</replaceable> DEFAULT <replaceable>cycle_mark_default</replaceable> ] USING <replaceable>cycle_path_col_name</replaceable> ]
 
@@ -230,10 +230,10 @@ TABLE [ ONLY ] <replaceable class="param
     The subqueries effectively act as temporary tables or views
     for the duration of the primary query.
     Each subquery can be a <command>SELECT</command>, <command>TABLE</command>, <command>VALUES</command>,
-    <command>INSERT</command>, <command>UPDATE</command> or
-    <command>DELETE</command> statement.
+    <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command> statement.
     When writing a data-modifying statement (<command>INSERT</command>,
-    <command>UPDATE</command> or <command>DELETE</command>) in
+    <command>UPDATE</command>, <command>DELETE</command>, or <command>MERGE</command>) in
     <literal>WITH</literal>, it is usual to include a <literal>RETURNING</literal> clause.
     It is the output of <literal>RETURNING</literal>, <emphasis>not</emphasis> the underlying
     table that the statement modifies, that forms the temporary table that is
@@ -2184,7 +2184,8 @@ SELECT 2+2;
 
    <para>
     <productname>PostgreSQL</productname> allows <command>INSERT</command>,
-    <command>UPDATE</command>, and <command>DELETE</command> to be used as <literal>WITH</literal>
+    <command>UPDATE</command>, <command>DELETE</command>, and
+    <command>MERGE</command> to be used as <literal>WITH</literal>
     queries.  This is not found in the SQL standard.
    </para>
   </refsect2>
diff --git a/doc/src/sgml/rowtypes.sgml b/doc/src/sgml/rowtypes.sgml
new file mode 100644
index 4d86f97..bbeac84
--- a/doc/src/sgml/rowtypes.sgml
+++ b/doc/src/sgml/rowtypes.sgml
@@ -348,7 +348,7 @@ SELECT m.* FROM some_table, LATERAL myfu
    column expansion of this kind when it appears at the top level of
    a <link linkend="queries-select-lists"><command>SELECT</command> output
    list</link>, a <link linkend="dml-returning"><literal>RETURNING</literal>
-   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>,
+   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>,
    a <link linkend="queries-values"><literal>VALUES</literal> clause</link>, or
    a <link linkend="sql-syntax-row-constructors">row constructor</link>.
    In all other contexts (including when nested inside one of those
diff --git a/doc/src/sgml/spi.sgml b/doc/src/sgml/spi.sgml
new file mode 100644
index 47f4b5b..bb37786
--- a/doc/src/sgml/spi.sgml
+++ b/doc/src/sgml/spi.sgml
@@ -301,8 +301,9 @@ SPI_execute("INSERT INTO foo SELECT * FR
    is returned in the global variable <varname>SPI_processed</varname>.
    If the return value of the function is <symbol>SPI_OK_SELECT</symbol>,
    <symbol>SPI_OK_INSERT_RETURNING</symbol>,
-   <symbol>SPI_OK_DELETE_RETURNING</symbol>, or
-   <symbol>SPI_OK_UPDATE_RETURNING</symbol>,
+   <symbol>SPI_OK_DELETE_RETURNING</symbol>,
+   <symbol>SPI_OK_UPDATE_RETURNING</symbol>, or
+   <symbol>SPI_OK_MERGE_RETURNING</symbol>,
    then you can use the
    global pointer <literal>SPITupleTable *SPI_tuptable</literal> to
    access the result rows.  Some utility commands (such as
@@ -472,6 +473,15 @@ typedef struct SPITupleTable
       </para>
      </listitem>
     </varlistentry>
+
+    <varlistentry>
+     <term><symbol>SPI_OK_MERGE_RETURNING</symbol></term>
+     <listitem>
+      <para>
+       if a <command>MERGE RETURNING</command> was executed
+      </para>
+     </listitem>
+    </varlistentry>
 
     <varlistentry>
      <term><symbol>SPI_OK_UTILITY</symbol></term>
diff --git a/doc/src/sgml/xfunc.sgml b/doc/src/sgml/xfunc.sgml
new file mode 100644
index 89116ae..cb0aa55
--- a/doc/src/sgml/xfunc.sgml
+++ b/doc/src/sgml/xfunc.sgml
@@ -177,7 +177,8 @@
     statements separated by semicolons.  A semicolon after the last
     statement is optional.  Unless the function is declared to return
     <type>void</type>, the last statement must be a <command>SELECT</command>,
-    or an <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>
+    or an <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command>
     that has a <literal>RETURNING</literal> clause.
    </para>
 
@@ -1236,8 +1237,9 @@ SELECT x, CASE WHEN x &gt; 0 THEN genera
 
     <note>
      <para>
-      If a function's last command is <command>INSERT</command>, <command>UPDATE</command>,
-      or <command>DELETE</command> with <literal>RETURNING</literal>, that command will
+      If a function's last command is <command>INSERT</command>,
+      <command>UPDATE</command>, <command>DELETE</command>, or
+      <command>MERGE</command> with <literal>RETURNING</literal>, that command will
       always be executed to completion, even if the function is not declared
       with <literal>SETOF</literal> or the calling query does not fetch all the
       result rows.  Any extra rows produced by the <literal>RETURNING</literal>
diff --git a/src/backend/commands/copy.c b/src/backend/commands/copy.c
new file mode 100644
index c36d7f1..7bcf923
--- a/src/backend/commands/copy.c
+++ b/src/backend/commands/copy.c
@@ -283,12 +283,6 @@ DoCopy(ParseState *pstate, const CopyStm
 	{
 		Assert(stmt->query);
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(stmt->query, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in COPY"));
-
 		query = makeNode(RawStmt);
 		query->stmt = stmt->query;
 		query->stmt_location = stmt_location;
diff --git a/src/backend/commands/copyto.c b/src/backend/commands/copyto.c
new file mode 100644
index d3dc3fc..c29fecd
--- a/src/backend/commands/copyto.c
+++ b/src/backend/commands/copyto.c
@@ -510,7 +510,8 @@ BeginCopyTo(ParseState *pstate,
 		{
 			Assert(query->commandType == CMD_INSERT ||
 				   query->commandType == CMD_UPDATE ||
-				   query->commandType == CMD_DELETE);
+				   query->commandType == CMD_DELETE ||
+				   query->commandType == CMD_MERGE);
 
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
diff --git a/src/backend/executor/execExpr.c b/src/backend/executor/execExpr.c
new file mode 100644
index 91df200..7744be8
--- a/src/backend/executor/execExpr.c
+++ b/src/backend/executor/execExpr.c
@@ -1108,6 +1108,21 @@ ExecInitExprRec(Expr *node, ExprState *s
 				break;
 			}
 
+		case T_MergingFunc:
+			{
+				MergingFunc *mf_node = (MergingFunc *) node;
+
+				if (!state->parent ||
+					!IsA(state->parent, ModifyTableState) ||
+					((ModifyTableState *) state->parent)->operation != CMD_MERGE)
+					elog(ERROR, "MergingFunc found in non-merge plan node");
+
+				scratch.opcode = EEOP_MERGING_FUNC;
+				scratch.d.merging_func.mfop = mf_node->mfop;
+				ExprEvalPushStep(state, &scratch);
+				break;
+			}
+
 		case T_SubscriptingRef:
 			{
 				SubscriptingRef *sbsref = (SubscriptingRef *) node;
diff --git a/src/backend/executor/execExprInterp.c b/src/backend/executor/execExprInterp.c
new file mode 100644
index 3c17cc6..11ceeab
--- a/src/backend/executor/execExprInterp.c
+++ b/src/backend/executor/execExprInterp.c
@@ -483,6 +483,7 @@ ExecInterpExpr(ExprState *state, ExprCon
 		&&CASE_EEOP_AGGREF,
 		&&CASE_EEOP_GROUPING_FUNC,
 		&&CASE_EEOP_WINDOW_FUNC,
+		&&CASE_EEOP_MERGING_FUNC,
 		&&CASE_EEOP_SUBPLAN,
 		&&CASE_EEOP_AGG_STRICT_DESERIALIZE,
 		&&CASE_EEOP_AGG_DESERIALIZE,
@@ -1582,6 +1583,14 @@ ExecInterpExpr(ExprState *state, ExprCon
 			EEO_NEXT();
 		}
 
+		EEO_CASE(EEOP_MERGING_FUNC)
+		{
+			/* too complex/uncommon for an inline implementation */
+			ExecEvalMergingFunc(state, op, econtext);
+
+			EEO_NEXT();
+		}
+
 		EEO_CASE(EEOP_SUBPLAN)
 		{
 			/* too complex for an inline implementation */
@@ -4171,6 +4180,59 @@ ExecEvalGroupingFunc(ExprState *state, E
 }
 
 /*
+ * ExecEvalMergingFunc
+ *
+ * Returns information about the current MERGE action for its RETURNING list.
+ */
+void
+ExecEvalMergingFunc(ExprState *state, ExprEvalStep *op, ExprContext *econtext)
+{
+	MergingFuncOp mfop = op->d.merging_func.mfop;
+	ModifyTableState *mtstate = castNode(ModifyTableState, state->parent);
+	MergeActionState *relaction = mtstate->mt_merge_action;
+
+	if (!relaction)
+		elog(ERROR, "no merge action in progress");
+
+	switch (mfop)
+	{
+		case MERGING_ACTION:
+			/* Return the MERGE action ("INSERT", "UPDATE", or "DELETE") */
+			switch (relaction->mas_action->commandType)
+			{
+				case CMD_INSERT:
+					*op->resvalue = PointerGetDatum(cstring_to_text_with_len("INSERT", 6));
+					*op->resnull = false;
+					break;
+				case CMD_UPDATE:
+					*op->resvalue = PointerGetDatum(cstring_to_text_with_len("UPDATE", 6));
+					*op->resnull = false;
+					break;
+				case CMD_DELETE:
+					*op->resvalue = PointerGetDatum(cstring_to_text_with_len("DELETE", 6));
+					*op->resnull = false;
+					break;
+				case CMD_NOTHING:
+					elog(ERROR, "unexpected merge action: DO NOTHING");
+					break;
+				default:
+					elog(ERROR, "unrecognized commandType: %d",
+						 (int) relaction->mas_action->commandType);
+			}
+			break;
+
+		case MERGING_CLAUSE_NUMBER:
+			/* Return the 1-based index of the MERGE action */
+			*op->resvalue = Int32GetDatum(relaction->mas_action->index);
+			*op->resnull = false;
+			break;
+
+		default:
+			elog(ERROR, "unrecognized MergingFuncOp: %d", (int) mfop);
+	}
+}
+
+/*
  * Hand off evaluation of a subplan to nodeSubplan.c
  */
 void
diff --git a/src/backend/executor/execPartition.c b/src/backend/executor/execPartition.c
new file mode 100644
index b22040a..9c56a10
--- a/src/backend/executor/execPartition.c
+++ b/src/backend/executor/execPartition.c
@@ -612,8 +612,8 @@ ExecInitPartitionInfo(ModifyTableState *
 	 * Build the RETURNING projection for the partition.  Note that we didn't
 	 * build the returningList for partitions within the planner, but simple
 	 * translation of varattnos will suffice.  This only occurs for the INSERT
-	 * case or in the case of UPDATE tuple routing where we didn't find a
-	 * result rel to reuse.
+	 * case or in the case of UPDATE/MERGE tuple routing where we didn't find
+	 * a result rel to reuse.
 	 */
 	if (node && node->returningLists != NIL)
 	{
@@ -622,12 +622,14 @@ ExecInitPartitionInfo(ModifyTableState *
 		List	   *returningList;
 
 		/* See the comment above for WCO lists. */
-		/* (except no RETURNING support for MERGE yet) */
 		Assert((node->operation == CMD_INSERT &&
 				list_length(node->returningLists) == 1 &&
 				list_length(node->resultRelations) == 1) ||
 			   (node->operation == CMD_UPDATE &&
 				list_length(node->returningLists) ==
+				list_length(node->resultRelations)) ||
+			   (node->operation == CMD_MERGE &&
+				list_length(node->returningLists) ==
 				list_length(node->resultRelations)));
 
 		/*
diff --git a/src/backend/executor/functions.c b/src/backend/executor/functions.c
new file mode 100644
index 0f811fd..7da02b0
--- a/src/backend/executor/functions.c
+++ b/src/backend/executor/functions.c
@@ -1660,8 +1660,8 @@ check_sql_fn_retval(List *queryTreeLists
 
 	/*
 	 * If it's a plain SELECT, it returns whatever the targetlist says.
-	 * Otherwise, if it's INSERT/UPDATE/DELETE with RETURNING, it returns
-	 * that. Otherwise, the function return type must be VOID.
+	 * Otherwise, if it's INSERT/UPDATE/DELETE/MERGE with RETURNING, it
+	 * returns that. Otherwise, the function return type must be VOID.
 	 *
 	 * Note: eventually replace this test with QueryReturnsTuples?	We'd need
 	 * a more general method of determining the output type, though.  Also, it
@@ -1679,7 +1679,8 @@ check_sql_fn_retval(List *queryTreeLists
 	else if (parse &&
 			 (parse->commandType == CMD_INSERT ||
 			  parse->commandType == CMD_UPDATE ||
-			  parse->commandType == CMD_DELETE) &&
+			  parse->commandType == CMD_DELETE ||
+			  parse->commandType == CMD_MERGE) &&
 			 parse->returningList)
 	{
 		tlist = parse->returningList;
@@ -1693,7 +1694,7 @@ check_sql_fn_retval(List *queryTreeLists
 				(errcode(ERRCODE_INVALID_FUNCTION_DEFINITION),
 				 errmsg("return type mismatch in function declared to return %s",
 						format_type_be(rettype)),
-				 errdetail("Function's final statement must be SELECT or INSERT/UPDATE/DELETE RETURNING.")));
+				 errdetail("Function's final statement must be SELECT or INSERT/UPDATE/DELETE/MERGE RETURNING.")));
 		return false;			/* keep compiler quiet */
 	}
 
diff --git a/src/backend/executor/nodeModifyTable.c b/src/backend/executor/nodeModifyTable.c
new file mode 100644
index 9fc5abf..01ef113
--- a/src/backend/executor/nodeModifyTable.c
+++ b/src/backend/executor/nodeModifyTable.c
@@ -36,8 +36,7 @@
  *		RETURNING tuple after completing each row insert, update, or delete.
  *		It must be called again to continue the operation.  Without RETURNING,
  *		we just loop within the node until all the work is done, then
- *		return NULL.  This avoids useless call/return overhead.  (MERGE does
- *		not support RETURNING.)
+ *		return NULL.  This avoids useless call/return overhead.
  */
 
 #include "postgres.h"
@@ -88,9 +87,6 @@ typedef struct ModifyTableContext
 	 */
 	TupleTableSlot *planSlot;
 
-	/* MERGE specific */
-	MergeActionState *relaction;	/* MERGE action in progress */
-
 	/*
 	 * Information about the changes that were made concurrently to a tuple
 	 * being updated or deleted
@@ -153,13 +149,14 @@ static TupleTableSlot *ExecMerge(ModifyT
 								 ItemPointer tupleid,
 								 bool canSetTag);
 static void ExecInitMerge(ModifyTableState *mtstate, EState *estate);
-static bool ExecMergeMatched(ModifyTableContext *context,
-							 ResultRelInfo *resultRelInfo,
-							 ItemPointer tupleid,
-							 bool canSetTag);
-static void ExecMergeNotMatched(ModifyTableContext *context,
-								ResultRelInfo *resultRelInfo,
-								bool canSetTag);
+static TupleTableSlot *ExecMergeMatched(ModifyTableContext *context,
+										ResultRelInfo *resultRelInfo,
+										ItemPointer tupleid,
+										bool canSetTag,
+										bool *matched);
+static TupleTableSlot *ExecMergeNotMatched(ModifyTableContext *context,
+										   ResultRelInfo *resultRelInfo,
+										   bool canSetTag);
 
 
 /*
@@ -979,7 +976,7 @@ ExecInsert(ModifyTableContext *context,
 		if (mtstate->operation == CMD_UPDATE)
 			wco_kind = WCO_RLS_UPDATE_CHECK;
 		else if (mtstate->operation == CMD_MERGE)
-			wco_kind = (context->relaction->mas_action->commandType == CMD_UPDATE) ?
+			wco_kind = (mtstate->mt_merge_action->mas_action->commandType == CMD_UPDATE) ?
 				WCO_RLS_UPDATE_CHECK : WCO_RLS_INSERT_CHECK;
 		else
 			wco_kind = WCO_RLS_INSERT_CHECK;
@@ -1833,7 +1830,7 @@ ExecCrossPartitionUpdate(ModifyTableCont
 		 * additional rechecking, and might end up executing a different
 		 * action entirely).
 		 */
-		if (context->relaction != NULL)
+		if (mtstate->mt_merge_action != NULL)
 			return *tmresult == TM_Ok;
 		else if (TupIsNull(epqslot))
 			return true;
@@ -2075,7 +2072,7 @@ lreplace:
 		 * No luck, a retry is needed.  If running MERGE, we do not do so
 		 * here; instead let it handle that on its own rules.
 		 */
-		if (context->relaction != NULL)
+		if (context->mtstate->mt_merge_action != NULL)
 			return result;
 
 		/*
@@ -2718,6 +2715,7 @@ static TupleTableSlot *
 ExecMerge(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 		  ItemPointer tupleid, bool canSetTag)
 {
+	TupleTableSlot *rslot = NULL;
 	bool		matched;
 
 	/*-----
@@ -2765,18 +2763,18 @@ ExecMerge(ModifyTableContext *context, R
 	 */
 	matched = tupleid != NULL;
 	if (matched)
-		matched = ExecMergeMatched(context, resultRelInfo, tupleid, canSetTag);
+		rslot = ExecMergeMatched(context, resultRelInfo, tupleid, canSetTag,
+								 &matched);
 
 	/*
-	 * Either we were dealing with a NOT MATCHED tuple or ExecMergeMatched()
-	 * returned "false", indicating the previously MATCHED tuple no longer
-	 * matches.
+	 * Deal with the NOT MATCHED case (either a NOT MATCHED tuple from the
+	 * join, or a previously MATCHED tuple for which ExecMergeMatched() set
+	 * "matched" to false, indicating that it no longer matches).
 	 */
 	if (!matched)
-		ExecMergeNotMatched(context, resultRelInfo, canSetTag);
+		rslot = ExecMergeNotMatched(context, resultRelInfo, canSetTag);
 
-	/* No RETURNING support yet */
-	return NULL;
+	return rslot;
 }
 
 /*
@@ -2786,8 +2784,8 @@ ExecMerge(ModifyTableContext *context, R
  * We start from the first WHEN MATCHED action and check if the WHEN quals
  * pass, if any. If the WHEN quals for the first action do not pass, we
  * check the second, then the third and so on. If we reach to the end, no
- * action is taken and we return true, indicating that no further action is
- * required for this tuple.
+ * action is taken and "matched" is set to true, indicating that no further
+ * action is required for this tuple.
  *
  * If we do find a qualifying action, then we attempt to execute the action.
  *
@@ -2796,16 +2794,17 @@ ExecMerge(ModifyTableContext *context, R
  * with individual actions are evaluated by this routine via ExecQual, while
  * EvalPlanQual checks for the join quals. If EvalPlanQual tells us that the
  * updated tuple still passes the join quals, then we restart from the first
- * action to look for a qualifying action. Otherwise, we return false --
- * meaning that a NOT MATCHED action must now be executed for the current
- * source tuple.
+ * action to look for a qualifying action. Otherwise, "matched" is set to
+ * false -- meaning that a NOT MATCHED action must now be executed for the
+ * current source tuple.
  */
-static bool
+static TupleTableSlot *
 ExecMergeMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
-				 ItemPointer tupleid, bool canSetTag)
+				 ItemPointer tupleid, bool canSetTag, bool *matched)
 {
 	ModifyTableState *mtstate = context->mtstate;
-	TupleTableSlot *newslot;
+	TupleTableSlot *rslot = NULL;
+	TupleTableSlot *newslot = NULL;
 	EState	   *estate = context->estate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
 	bool		isNull;
@@ -2816,7 +2815,10 @@ ExecMergeMatched(ModifyTableContext *con
 	 * If there are no WHEN MATCHED actions, we are done.
 	 */
 	if (resultRelInfo->ri_matchedMergeAction == NIL)
-		return true;
+	{
+		*matched = true;
+		return NULL;
+	}
 
 	/*
 	 * Make tuple and any needed join variables available to ExecQual and
@@ -2895,12 +2897,15 @@ lmerge_matched:
 				 */
 				newslot = ExecProject(relaction->mas_proj);
 
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				if (!ExecUpdatePrologue(context, resultRelInfo,
 										tupleid, NULL, newslot, &result))
 				{
 					if (result == TM_Ok)
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 					break;		/* concurrent update/delete */
 				}
 				result = ExecUpdateAct(context, resultRelInfo, tupleid, NULL,
@@ -2911,12 +2916,13 @@ lmerge_matched:
 				 * cross-partition update was done, then there's nothing else
 				 * for us to do --- the UPDATE has been turned into a DELETE
 				 * and an INSERT, and we must not perform any of the usual
-				 * post-update tasks.
+				 * post-update tasks.  Also, the RETURNING tuple (if any) has
+				 * been projected, so we can just return that.
 				 */
 				if (updateCxt.crossPartUpdate)
 				{
 					mtstate->mt_merge_updated += 1;
-					return true;
+					return context->cpUpdateReturningSlot;
 				}
 
 				if (result == TM_Ok && updateCxt.updated)
@@ -2928,12 +2934,15 @@ lmerge_matched:
 				break;
 
 			case CMD_DELETE:
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				if (!ExecDeletePrologue(context, resultRelInfo, tupleid,
 										NULL, NULL, &result))
 				{
 					if (result == TM_Ok)
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 					break;		/* concurrent update/delete */
 				}
 				result = ExecDeleteAct(context, resultRelInfo, tupleid, false);
@@ -2989,7 +2998,8 @@ lmerge_matched:
 				 * If the tuple was already deleted, return to let caller
 				 * handle it under NOT MATCHED clauses.
 				 */
-				return false;
+				*matched = false;
+				return NULL;
 
 			case TM_Updated:
 				{
@@ -3035,13 +3045,19 @@ lmerge_matched:
 							 * NOT MATCHED actions.
 							 */
 							if (TupIsNull(epqslot))
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							(void) ExecGetJunkAttribute(epqslot,
 														resultRelInfo->ri_RowIdAttNo,
 														&isNull);
 							if (isNull)
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							/*
 							 * When a tuple was updated and migrated to
@@ -3076,7 +3092,8 @@ lmerge_matched:
 							 * tuple already deleted; tell caller to run NOT
 							 * MATCHED actions
 							 */
-							return false;
+							*matched = false;
+							return NULL;
 
 						case TM_SelfModified:
 
@@ -3096,13 +3113,14 @@ lmerge_matched:
 										(errcode(ERRCODE_TRIGGERED_DATA_CHANGE_VIOLATION),
 										 errmsg("tuple to be updated or deleted was already modified by an operation triggered by the current command"),
 										 errhint("Consider using an AFTER trigger instead of a BEFORE trigger to propagate changes to other rows.")));
-							return false;
+							*matched = false;
+							return NULL;
 
 						default:
 							/* see table_tuple_lock call in ExecDelete() */
 							elog(ERROR, "unexpected table_tuple_lock status: %u",
 								 result);
-							return false;
+							return NULL;
 					}
 				}
 
@@ -3114,6 +3132,31 @@ lmerge_matched:
 				break;
 		}
 
+		/* Process RETURNING if present */
+		if (resultRelInfo->ri_projectReturning)
+		{
+			switch (commandType)
+			{
+				case CMD_UPDATE:
+					rslot = ExecProcessReturning(resultRelInfo, newslot,
+												 context->planSlot);
+					break;
+
+				case CMD_DELETE:
+					rslot = ExecProcessReturning(resultRelInfo,
+												 resultRelInfo->ri_oldTupleSlot,
+												 context->planSlot);
+					break;
+
+				case CMD_NOTHING:
+					break;
+
+				default:
+					elog(ERROR, "unrecognized commandType: %d",
+						 (int) commandType);
+			}
+		}
+
 		/*
 		 * We've activated one of the WHEN clauses, so we don't search
 		 * further. This is required behaviour, not an optimization.
@@ -3124,19 +3167,22 @@ lmerge_matched:
 	/*
 	 * Successfully executed an action or no qualifying action was found.
 	 */
-	return true;
+	*matched = true;
+
+	return rslot;
 }
 
 /*
  * Execute the first qualifying NOT MATCHED action.
  */
-static void
+static TupleTableSlot *
 ExecMergeNotMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 					bool canSetTag)
 {
 	ModifyTableState *mtstate = context->mtstate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
 	List	   *actionStates = NIL;
+	TupleTableSlot *rslot = NULL;
 	ListCell   *l;
 
 	/*
@@ -3186,10 +3232,10 @@ ExecMergeNotMatched(ModifyTableContext *
 				 * so we don't need to map the tuple here.
 				 */
 				newslot = ExecProject(action->mas_proj);
-				context->relaction = action;
+				mtstate->mt_merge_action = action;
 
-				(void) ExecInsert(context, mtstate->rootResultRelInfo, newslot,
-								  canSetTag, NULL, NULL);
+				rslot = ExecInsert(context, mtstate->rootResultRelInfo,
+								   newslot, canSetTag, NULL, NULL);
 				mtstate->mt_merge_inserted += 1;
 				break;
 			case CMD_NOTHING:
@@ -3205,6 +3251,8 @@ ExecMergeNotMatched(ModifyTableContext *
 		 */
 		break;
 	}
+
+	return rslot;
 }
 
 /*
@@ -3667,8 +3715,17 @@ ExecModifyTable(PlanState *pstate)
 				{
 					EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-					ExecMerge(&context, node->resultRelInfo, NULL, node->canSetTag);
-					continue;	/* no RETURNING support yet */
+					slot = ExecMerge(&context, node->resultRelInfo, NULL,
+									 node->canSetTag);
+
+					/*
+					 * If we got a RETURNING result, return it to the caller.
+					 * We'll continue the work on next call.
+					 */
+					if (slot)
+						return slot;
+
+					continue;	/* continue with the next tuple */
 				}
 
 				elog(ERROR, "tableoid is NULL");
@@ -3745,8 +3802,17 @@ ExecModifyTable(PlanState *pstate)
 					{
 						EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-						ExecMerge(&context, node->resultRelInfo, NULL, node->canSetTag);
-						continue;	/* no RETURNING support yet */
+						slot = ExecMerge(&context, node->resultRelInfo, NULL,
+										 node->canSetTag);
+
+						/*
+						 * If we got a RETURNING result, return it to the
+						 * caller.  We'll continue the work on next call.
+						 */
+						if (slot)
+							return slot;
+
+						continue;	/* continue with the next tuple */
 					}
 
 					elog(ERROR, "ctid is NULL");
@@ -3838,7 +3904,7 @@ ExecModifyTable(PlanState *pstate)
 				}
 				slot = ExecGetUpdateNewTuple(resultRelInfo, context.planSlot,
 											 oldSlot);
-				context.relaction = NULL;
+				node->mt_merge_action = NULL;
 
 				/* Now apply the update. */
 				slot = ExecUpdate(&context, resultRelInfo, tupleid, oldtuple,
diff --git a/src/backend/executor/spi.c b/src/backend/executor/spi.c
new file mode 100644
index d71484c..abb8b0d
--- a/src/backend/executor/spi.c
+++ b/src/backend/executor/spi.c
@@ -2033,6 +2033,8 @@ SPI_result_code_string(int code)
 			return "SPI_OK_TD_REGISTER";
 		case SPI_OK_MERGE:
 			return "SPI_OK_MERGE";
+		case SPI_OK_MERGE_RETURNING:
+			return "SPI_OK_MERGE_RETURNING";
 	}
 	/* Unrecognized code ... return something useful ... */
 	sprintf(buf, "Unrecognized SPI code %d", code);
@@ -2886,7 +2888,10 @@ _SPI_pquery(QueryDesc *queryDesc, bool f
 				res = SPI_OK_UPDATE;
 			break;
 		case CMD_MERGE:
-			res = SPI_OK_MERGE;
+			if (queryDesc->plannedstmt->hasReturning)
+				res = SPI_OK_MERGE_RETURNING;
+			else
+				res = SPI_OK_MERGE;
 			break;
 		default:
 			return SPI_ERROR_OPUNKNOWN;
diff --git a/src/backend/jit/llvm/llvmjit_expr.c b/src/backend/jit/llvm/llvmjit_expr.c
new file mode 100644
index 33161d8..bbbaa93
--- a/src/backend/jit/llvm/llvmjit_expr.c
+++ b/src/backend/jit/llvm/llvmjit_expr.c
@@ -1980,6 +1980,12 @@ llvm_compile_expr(ExprState *state)
 					break;
 				}
 
+			case EEOP_MERGING_FUNC:
+				build_EvalXFunc(b, mod, "ExecEvalMergingFunc",
+								v_state, op, v_econtext);
+				LLVMBuildBr(b, opblocks[opno + 1]);
+				break;
+
 			case EEOP_SUBPLAN:
 				build_EvalXFunc(b, mod, "ExecEvalSubPlan",
 								v_state, op, v_econtext);
diff --git a/src/backend/jit/llvm/llvmjit_types.c b/src/backend/jit/llvm/llvmjit_types.c
new file mode 100644
index 5212f52..61fa503
--- a/src/backend/jit/llvm/llvmjit_types.c
+++ b/src/backend/jit/llvm/llvmjit_types.c
@@ -155,6 +155,7 @@ void	   *referenced_functions[] =
 	ExecEvalFuncExprFusage,
 	ExecEvalFuncExprStrictFusage,
 	ExecEvalGroupingFunc,
+	ExecEvalMergingFunc,
 	ExecEvalMinMax,
 	ExecEvalNextValueExpr,
 	ExecEvalParamExec,
diff --git a/src/backend/nodes/nodeFuncs.c b/src/backend/nodes/nodeFuncs.c
new file mode 100644
index e1a5bc7..2468063
--- a/src/backend/nodes/nodeFuncs.c
+++ b/src/backend/nodes/nodeFuncs.c
@@ -67,6 +67,9 @@ exprType(const Node *expr)
 		case T_WindowFunc:
 			type = ((const WindowFunc *) expr)->wintype;
 			break;
+		case T_MergingFunc:
+			type = ((const MergingFunc *) expr)->mftype;
+			break;
 		case T_SubscriptingRef:
 			type = ((const SubscriptingRef *) expr)->refrestype;
 			break;
@@ -810,6 +813,9 @@ exprCollation(const Node *expr)
 		case T_WindowFunc:
 			coll = ((const WindowFunc *) expr)->wincollid;
 			break;
+		case T_MergingFunc:
+			coll = ((const MergingFunc *) expr)->mfcollid;
+			break;
 		case T_SubscriptingRef:
 			coll = ((const SubscriptingRef *) expr)->refcollid;
 			break;
@@ -1085,6 +1091,9 @@ exprSetCollation(Node *expr, Oid collati
 		case T_WindowFunc:
 			((WindowFunc *) expr)->wincollid = collation;
 			break;
+		case T_MergingFunc:
+			((MergingFunc *) expr)->mfcollid = collation;
+			break;
 		case T_SubscriptingRef:
 			((SubscriptingRef *) expr)->refcollid = collation;
 			break;
@@ -1343,6 +1352,9 @@ exprLocation(const Node *expr)
 			/* function name should always be the first thing */
 			loc = ((const WindowFunc *) expr)->location;
 			break;
+		case T_MergingFunc:
+			loc = ((const MergingFunc *) expr)->location;
+			break;
 		case T_SubscriptingRef:
 			/* just use container argument's location */
 			loc = exprLocation((Node *) ((const SubscriptingRef *) expr)->refexpr);
@@ -2035,6 +2047,7 @@ expression_tree_walker_impl(Node *node,
 		case T_RangeTblRef:
 		case T_SortGroupClause:
 		case T_CTESearchClause:
+		case T_MergingFunc:
 			/* primitive node types with no expression subnodes */
 			break;
 		case T_WithCheckOption:
@@ -2869,6 +2882,7 @@ expression_tree_mutator_impl(Node *node,
 		case T_RangeTblRef:
 		case T_SortGroupClause:
 		case T_CTESearchClause:
+		case T_MergingFunc:
 			return (Node *) copyObject(node);
 		case T_WithCheckOption:
 			{
@@ -3833,6 +3847,7 @@ raw_expression_tree_walker_impl(Node *no
 		case T_ParamRef:
 		case T_A_Const:
 		case T_A_Star:
+		case T_MergingFunc:
 			/* primitive node types with no subnodes */
 			break;
 		case T_Alias:
@@ -4053,6 +4068,8 @@ raw_expression_tree_walker_impl(Node *no
 					return true;
 				if (WALK(stmt->mergeWhenClauses))
 					return true;
+				if (WALK(stmt->returningList))
+					return true;
 				if (WALK(stmt->withClause))
 					return true;
 			}
diff --git a/src/backend/optimizer/plan/subselect.c b/src/backend/optimizer/plan/subselect.c
new file mode 100644
index 3115d79..a3646c0
--- a/src/backend/optimizer/plan/subselect.c
+++ b/src/backend/optimizer/plan/subselect.c
@@ -1846,7 +1846,8 @@ convert_EXISTS_to_ANY(PlannerInfo *root,
 /*
  * Replace correlation vars (uplevel vars) with Params.
  *
- * Uplevel PlaceHolderVars and aggregates are replaced, too.
+ * Uplevel PlaceHolderVars, aggregates, GROUPING() expressions, and MERGING()
+ * expressions are replaced, too.
  *
  * Note: it is critical that this runs immediately after SS_process_sublinks.
  * Since we do not recurse into the arguments of uplevel PHVs and aggregates,
@@ -1900,6 +1901,11 @@ replace_correlation_vars_mutator(Node *n
 		if (((GroupingFunc *) node)->agglevelsup > 0)
 			return (Node *) replace_outer_grouping(root, (GroupingFunc *) node);
 	}
+	if (IsA(node, MergingFunc))
+	{
+		if (root->parse->commandType != CMD_MERGE)
+			return (Node *) replace_outer_merging(root, (MergingFunc *) node);
+	}
 	return expression_tree_mutator(node,
 								   replace_correlation_vars_mutator,
 								   (void *) root);
diff --git a/src/backend/optimizer/util/paramassign.c b/src/backend/optimizer/util/paramassign.c
new file mode 100644
index a711003..9d8c5f6
--- a/src/backend/optimizer/util/paramassign.c
+++ b/src/backend/optimizer/util/paramassign.c
@@ -307,6 +307,56 @@ replace_outer_grouping(PlannerInfo *root
 }
 
 /*
+ * Generate a Param node to replace the given MergingFunc expression which is
+ * expected to be in the RETURNING list of an upper-level MERGE query.  Record
+ * the need for the MergingFunc in the proper upper-level root->plan_params.
+ */
+Param *
+replace_outer_merging(PlannerInfo *root, MergingFunc *mf)
+{
+	Param	   *retval;
+	PlannerParamItem *pitem;
+	Oid			ptype = exprType((Node *) mf);
+
+	Assert(root->parse->commandType != CMD_MERGE);
+
+	/*
+	 * The parser should have ensured that the MergingFunc is in the RETURNING
+	 * list of an upper-level MERGE query, so find that query.
+	 */
+	do
+	{
+		root = root->parent_root;
+		if (root == NULL)
+			elog(ERROR, "MergingFunc found outside MERGE");
+	} while (root->parse->commandType != CMD_MERGE);
+
+	/*
+	 * It does not seem worthwhile to try to de-duplicate references to outer
+	 * MergingFunc expressions.  Just make a new slot every time.
+	 */
+	mf = copyObject(mf);
+
+	pitem = makeNode(PlannerParamItem);
+	pitem->item = (Node *) mf;
+	pitem->paramId = list_length(root->glob->paramExecTypes);
+	root->glob->paramExecTypes = lappend_oid(root->glob->paramExecTypes,
+											 ptype);
+
+	root->plan_params = lappend(root->plan_params, pitem);
+
+	retval = makeNode(Param);
+	retval->paramkind = PARAM_EXEC;
+	retval->paramid = pitem->paramId;
+	retval->paramtype = ptype;
+	retval->paramtypmod = -1;
+	retval->paramcollid = InvalidOid;
+	retval->location = mf->location;
+
+	return retval;
+}
+
+/*
  * Generate a Param node to replace the given Var,
  * which is expected to come from some upper NestLoop plan node.
  * Record the need for the Var in root->curOuterParams.
diff --git a/src/backend/parser/analyze.c b/src/backend/parser/analyze.c
new file mode 100644
index 06fc8ce..f2b32ba
--- a/src/backend/parser/analyze.c
+++ b/src/backend/parser/analyze.c
@@ -74,7 +74,6 @@ static void determineRecursiveColTypes(P
 									   Node *larg, List *nrtargetlist);
 static Query *transformReturnStmt(ParseState *pstate, ReturnStmt *stmt);
 static Query *transformUpdateStmt(ParseState *pstate, UpdateStmt *stmt);
-static List *transformReturningList(ParseState *pstate, List *returningList);
 static Query *transformPLAssignStmt(ParseState *pstate,
 									PLAssignStmt *stmt);
 static Query *transformDeclareCursorStmt(ParseState *pstate,
@@ -553,7 +552,8 @@ transformDeleteStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -980,7 +980,8 @@ transformInsertStmt(ParseState *pstate,
 	/* Process RETURNING, if any. */
 	if (stmt->returningList)
 		qry->returningList = transformReturningList(pstate,
-													stmt->returningList);
+													stmt->returningList,
+													EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -2445,7 +2446,8 @@ transformUpdateStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/*
 	 * Now we are done with SELECT-like processing, and can get on with
@@ -2539,10 +2541,11 @@ transformUpdateTargetList(ParseState *ps
 
 /*
  * transformReturningList -
- *	handle a RETURNING clause in INSERT/UPDATE/DELETE
+ *	handle a RETURNING clause in INSERT/UPDATE/DELETE/MERGE
  */
-static List *
-transformReturningList(ParseState *pstate, List *returningList)
+List *
+transformReturningList(ParseState *pstate, List *returningList,
+					   ParseExprKind exprKind)
 {
 	List	   *rlist;
 	int			save_next_resno;
@@ -2559,7 +2562,7 @@ transformReturningList(ParseState *pstat
 	pstate->p_next_resno = 1;
 
 	/* transform RETURNING identically to a SELECT targetlist */
-	rlist = transformTargetList(pstate, returningList, EXPR_KIND_RETURNING);
+	rlist = transformTargetList(pstate, returningList, exprKind);
 
 	/*
 	 * Complain if the nonempty tlist expanded to nothing (which is possible
diff --git a/src/backend/parser/gram.y b/src/backend/parser/gram.y
new file mode 100644
index 3460fea..2b24b1b
--- a/src/backend/parser/gram.y
+++ b/src/backend/parser/gram.y
@@ -694,7 +694,7 @@ static Node *makeRecursiveViewSelect(cha
 	BOOLEAN_P BOTH BREADTH BY
 
 	CACHE CALL CALLED CASCADE CASCADED CASE CAST CATALOG_P CHAIN CHAR_P
-	CHARACTER CHARACTERISTICS CHECK CHECKPOINT CLASS CLOSE
+	CHARACTER CHARACTERISTICS CHECK CHECKPOINT CLASS CLAUSE_NUMBER CLOSE
 	CLUSTER COALESCE COLLATE COLLATION COLUMN COLUMNS COMMENT COMMENTS COMMIT
 	COMMITTED COMPRESSION CONCURRENTLY CONFIGURATION CONFLICT
 	CONNECTION CONSTRAINT CONSTRAINTS CONTENT_P CONTINUE_P CONVERSION_P COPY
@@ -732,7 +732,7 @@ static Node *makeRecursiveViewSelect(cha
 	LEADING LEAKPROOF LEAST LEFT LEVEL LIKE LIMIT LISTEN LOAD LOCAL
 	LOCALTIME LOCALTIMESTAMP LOCATION LOCK_P LOCKED LOGGED
 
-	MAPPING MATCH MATCHED MATERIALIZED MAXVALUE MERGE METHOD
+	MAPPING MATCH MATCHED MATERIALIZED MAXVALUE MERGE MERGING METHOD
 	MINUTE_P MINVALUE MODE MONTH_P MOVE
 
 	NAME_P NAMES NATIONAL NATURAL NCHAR NEW NEXT NFC NFD NFKC NFKD NO NONE
@@ -12361,6 +12361,7 @@ MergeStmt:
 			USING table_ref
 			ON a_expr
 			merge_when_list
+			returning_clause
 				{
 					MergeStmt  *m = makeNode(MergeStmt);
 
@@ -12369,6 +12370,7 @@ MergeStmt:
 					m->sourceRelation = $6;
 					m->joinCondition = $8;
 					m->mergeWhenClauses = $9;
+					m->returningList = $10;
 
 					$$ = (Node *) m;
 				}
@@ -15782,6 +15784,24 @@ func_expr_common_subexpr:
 					n->location = @1;
 					$$ = (Node *) n;
 				}
+			| MERGING '(' ACTION ')'
+				{
+					MergingFunc *m = makeNode(MergingFunc);
+
+					m->mfop = MERGING_ACTION;
+					m->mftype = TEXTOID;
+					m->location = @1;
+					$$ = (Node *) m;
+				}
+			| MERGING '(' CLAUSE_NUMBER ')'
+				{
+					MergingFunc *m = makeNode(MergingFunc);
+
+					m->mfop = MERGING_CLAUSE_NUMBER;
+					m->mftype = INT4OID;
+					m->location = @1;
+					$$ = (Node *) m;
+				}
 			;
 
 
@@ -17160,6 +17180,7 @@ unreserved_keyword:
 			| CHARACTERISTICS
 			| CHECKPOINT
 			| CLASS
+			| CLAUSE_NUMBER
 			| CLOSE
 			| CLUSTER
 			| COLUMNS
@@ -17479,6 +17500,7 @@ col_name_keyword:
 			| JSON_SCALAR
 			| JSON_SERIALIZE
 			| LEAST
+			| MERGING
 			| NATIONAL
 			| NCHAR
 			| NONE
@@ -17697,6 +17719,7 @@ bare_label_keyword:
 			| CHECK
 			| CHECKPOINT
 			| CLASS
+			| CLAUSE_NUMBER
 			| CLOSE
 			| CLUSTER
 			| COALESCE
@@ -17868,6 +17891,7 @@ bare_label_keyword:
 			| MATERIALIZED
 			| MAXVALUE
 			| MERGE
+			| MERGING
 			| METHOD
 			| MINVALUE
 			| MODE
diff --git a/src/backend/parser/parse_agg.c b/src/backend/parser/parse_agg.c
new file mode 100644
index 7b211a7..f143310
--- a/src/backend/parser/parse_agg.c
+++ b/src/backend/parser/parse_agg.c
@@ -467,6 +467,7 @@ check_agglevels_and_constraints(ParseSta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
@@ -914,6 +915,7 @@ transformWindowFuncCall(ParseState *psta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_cte.c b/src/backend/parser/parse_cte.c
new file mode 100644
index 3c88c9a..6826d4f
--- a/src/backend/parser/parse_cte.c
+++ b/src/backend/parser/parse_cte.c
@@ -126,13 +126,6 @@ transformWithClause(ParseState *pstate,
 		CommonTableExpr *cte = (CommonTableExpr *) lfirst(lc);
 		ListCell   *rest;
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(cte->ctequery, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in WITH query"),
-					parser_errposition(pstate, cte->location));
-
 		for_each_cell(rest, withClause->ctes, lnext(withClause->ctes, lc))
 		{
 			CommonTableExpr *cte2 = (CommonTableExpr *) lfirst(rest);
@@ -153,7 +146,8 @@ transformWithClause(ParseState *pstate,
 			/* must be a data-modifying statement */
 			Assert(IsA(cte->ctequery, InsertStmt) ||
 				   IsA(cte->ctequery, UpdateStmt) ||
-				   IsA(cte->ctequery, DeleteStmt));
+				   IsA(cte->ctequery, DeleteStmt) ||
+				   IsA(cte->ctequery, MergeStmt));
 
 			pstate->p_hasModifyingCTE = true;
 		}
diff --git a/src/backend/parser/parse_expr.c b/src/backend/parser/parse_expr.c
new file mode 100644
index 9300c7b..ab7e564
--- a/src/backend/parser/parse_expr.c
+++ b/src/backend/parser/parse_expr.c
@@ -54,6 +54,7 @@ static Node *transformAExprDistinct(Pars
 static Node *transformAExprNullIf(ParseState *pstate, A_Expr *a);
 static Node *transformAExprIn(ParseState *pstate, A_Expr *a);
 static Node *transformAExprBetween(ParseState *pstate, A_Expr *a);
+static Node *transformMergingFunc(ParseState *pstate, MergingFunc *f);
 static Node *transformBoolExpr(ParseState *pstate, BoolExpr *a);
 static Node *transformFuncCall(ParseState *pstate, FuncCall *fn);
 static Node *transformMultiAssignRef(ParseState *pstate, MultiAssignRef *maref);
@@ -227,6 +228,10 @@ transformExprRecurse(ParseState *pstate,
 			result = transformGroupingFunc(pstate, (GroupingFunc *) expr);
 			break;
 
+		case T_MergingFunc:
+			result = transformMergingFunc(pstate, (MergingFunc *) expr);
+			break;
+
 		case T_NamedArgExpr:
 			{
 				NamedArgExpr *na = (NamedArgExpr *) expr;
@@ -541,6 +546,7 @@ transformColumnRef(ParseState *pstate, C
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CHECK_CONSTRAINT:
@@ -1354,6 +1360,32 @@ transformAExprBetween(ParseState *pstate
 }
 
 static Node *
+transformMergingFunc(ParseState *pstate, MergingFunc *f)
+{
+	/*
+	 * All we need to do is check that we're in the RETURNING list of a MERGE
+	 * command.  If so, we just return the node as-is.
+	 */
+	if (pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+	{
+		ParseState *parent_pstate = pstate->parentParseState;
+
+		while (parent_pstate &&
+			   parent_pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+			parent_pstate = parent_pstate->parentParseState;
+
+		if (!parent_pstate ||
+			parent_pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+			ereport(ERROR,
+					errcode(ERRCODE_SYNTAX_ERROR),
+					errmsg("MERGING() can only be used in the RETURNING list of a MERGE command"),
+					parser_errposition(pstate, f->location));
+	}
+
+	return (Node *) f;
+}
+
+static Node *
 transformBoolExpr(ParseState *pstate, BoolExpr *a)
 {
 	List	   *args = NIL;
@@ -1767,6 +1799,7 @@ transformSubLink(ParseState *pstate, Sub
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CYCLE_MARK:
@@ -3115,6 +3148,7 @@ ParseExprKindName(ParseExprKind exprKind
 		case EXPR_KIND_OFFSET:
 			return "OFFSET";
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			return "RETURNING";
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
diff --git a/src/backend/parser/parse_func.c b/src/backend/parser/parse_func.c
new file mode 100644
index fdb3e6d..0cbc950
--- a/src/backend/parser/parse_func.c
+++ b/src/backend/parser/parse_func.c
@@ -2599,6 +2599,7 @@ check_srf_call_placement(ParseState *pst
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_merge.c b/src/backend/parser/parse_merge.c
new file mode 100644
index 5f6a683..55936bb
--- a/src/backend/parser/parse_merge.c
+++ b/src/backend/parser/parse_merge.c
@@ -233,6 +233,10 @@ transformMergeStmt(ParseState *pstate, M
 	 */
 	qry->jointree = makeFromExpr(pstate->p_joinlist, joinExpr);
 
+	/* Transform the RETURNING list, if any */
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_MERGE_RETURNING);
+
 	/*
 	 * We now have a good query shape, so now look at the WHEN conditions and
 	 * action targetlists.
@@ -254,6 +258,7 @@ transformMergeStmt(ParseState *pstate, M
 		MergeAction *action;
 
 		action = makeNode(MergeAction);
+		action->index = foreach_current_index(l) + 1;
 		action->commandType = mergeWhenClause->commandType;
 		action->matched = mergeWhenClause->matched;
 
@@ -390,9 +395,6 @@ transformMergeStmt(ParseState *pstate, M
 
 	qry->mergeActionList = mergeActionList;
 
-	/* RETURNING could potentially be added in the future, but not in SQL std */
-	qry->returningList = NULL;
-
 	qry->hasTargetSRFs = false;
 	qry->hasSubLinks = pstate->p_hasSubLinks;
 
diff --git a/src/backend/parser/parse_relation.c b/src/backend/parser/parse_relation.c
new file mode 100644
index 34a0ec5..60b7e4f
--- a/src/backend/parser/parse_relation.c
+++ b/src/backend/parser/parse_relation.c
@@ -2346,9 +2346,10 @@ addRangeTableEntryForCTE(ParseState *pst
 		cte->cterefcount++;
 
 	/*
-	 * We throw error if the CTE is INSERT/UPDATE/DELETE without RETURNING.
-	 * This won't get checked in case of a self-reference, but that's OK
-	 * because data-modifying CTEs aren't allowed to be recursive anyhow.
+	 * We throw error if the CTE is INSERT/UPDATE/DELETE/MERGE without
+	 * RETURNING.  This won't get checked in case of a self-reference, but
+	 * that's OK because data-modifying CTEs aren't allowed to be recursive
+	 * anyhow.
 	 */
 	if (IsA(cte->ctequery, Query))
 	{
diff --git a/src/backend/parser/parse_target.c b/src/backend/parser/parse_target.c
new file mode 100644
index 0cd904f..eab78f4
--- a/src/backend/parser/parse_target.c
+++ b/src/backend/parser/parse_target.c
@@ -1807,6 +1807,20 @@ FigureColnameInternal(Node *node, char *
 			/* make GROUPING() act like a regular function */
 			*name = "grouping";
 			return 2;
+		case T_MergingFunc:
+			switch (((MergingFunc *) node)->mfop)
+			{
+				case MERGING_ACTION:
+					*name = "action";
+					return 2;
+				case MERGING_CLAUSE_NUMBER:
+					*name = "clause_number";
+					return 2;
+				default:
+					elog(ERROR, "unrecognized MergingFuncOp: %d",
+						 (int) ((MergingFunc *) node)->mfop);
+			}
+			break;
 		case T_SubLink:
 			switch (((SubLink *) node)->subLinkType)
 			{
diff --git a/src/backend/rewrite/rewriteHandler.c b/src/backend/rewrite/rewriteHandler.c
new file mode 100644
index f60b34d..d719d07
--- a/src/backend/rewrite/rewriteHandler.c
+++ b/src/backend/rewrite/rewriteHandler.c
@@ -3632,9 +3632,9 @@ RewriteQuery(Query *parsetree, List *rew
 	ListCell   *lc1;
 
 	/*
-	 * First, recursively process any insert/update/delete statements in WITH
-	 * clauses.  (We have to do this first because the WITH clauses may get
-	 * copied into rule actions below.)
+	 * First, recursively process any insert/update/delete/merge statements in
+	 * WITH clauses.  (We have to do this first because the WITH clauses may
+	 * get copied into rule actions below.)
 	 */
 	foreach(lc1, parsetree->cteList)
 	{
@@ -3659,7 +3659,8 @@ RewriteQuery(Query *parsetree, List *rew
 			if (!(ctequery->commandType == CMD_SELECT ||
 				  ctequery->commandType == CMD_UPDATE ||
 				  ctequery->commandType == CMD_INSERT ||
-				  ctequery->commandType == CMD_DELETE))
+				  ctequery->commandType == CMD_DELETE ||
+				  ctequery->commandType == CMD_MERGE))
 			{
 				/*
 				 * Currently it could only be NOTIFY; this error message will
diff --git a/src/backend/rewrite/rowsecurity.c b/src/backend/rewrite/rowsecurity.c
new file mode 100644
index 94268a2..65ce579
--- a/src/backend/rewrite/rowsecurity.c
+++ b/src/backend/rewrite/rowsecurity.c
@@ -395,10 +395,10 @@ get_row_security_policies(Query *root, R
 	 * on the final action we take.
 	 *
 	 * We already fetched the SELECT policies above, to check existing rows,
-	 * but we must also check that new rows created by UPDATE actions are
-	 * visible, if SELECT rights are required for this relation. We don't do
-	 * this for INSERT actions, since an INSERT command would only do this
-	 * check if it had a RETURNING list, and MERGE does not support RETURNING.
+	 * but we must also check that new rows created by INSERT/UPDATE actions
+	 * are visible, if SELECT rights are required. For INSERT actions, we only
+	 * do this if RETURNING is specified, to be consistent with a plain INSERT
+	 * command, which can only require SELECT rights when RETURNING is used.
 	 *
 	 * We don't push the UPDATE/DELETE USING quals to the RTE because we don't
 	 * really want to apply them while scanning the relation since we don't
@@ -420,6 +420,8 @@ get_row_security_policies(Query *root, R
 		List	   *merge_delete_restrictive_policies;
 		List	   *merge_insert_permissive_policies;
 		List	   *merge_insert_restrictive_policies;
+		List	   *merge_select_permissive_policies = NIL;
+		List	   *merge_select_restrictive_policies = NIL;
 
 		/*
 		 * Fetch the UPDATE policies and set them up to execute on the
@@ -457,9 +459,6 @@ get_row_security_policies(Query *root, R
 		 */
 		if (perminfo->requiredPerms & ACL_SELECT)
 		{
-			List	   *merge_select_permissive_policies;
-			List	   *merge_select_restrictive_policies;
-
 			get_policies_for_relation(rel, CMD_SELECT, user_id,
 									  &merge_select_permissive_policies,
 									  &merge_select_restrictive_policies);
@@ -508,6 +507,21 @@ get_row_security_policies(Query *root, R
 							   withCheckOptions,
 							   hasSubLinks,
 							   false);
+
+		/*
+		 * Add ALL/SELECT policies as WCO_RLS_INSERT_CHECK WCOs, to ensure
+		 * that the inserted row is visible when executing an INSERT action,
+		 * if RETURNING is specified and SELECT rights are required for this
+		 * relation.
+		 */
+		if (perminfo->requiredPerms & ACL_SELECT && root->returningList)
+			add_with_check_options(rel, rt_index,
+								   WCO_RLS_INSERT_CHECK,
+								   merge_select_permissive_policies,
+								   merge_select_restrictive_policies,
+								   withCheckOptions,
+								   hasSubLinks,
+								   true);
 	}
 
 	table_close(rel, NoLock);
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
new file mode 100644
index 8de821f..f84a943
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -2145,11 +2145,10 @@ QueryReturnsTuples(Query *parsetree)
 		case CMD_SELECT:
 			/* returns tuples */
 			return true;
-		case CMD_MERGE:
-			return false;
 		case CMD_INSERT:
 		case CMD_UPDATE:
 		case CMD_DELETE:
+		case CMD_MERGE:
 			/* the forms with RETURNING return tuples */
 			if (parsetree->returningList)
 				return true;
diff --git a/src/backend/utils/adt/ruleutils.c b/src/backend/utils/adt/ruleutils.c
new file mode 100644
index 0b2a164..6775f91
--- a/src/backend/utils/adt/ruleutils.c
+++ b/src/backend/utils/adt/ruleutils.c
@@ -7197,8 +7197,13 @@ get_merge_query_def(Query *query, depars
 			appendStringInfoString(buf, "DO NOTHING");
 	}
 
-	/* No RETURNING support in MERGE yet */
-	Assert(query->returningList == NIL);
+	/* Add RETURNING if present */
+	if (query->returningList)
+	{
+		appendContextKeyword(context, " RETURNING",
+							 -PRETTYINDENT_STD, PRETTYINDENT_STD, 1);
+		get_target_list(query->returningList, context, NULL, colNamesVisible);
+	}
 }
 
 
@@ -8298,6 +8303,7 @@ isSimpleNode(Node *node, Node *parentNod
 		case T_Aggref:
 		case T_GroupingFunc:
 		case T_WindowFunc:
+		case T_MergingFunc:
 		case T_FuncExpr:
 		case T_JsonConstructorExpr:
 			/* function-like: name(..) or name[..] */
@@ -8652,6 +8658,27 @@ get_rule_expr(Node *node, deparse_contex
 			get_windowfunc_expr((WindowFunc *) node, context);
 			break;
 
+		case T_MergingFunc:
+			{
+				MergingFunc *mexpr = (MergingFunc *) node;
+
+				appendStringInfoString(buf, "MERGING(");
+				switch (mexpr->mfop)
+				{
+					case MERGING_ACTION:
+						appendStringInfoString(buf, "ACTION");
+						break;
+					case MERGING_CLAUSE_NUMBER:
+						appendStringInfoString(buf, "CLAUSE_NUMBER");
+						break;
+					default:
+						elog(ERROR, "unrecognized MergingFuncOp: %d",
+							 (int) mexpr->mfop);
+				}
+				appendStringInfoChar(buf, ')');
+			}
+			break;
+
 		case T_SubscriptingRef:
 			{
 				SubscriptingRef *sbsref = (SubscriptingRef *) node;
diff --git a/src/bin/psql/common.c b/src/bin/psql/common.c
new file mode 100644
index 76e01b0..2830bde
--- a/src/bin/psql/common.c
+++ b/src/bin/psql/common.c
@@ -983,13 +983,17 @@ PrintQueryResult(PGresult *result, bool
 			else
 				success = true;
 
-			/* if it's INSERT/UPDATE/DELETE RETURNING, also print status */
+			/*
+			 * If it's INSERT/UPDATE/DELETE/MERGE RETURNING, also print
+			 * status.
+			 */
 			if (last || pset.show_all_results)
 			{
 				cmdstatus = PQcmdStatus(result);
 				if (strncmp(cmdstatus, "INSERT", 6) == 0 ||
 					strncmp(cmdstatus, "UPDATE", 6) == 0 ||
-					strncmp(cmdstatus, "DELETE", 6) == 0)
+					strncmp(cmdstatus, "DELETE", 6) == 0 ||
+					strncmp(cmdstatus, "MERGE", 5) == 0)
 					PrintQueryStatus(result, printStatusFout);
 			}
 
diff --git a/src/include/executor/execExpr.h b/src/include/executor/execExpr.h
new file mode 100644
index a20c539..a24aa65
--- a/src/include/executor/execExpr.h
+++ b/src/include/executor/execExpr.h
@@ -241,6 +241,7 @@ typedef enum ExprEvalOp
 	EEOP_AGGREF,
 	EEOP_GROUPING_FUNC,
 	EEOP_WINDOW_FUNC,
+	EEOP_MERGING_FUNC,
 	EEOP_SUBPLAN,
 
 	/* aggregation related nodes */
@@ -623,6 +624,12 @@ typedef struct ExprEvalStep
 			WindowFuncExprState *wfstate;
 		}			window_func;
 
+		/* for EEOP_MERGING_FUNC */
+		struct
+		{
+			MergingFuncOp mfop; /* the MERGING() operation to perform */
+		}			merging_func;
+
 		/* for EEOP_SUBPLAN */
 		struct
 		{
@@ -806,6 +813,8 @@ extern void ExecEvalJsonConstructor(Expr
 									ExprContext *econtext);
 extern void ExecEvalJsonIsPredicate(ExprState *state, ExprEvalStep *op);
 extern void ExecEvalGroupingFunc(ExprState *state, ExprEvalStep *op);
+extern void ExecEvalMergingFunc(ExprState *state, ExprEvalStep *op,
+								ExprContext *econtext);
 extern void ExecEvalSubPlan(ExprState *state, ExprEvalStep *op,
 							ExprContext *econtext);
 extern void ExecEvalWholeRowVar(ExprState *state, ExprEvalStep *op,
diff --git a/src/include/executor/spi.h b/src/include/executor/spi.h
new file mode 100644
index c13e1f6..48b8773
--- a/src/include/executor/spi.h
+++ b/src/include/executor/spi.h
@@ -97,6 +97,7 @@ typedef struct _SPI_plan *SPIPlanPtr;
 #define SPI_OK_REL_UNREGISTER	16
 #define SPI_OK_TD_REGISTER		17
 #define SPI_OK_MERGE			18
+#define SPI_OK_MERGE_RETURNING	19
 
 #define SPI_OPT_NONATOMIC		(1 << 0)
 
diff --git a/src/include/nodes/execnodes.h b/src/include/nodes/execnodes.h
new file mode 100644
index 561fdd9..75c4fde
--- a/src/include/nodes/execnodes.h
+++ b/src/include/nodes/execnodes.h
@@ -1318,6 +1318,9 @@ typedef struct ModifyTableState
 	/* Flags showing which subcommands are present INS/UPD/DEL/DO NOTHING */
 	int			mt_merge_subcommands;
 
+	/* For MERGE, the action currently being executed */
+	MergeActionState *mt_merge_action;
+
 	/* tuple counters for MERGE */
 	double		mt_merge_inserted;
 	double		mt_merge_updated;
diff --git a/src/include/nodes/parsenodes.h b/src/include/nodes/parsenodes.h
new file mode 100644
index b3181f3..89447ca
--- a/src/include/nodes/parsenodes.h
+++ b/src/include/nodes/parsenodes.h
@@ -1927,6 +1927,7 @@ typedef struct MergeStmt
 	Node	   *sourceRelation; /* source relation */
 	Node	   *joinCondition;	/* join condition between source and target */
 	List	   *mergeWhenClauses;	/* list of MergeWhenClause(es) */
+	List	   *returningList;	/* list of expressions to return */
 	WithClause *withClause;		/* WITH clause */
 } MergeStmt;
 
diff --git a/src/include/nodes/primnodes.h b/src/include/nodes/primnodes.h
new file mode 100644
index 4a15460..97806f5
--- a/src/include/nodes/primnodes.h
+++ b/src/include/nodes/primnodes.h
@@ -572,6 +572,39 @@ typedef struct WindowFunc
 } WindowFunc;
 
 /*
+ * MergingFunc
+ *
+ * A MergingFunc is a MERGING(...) expression that can only appear in the
+ * RETURNING list of a MERGE command.  It returns information about the
+ * currently executing merge action.  Possible operations are:
+ *
+ *	MERGING_ACTION:
+ *		Return the command string of the current merge action ("INSERT",
+ *		"UPDATE" or "DELETE").
+ *
+ *	MERGING_CLAUSE_NUMBER:
+ *		Return the 1-based index of the current merge WHEN clause.
+ */
+typedef enum MergingFuncOp
+{
+	MERGING_ACTION,
+	MERGING_CLAUSE_NUMBER,
+} MergingFuncOp;
+
+typedef struct MergingFunc
+{
+	Expr		xpr;
+	/* operation to perform */
+	MergingFuncOp mfop;
+	/* type Oid of result */
+	Oid			mftype pg_node_attr(query_jumble_ignore);
+	/* OID of collation, or InvalidOid if none */
+	Oid			mfcollid pg_node_attr(query_jumble_ignore);
+	/* token location, or -1 if unknown */
+	int			location;
+} MergingFunc;
+
+/*
  * SubscriptingRef: describes a subscripting operation over a container
  * (array, etc).
  *
@@ -1735,6 +1768,7 @@ typedef struct BooleanTest
 typedef struct MergeAction
 {
 	NodeTag		type;
+	int			index;			/* 1-based index of the clause */
 	bool		matched;		/* true=MATCHED, false=NOT MATCHED */
 	CmdType		commandType;	/* INSERT/UPDATE/DELETE/DO NOTHING */
 	/* OVERRIDING clause */
diff --git a/src/include/optimizer/paramassign.h b/src/include/optimizer/paramassign.h
new file mode 100644
index 501f728..1e95c8f
--- a/src/include/optimizer/paramassign.h
+++ b/src/include/optimizer/paramassign.h
@@ -20,6 +20,7 @@ extern Param *replace_outer_placeholderv
 										   PlaceHolderVar *phv);
 extern Param *replace_outer_agg(PlannerInfo *root, Aggref *agg);
 extern Param *replace_outer_grouping(PlannerInfo *root, GroupingFunc *grp);
+extern Param *replace_outer_merging(PlannerInfo *root, MergingFunc *mf);
 extern Param *replace_nestloop_param_var(PlannerInfo *root, Var *var);
 extern Param *replace_nestloop_param_placeholdervar(PlannerInfo *root,
 													PlaceHolderVar *phv);
diff --git a/src/include/parser/analyze.h b/src/include/parser/analyze.h
new file mode 100644
index 5ecdd10..28b66fc
--- a/src/include/parser/analyze.h
+++ b/src/include/parser/analyze.h
@@ -44,6 +44,8 @@ extern List *transformInsertRow(ParseSta
 								bool strip_indirection);
 extern List *transformUpdateTargetList(ParseState *pstate,
 									   List *origTlist);
+extern List *transformReturningList(ParseState *pstate, List *returningList,
+									ParseExprKind exprKind);
 extern Query *transformTopLevelStmt(ParseState *pstate, RawStmt *parseTree);
 extern Query *transformStmt(ParseState *pstate, Node *parseTree);
 
diff --git a/src/include/parser/kwlist.h b/src/include/parser/kwlist.h
new file mode 100644
index 2331aca..9c5d6ea
--- a/src/include/parser/kwlist.h
+++ b/src/include/parser/kwlist.h
@@ -80,6 +80,7 @@ PG_KEYWORD("characteristics", CHARACTERI
 PG_KEYWORD("check", CHECK, RESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("checkpoint", CHECKPOINT, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("class", CLASS, UNRESERVED_KEYWORD, BARE_LABEL)
+PG_KEYWORD("clause_number", CLAUSE_NUMBER, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("close", CLOSE, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("cluster", CLUSTER, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("coalesce", COALESCE, COL_NAME_KEYWORD, BARE_LABEL)
@@ -266,6 +267,7 @@ PG_KEYWORD("matched", MATCHED, UNRESERVE
 PG_KEYWORD("materialized", MATERIALIZED, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("maxvalue", MAXVALUE, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("merge", MERGE, UNRESERVED_KEYWORD, BARE_LABEL)
+PG_KEYWORD("merging", MERGING, COL_NAME_KEYWORD, BARE_LABEL)
 PG_KEYWORD("method", METHOD, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("minute", MINUTE_P, UNRESERVED_KEYWORD, AS_LABEL)
 PG_KEYWORD("minvalue", MINVALUE, UNRESERVED_KEYWORD, BARE_LABEL)
diff --git a/src/include/parser/parse_node.h b/src/include/parser/parse_node.h
new file mode 100644
index 99d6515..5b781d8
--- a/src/include/parser/parse_node.h
+++ b/src/include/parser/parse_node.h
@@ -61,7 +61,8 @@ typedef enum ParseExprKind
 	EXPR_KIND_DISTINCT_ON,		/* DISTINCT ON */
 	EXPR_KIND_LIMIT,			/* LIMIT */
 	EXPR_KIND_OFFSET,			/* OFFSET */
-	EXPR_KIND_RETURNING,		/* RETURNING */
+	EXPR_KIND_RETURNING,		/* RETURNING in INSERT/UPDATE/DELETE */
+	EXPR_KIND_MERGE_RETURNING,	/* RETURNING in MERGE */
 	EXPR_KIND_VALUES,			/* VALUES */
 	EXPR_KIND_VALUES_SINGLE,	/* single-row VALUES (in INSERT only) */
 	EXPR_KIND_CHECK_CONSTRAINT, /* CHECK constraint for a table */
diff --git a/src/pl/plpgsql/src/pl_exec.c b/src/pl/plpgsql/src/pl_exec.c
new file mode 100644
index 6d16913..51de094
--- a/src/pl/plpgsql/src/pl_exec.c
+++ b/src/pl/plpgsql/src/pl_exec.c
@@ -4267,9 +4267,9 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 	/*
 	 * If we have INTO, then we only need one row back ... but if we have INTO
 	 * STRICT or extra check too_many_rows, ask for two rows, so that we can
-	 * verify the statement returns only one.  INSERT/UPDATE/DELETE are always
-	 * treated strictly. Without INTO, just run the statement to completion
-	 * (tcount = 0).
+	 * verify the statement returns only one.  INSERT/UPDATE/DELETE/MERGE are
+	 * always treated strictly. Without INTO, just run the statement to
+	 * completion (tcount = 0).
 	 *
 	 * We could just ask for two rows always when using INTO, but there are
 	 * some cases where demanding the extra row costs significant time, eg by
@@ -4307,10 +4307,11 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 			Assert(stmt->mod_stmt);
 			exec_set_found(estate, (SPI_processed != 0));
 			break;
@@ -4489,10 +4490,11 @@ exec_stmt_dynexecute(PLpgSQL_execstate *
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 		case SPI_OK_UTILITY:
 		case SPI_OK_REWRITTEN:
 			break;
diff --git a/src/pl/tcl/pltcl.c b/src/pl/tcl/pltcl.c
new file mode 100644
index 6187e15..18d14a2
--- a/src/pl/tcl/pltcl.c
+++ b/src/pl/tcl/pltcl.c
@@ -2456,6 +2456,7 @@ pltcl_process_SPI_result(Tcl_Interp *int
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
+		case SPI_OK_MERGE_RETURNING:
 
 			/*
 			 * Process the tuples we got
diff --git a/src/test/regress/expected/merge.out b/src/test/regress/expected/merge.out
new file mode 100644
index f87905f..8baa0f4
--- a/src/test/regress/expected/merge.out
+++ b/src/test/regress/expected/merge.out
@@ -123,20 +123,20 @@ ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
 ERROR:  name "target" specified more than once
 DETAIL:  The name is used both as MERGE target table and data source.
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
-ERROR:  MERGE not supported in WITH query
-LINE 1: WITH foo AS (
-             ^
--- used in COPY
+ERROR:  WITH query "foo" does not have a RETURNING clause
+LINE 4: ) SELECT * FROM foo;
+                        ^
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) TO stdout;
-ERROR:  MERGE not supported in COPY
+ERROR:  COPY query must have a RETURNING clause
 -- unsupported relation types
 -- view
 CREATE VIEW tv AS SELECT * FROM target;
@@ -1256,7 +1256,7 @@ BEGIN;
 MERGE INTO sq_target
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1271,7 +1271,7 @@ INSERT INTO sq_source (sid, balance, del
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1295,7 +1295,7 @@ WITH targq AS (
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1304,20 +1304,191 @@ WHEN MATCHED AND tid < 2 THEN
 ROLLBACK;
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
+CREATE TABLE merge_actions(action text, abbrev text);
+INSERT INTO merge_actions VALUES ('INSERT', 'ins'), ('UPDATE', 'upd'), ('DELETE', 'del');
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
-ERROR:  syntax error at or near "RETURNING"
-LINE 10: RETURNING *;
-         ^
+    DELETE
+RETURNING MERGING(CLAUSE_NUMBER) AS clause,
+          (SELECT abbrev FROM merge_actions
+            WHERE action = MERGING(ACTION)) AS action,
+          t.*,
+          CASE MERGING(ACTION)
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ clause | action | tid | balance |     description     
+--------+--------+-----+---------+---------------------
+      3 | del    |   1 |     100 | Removed (1,100)
+      1 | upd    |   2 |     220 | Added 20 to balance
+      2 | ins    |   4 |      40 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- error when using MERGING() outside MERGE
+SELECT MERGING(ACTION) FROM sq_target;
+ERROR:  MERGING() can only be used in the RETURNING list of a MERGE command
+LINE 1: SELECT MERGING(ACTION) FROM sq_target;
+               ^
+UPDATE sq_target SET balance = balance + 1 RETURNING MERGING(CLAUSE_NUMBER);
+ERROR:  MERGING() can only be used in the RETURNING list of a MERGE command
+LINE 1: ...ATE sq_target SET balance = balance + 1 RETURNING MERGING(CL...
+                                                             ^
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(CLAUSE_NUMBER), MERGING(ACTION), t.*,
+              CASE MERGING(ACTION)
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING action, MERGING(ACTION) AS log_action, l.*
+)
+SELECT * FROM m2;
+ action | log_action | tid |     last_change     
+--------+------------+-----+---------------------
+ DELETE | UPDATE     |   1 | Removed (1,100)
+ UPDATE | INSERT     |   2 | Added 20 to balance
+ INSERT | INSERT     |   4 | Inserted (4,40)
+(3 rows)
+
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ tid |     last_change     
+-----+---------------------
+   1 | Removed (1,100)
+   2 | Added 20 to balance
+   4 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.*
+) TO stdout;
+DELETE	1	100
+UPDATE	2	220
+INSERT	4	40
+ROLLBACK;
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ action | tid | new_balance 
+--------+-----+-------------
+ DELETE |   1 |         100
+ UPDATE |   3 |         320
+ INSERT |   4 |         110
+(3 rows)
+
+ROLLBACK;
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ action | tid | balance 
+--------+-----+---------
+ DELETE |   1 |     100
+ UPDATE |   2 |     220
+ INSERT |   4 |      40
+(3 rows)
+
+ROLLBACK;
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ r_action | r_tid | r_balance 
+----------+-------+-----------
+ DELETE   |     1 |       100
+ UPDATE   |     3 |       320
+ INSERT   |     4 |       110
+(3 rows)
+
 ROLLBACK;
 -- EXPLAIN
 CREATE TABLE ex_mtarget (a int, b int)
@@ -1514,7 +1685,7 @@ SELECT * FROM sq_target WHERE tid = 1;
 (1 row)
 
 ROLLBACK;
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 NOTICE:  drop cascades to view v
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1640,6 +1811,32 @@ SELECT * FROM pa_target ORDER BY tid;
 (14 rows)
 
 ROLLBACK;
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING MERGING(ACTION), MERGING(CLAUSE_NUMBER), t.*;
+ action | clause_number | tid | balance |           val            
+--------+---------------+-----+---------+--------------------------
+ UPDATE |             1 |   2 |     110 | initial updated by merge
+(1 row)
+
+SELECT * FROM pa_target ORDER BY tid;
+ tid | balance |           val            
+-----+---------+--------------------------
+   2 |     110 | initial updated by merge
+   3 |     300 | initial
+   5 |     500 | initial
+   7 |     700 | initial
+   9 |     900 | initial
+  11 |    1100 | initial
+  13 |    1300 | initial
+(7 rows)
+
+ROLLBACK;
 DROP TABLE pa_target CASCADE;
 -- The target table is partitioned in the same way, but this time by attaching
 -- partitions which have columns in different order, dropped columns etc.
@@ -1889,7 +2086,21 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING MERGING(ACTION), MERGING(CLAUSE_NUMBER), t.*;
+ action | clause_number |          logts           | tid | balance |           val            
+--------+---------------+--------------------------+-----+---------+--------------------------
+ UPDATE |             1 | Tue Jan 31 00:00:00 2017 |   1 |     110 | initial updated by merge
+ UPDATE |             1 | Tue Feb 28 00:00:00 2017 |   2 |     220 | initial updated by merge
+ INSERT |             2 | Sun Jan 15 00:00:00 2017 |   3 |      30 | inserted by merge
+ UPDATE |             1 | Tue Jan 31 00:00:00 2017 |   4 |     440 | initial updated by merge
+ UPDATE |             1 | Tue Feb 28 00:00:00 2017 |   5 |     550 | initial updated by merge
+ INSERT |             2 | Sun Jan 15 00:00:00 2017 |   6 |      60 | inserted by merge
+ UPDATE |             1 | Tue Jan 31 00:00:00 2017 |   7 |     770 | initial updated by merge
+ UPDATE |             1 | Tue Feb 28 00:00:00 2017 |   8 |     880 | initial updated by merge
+ INSERT |             2 | Sun Jan 15 00:00:00 2017 |   9 |      90 | inserted by merge
+(9 rows)
+
 SELECT * FROM pa_target ORDER BY tid;
           logts           | tid | balance |           val            
 --------------------------+-----+---------+--------------------------
diff --git a/src/test/regress/expected/rowsecurity.out b/src/test/regress/expected/rowsecurity.out
new file mode 100644
index 6988128..3fd4fab
--- a/src/test/regress/expected/rowsecurity.out
+++ b/src/test/regress/expected/rowsecurity.out
@@ -2302,6 +2302,35 @@ WHEN MATCHED THEN
 	UPDATE SET dnotes = dnotes || ' notes added by merge8 '
 WHEN NOT MATCHED THEN
 	INSERT VALUES (13, 44, 1, 'regress_rls_bob', 'new manga');
+SELECT * FROM document WHERE did = 13;
+ did | cid | dlevel | dauthor | dtitle | dnotes 
+-----+-----+--------+---------+--------+--------
+(0 rows)
+
+-- but not OK if RETURNING is used
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge9 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 44, 1, 'regress_rls_bob', 'new manga')
+RETURNING *;
+ERROR:  new row violates row-level security policy for table "document"
+-- but OK if new row is visible
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge10 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 11, 1, 'regress_rls_bob', 'new novel')
+RETURNING *;
+ sdid | did | cid | dlevel |     dauthor     |  dtitle   | dnotes 
+------+-----+-----+--------+-----------------+-----------+--------
+   14 |  14 |  11 |      1 | regress_rls_bob | new novel | 
+(1 row)
+
 RESET SESSION AUTHORIZATION;
 -- drop the restrictive SELECT policy so that we can look at the
 -- final state of the table
@@ -2325,7 +2354,8 @@ SELECT * FROM document;
   12 |  11 |      1 | regress_rls_bob   | another novel                    | 
    1 |  11 |      1 | regress_rls_bob   | my first novel                   |  notes added by merge2  notes added by merge3  notes added by merge4  notes added by merge7 
   13 |  44 |      1 | regress_rls_bob   | new manga                        | 
-(15 rows)
+  14 |  11 |      1 | regress_rls_bob   | new novel                        | 
+(16 rows)
 
 --
 -- ROLE/GROUP
diff --git a/src/test/regress/expected/rules.out b/src/test/regress/expected/rules.out
new file mode 100644
index 55f2e95..57fb335
--- a/src/test/regress/expected/rules.out
+++ b/src/test/regress/expected/rules.out
@@ -3623,7 +3623,8 @@ MERGE INTO rule_merge2 t USING (SELECT 1
 -- test deparsing
 CREATE TABLE sf_target(id int, data text, filling int[]);
 CREATE FUNCTION merge_sf_test()
- RETURNS void
+ RETURNS TABLE(action text, clause_number int,
+               a int, b text, id int, data text, filling int[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -3660,11 +3661,13 @@ WHEN NOT MATCHED
    VALUES (s.a, s.b, DEFAULT)
 WHEN NOT MATCHED
    THEN INSERT (filling[1], id)
-   VALUES (s.a, s.a);
+   VALUES (s.a, s.a)
+RETURNING
+   merging(action), merging(clause_number) AS clause_number, *;
 END;
 \sf merge_sf_test
 CREATE OR REPLACE FUNCTION public.merge_sf_test()
- RETURNS void
+ RETURNS TABLE(action text, clause_number integer, a integer, b text, id integer, data text, filling integer[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -3701,7 +3704,14 @@ BEGIN ATOMIC
       VALUES (s.a, s.b, DEFAULT)
     WHEN NOT MATCHED
      THEN INSERT (filling[1], id)
-      VALUES (s.a, s.a);
+      VALUES (s.a, s.a)
+   RETURNING MERGING(ACTION) AS action,
+     MERGING(CLAUSE_NUMBER) AS clause_number,
+     s.a,
+     s.b,
+     t.id,
+     t.data,
+     t.filling;
 END
 DROP FUNCTION merge_sf_test;
 DROP TABLE sf_target;
diff --git a/src/test/regress/expected/with.out b/src/test/regress/expected/with.out
new file mode 100644
index 7d796ea..4666873
--- a/src/test/regress/expected/with.out
+++ b/src/test/regress/expected/with.out
@@ -3493,6 +3493,16 @@ SELECT * FROM t;
 ERROR:  WITH query "t" does not have a RETURNING clause
 LINE 4: SELECT * FROM t;
                       ^
+-- RETURNING tries to return its own output
+WITH RECURSIVE t(action, a) AS (
+	MERGE INTO y USING (VALUES (11)) v(a) ON y.a = v.a
+		WHEN NOT MATCHED THEN INSERT VALUES (v.a)
+		RETURNING MERGING(ACTION), (SELECT a FROM t)
+)
+SELECT * FROM t;
+ERROR:  recursive query "t" must not contain data-modifying statements
+LINE 1: WITH RECURSIVE t(action, a) AS (
+                       ^
 -- data-modifying WITH allowed only at the top level
 SELECT * FROM (
 	WITH t AS (UPDATE y SET a=a+1 RETURNING *)
diff --git a/src/test/regress/sql/merge.sql b/src/test/regress/sql/merge.sql
new file mode 100644
index 66cb75a..daa9dde
--- a/src/test/regress/sql/merge.sql
+++ b/src/test/regress/sql/merge.sql
@@ -86,12 +86,12 @@ MERGE INTO target
 USING target
 ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
--- used in COPY
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
@@ -816,7 +816,7 @@ BEGIN;
 MERGE INTO sq_target
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -829,7 +829,7 @@ INSERT INTO sq_source (sid, balance, del
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -847,7 +847,7 @@ WITH targq AS (
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -857,17 +857,146 @@ ROLLBACK;
 
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
+CREATE TABLE merge_actions(action text, abbrev text);
+INSERT INTO merge_actions VALUES ('INSERT', 'ins'), ('UPDATE', 'upd'), ('DELETE', 'del');
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
+    DELETE
+RETURNING MERGING(CLAUSE_NUMBER) AS clause,
+          (SELECT abbrev FROM merge_actions
+            WHERE action = MERGING(ACTION)) AS action,
+          t.*,
+          CASE MERGING(ACTION)
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ROLLBACK;
+
+-- error when using MERGING() outside MERGE
+SELECT MERGING(ACTION) FROM sq_target;
+UPDATE sq_target SET balance = balance + 1 RETURNING MERGING(CLAUSE_NUMBER);
+
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(CLAUSE_NUMBER), MERGING(ACTION), t.*,
+              CASE MERGING(ACTION)
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING action, MERGING(ACTION) AS log_action, l.*
+)
+SELECT * FROM m2;
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ROLLBACK;
+
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.*
+) TO stdout;
+ROLLBACK;
+
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ROLLBACK;
+
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ROLLBACK;
+
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
 ROLLBACK;
 
 -- EXPLAIN
@@ -966,7 +1095,7 @@ WHEN MATCHED THEN
 SELECT * FROM sq_target WHERE tid = 1;
 ROLLBACK;
 
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1033,6 +1162,17 @@ SELECT merge_func();
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING MERGING(ACTION), MERGING(CLAUSE_NUMBER), t.*;
+SELECT * FROM pa_target ORDER BY tid;
+ROLLBACK;
+
 DROP TABLE pa_target CASCADE;
 
 -- The target table is partitioned in the same way, but this time by attaching
@@ -1209,7 +1349,8 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING MERGING(ACTION), MERGING(CLAUSE_NUMBER), t.*;
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
diff --git a/src/test/regress/sql/rowsecurity.sql b/src/test/regress/sql/rowsecurity.sql
new file mode 100644
index dec7340..1d5ed0a
--- a/src/test/regress/sql/rowsecurity.sql
+++ b/src/test/regress/sql/rowsecurity.sql
@@ -992,6 +992,27 @@ WHEN MATCHED THEN
 	UPDATE SET dnotes = dnotes || ' notes added by merge8 '
 WHEN NOT MATCHED THEN
 	INSERT VALUES (13, 44, 1, 'regress_rls_bob', 'new manga');
+SELECT * FROM document WHERE did = 13;
+
+-- but not OK if RETURNING is used
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge9 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 44, 1, 'regress_rls_bob', 'new manga')
+RETURNING *;
+
+-- but OK if new row is visible
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge10 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 11, 1, 'regress_rls_bob', 'new novel')
+RETURNING *;
 
 RESET SESSION AUTHORIZATION;
 -- drop the restrictive SELECT policy so that we can look at the
diff --git a/src/test/regress/sql/rules.sql b/src/test/regress/sql/rules.sql
new file mode 100644
index 93aff4e..4b15a13
--- a/src/test/regress/sql/rules.sql
+++ b/src/test/regress/sql/rules.sql
@@ -1281,7 +1281,8 @@ MERGE INTO rule_merge2 t USING (SELECT 1
 CREATE TABLE sf_target(id int, data text, filling int[]);
 
 CREATE FUNCTION merge_sf_test()
- RETURNS void
+ RETURNS TABLE(action text, clause_number int,
+               a int, b text, id int, data text, filling int[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -1318,7 +1319,9 @@ WHEN NOT MATCHED
    VALUES (s.a, s.b, DEFAULT)
 WHEN NOT MATCHED
    THEN INSERT (filling[1], id)
-   VALUES (s.a, s.a);
+   VALUES (s.a, s.a)
+RETURNING
+   merging(action), merging(clause_number) AS clause_number, *;
 END;
 
 \sf merge_sf_test
diff --git a/src/test/regress/sql/with.sql b/src/test/regress/sql/with.sql
new file mode 100644
index f8a213e..1d82d4a
--- a/src/test/regress/sql/with.sql
+++ b/src/test/regress/sql/with.sql
@@ -1591,6 +1591,14 @@ WITH t AS (
 )
 SELECT * FROM t;
 
+-- RETURNING tries to return its own output
+WITH RECURSIVE t(action, a) AS (
+	MERGE INTO y USING (VALUES (11)) v(a) ON y.a = v.a
+		WHEN NOT MATCHED THEN INSERT VALUES (v.a)
+		RETURNING MERGING(ACTION), (SELECT a FROM t)
+)
+SELECT * FROM t;
+
 -- data-modifying WITH allowed only at the top level
 SELECT * FROM (
 	WITH t AS (UPDATE y SET a=a+1 RETURNING *)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
new file mode 100644
index 29fd1ca..82e1747
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -1570,6 +1570,8 @@ MergePath
 MergeScanSelCache
 MergeStmt
 MergeWhenClause
+MergingFunc
+MergingFuncOp
 MetaCommand
 MinMaxAggInfo
 MinMaxAggPath

diff --git a/doc/src/sgml/dml.sgml b/doc/src/sgml/dml.sgml
new file mode 100644
index cbbc5e2..3d95bdb
--- a/doc/src/sgml/dml.sgml
+++ b/doc/src/sgml/dml.sgml
@@ -283,10 +283,15 @@ DELETE FROM products;
    <secondary>RETURNING</secondary>
   </indexterm>
 
+  <indexterm zone="dml-returning">
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
   <para>
    Sometimes it is useful to obtain data from modified rows while they are
    being manipulated.  The <command>INSERT</command>, <command>UPDATE</command>,
-   and <command>DELETE</command> commands all have an
+   <command>DELETE</command>, and <command>MERGE</command> commands all have an
    optional <literal>RETURNING</literal> clause that supports this.  Use
    of <literal>RETURNING</literal> avoids performing an extra database query to
    collect the data, and is especially valuable when it would otherwise be
@@ -339,6 +344,21 @@ DELETE FROM products
 </programlisting>
   </para>
 
+  <para>
+   In a <command>MERGE</command>, the data available to <literal>RETURNING</literal> is
+   the content of the source row plus the content of the inserted, updated, or
+   deleted target row.  Since it is quite common for the source and target to
+   have many of the same columns, specifying <literal>RETURNING *</literal>
+   can lead to a lot of duplicated columns, so it is often more useful to
+   qualify it so as to return just the source or target row.  For example:
+<programlisting>
+MERGE INTO products p USING new_products n ON p.product_no = n.product_no
+  WHEN NOT MATCHED THEN INSERT VALUES (n.product_no, n.name, n.price)
+  WHEN MATCHED THEN UPDATE SET name = n.name, price = n.price
+  RETURNING p.*;
+</programlisting>
+  </para>
+
   <para>
    If there are triggers (<xref linkend="triggers"/>) on the target table,
    the data available to <literal>RETURNING</literal> is the row as modified by
diff --git a/doc/src/sgml/func.sgml b/doc/src/sgml/func.sgml
new file mode 100644
index 6788ba8..a26a06d
--- a/doc/src/sgml/func.sgml
+++ b/doc/src/sgml/func.sgml
@@ -22417,6 +22417,84 @@ SELECT count(*) FROM sometable;
 
  </sect1>
 
+ <sect1 id="functions-merge-support">
+  <title>Merge Support Functions</title>
+
+  <indexterm>
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
+  <indexterm>
+   <primary>MERGING</primary>
+  </indexterm>
+
+  <para>
+   <productname>PostgreSQL</productname> includes one merge support function
+   that may be used in the <literal>RETURNING</literal> list of a
+   <xref linkend="sql-merge"/> command to return additional information about
+   the action taken for each row:
+<synopsis>
+<function id="function-merging">MERGING</function> ( <parameter>property</parameter> )
+</synopsis>
+   The <parameter>property</parameter> key word specifies what to return:
+
+   <variablelist>
+    <varlistentry>
+     <term><literal>ACTION</literal></term>
+     <listitem>
+      <para>
+       The merge action command executed for the current row.  This will be
+       <literal>'INSERT'</literal>, <literal>'UPDATE'</literal>, or
+       <literal>'DELETE'</literal>.
+      </para>
+     </listitem>
+    </varlistentry>
+
+    <varlistentry>
+     <term><literal>CLAUSE_NUMBER</literal></term>
+     <listitem>
+      <para>
+       The 1-based index of the <literal>WHEN</literal> clause executed for
+       the current row.
+      </para>
+     </listitem>
+    </varlistentry>
+   </variablelist>
+  </para>
+
+  <para>
+   The following example illustrates how this may be used to determine
+   which actions were performed for each row affected by the
+   <command>MERGE</command> command:
+<screen><![CDATA[
+MERGE INTO products p
+  USING stock s ON p.product_id = s.product_id
+  WHEN MATCHED AND s.quantity > 0 THEN
+    UPDATE SET in_stock = true, quantity = s.quantity
+  WHEN MATCHED THEN
+    UPDATE SET in_stock = false, quantity = 0
+  WHEN NOT MATCHED THEN
+    INSERT (product_id, in_stock, quantity)
+      VALUES (s.product_id, true, s.quantity)
+  RETURNING MERGING(ACTION), MERGING(CLAUSE_NUMBER), p.*;
+
+ action | clause_number | product_id | in_stock | quantity
+--------+---------------+------------+----------+----------
+ UPDATE |             1 |       1001 | t        |       50
+ UPDATE |             2 |       1002 | f        |        0
+ INSERT |             3 |       1003 | t        |       10
+]]></screen>
+  </para>
+
+  <para>
+   Note that this function can only be used in the <literal>RETURNING</literal>
+   list of a <command>MERGE</command> command.  It is an error to use it in any
+   other part of a query.
+  </para>
+
+ </sect1>
+
  <sect1 id="functions-subquery">
   <title>Subquery Expressions</title>
 
diff --git a/doc/src/sgml/glossary.sgml b/doc/src/sgml/glossary.sgml
new file mode 100644
index 8c2f114..a81c17a
--- a/doc/src/sgml/glossary.sgml
+++ b/doc/src/sgml/glossary.sgml
@@ -1442,9 +1442,9 @@
      to a <glossterm linkend="glossary-client">client</glossterm> upon the
      completion of an <acronym>SQL</acronym> command, usually a
      <command>SELECT</command> but it can be an
-     <command>INSERT</command>, <command>UPDATE</command>, or
-     <command>DELETE</command> command if the <literal>RETURNING</literal>
-     clause is specified.
+     <command>INSERT</command>, <command>UPDATE</command>,
+     <command>DELETE</command>, or <command>MERGE</command> command if the
+     <literal>RETURNING</literal> clause is specified.
     </para>
     <para>
      The fact that a result set is a relation means that a query can be used
diff --git a/doc/src/sgml/plpgsql.sgml b/doc/src/sgml/plpgsql.sgml
new file mode 100644
index c2b9c6a..406834e
--- a/doc/src/sgml/plpgsql.sgml
+++ b/doc/src/sgml/plpgsql.sgml
@@ -1043,8 +1043,8 @@ INSERT INTO mytable VALUES (1,'one'), (2
     </para>
 
     <para>
-     If the command does return rows (for example <command>SELECT</command>,
-     or <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>
+     If the command does return rows (for example <command>SELECT</command>, or
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
      with <literal>RETURNING</literal>), there are two ways to proceed.
      When the command will return at most one row, or you only care about
      the first row of output, write the command as usual but add
@@ -1172,6 +1172,7 @@ SELECT <replaceable>select_expressions</
 INSERT ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 UPDATE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 DELETE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
+MERGE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 </synopsis>
 
      where <replaceable>target</replaceable> can be a record variable, a row
@@ -1182,8 +1183,8 @@ DELETE ... RETURNING <replaceable>expres
      <literal>INTO</literal> clause) just as described above,
      and the plan is cached in the same way.
      This works for <command>SELECT</command>,
-     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
-     <literal>RETURNING</literal>, and certain utility commands
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
+     with <literal>RETURNING</literal>, and certain utility commands
      that return row sets, such as <command>EXPLAIN</command>.
      Except for the <literal>INTO</literal> clause, the SQL command is the same
      as it would be written outside <application>PL/pgSQL</application>.
@@ -1259,7 +1260,7 @@ END;
     </para>
 
     <para>
-     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
+     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command> with
      <literal>RETURNING</literal>, <application>PL/pgSQL</application> reports
      an error for more than one returned row, even when
      <literal>STRICT</literal> is not specified.  This is because there
diff --git a/doc/src/sgml/queries.sgml b/doc/src/sgml/queries.sgml
new file mode 100644
index 648b283..372cce1
--- a/doc/src/sgml/queries.sgml
+++ b/doc/src/sgml/queries.sgml
@@ -2063,9 +2063,10 @@ SELECT <replaceable>select_list</replace
    Table Expressions or <acronym>CTE</acronym>s, can be thought of as defining
    temporary tables that exist just for one query.  Each auxiliary statement
    in a <literal>WITH</literal> clause can be a <command>SELECT</command>,
-   <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>; and the
+   <command>INSERT</command>, <command>UPDATE</command>, <command>DELETE</command>,
+   or <command>MERGE</command>; and the
    <literal>WITH</literal> clause itself is attached to a primary statement that can
-   be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
+   also be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
    <command>DELETE</command>, or <command>MERGE</command>.
   </para>
 
@@ -2599,8 +2600,8 @@ SELECT * FROM w AS w1 JOIN w AS w2 ON w1
    <title>Data-Modifying Statements in <literal>WITH</literal></title>
 
    <para>
-    You can use most data-modifying statements (<command>INSERT</command>,
-    <command>UPDATE</command>, or <command>DELETE</command>, but not
+    You can use data-modifying statements (<command>INSERT</command>,
+    <command>UPDATE</command>, <command>DELETE</command>, or
     <command>MERGE</command>) in <literal>WITH</literal>.  This
     allows you to perform several different operations in the same query.
     An example is:
diff --git a/doc/src/sgml/ref/copy.sgml b/doc/src/sgml/ref/copy.sgml
new file mode 100644
index 21a5c4a..d64db80
--- a/doc/src/sgml/ref/copy.sgml
+++ b/doc/src/sgml/ref/copy.sgml
@@ -123,16 +123,19 @@ COPY { <replaceable class="parameter">ta
       A <link linkend="sql-select"><command>SELECT</command></link>,
       <link linkend="sql-values"><command>VALUES</command></link>,
       <link linkend="sql-insert"><command>INSERT</command></link>,
-      <link linkend="sql-update"><command>UPDATE</command></link>, or
-      <link linkend="sql-delete"><command>DELETE</command></link> command whose results are to be
-      copied.  Note that parentheses are required around the query.
+      <link linkend="sql-update"><command>UPDATE</command></link>,
+      <link linkend="sql-delete"><command>DELETE</command></link>, or
+      <link linkend="sql-merge"><command>MERGE</command></link> command
+      whose results are to be copied.  Note that parentheses are required
+      around the query.
      </para>
      <para>
-      For <command>INSERT</command>, <command>UPDATE</command> and
-      <command>DELETE</command> queries a <literal>RETURNING</literal> clause
-      must be provided, and the target relation must not have a conditional
-      rule, nor an <literal>ALSO</literal> rule, nor an
-      <literal>INSTEAD</literal> rule that expands to multiple statements.
+      For <command>INSERT</command>, <command>UPDATE</command>,
+      <command>DELETE</command>, and <command>MERGE</command> queries a
+      <literal>RETURNING</literal> clause must be provided, and the target
+      relation must not have a conditional rule, nor an
+      <literal>ALSO</literal> rule, nor an <literal>INSTEAD</literal> rule
+      that expands to multiple statements.
      </para>
     </listitem>
    </varlistentry>
diff --git a/doc/src/sgml/ref/merge.sgml b/doc/src/sgml/ref/merge.sgml
new file mode 100644
index 655f7dc..44b3fa4
--- a/doc/src/sgml/ref/merge.sgml
+++ b/doc/src/sgml/ref/merge.sgml
@@ -25,6 +25,7 @@ PostgreSQL documentation
 MERGE INTO [ ONLY ] <replaceable class="parameter">target_table_name</replaceable> [ * ] [ [ AS ] <replaceable class="parameter">target_alias</replaceable> ]
 USING <replaceable class="parameter">data_source</replaceable> ON <replaceable class="parameter">join_condition</replaceable>
 <replaceable class="parameter">when_clause</replaceable> [...]
+[ RETURNING * | <replaceable class="parameter">output_expression</replaceable> [ [ AS ] <replaceable class="parameter">output_name</replaceable> ] [, ...] ]
 
 <phrase>where <replaceable class="parameter">data_source</replaceable> is:</phrase>
 
@@ -95,6 +96,18 @@ DELETE
   </para>
 
   <para>
+   The optional <literal>RETURNING</literal> clause causes <command>MERGE</command>
+   to compute and return value(s) based on each row inserted, updated, or
+   deleted.  Any expression using the source or target table's columns, or
+   the <xref linkend="function-merging"/> function can be computed.  When an
+   <command>INSERT</command> or <command>UPDATE</command> action is performed,
+   the new values of the target table's columns are used.  When a
+   <command>DELETE</command> is performed, the old values of the target table's
+   columns are used.  The syntax of the <literal>RETURNING</literal> list is
+   identical to that of the output list of <command>SELECT</command>.
+  </para>
+
+  <para>
    There is no separate <literal>MERGE</literal> privilege.
    If you specify an update action, you must have the
    <literal>UPDATE</literal> privilege on the column(s)
@@ -125,6 +138,18 @@ DELETE
 
   <variablelist>
    <varlistentry>
+    <term><replaceable class="parameter">with_query</replaceable></term>
+    <listitem>
+     <para>
+      The <literal>WITH</literal> clause allows you to specify one or more
+      subqueries that can be referenced by name in the <command>MERGE</command>
+      query. See <xref linkend="queries-with"/> and <xref linkend="sql-select"/>
+      for details.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
     <term><replaceable class="parameter">target_table_name</replaceable></term>
     <listitem>
      <para>
@@ -399,13 +424,30 @@ DELETE
    </varlistentry>
 
    <varlistentry>
-    <term><replaceable class="parameter">with_query</replaceable></term>
+    <term><replaceable class="parameter">output_expression</replaceable></term>
     <listitem>
      <para>
-      The <literal>WITH</literal> clause allows you to specify one or more
-      subqueries that can be referenced by name in the <command>MERGE</command>
-      query. See <xref linkend="queries-with"/> and <xref linkend="sql-select"/>
-      for details.
+      An expression to be computed and returned by the <command>MERGE</command>
+      command after each row is merged.  The expression can use any columns of
+      the source or target tables, or the <xref linkend="function-merging"/>
+      function to return additional information about the action executed.
+     </para>
+     <para>
+      Writing <literal>*</literal> will return all columns from the source
+      table, followed by all columns from the target table.  Often this will
+      lead to a lot of duplication, since it is common for the source and
+      target tables to have a lot of the same columns.  This can be avoided by
+      qualifying the <literal>*</literal> with the name or alias of the source
+      or target table.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
+    <term><replaceable class="parameter">output_name</replaceable></term>
+    <listitem>
+     <para>
+      A name to use for a returned column.
      </para>
     </listitem>
    </varlistentry>
@@ -428,6 +470,13 @@ MERGE <replaceable class="parameter">tot
    were changed in any way.
   </para>
 
+  <para>
+   If the <command>MERGE</command> command contains a <literal>RETURNING</literal>
+   clause, the result will be similar to that of a <command>SELECT</command>
+   statement containing the columns and values defined in the
+   <literal>RETURNING</literal> list, computed over the row(s) inserted, updated,
+   or deleted by the command.
+  </para>
  </refsect1>
 
  <refsect1>
@@ -544,13 +593,6 @@ MERGE <replaceable class="parameter">tot
   </para>
 
   <para>
-   There is no <literal>RETURNING</literal> clause with
-   <command>MERGE</command>.  Actions of <command>INSERT</command>,
-   <command>UPDATE</command> and <command>DELETE</command> cannot contain
-   <literal>RETURNING</literal> or <literal>WITH</literal> clauses.
-  </para>
-
-  <para>
    When <command>MERGE</command> is run concurrently with other commands
    that modify the target table, the usual transaction isolation rules
    apply; see <xref linkend="transaction-iso"/> for an explanation
@@ -602,7 +644,8 @@ WHEN NOT MATCHED THEN
   <para>
    Attempt to insert a new stock item along with the quantity of stock. If
    the item already exists, instead update the stock count of the existing
-   item. Don't allow entries that have zero stock.
+   item. Don't allow entries that have zero stock. Return details of all
+   changes made.
 <programlisting>
 MERGE INTO wines w
 USING wine_stock_changes s
@@ -612,7 +655,8 @@ WHEN NOT MATCHED AND s.stock_delta > 0 T
 WHEN MATCHED AND w.stock + s.stock_delta > 0 THEN
   UPDATE SET stock = w.stock + s.stock_delta
 WHEN MATCHED THEN
-  DELETE;
+  DELETE
+RETURNING MERGING(ACTION), w.*;
 </programlisting>
 
    The <literal>wine_stock_changes</literal> table might be, for example, a
@@ -627,8 +671,9 @@ WHEN MATCHED THEN
     This command conforms to the <acronym>SQL</acronym> standard.
   </para>
    <para>
-    The <literal>WITH</literal> clause and <literal>DO NOTHING</literal>
-    action are extensions to the <acronym>SQL</acronym> standard.
+    The <literal>WITH</literal> clause, <literal>DO NOTHING</literal> action,
+    and <literal>RETURNING</literal> clause are extensions to the
+    <acronym>SQL</acronym> standard.
   </para>
  </refsect1>
 </refentry>
diff --git a/doc/src/sgml/ref/select.sgml b/doc/src/sgml/ref/select.sgml
new file mode 100644
index 9917df7..066aed4
--- a/doc/src/sgml/ref/select.sgml
+++ b/doc/src/sgml/ref/select.sgml
@@ -74,7 +74,7 @@ SELECT [ ALL | DISTINCT [ ON ( <replacea
 
 <phrase>and <replaceable class="parameter">with_query</replaceable> is:</phrase>
 
-    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> )
+    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> | <replaceable class="parameter">merge</replaceable> )
         [ SEARCH { BREADTH | DEPTH } FIRST BY <replaceable>column_name</replaceable> [, ...] SET <replaceable>search_seq_col_name</replaceable> ]
         [ CYCLE <replaceable>column_name</replaceable> [, ...] SET <replaceable>cycle_mark_col_name</replaceable> [ TO <replaceable>cycle_mark_value</replaceable> DEFAULT <replaceable>cycle_mark_default</replaceable> ] USING <replaceable>cycle_path_col_name</replaceable> ]
 
@@ -230,10 +230,10 @@ TABLE [ ONLY ] <replaceable class="param
     The subqueries effectively act as temporary tables or views
     for the duration of the primary query.
     Each subquery can be a <command>SELECT</command>, <command>TABLE</command>, <command>VALUES</command>,
-    <command>INSERT</command>, <command>UPDATE</command> or
-    <command>DELETE</command> statement.
+    <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command> statement.
     When writing a data-modifying statement (<command>INSERT</command>,
-    <command>UPDATE</command> or <command>DELETE</command>) in
+    <command>UPDATE</command>, <command>DELETE</command>, or <command>MERGE</command>) in
     <literal>WITH</literal>, it is usual to include a <literal>RETURNING</literal> clause.
     It is the output of <literal>RETURNING</literal>, <emphasis>not</emphasis> the underlying
     table that the statement modifies, that forms the temporary table that is
@@ -2184,7 +2184,8 @@ SELECT 2+2;
 
    <para>
     <productname>PostgreSQL</productname> allows <command>INSERT</command>,
-    <command>UPDATE</command>, and <command>DELETE</command> to be used as <literal>WITH</literal>
+    <command>UPDATE</command>, <command>DELETE</command>, and
+    <command>MERGE</command> to be used as <literal>WITH</literal>
     queries.  This is not found in the SQL standard.
    </para>
   </refsect2>
diff --git a/doc/src/sgml/rowtypes.sgml b/doc/src/sgml/rowtypes.sgml
new file mode 100644
index 4d86f97..bbeac84
--- a/doc/src/sgml/rowtypes.sgml
+++ b/doc/src/sgml/rowtypes.sgml
@@ -348,7 +348,7 @@ SELECT m.* FROM some_table, LATERAL myfu
    column expansion of this kind when it appears at the top level of
    a <link linkend="queries-select-lists"><command>SELECT</command> output
    list</link>, a <link linkend="dml-returning"><literal>RETURNING</literal>
-   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>,
+   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>,
    a <link linkend="queries-values"><literal>VALUES</literal> clause</link>, or
    a <link linkend="sql-syntax-row-constructors">row constructor</link>.
    In all other contexts (including when nested inside one of those
diff --git a/doc/src/sgml/spi.sgml b/doc/src/sgml/spi.sgml
new file mode 100644
index 47f4b5b..bb37786
--- a/doc/src/sgml/spi.sgml
+++ b/doc/src/sgml/spi.sgml
@@ -301,8 +301,9 @@ SPI_execute("INSERT INTO foo SELECT * FR
    is returned in the global variable <varname>SPI_processed</varname>.
    If the return value of the function is <symbol>SPI_OK_SELECT</symbol>,
    <symbol>SPI_OK_INSERT_RETURNING</symbol>,
-   <symbol>SPI_OK_DELETE_RETURNING</symbol>, or
-   <symbol>SPI_OK_UPDATE_RETURNING</symbol>,
+   <symbol>SPI_OK_DELETE_RETURNING</symbol>,
+   <symbol>SPI_OK_UPDATE_RETURNING</symbol>, or
+   <symbol>SPI_OK_MERGE_RETURNING</symbol>,
    then you can use the
    global pointer <literal>SPITupleTable *SPI_tuptable</literal> to
    access the result rows.  Some utility commands (such as
@@ -472,6 +473,15 @@ typedef struct SPITupleTable
       </para>
      </listitem>
     </varlistentry>
+
+    <varlistentry>
+     <term><symbol>SPI_OK_MERGE_RETURNING</symbol></term>
+     <listitem>
+      <para>
+       if a <command>MERGE RETURNING</command> was executed
+      </para>
+     </listitem>
+    </varlistentry>
 
     <varlistentry>
      <term><symbol>SPI_OK_UTILITY</symbol></term>
diff --git a/doc/src/sgml/xfunc.sgml b/doc/src/sgml/xfunc.sgml
new file mode 100644
index 8a79ad0..7d05369
--- a/doc/src/sgml/xfunc.sgml
+++ b/doc/src/sgml/xfunc.sgml
@@ -177,7 +177,8 @@
     statements separated by semicolons.  A semicolon after the last
     statement is optional.  Unless the function is declared to return
     <type>void</type>, the last statement must be a <command>SELECT</command>,
-    or an <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>
+    or an <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command>
     that has a <literal>RETURNING</literal> clause.
    </para>
 
@@ -1236,8 +1237,9 @@ SELECT x, CASE WHEN x &gt; 0 THEN genera
 
     <note>
      <para>
-      If a function's last command is <command>INSERT</command>, <command>UPDATE</command>,
-      or <command>DELETE</command> with <literal>RETURNING</literal>, that command will
+      If a function's last command is <command>INSERT</command>,
+      <command>UPDATE</command>, <command>DELETE</command>, or
+      <command>MERGE</command> with <literal>RETURNING</literal>, that command will
       always be executed to completion, even if the function is not declared
       with <literal>SETOF</literal> or the calling query does not fetch all the
       result rows.  Any extra rows produced by the <literal>RETURNING</literal>
diff --git a/src/backend/commands/copy.c b/src/backend/commands/copy.c
new file mode 100644
index cc0786c..e96067a
--- a/src/backend/commands/copy.c
+++ b/src/backend/commands/copy.c
@@ -283,12 +283,6 @@ DoCopy(ParseState *pstate, const CopyStm
 	{
 		Assert(stmt->query);
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(stmt->query, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in COPY"));
-
 		query = makeNode(RawStmt);
 		query->stmt = stmt->query;
 		query->stmt_location = stmt_location;
diff --git a/src/backend/commands/copyto.c b/src/backend/commands/copyto.c
new file mode 100644
index d3dc3fc..c29fecd
--- a/src/backend/commands/copyto.c
+++ b/src/backend/commands/copyto.c
@@ -510,7 +510,8 @@ BeginCopyTo(ParseState *pstate,
 		{
 			Assert(query->commandType == CMD_INSERT ||
 				   query->commandType == CMD_UPDATE ||
-				   query->commandType == CMD_DELETE);
+				   query->commandType == CMD_DELETE ||
+				   query->commandType == CMD_MERGE);
 
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
diff --git a/src/backend/executor/execExpr.c b/src/backend/executor/execExpr.c
new file mode 100644
index 3181b11..c4c2fb5
--- a/src/backend/executor/execExpr.c
+++ b/src/backend/executor/execExpr.c
@@ -1108,6 +1108,21 @@ ExecInitExprRec(Expr *node, ExprState *s
 				break;
 			}
 
+		case T_MergingFunc:
+			{
+				MergingFunc *mf_node = (MergingFunc *) node;
+
+				if (!state->parent ||
+					!IsA(state->parent, ModifyTableState) ||
+					((ModifyTableState *) state->parent)->operation != CMD_MERGE)
+					elog(ERROR, "MergingFunc found in non-merge plan node");
+
+				scratch.opcode = EEOP_MERGING_FUNC;
+				scratch.d.merging_func.mfop = mf_node->mfop;
+				ExprEvalPushStep(state, &scratch);
+				break;
+			}
+
 		case T_SubscriptingRef:
 			{
 				SubscriptingRef *sbsref = (SubscriptingRef *) node;
diff --git a/src/backend/executor/execExprInterp.c b/src/backend/executor/execExprInterp.c
new file mode 100644
index 3f20f1d..4bcad29
--- a/src/backend/executor/execExprInterp.c
+++ b/src/backend/executor/execExprInterp.c
@@ -485,6 +485,7 @@ ExecInterpExpr(ExprState *state, ExprCon
 		&&CASE_EEOP_AGGREF,
 		&&CASE_EEOP_GROUPING_FUNC,
 		&&CASE_EEOP_WINDOW_FUNC,
+		&&CASE_EEOP_MERGING_FUNC,
 		&&CASE_EEOP_SUBPLAN,
 		&&CASE_EEOP_AGG_STRICT_DESERIALIZE,
 		&&CASE_EEOP_AGG_DESERIALIZE,
@@ -1593,6 +1594,14 @@ ExecInterpExpr(ExprState *state, ExprCon
 			EEO_NEXT();
 		}
 
+		EEO_CASE(EEOP_MERGING_FUNC)
+		{
+			/* too complex/uncommon for an inline implementation */
+			ExecEvalMergingFunc(state, op, econtext);
+
+			EEO_NEXT();
+		}
+
 		EEO_CASE(EEOP_SUBPLAN)
 		{
 			/* too complex for an inline implementation */
@@ -4247,6 +4256,59 @@ ExecEvalGroupingFunc(ExprState *state, E
 }
 
 /*
+ * ExecEvalMergingFunc
+ *
+ * Returns information about the current MERGE action for its RETURNING list.
+ */
+void
+ExecEvalMergingFunc(ExprState *state, ExprEvalStep *op, ExprContext *econtext)
+{
+	MergingFuncOp mfop = op->d.merging_func.mfop;
+	ModifyTableState *mtstate = castNode(ModifyTableState, state->parent);
+	MergeActionState *relaction = mtstate->mt_merge_action;
+
+	if (!relaction)
+		elog(ERROR, "no merge action in progress");
+
+	switch (mfop)
+	{
+		case MERGING_ACTION:
+			/* Return the MERGE action ("INSERT", "UPDATE", or "DELETE") */
+			switch (relaction->mas_action->commandType)
+			{
+				case CMD_INSERT:
+					*op->resvalue = PointerGetDatum(cstring_to_text_with_len("INSERT", 6));
+					*op->resnull = false;
+					break;
+				case CMD_UPDATE:
+					*op->resvalue = PointerGetDatum(cstring_to_text_with_len("UPDATE", 6));
+					*op->resnull = false;
+					break;
+				case CMD_DELETE:
+					*op->resvalue = PointerGetDatum(cstring_to_text_with_len("DELETE", 6));
+					*op->resnull = false;
+					break;
+				case CMD_NOTHING:
+					elog(ERROR, "unexpected merge action: DO NOTHING");
+					break;
+				default:
+					elog(ERROR, "unrecognized commandType: %d",
+						 (int) relaction->mas_action->commandType);
+			}
+			break;
+
+		case MERGING_CLAUSE_NUMBER:
+			/* Return the 1-based index of the MERGE action */
+			*op->resvalue = Int32GetDatum(relaction->mas_action->index);
+			*op->resnull = false;
+			break;
+
+		default:
+			elog(ERROR, "unrecognized MergingFuncOp: %d", (int) mfop);
+	}
+}
+
+/*
  * Hand off evaluation of a subplan to nodeSubplan.c
  */
 void
diff --git a/src/backend/executor/execPartition.c b/src/backend/executor/execPartition.c
new file mode 100644
index b22040a..9c56a10
--- a/src/backend/executor/execPartition.c
+++ b/src/backend/executor/execPartition.c
@@ -612,8 +612,8 @@ ExecInitPartitionInfo(ModifyTableState *
 	 * Build the RETURNING projection for the partition.  Note that we didn't
 	 * build the returningList for partitions within the planner, but simple
 	 * translation of varattnos will suffice.  This only occurs for the INSERT
-	 * case or in the case of UPDATE tuple routing where we didn't find a
-	 * result rel to reuse.
+	 * case or in the case of UPDATE/MERGE tuple routing where we didn't find
+	 * a result rel to reuse.
 	 */
 	if (node && node->returningLists != NIL)
 	{
@@ -622,12 +622,14 @@ ExecInitPartitionInfo(ModifyTableState *
 		List	   *returningList;
 
 		/* See the comment above for WCO lists. */
-		/* (except no RETURNING support for MERGE yet) */
 		Assert((node->operation == CMD_INSERT &&
 				list_length(node->returningLists) == 1 &&
 				list_length(node->resultRelations) == 1) ||
 			   (node->operation == CMD_UPDATE &&
 				list_length(node->returningLists) ==
+				list_length(node->resultRelations)) ||
+			   (node->operation == CMD_MERGE &&
+				list_length(node->returningLists) ==
 				list_length(node->resultRelations)));
 
 		/*
diff --git a/src/backend/executor/functions.c b/src/backend/executor/functions.c
new file mode 100644
index 0f811fd..7da02b0
--- a/src/backend/executor/functions.c
+++ b/src/backend/executor/functions.c
@@ -1660,8 +1660,8 @@ check_sql_fn_retval(List *queryTreeLists
 
 	/*
 	 * If it's a plain SELECT, it returns whatever the targetlist says.
-	 * Otherwise, if it's INSERT/UPDATE/DELETE with RETURNING, it returns
-	 * that. Otherwise, the function return type must be VOID.
+	 * Otherwise, if it's INSERT/UPDATE/DELETE/MERGE with RETURNING, it
+	 * returns that. Otherwise, the function return type must be VOID.
 	 *
 	 * Note: eventually replace this test with QueryReturnsTuples?	We'd need
 	 * a more general method of determining the output type, though.  Also, it
@@ -1679,7 +1679,8 @@ check_sql_fn_retval(List *queryTreeLists
 	else if (parse &&
 			 (parse->commandType == CMD_INSERT ||
 			  parse->commandType == CMD_UPDATE ||
-			  parse->commandType == CMD_DELETE) &&
+			  parse->commandType == CMD_DELETE ||
+			  parse->commandType == CMD_MERGE) &&
 			 parse->returningList)
 	{
 		tlist = parse->returningList;
@@ -1693,7 +1694,7 @@ check_sql_fn_retval(List *queryTreeLists
 				(errcode(ERRCODE_INVALID_FUNCTION_DEFINITION),
 				 errmsg("return type mismatch in function declared to return %s",
 						format_type_be(rettype)),
-				 errdetail("Function's final statement must be SELECT or INSERT/UPDATE/DELETE RETURNING.")));
+				 errdetail("Function's final statement must be SELECT or INSERT/UPDATE/DELETE/MERGE RETURNING.")));
 		return false;			/* keep compiler quiet */
 	}
 
diff --git a/src/backend/executor/nodeModifyTable.c b/src/backend/executor/nodeModifyTable.c
new file mode 100644
index 9fc5abf..b11eb7b
--- a/src/backend/executor/nodeModifyTable.c
+++ b/src/backend/executor/nodeModifyTable.c
@@ -36,8 +36,7 @@
  *		RETURNING tuple after completing each row insert, update, or delete.
  *		It must be called again to continue the operation.  Without RETURNING,
  *		we just loop within the node until all the work is done, then
- *		return NULL.  This avoids useless call/return overhead.  (MERGE does
- *		not support RETURNING.)
+ *		return NULL.  This avoids useless call/return overhead.
  */
 
 #include "postgres.h"
@@ -88,9 +87,6 @@ typedef struct ModifyTableContext
 	 */
 	TupleTableSlot *planSlot;
 
-	/* MERGE specific */
-	MergeActionState *relaction;	/* MERGE action in progress */
-
 	/*
 	 * Information about the changes that were made concurrently to a tuple
 	 * being updated or deleted
@@ -153,13 +149,14 @@ static TupleTableSlot *ExecMerge(ModifyT
 								 ItemPointer tupleid,
 								 bool canSetTag);
 static void ExecInitMerge(ModifyTableState *mtstate, EState *estate);
-static bool ExecMergeMatched(ModifyTableContext *context,
-							 ResultRelInfo *resultRelInfo,
-							 ItemPointer tupleid,
-							 bool canSetTag);
-static void ExecMergeNotMatched(ModifyTableContext *context,
-								ResultRelInfo *resultRelInfo,
-								bool canSetTag);
+static TupleTableSlot *ExecMergeMatched(ModifyTableContext *context,
+										ResultRelInfo *resultRelInfo,
+										ItemPointer tupleid,
+										bool canSetTag,
+										bool *matched);
+static TupleTableSlot *ExecMergeNotMatched(ModifyTableContext *context,
+										   ResultRelInfo *resultRelInfo,
+										   bool canSetTag);
 
 
 /*
@@ -979,7 +976,7 @@ ExecInsert(ModifyTableContext *context,
 		if (mtstate->operation == CMD_UPDATE)
 			wco_kind = WCO_RLS_UPDATE_CHECK;
 		else if (mtstate->operation == CMD_MERGE)
-			wco_kind = (context->relaction->mas_action->commandType == CMD_UPDATE) ?
+			wco_kind = (mtstate->mt_merge_action->mas_action->commandType == CMD_UPDATE) ?
 				WCO_RLS_UPDATE_CHECK : WCO_RLS_INSERT_CHECK;
 		else
 			wco_kind = WCO_RLS_INSERT_CHECK;
@@ -1833,7 +1830,7 @@ ExecCrossPartitionUpdate(ModifyTableCont
 		 * additional rechecking, and might end up executing a different
 		 * action entirely).
 		 */
-		if (context->relaction != NULL)
+		if (mtstate->operation == CMD_MERGE)
 			return *tmresult == TM_Ok;
 		else if (TupIsNull(epqslot))
 			return true;
@@ -2075,7 +2072,7 @@ lreplace:
 		 * No luck, a retry is needed.  If running MERGE, we do not do so
 		 * here; instead let it handle that on its own rules.
 		 */
-		if (context->relaction != NULL)
+		if (context->mtstate->operation == CMD_MERGE)
 			return result;
 
 		/*
@@ -2718,6 +2715,7 @@ static TupleTableSlot *
 ExecMerge(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 		  ItemPointer tupleid, bool canSetTag)
 {
+	TupleTableSlot *rslot = NULL;
 	bool		matched;
 
 	/*-----
@@ -2765,18 +2763,18 @@ ExecMerge(ModifyTableContext *context, R
 	 */
 	matched = tupleid != NULL;
 	if (matched)
-		matched = ExecMergeMatched(context, resultRelInfo, tupleid, canSetTag);
+		rslot = ExecMergeMatched(context, resultRelInfo, tupleid, canSetTag,
+								 &matched);
 
 	/*
-	 * Either we were dealing with a NOT MATCHED tuple or ExecMergeMatched()
-	 * returned "false", indicating the previously MATCHED tuple no longer
-	 * matches.
+	 * Deal with the NOT MATCHED case (either a NOT MATCHED tuple from the
+	 * join, or a previously MATCHED tuple for which ExecMergeMatched() set
+	 * "matched" to false, indicating that it no longer matches).
 	 */
 	if (!matched)
-		ExecMergeNotMatched(context, resultRelInfo, canSetTag);
+		rslot = ExecMergeNotMatched(context, resultRelInfo, canSetTag);
 
-	/* No RETURNING support yet */
-	return NULL;
+	return rslot;
 }
 
 /*
@@ -2786,8 +2784,8 @@ ExecMerge(ModifyTableContext *context, R
  * We start from the first WHEN MATCHED action and check if the WHEN quals
  * pass, if any. If the WHEN quals for the first action do not pass, we
  * check the second, then the third and so on. If we reach to the end, no
- * action is taken and we return true, indicating that no further action is
- * required for this tuple.
+ * action is taken and "matched" is set to true, indicating that no further
+ * action is required for this tuple.
  *
  * If we do find a qualifying action, then we attempt to execute the action.
  *
@@ -2796,16 +2794,17 @@ ExecMerge(ModifyTableContext *context, R
  * with individual actions are evaluated by this routine via ExecQual, while
  * EvalPlanQual checks for the join quals. If EvalPlanQual tells us that the
  * updated tuple still passes the join quals, then we restart from the first
- * action to look for a qualifying action. Otherwise, we return false --
- * meaning that a NOT MATCHED action must now be executed for the current
- * source tuple.
+ * action to look for a qualifying action. Otherwise, "matched" is set to
+ * false -- meaning that a NOT MATCHED action must now be executed for the
+ * current source tuple.
  */
-static bool
+static TupleTableSlot *
 ExecMergeMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
-				 ItemPointer tupleid, bool canSetTag)
+				 ItemPointer tupleid, bool canSetTag, bool *matched)
 {
 	ModifyTableState *mtstate = context->mtstate;
-	TupleTableSlot *newslot;
+	TupleTableSlot *rslot = NULL;
+	TupleTableSlot *newslot = NULL;
 	EState	   *estate = context->estate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
 	bool		isNull;
@@ -2816,7 +2815,10 @@ ExecMergeMatched(ModifyTableContext *con
 	 * If there are no WHEN MATCHED actions, we are done.
 	 */
 	if (resultRelInfo->ri_matchedMergeAction == NIL)
-		return true;
+	{
+		*matched = true;
+		return NULL;
+	}
 
 	/*
 	 * Make tuple and any needed join variables available to ExecQual and
@@ -2895,12 +2897,15 @@ lmerge_matched:
 				 */
 				newslot = ExecProject(relaction->mas_proj);
 
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				if (!ExecUpdatePrologue(context, resultRelInfo,
 										tupleid, NULL, newslot, &result))
 				{
 					if (result == TM_Ok)
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 					break;		/* concurrent update/delete */
 				}
 				result = ExecUpdateAct(context, resultRelInfo, tupleid, NULL,
@@ -2911,12 +2916,13 @@ lmerge_matched:
 				 * cross-partition update was done, then there's nothing else
 				 * for us to do --- the UPDATE has been turned into a DELETE
 				 * and an INSERT, and we must not perform any of the usual
-				 * post-update tasks.
+				 * post-update tasks.  Also, the RETURNING tuple (if any) has
+				 * been projected, so we can just return that.
 				 */
 				if (updateCxt.crossPartUpdate)
 				{
 					mtstate->mt_merge_updated += 1;
-					return true;
+					return context->cpUpdateReturningSlot;
 				}
 
 				if (result == TM_Ok && updateCxt.updated)
@@ -2928,12 +2934,15 @@ lmerge_matched:
 				break;
 
 			case CMD_DELETE:
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				if (!ExecDeletePrologue(context, resultRelInfo, tupleid,
 										NULL, NULL, &result))
 				{
 					if (result == TM_Ok)
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 					break;		/* concurrent update/delete */
 				}
 				result = ExecDeleteAct(context, resultRelInfo, tupleid, false);
@@ -2989,7 +2998,8 @@ lmerge_matched:
 				 * If the tuple was already deleted, return to let caller
 				 * handle it under NOT MATCHED clauses.
 				 */
-				return false;
+				*matched = false;
+				return NULL;
 
 			case TM_Updated:
 				{
@@ -3035,13 +3045,19 @@ lmerge_matched:
 							 * NOT MATCHED actions.
 							 */
 							if (TupIsNull(epqslot))
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							(void) ExecGetJunkAttribute(epqslot,
 														resultRelInfo->ri_RowIdAttNo,
 														&isNull);
 							if (isNull)
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							/*
 							 * When a tuple was updated and migrated to
@@ -3076,7 +3092,8 @@ lmerge_matched:
 							 * tuple already deleted; tell caller to run NOT
 							 * MATCHED actions
 							 */
-							return false;
+							*matched = false;
+							return NULL;
 
 						case TM_SelfModified:
 
@@ -3096,13 +3113,14 @@ lmerge_matched:
 										(errcode(ERRCODE_TRIGGERED_DATA_CHANGE_VIOLATION),
 										 errmsg("tuple to be updated or deleted was already modified by an operation triggered by the current command"),
 										 errhint("Consider using an AFTER trigger instead of a BEFORE trigger to propagate changes to other rows.")));
-							return false;
+							*matched = false;
+							return NULL;
 
 						default:
 							/* see table_tuple_lock call in ExecDelete() */
 							elog(ERROR, "unexpected table_tuple_lock status: %u",
 								 result);
-							return false;
+							return NULL;
 					}
 				}
 
@@ -3114,6 +3132,31 @@ lmerge_matched:
 				break;
 		}
 
+		/* Process RETURNING if present */
+		if (resultRelInfo->ri_projectReturning)
+		{
+			switch (commandType)
+			{
+				case CMD_UPDATE:
+					rslot = ExecProcessReturning(resultRelInfo, newslot,
+												 context->planSlot);
+					break;
+
+				case CMD_DELETE:
+					rslot = ExecProcessReturning(resultRelInfo,
+												 resultRelInfo->ri_oldTupleSlot,
+												 context->planSlot);
+					break;
+
+				case CMD_NOTHING:
+					break;
+
+				default:
+					elog(ERROR, "unrecognized commandType: %d",
+						 (int) commandType);
+			}
+		}
+
 		/*
 		 * We've activated one of the WHEN clauses, so we don't search
 		 * further. This is required behaviour, not an optimization.
@@ -3124,19 +3167,22 @@ lmerge_matched:
 	/*
 	 * Successfully executed an action or no qualifying action was found.
 	 */
-	return true;
+	*matched = true;
+
+	return rslot;
 }
 
 /*
  * Execute the first qualifying NOT MATCHED action.
  */
-static void
+static TupleTableSlot *
 ExecMergeNotMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 					bool canSetTag)
 {
 	ModifyTableState *mtstate = context->mtstate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
 	List	   *actionStates = NIL;
+	TupleTableSlot *rslot = NULL;
 	ListCell   *l;
 
 	/*
@@ -3186,10 +3232,10 @@ ExecMergeNotMatched(ModifyTableContext *
 				 * so we don't need to map the tuple here.
 				 */
 				newslot = ExecProject(action->mas_proj);
-				context->relaction = action;
+				mtstate->mt_merge_action = action;
 
-				(void) ExecInsert(context, mtstate->rootResultRelInfo, newslot,
-								  canSetTag, NULL, NULL);
+				rslot = ExecInsert(context, mtstate->rootResultRelInfo,
+								   newslot, canSetTag, NULL, NULL);
 				mtstate->mt_merge_inserted += 1;
 				break;
 			case CMD_NOTHING:
@@ -3205,6 +3251,8 @@ ExecMergeNotMatched(ModifyTableContext *
 		 */
 		break;
 	}
+
+	return rslot;
 }
 
 /*
@@ -3667,8 +3715,17 @@ ExecModifyTable(PlanState *pstate)
 				{
 					EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-					ExecMerge(&context, node->resultRelInfo, NULL, node->canSetTag);
-					continue;	/* no RETURNING support yet */
+					slot = ExecMerge(&context, node->resultRelInfo, NULL,
+									 node->canSetTag);
+
+					/*
+					 * If we got a RETURNING result, return it to the caller.
+					 * We'll continue the work on next call.
+					 */
+					if (slot)
+						return slot;
+
+					continue;	/* continue with the next tuple */
 				}
 
 				elog(ERROR, "tableoid is NULL");
@@ -3745,8 +3802,17 @@ ExecModifyTable(PlanState *pstate)
 					{
 						EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-						ExecMerge(&context, node->resultRelInfo, NULL, node->canSetTag);
-						continue;	/* no RETURNING support yet */
+						slot = ExecMerge(&context, node->resultRelInfo, NULL,
+										 node->canSetTag);
+
+						/*
+						 * If we got a RETURNING result, return it to the
+						 * caller.  We'll continue the work on next call.
+						 */
+						if (slot)
+							return slot;
+
+						continue;	/* continue with the next tuple */
 					}
 
 					elog(ERROR, "ctid is NULL");
@@ -3838,7 +3904,6 @@ ExecModifyTable(PlanState *pstate)
 				}
 				slot = ExecGetUpdateNewTuple(resultRelInfo, context.planSlot,
 											 oldSlot);
-				context.relaction = NULL;
 
 				/* Now apply the update. */
 				slot = ExecUpdate(&context, resultRelInfo, tupleid, oldtuple,
diff --git a/src/backend/executor/spi.c b/src/backend/executor/spi.c
new file mode 100644
index d71484c..abb8b0d
--- a/src/backend/executor/spi.c
+++ b/src/backend/executor/spi.c
@@ -2033,6 +2033,8 @@ SPI_result_code_string(int code)
 			return "SPI_OK_TD_REGISTER";
 		case SPI_OK_MERGE:
 			return "SPI_OK_MERGE";
+		case SPI_OK_MERGE_RETURNING:
+			return "SPI_OK_MERGE_RETURNING";
 	}
 	/* Unrecognized code ... return something useful ... */
 	sprintf(buf, "Unrecognized SPI code %d", code);
@@ -2886,7 +2888,10 @@ _SPI_pquery(QueryDesc *queryDesc, bool f
 				res = SPI_OK_UPDATE;
 			break;
 		case CMD_MERGE:
-			res = SPI_OK_MERGE;
+			if (queryDesc->plannedstmt->hasReturning)
+				res = SPI_OK_MERGE_RETURNING;
+			else
+				res = SPI_OK_MERGE;
 			break;
 		default:
 			return SPI_ERROR_OPUNKNOWN;
diff --git a/src/backend/jit/llvm/llvmjit_expr.c b/src/backend/jit/llvm/llvmjit_expr.c
new file mode 100644
index 0c44842..a5cd6fb
--- a/src/backend/jit/llvm/llvmjit_expr.c
+++ b/src/backend/jit/llvm/llvmjit_expr.c
@@ -1986,6 +1986,12 @@ llvm_compile_expr(ExprState *state)
 					break;
 				}
 
+			case EEOP_MERGING_FUNC:
+				build_EvalXFunc(b, mod, "ExecEvalMergingFunc",
+								v_state, op, v_econtext);
+				LLVMBuildBr(b, opblocks[opno + 1]);
+				break;
+
 			case EEOP_SUBPLAN:
 				build_EvalXFunc(b, mod, "ExecEvalSubPlan",
 								v_state, op, v_econtext);
diff --git a/src/backend/jit/llvm/llvmjit_types.c b/src/backend/jit/llvm/llvmjit_types.c
new file mode 100644
index 47c9daf..912d9ad
--- a/src/backend/jit/llvm/llvmjit_types.c
+++ b/src/backend/jit/llvm/llvmjit_types.c
@@ -155,6 +155,7 @@ void	   *referenced_functions[] =
 	ExecEvalFuncExprFusage,
 	ExecEvalFuncExprStrictFusage,
 	ExecEvalGroupingFunc,
+	ExecEvalMergingFunc,
 	ExecEvalMinMax,
 	ExecEvalNextValueExpr,
 	ExecEvalParamExec,
diff --git a/src/backend/nodes/nodeFuncs.c b/src/backend/nodes/nodeFuncs.c
new file mode 100644
index e1a5bc7..2468063
--- a/src/backend/nodes/nodeFuncs.c
+++ b/src/backend/nodes/nodeFuncs.c
@@ -67,6 +67,9 @@ exprType(const Node *expr)
 		case T_WindowFunc:
 			type = ((const WindowFunc *) expr)->wintype;
 			break;
+		case T_MergingFunc:
+			type = ((const MergingFunc *) expr)->mftype;
+			break;
 		case T_SubscriptingRef:
 			type = ((const SubscriptingRef *) expr)->refrestype;
 			break;
@@ -810,6 +813,9 @@ exprCollation(const Node *expr)
 		case T_WindowFunc:
 			coll = ((const WindowFunc *) expr)->wincollid;
 			break;
+		case T_MergingFunc:
+			coll = ((const MergingFunc *) expr)->mfcollid;
+			break;
 		case T_SubscriptingRef:
 			coll = ((const SubscriptingRef *) expr)->refcollid;
 			break;
@@ -1085,6 +1091,9 @@ exprSetCollation(Node *expr, Oid collati
 		case T_WindowFunc:
 			((WindowFunc *) expr)->wincollid = collation;
 			break;
+		case T_MergingFunc:
+			((MergingFunc *) expr)->mfcollid = collation;
+			break;
 		case T_SubscriptingRef:
 			((SubscriptingRef *) expr)->refcollid = collation;
 			break;
@@ -1343,6 +1352,9 @@ exprLocation(const Node *expr)
 			/* function name should always be the first thing */
 			loc = ((const WindowFunc *) expr)->location;
 			break;
+		case T_MergingFunc:
+			loc = ((const MergingFunc *) expr)->location;
+			break;
 		case T_SubscriptingRef:
 			/* just use container argument's location */
 			loc = exprLocation((Node *) ((const SubscriptingRef *) expr)->refexpr);
@@ -2035,6 +2047,7 @@ expression_tree_walker_impl(Node *node,
 		case T_RangeTblRef:
 		case T_SortGroupClause:
 		case T_CTESearchClause:
+		case T_MergingFunc:
 			/* primitive node types with no expression subnodes */
 			break;
 		case T_WithCheckOption:
@@ -2869,6 +2882,7 @@ expression_tree_mutator_impl(Node *node,
 		case T_RangeTblRef:
 		case T_SortGroupClause:
 		case T_CTESearchClause:
+		case T_MergingFunc:
 			return (Node *) copyObject(node);
 		case T_WithCheckOption:
 			{
@@ -3833,6 +3847,7 @@ raw_expression_tree_walker_impl(Node *no
 		case T_ParamRef:
 		case T_A_Const:
 		case T_A_Star:
+		case T_MergingFunc:
 			/* primitive node types with no subnodes */
 			break;
 		case T_Alias:
@@ -4053,6 +4068,8 @@ raw_expression_tree_walker_impl(Node *no
 					return true;
 				if (WALK(stmt->mergeWhenClauses))
 					return true;
+				if (WALK(stmt->returningList))
+					return true;
 				if (WALK(stmt->withClause))
 					return true;
 			}
diff --git a/src/backend/optimizer/plan/subselect.c b/src/backend/optimizer/plan/subselect.c
new file mode 100644
index 3115d79..a3646c0
--- a/src/backend/optimizer/plan/subselect.c
+++ b/src/backend/optimizer/plan/subselect.c
@@ -1846,7 +1846,8 @@ convert_EXISTS_to_ANY(PlannerInfo *root,
 /*
  * Replace correlation vars (uplevel vars) with Params.
  *
- * Uplevel PlaceHolderVars and aggregates are replaced, too.
+ * Uplevel PlaceHolderVars, aggregates, GROUPING() expressions, and MERGING()
+ * expressions are replaced, too.
  *
  * Note: it is critical that this runs immediately after SS_process_sublinks.
  * Since we do not recurse into the arguments of uplevel PHVs and aggregates,
@@ -1900,6 +1901,11 @@ replace_correlation_vars_mutator(Node *n
 		if (((GroupingFunc *) node)->agglevelsup > 0)
 			return (Node *) replace_outer_grouping(root, (GroupingFunc *) node);
 	}
+	if (IsA(node, MergingFunc))
+	{
+		if (root->parse->commandType != CMD_MERGE)
+			return (Node *) replace_outer_merging(root, (MergingFunc *) node);
+	}
 	return expression_tree_mutator(node,
 								   replace_correlation_vars_mutator,
 								   (void *) root);
diff --git a/src/backend/optimizer/util/paramassign.c b/src/backend/optimizer/util/paramassign.c
new file mode 100644
index a58da7c..6ea6db2
--- a/src/backend/optimizer/util/paramassign.c
+++ b/src/backend/optimizer/util/paramassign.c
@@ -308,6 +308,56 @@ replace_outer_grouping(PlannerInfo *root
 }
 
 /*
+ * Generate a Param node to replace the given MergingFunc expression which is
+ * expected to be in the RETURNING list of an upper-level MERGE query.  Record
+ * the need for the MergingFunc in the proper upper-level root->plan_params.
+ */
+Param *
+replace_outer_merging(PlannerInfo *root, MergingFunc *mf)
+{
+	Param	   *retval;
+	PlannerParamItem *pitem;
+	Oid			ptype = exprType((Node *) mf);
+
+	Assert(root->parse->commandType != CMD_MERGE);
+
+	/*
+	 * The parser should have ensured that the MergingFunc is in the RETURNING
+	 * list of an upper-level MERGE query, so find that query.
+	 */
+	do
+	{
+		root = root->parent_root;
+		if (root == NULL)
+			elog(ERROR, "MergingFunc found outside MERGE");
+	} while (root->parse->commandType != CMD_MERGE);
+
+	/*
+	 * It does not seem worthwhile to try to de-duplicate references to outer
+	 * MergingFunc expressions.  Just make a new slot every time.
+	 */
+	mf = copyObject(mf);
+
+	pitem = makeNode(PlannerParamItem);
+	pitem->item = (Node *) mf;
+	pitem->paramId = list_length(root->glob->paramExecTypes);
+	root->glob->paramExecTypes = lappend_oid(root->glob->paramExecTypes,
+											 ptype);
+
+	root->plan_params = lappend(root->plan_params, pitem);
+
+	retval = makeNode(Param);
+	retval->paramkind = PARAM_EXEC;
+	retval->paramid = pitem->paramId;
+	retval->paramtype = ptype;
+	retval->paramtypmod = -1;
+	retval->paramcollid = InvalidOid;
+	retval->location = mf->location;
+
+	return retval;
+}
+
+/*
  * Generate a Param node to replace the given Var,
  * which is expected to come from some upper NestLoop plan node.
  * Record the need for the Var in root->curOuterParams.
diff --git a/src/backend/parser/analyze.c b/src/backend/parser/analyze.c
new file mode 100644
index dbdf6bf..44893d8
--- a/src/backend/parser/analyze.c
+++ b/src/backend/parser/analyze.c
@@ -74,7 +74,6 @@ static void determineRecursiveColTypes(P
 									   Node *larg, List *nrtargetlist);
 static Query *transformReturnStmt(ParseState *pstate, ReturnStmt *stmt);
 static Query *transformUpdateStmt(ParseState *pstate, UpdateStmt *stmt);
-static List *transformReturningList(ParseState *pstate, List *returningList);
 static Query *transformPLAssignStmt(ParseState *pstate,
 									PLAssignStmt *stmt);
 static Query *transformDeclareCursorStmt(ParseState *pstate,
@@ -553,7 +552,8 @@ transformDeleteStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -980,7 +980,8 @@ transformInsertStmt(ParseState *pstate,
 	/* Process RETURNING, if any. */
 	if (stmt->returningList)
 		qry->returningList = transformReturningList(pstate,
-													stmt->returningList);
+													stmt->returningList,
+													EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -2445,7 +2446,8 @@ transformUpdateStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/*
 	 * Now we are done with SELECT-like processing, and can get on with
@@ -2542,10 +2544,11 @@ transformUpdateTargetList(ParseState *ps
 
 /*
  * transformReturningList -
- *	handle a RETURNING clause in INSERT/UPDATE/DELETE
+ *	handle a RETURNING clause in INSERT/UPDATE/DELETE/MERGE
  */
-static List *
-transformReturningList(ParseState *pstate, List *returningList)
+List *
+transformReturningList(ParseState *pstate, List *returningList,
+					   ParseExprKind exprKind)
 {
 	List	   *rlist;
 	int			save_next_resno;
@@ -2562,7 +2565,7 @@ transformReturningList(ParseState *pstat
 	pstate->p_next_resno = 1;
 
 	/* transform RETURNING identically to a SELECT targetlist */
-	rlist = transformTargetList(pstate, returningList, EXPR_KIND_RETURNING);
+	rlist = transformTargetList(pstate, returningList, exprKind);
 
 	/*
 	 * Complain if the nonempty tlist expanded to nothing (which is possible
diff --git a/src/backend/parser/gram.y b/src/backend/parser/gram.y
new file mode 100644
index 130f7fc..04fee0d
--- a/src/backend/parser/gram.y
+++ b/src/backend/parser/gram.y
@@ -694,7 +694,7 @@ static Node *makeRecursiveViewSelect(cha
 	BOOLEAN_P BOTH BREADTH BY
 
 	CACHE CALL CALLED CASCADE CASCADED CASE CAST CATALOG_P CHAIN CHAR_P
-	CHARACTER CHARACTERISTICS CHECK CHECKPOINT CLASS CLOSE
+	CHARACTER CHARACTERISTICS CHECK CHECKPOINT CLASS CLAUSE_NUMBER CLOSE
 	CLUSTER COALESCE COLLATE COLLATION COLUMN COLUMNS COMMENT COMMENTS COMMIT
 	COMMITTED COMPRESSION CONCURRENTLY CONFIGURATION CONFLICT
 	CONNECTION CONSTRAINT CONSTRAINTS CONTENT_P CONTINUE_P CONVERSION_P COPY
@@ -732,7 +732,7 @@ static Node *makeRecursiveViewSelect(cha
 	LEADING LEAKPROOF LEAST LEFT LEVEL LIKE LIMIT LISTEN LOAD LOCAL
 	LOCALTIME LOCALTIMESTAMP LOCATION LOCK_P LOCKED LOGGED
 
-	MAPPING MATCH MATCHED MATERIALIZED MAXVALUE MERGE METHOD
+	MAPPING MATCH MATCHED MATERIALIZED MAXVALUE MERGE MERGING METHOD
 	MINUTE_P MINVALUE MODE MONTH_P MOVE
 
 	NAME_P NAMES NATIONAL NATURAL NCHAR NEW NEXT NFC NFD NFKC NFKD NO NONE
@@ -12368,6 +12368,7 @@ MergeStmt:
 			USING table_ref
 			ON a_expr
 			merge_when_list
+			returning_clause
 				{
 					MergeStmt  *m = makeNode(MergeStmt);
 
@@ -12376,6 +12377,7 @@ MergeStmt:
 					m->sourceRelation = $6;
 					m->joinCondition = $8;
 					m->mergeWhenClauses = $9;
+					m->returningList = $10;
 
 					$$ = (Node *) m;
 				}
@@ -15789,6 +15791,24 @@ func_expr_common_subexpr:
 					n->location = @1;
 					$$ = (Node *) n;
 				}
+			| MERGING '(' ACTION ')'
+				{
+					MergingFunc *m = makeNode(MergingFunc);
+
+					m->mfop = MERGING_ACTION;
+					m->mftype = TEXTOID;
+					m->location = @1;
+					$$ = (Node *) m;
+				}
+			| MERGING '(' CLAUSE_NUMBER ')'
+				{
+					MergingFunc *m = makeNode(MergingFunc);
+
+					m->mfop = MERGING_CLAUSE_NUMBER;
+					m->mftype = INT4OID;
+					m->location = @1;
+					$$ = (Node *) m;
+				}
 			;
 
 
@@ -17167,6 +17187,7 @@ unreserved_keyword:
 			| CHARACTERISTICS
 			| CHECKPOINT
 			| CLASS
+			| CLAUSE_NUMBER
 			| CLOSE
 			| CLUSTER
 			| COLUMNS
@@ -17486,6 +17507,7 @@ col_name_keyword:
 			| JSON_SCALAR
 			| JSON_SERIALIZE
 			| LEAST
+			| MERGING
 			| NATIONAL
 			| NCHAR
 			| NONE
@@ -17704,6 +17726,7 @@ bare_label_keyword:
 			| CHECK
 			| CHECKPOINT
 			| CLASS
+			| CLAUSE_NUMBER
 			| CLOSE
 			| CLUSTER
 			| COALESCE
@@ -17875,6 +17898,7 @@ bare_label_keyword:
 			| MATERIALIZED
 			| MAXVALUE
 			| MERGE
+			| MERGING
 			| METHOD
 			| MINVALUE
 			| MODE
diff --git a/src/backend/parser/parse_agg.c b/src/backend/parser/parse_agg.c
new file mode 100644
index 7b211a7..f143310
--- a/src/backend/parser/parse_agg.c
+++ b/src/backend/parser/parse_agg.c
@@ -467,6 +467,7 @@ check_agglevels_and_constraints(ParseSta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
@@ -914,6 +915,7 @@ transformWindowFuncCall(ParseState *psta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_cte.c b/src/backend/parser/parse_cte.c
new file mode 100644
index 3c88c9a..6826d4f
--- a/src/backend/parser/parse_cte.c
+++ b/src/backend/parser/parse_cte.c
@@ -126,13 +126,6 @@ transformWithClause(ParseState *pstate,
 		CommonTableExpr *cte = (CommonTableExpr *) lfirst(lc);
 		ListCell   *rest;
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(cte->ctequery, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in WITH query"),
-					parser_errposition(pstate, cte->location));
-
 		for_each_cell(rest, withClause->ctes, lnext(withClause->ctes, lc))
 		{
 			CommonTableExpr *cte2 = (CommonTableExpr *) lfirst(rest);
@@ -153,7 +146,8 @@ transformWithClause(ParseState *pstate,
 			/* must be a data-modifying statement */
 			Assert(IsA(cte->ctequery, InsertStmt) ||
 				   IsA(cte->ctequery, UpdateStmt) ||
-				   IsA(cte->ctequery, DeleteStmt));
+				   IsA(cte->ctequery, DeleteStmt) ||
+				   IsA(cte->ctequery, MergeStmt));
 
 			pstate->p_hasModifyingCTE = true;
 		}
diff --git a/src/backend/parser/parse_expr.c b/src/backend/parser/parse_expr.c
new file mode 100644
index 9300c7b..ab7e564
--- a/src/backend/parser/parse_expr.c
+++ b/src/backend/parser/parse_expr.c
@@ -54,6 +54,7 @@ static Node *transformAExprDistinct(Pars
 static Node *transformAExprNullIf(ParseState *pstate, A_Expr *a);
 static Node *transformAExprIn(ParseState *pstate, A_Expr *a);
 static Node *transformAExprBetween(ParseState *pstate, A_Expr *a);
+static Node *transformMergingFunc(ParseState *pstate, MergingFunc *f);
 static Node *transformBoolExpr(ParseState *pstate, BoolExpr *a);
 static Node *transformFuncCall(ParseState *pstate, FuncCall *fn);
 static Node *transformMultiAssignRef(ParseState *pstate, MultiAssignRef *maref);
@@ -227,6 +228,10 @@ transformExprRecurse(ParseState *pstate,
 			result = transformGroupingFunc(pstate, (GroupingFunc *) expr);
 			break;
 
+		case T_MergingFunc:
+			result = transformMergingFunc(pstate, (MergingFunc *) expr);
+			break;
+
 		case T_NamedArgExpr:
 			{
 				NamedArgExpr *na = (NamedArgExpr *) expr;
@@ -541,6 +546,7 @@ transformColumnRef(ParseState *pstate, C
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CHECK_CONSTRAINT:
@@ -1354,6 +1360,32 @@ transformAExprBetween(ParseState *pstate
 }
 
 static Node *
+transformMergingFunc(ParseState *pstate, MergingFunc *f)
+{
+	/*
+	 * All we need to do is check that we're in the RETURNING list of a MERGE
+	 * command.  If so, we just return the node as-is.
+	 */
+	if (pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+	{
+		ParseState *parent_pstate = pstate->parentParseState;
+
+		while (parent_pstate &&
+			   parent_pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+			parent_pstate = parent_pstate->parentParseState;
+
+		if (!parent_pstate ||
+			parent_pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+			ereport(ERROR,
+					errcode(ERRCODE_SYNTAX_ERROR),
+					errmsg("MERGING() can only be used in the RETURNING list of a MERGE command"),
+					parser_errposition(pstate, f->location));
+	}
+
+	return (Node *) f;
+}
+
+static Node *
 transformBoolExpr(ParseState *pstate, BoolExpr *a)
 {
 	List	   *args = NIL;
@@ -1767,6 +1799,7 @@ transformSubLink(ParseState *pstate, Sub
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CYCLE_MARK:
@@ -3115,6 +3148,7 @@ ParseExprKindName(ParseExprKind exprKind
 		case EXPR_KIND_OFFSET:
 			return "OFFSET";
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			return "RETURNING";
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
diff --git a/src/backend/parser/parse_func.c b/src/backend/parser/parse_func.c
new file mode 100644
index fdb3e6d..0cbc950
--- a/src/backend/parser/parse_func.c
+++ b/src/backend/parser/parse_func.c
@@ -2599,6 +2599,7 @@ check_srf_call_placement(ParseState *pst
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_merge.c b/src/backend/parser/parse_merge.c
new file mode 100644
index 5f6a683..55936bb
--- a/src/backend/parser/parse_merge.c
+++ b/src/backend/parser/parse_merge.c
@@ -233,6 +233,10 @@ transformMergeStmt(ParseState *pstate, M
 	 */
 	qry->jointree = makeFromExpr(pstate->p_joinlist, joinExpr);
 
+	/* Transform the RETURNING list, if any */
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_MERGE_RETURNING);
+
 	/*
 	 * We now have a good query shape, so now look at the WHEN conditions and
 	 * action targetlists.
@@ -254,6 +258,7 @@ transformMergeStmt(ParseState *pstate, M
 		MergeAction *action;
 
 		action = makeNode(MergeAction);
+		action->index = foreach_current_index(l) + 1;
 		action->commandType = mergeWhenClause->commandType;
 		action->matched = mergeWhenClause->matched;
 
@@ -390,9 +395,6 @@ transformMergeStmt(ParseState *pstate, M
 
 	qry->mergeActionList = mergeActionList;
 
-	/* RETURNING could potentially be added in the future, but not in SQL std */
-	qry->returningList = NULL;
-
 	qry->hasTargetSRFs = false;
 	qry->hasSubLinks = pstate->p_hasSubLinks;
 
diff --git a/src/backend/parser/parse_relation.c b/src/backend/parser/parse_relation.c
new file mode 100644
index 34a0ec5..60b7e4f
--- a/src/backend/parser/parse_relation.c
+++ b/src/backend/parser/parse_relation.c
@@ -2346,9 +2346,10 @@ addRangeTableEntryForCTE(ParseState *pst
 		cte->cterefcount++;
 
 	/*
-	 * We throw error if the CTE is INSERT/UPDATE/DELETE without RETURNING.
-	 * This won't get checked in case of a self-reference, but that's OK
-	 * because data-modifying CTEs aren't allowed to be recursive anyhow.
+	 * We throw error if the CTE is INSERT/UPDATE/DELETE/MERGE without
+	 * RETURNING.  This won't get checked in case of a self-reference, but
+	 * that's OK because data-modifying CTEs aren't allowed to be recursive
+	 * anyhow.
 	 */
 	if (IsA(cte->ctequery, Query))
 	{
diff --git a/src/backend/parser/parse_target.c b/src/backend/parser/parse_target.c
new file mode 100644
index 0cd904f..eab78f4
--- a/src/backend/parser/parse_target.c
+++ b/src/backend/parser/parse_target.c
@@ -1807,6 +1807,20 @@ FigureColnameInternal(Node *node, char *
 			/* make GROUPING() act like a regular function */
 			*name = "grouping";
 			return 2;
+		case T_MergingFunc:
+			switch (((MergingFunc *) node)->mfop)
+			{
+				case MERGING_ACTION:
+					*name = "action";
+					return 2;
+				case MERGING_CLAUSE_NUMBER:
+					*name = "clause_number";
+					return 2;
+				default:
+					elog(ERROR, "unrecognized MergingFuncOp: %d",
+						 (int) ((MergingFunc *) node)->mfop);
+			}
+			break;
 		case T_SubLink:
 			switch (((SubLink *) node)->subLinkType)
 			{
diff --git a/src/backend/rewrite/rewriteHandler.c b/src/backend/rewrite/rewriteHandler.c
new file mode 100644
index f60b34d..d719d07
--- a/src/backend/rewrite/rewriteHandler.c
+++ b/src/backend/rewrite/rewriteHandler.c
@@ -3632,9 +3632,9 @@ RewriteQuery(Query *parsetree, List *rew
 	ListCell   *lc1;
 
 	/*
-	 * First, recursively process any insert/update/delete statements in WITH
-	 * clauses.  (We have to do this first because the WITH clauses may get
-	 * copied into rule actions below.)
+	 * First, recursively process any insert/update/delete/merge statements in
+	 * WITH clauses.  (We have to do this first because the WITH clauses may
+	 * get copied into rule actions below.)
 	 */
 	foreach(lc1, parsetree->cteList)
 	{
@@ -3659,7 +3659,8 @@ RewriteQuery(Query *parsetree, List *rew
 			if (!(ctequery->commandType == CMD_SELECT ||
 				  ctequery->commandType == CMD_UPDATE ||
 				  ctequery->commandType == CMD_INSERT ||
-				  ctequery->commandType == CMD_DELETE))
+				  ctequery->commandType == CMD_DELETE ||
+				  ctequery->commandType == CMD_MERGE))
 			{
 				/*
 				 * Currently it could only be NOTIFY; this error message will
diff --git a/src/backend/rewrite/rowsecurity.c b/src/backend/rewrite/rowsecurity.c
new file mode 100644
index 94268a2..65ce579
--- a/src/backend/rewrite/rowsecurity.c
+++ b/src/backend/rewrite/rowsecurity.c
@@ -395,10 +395,10 @@ get_row_security_policies(Query *root, R
 	 * on the final action we take.
 	 *
 	 * We already fetched the SELECT policies above, to check existing rows,
-	 * but we must also check that new rows created by UPDATE actions are
-	 * visible, if SELECT rights are required for this relation. We don't do
-	 * this for INSERT actions, since an INSERT command would only do this
-	 * check if it had a RETURNING list, and MERGE does not support RETURNING.
+	 * but we must also check that new rows created by INSERT/UPDATE actions
+	 * are visible, if SELECT rights are required. For INSERT actions, we only
+	 * do this if RETURNING is specified, to be consistent with a plain INSERT
+	 * command, which can only require SELECT rights when RETURNING is used.
 	 *
 	 * We don't push the UPDATE/DELETE USING quals to the RTE because we don't
 	 * really want to apply them while scanning the relation since we don't
@@ -420,6 +420,8 @@ get_row_security_policies(Query *root, R
 		List	   *merge_delete_restrictive_policies;
 		List	   *merge_insert_permissive_policies;
 		List	   *merge_insert_restrictive_policies;
+		List	   *merge_select_permissive_policies = NIL;
+		List	   *merge_select_restrictive_policies = NIL;
 
 		/*
 		 * Fetch the UPDATE policies and set them up to execute on the
@@ -457,9 +459,6 @@ get_row_security_policies(Query *root, R
 		 */
 		if (perminfo->requiredPerms & ACL_SELECT)
 		{
-			List	   *merge_select_permissive_policies;
-			List	   *merge_select_restrictive_policies;
-
 			get_policies_for_relation(rel, CMD_SELECT, user_id,
 									  &merge_select_permissive_policies,
 									  &merge_select_restrictive_policies);
@@ -508,6 +507,21 @@ get_row_security_policies(Query *root, R
 							   withCheckOptions,
 							   hasSubLinks,
 							   false);
+
+		/*
+		 * Add ALL/SELECT policies as WCO_RLS_INSERT_CHECK WCOs, to ensure
+		 * that the inserted row is visible when executing an INSERT action,
+		 * if RETURNING is specified and SELECT rights are required for this
+		 * relation.
+		 */
+		if (perminfo->requiredPerms & ACL_SELECT && root->returningList)
+			add_with_check_options(rel, rt_index,
+								   WCO_RLS_INSERT_CHECK,
+								   merge_select_permissive_policies,
+								   merge_select_restrictive_policies,
+								   withCheckOptions,
+								   hasSubLinks,
+								   true);
 	}
 
 	table_close(rel, NoLock);
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
new file mode 100644
index 8de821f..f84a943
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -2145,11 +2145,10 @@ QueryReturnsTuples(Query *parsetree)
 		case CMD_SELECT:
 			/* returns tuples */
 			return true;
-		case CMD_MERGE:
-			return false;
 		case CMD_INSERT:
 		case CMD_UPDATE:
 		case CMD_DELETE:
+		case CMD_MERGE:
 			/* the forms with RETURNING return tuples */
 			if (parsetree->returningList)
 				return true;
diff --git a/src/backend/utils/adt/ruleutils.c b/src/backend/utils/adt/ruleutils.c
new file mode 100644
index b625f47..e710d02
--- a/src/backend/utils/adt/ruleutils.c
+++ b/src/backend/utils/adt/ruleutils.c
@@ -7199,8 +7199,13 @@ get_merge_query_def(Query *query, depars
 			appendStringInfoString(buf, "DO NOTHING");
 	}
 
-	/* No RETURNING support in MERGE yet */
-	Assert(query->returningList == NIL);
+	/* Add RETURNING if present */
+	if (query->returningList)
+	{
+		appendContextKeyword(context, " RETURNING",
+							 -PRETTYINDENT_STD, PRETTYINDENT_STD, 1);
+		get_target_list(query->returningList, context, NULL, colNamesVisible);
+	}
 }
 
 
@@ -8300,6 +8305,7 @@ isSimpleNode(Node *node, Node *parentNod
 		case T_Aggref:
 		case T_GroupingFunc:
 		case T_WindowFunc:
+		case T_MergingFunc:
 		case T_FuncExpr:
 		case T_JsonConstructorExpr:
 			/* function-like: name(..) or name[..] */
@@ -8654,6 +8660,27 @@ get_rule_expr(Node *node, deparse_contex
 			get_windowfunc_expr((WindowFunc *) node, context);
 			break;
 
+		case T_MergingFunc:
+			{
+				MergingFunc *mexpr = (MergingFunc *) node;
+
+				appendStringInfoString(buf, "MERGING(");
+				switch (mexpr->mfop)
+				{
+					case MERGING_ACTION:
+						appendStringInfoString(buf, "ACTION");
+						break;
+					case MERGING_CLAUSE_NUMBER:
+						appendStringInfoString(buf, "CLAUSE_NUMBER");
+						break;
+					default:
+						elog(ERROR, "unrecognized MergingFuncOp: %d",
+							 (int) mexpr->mfop);
+				}
+				appendStringInfoChar(buf, ')');
+			}
+			break;
+
 		case T_SubscriptingRef:
 			{
 				SubscriptingRef *sbsref = (SubscriptingRef *) node;
diff --git a/src/bin/psql/common.c b/src/bin/psql/common.c
new file mode 100644
index 76e01b0..2830bde
--- a/src/bin/psql/common.c
+++ b/src/bin/psql/common.c
@@ -983,13 +983,17 @@ PrintQueryResult(PGresult *result, bool
 			else
 				success = true;
 
-			/* if it's INSERT/UPDATE/DELETE RETURNING, also print status */
+			/*
+			 * If it's INSERT/UPDATE/DELETE/MERGE RETURNING, also print
+			 * status.
+			 */
 			if (last || pset.show_all_results)
 			{
 				cmdstatus = PQcmdStatus(result);
 				if (strncmp(cmdstatus, "INSERT", 6) == 0 ||
 					strncmp(cmdstatus, "UPDATE", 6) == 0 ||
-					strncmp(cmdstatus, "DELETE", 6) == 0)
+					strncmp(cmdstatus, "DELETE", 6) == 0 ||
+					strncmp(cmdstatus, "MERGE", 5) == 0)
 					PrintQueryStatus(result, printStatusFout);
 			}
 
diff --git a/src/include/executor/execExpr.h b/src/include/executor/execExpr.h
new file mode 100644
index a28ddcd..abb0fb4
--- a/src/include/executor/execExpr.h
+++ b/src/include/executor/execExpr.h
@@ -243,6 +243,7 @@ typedef enum ExprEvalOp
 	EEOP_AGGREF,
 	EEOP_GROUPING_FUNC,
 	EEOP_WINDOW_FUNC,
+	EEOP_MERGING_FUNC,
 	EEOP_SUBPLAN,
 
 	/* aggregation related nodes */
@@ -626,6 +627,12 @@ typedef struct ExprEvalStep
 			WindowFuncExprState *wfstate;
 		}			window_func;
 
+		/* for EEOP_MERGING_FUNC */
+		struct
+		{
+			MergingFuncOp mfop; /* the MERGING() operation to perform */
+		}			merging_func;
+
 		/* for EEOP_SUBPLAN */
 		struct
 		{
@@ -810,6 +817,8 @@ extern void ExecEvalJsonConstructor(Expr
 									ExprContext *econtext);
 extern void ExecEvalJsonIsPredicate(ExprState *state, ExprEvalStep *op);
 extern void ExecEvalGroupingFunc(ExprState *state, ExprEvalStep *op);
+extern void ExecEvalMergingFunc(ExprState *state, ExprEvalStep *op,
+								ExprContext *econtext);
 extern void ExecEvalSubPlan(ExprState *state, ExprEvalStep *op,
 							ExprContext *econtext);
 extern void ExecEvalWholeRowVar(ExprState *state, ExprEvalStep *op,
diff --git a/src/include/executor/spi.h b/src/include/executor/spi.h
new file mode 100644
index c13e1f6..48b8773
--- a/src/include/executor/spi.h
+++ b/src/include/executor/spi.h
@@ -97,6 +97,7 @@ typedef struct _SPI_plan *SPIPlanPtr;
 #define SPI_OK_REL_UNREGISTER	16
 #define SPI_OK_TD_REGISTER		17
 #define SPI_OK_MERGE			18
+#define SPI_OK_MERGE_RETURNING	19
 
 #define SPI_OPT_NONATOMIC		(1 << 0)
 
diff --git a/src/include/nodes/execnodes.h b/src/include/nodes/execnodes.h
new file mode 100644
index 444a5f0..57fdfca
--- a/src/include/nodes/execnodes.h
+++ b/src/include/nodes/execnodes.h
@@ -1325,6 +1325,9 @@ typedef struct ModifyTableState
 	/* Flags showing which subcommands are present INS/UPD/DEL/DO NOTHING */
 	int			mt_merge_subcommands;
 
+	/* For MERGE, the action currently being executed */
+	MergeActionState *mt_merge_action;
+
 	/* tuple counters for MERGE */
 	double		mt_merge_inserted;
 	double		mt_merge_updated;
diff --git a/src/include/nodes/parsenodes.h b/src/include/nodes/parsenodes.h
new file mode 100644
index 476d55d..eb42475
--- a/src/include/nodes/parsenodes.h
+++ b/src/include/nodes/parsenodes.h
@@ -1927,6 +1927,7 @@ typedef struct MergeStmt
 	Node	   *sourceRelation; /* source relation */
 	Node	   *joinCondition;	/* join condition between source and target */
 	List	   *mergeWhenClauses;	/* list of MergeWhenClause(es) */
+	List	   *returningList;	/* list of expressions to return */
 	WithClause *withClause;		/* WITH clause */
 } MergeStmt;
 
diff --git a/src/include/nodes/primnodes.h b/src/include/nodes/primnodes.h
new file mode 100644
index 4a15460..97806f5
--- a/src/include/nodes/primnodes.h
+++ b/src/include/nodes/primnodes.h
@@ -572,6 +572,39 @@ typedef struct WindowFunc
 } WindowFunc;
 
 /*
+ * MergingFunc
+ *
+ * A MergingFunc is a MERGING(...) expression that can only appear in the
+ * RETURNING list of a MERGE command.  It returns information about the
+ * currently executing merge action.  Possible operations are:
+ *
+ *	MERGING_ACTION:
+ *		Return the command string of the current merge action ("INSERT",
+ *		"UPDATE" or "DELETE").
+ *
+ *	MERGING_CLAUSE_NUMBER:
+ *		Return the 1-based index of the current merge WHEN clause.
+ */
+typedef enum MergingFuncOp
+{
+	MERGING_ACTION,
+	MERGING_CLAUSE_NUMBER,
+} MergingFuncOp;
+
+typedef struct MergingFunc
+{
+	Expr		xpr;
+	/* operation to perform */
+	MergingFuncOp mfop;
+	/* type Oid of result */
+	Oid			mftype pg_node_attr(query_jumble_ignore);
+	/* OID of collation, or InvalidOid if none */
+	Oid			mfcollid pg_node_attr(query_jumble_ignore);
+	/* token location, or -1 if unknown */
+	int			location;
+} MergingFunc;
+
+/*
  * SubscriptingRef: describes a subscripting operation over a container
  * (array, etc).
  *
@@ -1735,6 +1768,7 @@ typedef struct BooleanTest
 typedef struct MergeAction
 {
 	NodeTag		type;
+	int			index;			/* 1-based index of the clause */
 	bool		matched;		/* true=MATCHED, false=NOT MATCHED */
 	CmdType		commandType;	/* INSERT/UPDATE/DELETE/DO NOTHING */
 	/* OVERRIDING clause */
diff --git a/src/include/optimizer/paramassign.h b/src/include/optimizer/paramassign.h
new file mode 100644
index 501f728..1e95c8f
--- a/src/include/optimizer/paramassign.h
+++ b/src/include/optimizer/paramassign.h
@@ -20,6 +20,7 @@ extern Param *replace_outer_placeholderv
 										   PlaceHolderVar *phv);
 extern Param *replace_outer_agg(PlannerInfo *root, Aggref *agg);
 extern Param *replace_outer_grouping(PlannerInfo *root, GroupingFunc *grp);
+extern Param *replace_outer_merging(PlannerInfo *root, MergingFunc *mf);
 extern Param *replace_nestloop_param_var(PlannerInfo *root, Var *var);
 extern Param *replace_nestloop_param_placeholdervar(PlannerInfo *root,
 													PlaceHolderVar *phv);
diff --git a/src/include/parser/analyze.h b/src/include/parser/analyze.h
new file mode 100644
index 5ecdd10..28b66fc
--- a/src/include/parser/analyze.h
+++ b/src/include/parser/analyze.h
@@ -44,6 +44,8 @@ extern List *transformInsertRow(ParseSta
 								bool strip_indirection);
 extern List *transformUpdateTargetList(ParseState *pstate,
 									   List *origTlist);
+extern List *transformReturningList(ParseState *pstate, List *returningList,
+									ParseExprKind exprKind);
 extern Query *transformTopLevelStmt(ParseState *pstate, RawStmt *parseTree);
 extern Query *transformStmt(ParseState *pstate, Node *parseTree);
 
diff --git a/src/include/parser/kwlist.h b/src/include/parser/kwlist.h
new file mode 100644
index 2331aca..9c5d6ea
--- a/src/include/parser/kwlist.h
+++ b/src/include/parser/kwlist.h
@@ -80,6 +80,7 @@ PG_KEYWORD("characteristics", CHARACTERI
 PG_KEYWORD("check", CHECK, RESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("checkpoint", CHECKPOINT, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("class", CLASS, UNRESERVED_KEYWORD, BARE_LABEL)
+PG_KEYWORD("clause_number", CLAUSE_NUMBER, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("close", CLOSE, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("cluster", CLUSTER, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("coalesce", COALESCE, COL_NAME_KEYWORD, BARE_LABEL)
@@ -266,6 +267,7 @@ PG_KEYWORD("matched", MATCHED, UNRESERVE
 PG_KEYWORD("materialized", MATERIALIZED, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("maxvalue", MAXVALUE, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("merge", MERGE, UNRESERVED_KEYWORD, BARE_LABEL)
+PG_KEYWORD("merging", MERGING, COL_NAME_KEYWORD, BARE_LABEL)
 PG_KEYWORD("method", METHOD, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("minute", MINUTE_P, UNRESERVED_KEYWORD, AS_LABEL)
 PG_KEYWORD("minvalue", MINVALUE, UNRESERVED_KEYWORD, BARE_LABEL)
diff --git a/src/include/parser/parse_node.h b/src/include/parser/parse_node.h
new file mode 100644
index 99d6515..5b781d8
--- a/src/include/parser/parse_node.h
+++ b/src/include/parser/parse_node.h
@@ -61,7 +61,8 @@ typedef enum ParseExprKind
 	EXPR_KIND_DISTINCT_ON,		/* DISTINCT ON */
 	EXPR_KIND_LIMIT,			/* LIMIT */
 	EXPR_KIND_OFFSET,			/* OFFSET */
-	EXPR_KIND_RETURNING,		/* RETURNING */
+	EXPR_KIND_RETURNING,		/* RETURNING in INSERT/UPDATE/DELETE */
+	EXPR_KIND_MERGE_RETURNING,	/* RETURNING in MERGE */
 	EXPR_KIND_VALUES,			/* VALUES */
 	EXPR_KIND_VALUES_SINGLE,	/* single-row VALUES (in INSERT only) */
 	EXPR_KIND_CHECK_CONSTRAINT, /* CHECK constraint for a table */
diff --git a/src/pl/plpgsql/src/pl_exec.c b/src/pl/plpgsql/src/pl_exec.c
new file mode 100644
index 6d16913..51de094
--- a/src/pl/plpgsql/src/pl_exec.c
+++ b/src/pl/plpgsql/src/pl_exec.c
@@ -4267,9 +4267,9 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 	/*
 	 * If we have INTO, then we only need one row back ... but if we have INTO
 	 * STRICT or extra check too_many_rows, ask for two rows, so that we can
-	 * verify the statement returns only one.  INSERT/UPDATE/DELETE are always
-	 * treated strictly. Without INTO, just run the statement to completion
-	 * (tcount = 0).
+	 * verify the statement returns only one.  INSERT/UPDATE/DELETE/MERGE are
+	 * always treated strictly. Without INTO, just run the statement to
+	 * completion (tcount = 0).
 	 *
 	 * We could just ask for two rows always when using INTO, but there are
 	 * some cases where demanding the extra row costs significant time, eg by
@@ -4307,10 +4307,11 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 			Assert(stmt->mod_stmt);
 			exec_set_found(estate, (SPI_processed != 0));
 			break;
@@ -4489,10 +4490,11 @@ exec_stmt_dynexecute(PLpgSQL_execstate *
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 		case SPI_OK_UTILITY:
 		case SPI_OK_REWRITTEN:
 			break;
diff --git a/src/pl/tcl/pltcl.c b/src/pl/tcl/pltcl.c
new file mode 100644
index 6187e15..18d14a2
--- a/src/pl/tcl/pltcl.c
+++ b/src/pl/tcl/pltcl.c
@@ -2456,6 +2456,7 @@ pltcl_process_SPI_result(Tcl_Interp *int
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
+		case SPI_OK_MERGE_RETURNING:
 
 			/*
 			 * Process the tuples we got
diff --git a/src/test/regress/expected/merge.out b/src/test/regress/expected/merge.out
new file mode 100644
index f87905f..8baa0f4
--- a/src/test/regress/expected/merge.out
+++ b/src/test/regress/expected/merge.out
@@ -123,20 +123,20 @@ ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
 ERROR:  name "target" specified more than once
 DETAIL:  The name is used both as MERGE target table and data source.
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
-ERROR:  MERGE not supported in WITH query
-LINE 1: WITH foo AS (
-             ^
--- used in COPY
+ERROR:  WITH query "foo" does not have a RETURNING clause
+LINE 4: ) SELECT * FROM foo;
+                        ^
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) TO stdout;
-ERROR:  MERGE not supported in COPY
+ERROR:  COPY query must have a RETURNING clause
 -- unsupported relation types
 -- view
 CREATE VIEW tv AS SELECT * FROM target;
@@ -1256,7 +1256,7 @@ BEGIN;
 MERGE INTO sq_target
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1271,7 +1271,7 @@ INSERT INTO sq_source (sid, balance, del
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1295,7 +1295,7 @@ WITH targq AS (
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1304,20 +1304,191 @@ WHEN MATCHED AND tid < 2 THEN
 ROLLBACK;
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
+CREATE TABLE merge_actions(action text, abbrev text);
+INSERT INTO merge_actions VALUES ('INSERT', 'ins'), ('UPDATE', 'upd'), ('DELETE', 'del');
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
-ERROR:  syntax error at or near "RETURNING"
-LINE 10: RETURNING *;
-         ^
+    DELETE
+RETURNING MERGING(CLAUSE_NUMBER) AS clause,
+          (SELECT abbrev FROM merge_actions
+            WHERE action = MERGING(ACTION)) AS action,
+          t.*,
+          CASE MERGING(ACTION)
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ clause | action | tid | balance |     description     
+--------+--------+-----+---------+---------------------
+      3 | del    |   1 |     100 | Removed (1,100)
+      1 | upd    |   2 |     220 | Added 20 to balance
+      2 | ins    |   4 |      40 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- error when using MERGING() outside MERGE
+SELECT MERGING(ACTION) FROM sq_target;
+ERROR:  MERGING() can only be used in the RETURNING list of a MERGE command
+LINE 1: SELECT MERGING(ACTION) FROM sq_target;
+               ^
+UPDATE sq_target SET balance = balance + 1 RETURNING MERGING(CLAUSE_NUMBER);
+ERROR:  MERGING() can only be used in the RETURNING list of a MERGE command
+LINE 1: ...ATE sq_target SET balance = balance + 1 RETURNING MERGING(CL...
+                                                             ^
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(CLAUSE_NUMBER), MERGING(ACTION), t.*,
+              CASE MERGING(ACTION)
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING action, MERGING(ACTION) AS log_action, l.*
+)
+SELECT * FROM m2;
+ action | log_action | tid |     last_change     
+--------+------------+-----+---------------------
+ DELETE | UPDATE     |   1 | Removed (1,100)
+ UPDATE | INSERT     |   2 | Added 20 to balance
+ INSERT | INSERT     |   4 | Inserted (4,40)
+(3 rows)
+
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ tid |     last_change     
+-----+---------------------
+   1 | Removed (1,100)
+   2 | Added 20 to balance
+   4 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.*
+) TO stdout;
+DELETE	1	100
+UPDATE	2	220
+INSERT	4	40
+ROLLBACK;
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ action | tid | new_balance 
+--------+-----+-------------
+ DELETE |   1 |         100
+ UPDATE |   3 |         320
+ INSERT |   4 |         110
+(3 rows)
+
+ROLLBACK;
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ action | tid | balance 
+--------+-----+---------
+ DELETE |   1 |     100
+ UPDATE |   2 |     220
+ INSERT |   4 |      40
+(3 rows)
+
+ROLLBACK;
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ r_action | r_tid | r_balance 
+----------+-------+-----------
+ DELETE   |     1 |       100
+ UPDATE   |     3 |       320
+ INSERT   |     4 |       110
+(3 rows)
+
 ROLLBACK;
 -- EXPLAIN
 CREATE TABLE ex_mtarget (a int, b int)
@@ -1514,7 +1685,7 @@ SELECT * FROM sq_target WHERE tid = 1;
 (1 row)
 
 ROLLBACK;
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 NOTICE:  drop cascades to view v
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1640,6 +1811,32 @@ SELECT * FROM pa_target ORDER BY tid;
 (14 rows)
 
 ROLLBACK;
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING MERGING(ACTION), MERGING(CLAUSE_NUMBER), t.*;
+ action | clause_number | tid | balance |           val            
+--------+---------------+-----+---------+--------------------------
+ UPDATE |             1 |   2 |     110 | initial updated by merge
+(1 row)
+
+SELECT * FROM pa_target ORDER BY tid;
+ tid | balance |           val            
+-----+---------+--------------------------
+   2 |     110 | initial updated by merge
+   3 |     300 | initial
+   5 |     500 | initial
+   7 |     700 | initial
+   9 |     900 | initial
+  11 |    1100 | initial
+  13 |    1300 | initial
+(7 rows)
+
+ROLLBACK;
 DROP TABLE pa_target CASCADE;
 -- The target table is partitioned in the same way, but this time by attaching
 -- partitions which have columns in different order, dropped columns etc.
@@ -1889,7 +2086,21 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING MERGING(ACTION), MERGING(CLAUSE_NUMBER), t.*;
+ action | clause_number |          logts           | tid | balance |           val            
+--------+---------------+--------------------------+-----+---------+--------------------------
+ UPDATE |             1 | Tue Jan 31 00:00:00 2017 |   1 |     110 | initial updated by merge
+ UPDATE |             1 | Tue Feb 28 00:00:00 2017 |   2 |     220 | initial updated by merge
+ INSERT |             2 | Sun Jan 15 00:00:00 2017 |   3 |      30 | inserted by merge
+ UPDATE |             1 | Tue Jan 31 00:00:00 2017 |   4 |     440 | initial updated by merge
+ UPDATE |             1 | Tue Feb 28 00:00:00 2017 |   5 |     550 | initial updated by merge
+ INSERT |             2 | Sun Jan 15 00:00:00 2017 |   6 |      60 | inserted by merge
+ UPDATE |             1 | Tue Jan 31 00:00:00 2017 |   7 |     770 | initial updated by merge
+ UPDATE |             1 | Tue Feb 28 00:00:00 2017 |   8 |     880 | initial updated by merge
+ INSERT |             2 | Sun Jan 15 00:00:00 2017 |   9 |      90 | inserted by merge
+(9 rows)
+
 SELECT * FROM pa_target ORDER BY tid;
           logts           | tid | balance |           val            
 --------------------------+-----+---------+--------------------------
diff --git a/src/test/regress/expected/rowsecurity.out b/src/test/regress/expected/rowsecurity.out
new file mode 100644
index 6988128..3fd4fab
--- a/src/test/regress/expected/rowsecurity.out
+++ b/src/test/regress/expected/rowsecurity.out
@@ -2302,6 +2302,35 @@ WHEN MATCHED THEN
 	UPDATE SET dnotes = dnotes || ' notes added by merge8 '
 WHEN NOT MATCHED THEN
 	INSERT VALUES (13, 44, 1, 'regress_rls_bob', 'new manga');
+SELECT * FROM document WHERE did = 13;
+ did | cid | dlevel | dauthor | dtitle | dnotes 
+-----+-----+--------+---------+--------+--------
+(0 rows)
+
+-- but not OK if RETURNING is used
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge9 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 44, 1, 'regress_rls_bob', 'new manga')
+RETURNING *;
+ERROR:  new row violates row-level security policy for table "document"
+-- but OK if new row is visible
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge10 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 11, 1, 'regress_rls_bob', 'new novel')
+RETURNING *;
+ sdid | did | cid | dlevel |     dauthor     |  dtitle   | dnotes 
+------+-----+-----+--------+-----------------+-----------+--------
+   14 |  14 |  11 |      1 | regress_rls_bob | new novel | 
+(1 row)
+
 RESET SESSION AUTHORIZATION;
 -- drop the restrictive SELECT policy so that we can look at the
 -- final state of the table
@@ -2325,7 +2354,8 @@ SELECT * FROM document;
   12 |  11 |      1 | regress_rls_bob   | another novel                    | 
    1 |  11 |      1 | regress_rls_bob   | my first novel                   |  notes added by merge2  notes added by merge3  notes added by merge4  notes added by merge7 
   13 |  44 |      1 | regress_rls_bob   | new manga                        | 
-(15 rows)
+  14 |  11 |      1 | regress_rls_bob   | new novel                        | 
+(16 rows)
 
 --
 -- ROLE/GROUP
diff --git a/src/test/regress/expected/rules.out b/src/test/regress/expected/rules.out
new file mode 100644
index abc944e..9a0ca93
--- a/src/test/regress/expected/rules.out
+++ b/src/test/regress/expected/rules.out
@@ -3625,7 +3625,8 @@ MERGE INTO rule_merge2 t USING (SELECT 1
 -- test deparsing
 CREATE TABLE sf_target(id int, data text, filling int[]);
 CREATE FUNCTION merge_sf_test()
- RETURNS void
+ RETURNS TABLE(action text, clause_number int,
+               a int, b text, id int, data text, filling int[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -3662,11 +3663,13 @@ WHEN NOT MATCHED
    VALUES (s.a, s.b, DEFAULT)
 WHEN NOT MATCHED
    THEN INSERT (filling[1], id)
-   VALUES (s.a, s.a);
+   VALUES (s.a, s.a)
+RETURNING
+   merging(action), merging(clause_number) AS clause_number, *;
 END;
 \sf merge_sf_test
 CREATE OR REPLACE FUNCTION public.merge_sf_test()
- RETURNS void
+ RETURNS TABLE(action text, clause_number integer, a integer, b text, id integer, data text, filling integer[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -3703,7 +3706,14 @@ BEGIN ATOMIC
       VALUES (s.a, s.b, DEFAULT)
     WHEN NOT MATCHED
      THEN INSERT (filling[1], id)
-      VALUES (s.a, s.a);
+      VALUES (s.a, s.a)
+   RETURNING MERGING(ACTION) AS action,
+     MERGING(CLAUSE_NUMBER) AS clause_number,
+     s.a,
+     s.b,
+     t.id,
+     t.data,
+     t.filling;
 END
 DROP FUNCTION merge_sf_test;
 DROP TABLE sf_target;
diff --git a/src/test/regress/expected/with.out b/src/test/regress/expected/with.out
new file mode 100644
index 7d796ea..4666873
--- a/src/test/regress/expected/with.out
+++ b/src/test/regress/expected/with.out
@@ -3493,6 +3493,16 @@ SELECT * FROM t;
 ERROR:  WITH query "t" does not have a RETURNING clause
 LINE 4: SELECT * FROM t;
                       ^
+-- RETURNING tries to return its own output
+WITH RECURSIVE t(action, a) AS (
+	MERGE INTO y USING (VALUES (11)) v(a) ON y.a = v.a
+		WHEN NOT MATCHED THEN INSERT VALUES (v.a)
+		RETURNING MERGING(ACTION), (SELECT a FROM t)
+)
+SELECT * FROM t;
+ERROR:  recursive query "t" must not contain data-modifying statements
+LINE 1: WITH RECURSIVE t(action, a) AS (
+                       ^
 -- data-modifying WITH allowed only at the top level
 SELECT * FROM (
 	WITH t AS (UPDATE y SET a=a+1 RETURNING *)
diff --git a/src/test/regress/sql/merge.sql b/src/test/regress/sql/merge.sql
new file mode 100644
index 66cb75a..daa9dde
--- a/src/test/regress/sql/merge.sql
+++ b/src/test/regress/sql/merge.sql
@@ -86,12 +86,12 @@ MERGE INTO target
 USING target
 ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
--- used in COPY
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
@@ -816,7 +816,7 @@ BEGIN;
 MERGE INTO sq_target
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -829,7 +829,7 @@ INSERT INTO sq_source (sid, balance, del
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -847,7 +847,7 @@ WITH targq AS (
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -857,17 +857,146 @@ ROLLBACK;
 
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
+CREATE TABLE merge_actions(action text, abbrev text);
+INSERT INTO merge_actions VALUES ('INSERT', 'ins'), ('UPDATE', 'upd'), ('DELETE', 'del');
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
+    DELETE
+RETURNING MERGING(CLAUSE_NUMBER) AS clause,
+          (SELECT abbrev FROM merge_actions
+            WHERE action = MERGING(ACTION)) AS action,
+          t.*,
+          CASE MERGING(ACTION)
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ROLLBACK;
+
+-- error when using MERGING() outside MERGE
+SELECT MERGING(ACTION) FROM sq_target;
+UPDATE sq_target SET balance = balance + 1 RETURNING MERGING(CLAUSE_NUMBER);
+
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(CLAUSE_NUMBER), MERGING(ACTION), t.*,
+              CASE MERGING(ACTION)
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING action, MERGING(ACTION) AS log_action, l.*
+)
+SELECT * FROM m2;
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ROLLBACK;
+
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.*
+) TO stdout;
+ROLLBACK;
+
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ROLLBACK;
+
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ROLLBACK;
+
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
 ROLLBACK;
 
 -- EXPLAIN
@@ -966,7 +1095,7 @@ WHEN MATCHED THEN
 SELECT * FROM sq_target WHERE tid = 1;
 ROLLBACK;
 
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1033,6 +1162,17 @@ SELECT merge_func();
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING MERGING(ACTION), MERGING(CLAUSE_NUMBER), t.*;
+SELECT * FROM pa_target ORDER BY tid;
+ROLLBACK;
+
 DROP TABLE pa_target CASCADE;
 
 -- The target table is partitioned in the same way, but this time by attaching
@@ -1209,7 +1349,8 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING MERGING(ACTION), MERGING(CLAUSE_NUMBER), t.*;
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
diff --git a/src/test/regress/sql/rowsecurity.sql b/src/test/regress/sql/rowsecurity.sql
new file mode 100644
index dec7340..1d5ed0a
--- a/src/test/regress/sql/rowsecurity.sql
+++ b/src/test/regress/sql/rowsecurity.sql
@@ -992,6 +992,27 @@ WHEN MATCHED THEN
 	UPDATE SET dnotes = dnotes || ' notes added by merge8 '
 WHEN NOT MATCHED THEN
 	INSERT VALUES (13, 44, 1, 'regress_rls_bob', 'new manga');
+SELECT * FROM document WHERE did = 13;
+
+-- but not OK if RETURNING is used
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge9 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 44, 1, 'regress_rls_bob', 'new manga')
+RETURNING *;
+
+-- but OK if new row is visible
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge10 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 11, 1, 'regress_rls_bob', 'new novel')
+RETURNING *;
 
 RESET SESSION AUTHORIZATION;
 -- drop the restrictive SELECT policy so that we can look at the
diff --git a/src/test/regress/sql/rules.sql b/src/test/regress/sql/rules.sql
new file mode 100644
index 93aff4e..4b15a13
--- a/src/test/regress/sql/rules.sql
+++ b/src/test/regress/sql/rules.sql
@@ -1281,7 +1281,8 @@ MERGE INTO rule_merge2 t USING (SELECT 1
 CREATE TABLE sf_target(id int, data text, filling int[]);
 
 CREATE FUNCTION merge_sf_test()
- RETURNS void
+ RETURNS TABLE(action text, clause_number int,
+               a int, b text, id int, data text, filling int[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -1318,7 +1319,9 @@ WHEN NOT MATCHED
    VALUES (s.a, s.b, DEFAULT)
 WHEN NOT MATCHED
    THEN INSERT (filling[1], id)
-   VALUES (s.a, s.a);
+   VALUES (s.a, s.a)
+RETURNING
+   merging(action), merging(clause_number) AS clause_number, *;
 END;
 
 \sf merge_sf_test
diff --git a/src/test/regress/sql/with.sql b/src/test/regress/sql/with.sql
new file mode 100644
index f8a213e..1d82d4a
--- a/src/test/regress/sql/with.sql
+++ b/src/test/regress/sql/with.sql
@@ -1591,6 +1591,14 @@ WITH t AS (
 )
 SELECT * FROM t;
 
+-- RETURNING tries to return its own output
+WITH RECURSIVE t(action, a) AS (
+	MERGE INTO y USING (VALUES (11)) v(a) ON y.a = v.a
+		WHEN NOT MATCHED THEN INSERT VALUES (v.a)
+		RETURNING MERGING(ACTION), (SELECT a FROM t)
+)
+SELECT * FROM t;
+
 -- data-modifying WITH allowed only at the top level
 SELECT * FROM (
 	WITH t AS (UPDATE y SET a=a+1 RETURNING *)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
new file mode 100644
index 90b37b9..fe1e1c2
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -1575,6 +1575,8 @@ MergePath
 MergeScanSelCache
 MergeStmt
 MergeWhenClause
+MergingFunc
+MergingFuncOp
 MetaCommand
 MinMaxAggInfo
 MinMaxAggPath

diff --git a/doc/src/sgml/dml.sgml b/doc/src/sgml/dml.sgml
new file mode 100644
index cbbc5e2..3d95bdb
--- a/doc/src/sgml/dml.sgml
+++ b/doc/src/sgml/dml.sgml
@@ -283,10 +283,15 @@ DELETE FROM products;
    <secondary>RETURNING</secondary>
   </indexterm>
 
+  <indexterm zone="dml-returning">
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
   <para>
    Sometimes it is useful to obtain data from modified rows while they are
    being manipulated.  The <command>INSERT</command>, <command>UPDATE</command>,
-   and <command>DELETE</command> commands all have an
+   <command>DELETE</command>, and <command>MERGE</command> commands all have an
    optional <literal>RETURNING</literal> clause that supports this.  Use
    of <literal>RETURNING</literal> avoids performing an extra database query to
    collect the data, and is especially valuable when it would otherwise be
@@ -339,6 +344,21 @@ DELETE FROM products
 </programlisting>
   </para>
 
+  <para>
+   In a <command>MERGE</command>, the data available to <literal>RETURNING</literal> is
+   the content of the source row plus the content of the inserted, updated, or
+   deleted target row.  Since it is quite common for the source and target to
+   have many of the same columns, specifying <literal>RETURNING *</literal>
+   can lead to a lot of duplicated columns, so it is often more useful to
+   qualify it so as to return just the source or target row.  For example:
+<programlisting>
+MERGE INTO products p USING new_products n ON p.product_no = n.product_no
+  WHEN NOT MATCHED THEN INSERT VALUES (n.product_no, n.name, n.price)
+  WHEN MATCHED THEN UPDATE SET name = n.name, price = n.price
+  RETURNING p.*;
+</programlisting>
+  </para>
+
   <para>
    If there are triggers (<xref linkend="triggers"/>) on the target table,
    the data available to <literal>RETURNING</literal> is the row as modified by
diff --git a/doc/src/sgml/func.sgml b/doc/src/sgml/func.sgml
new file mode 100644
index e5fa82c..d3b87e5
--- a/doc/src/sgml/func.sgml
+++ b/doc/src/sgml/func.sgml
@@ -22421,6 +22421,84 @@ SELECT count(*) FROM sometable;
 
  </sect1>
 
+ <sect1 id="functions-merge-support">
+  <title>Merge Support Functions</title>
+
+  <indexterm>
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
+  <indexterm>
+   <primary>MERGING</primary>
+  </indexterm>
+
+  <para>
+   <productname>PostgreSQL</productname> includes one merge support function
+   that may be used in the <literal>RETURNING</literal> list of a
+   <xref linkend="sql-merge"/> command to return additional information about
+   the action taken for each row:
+<synopsis>
+<function id="function-merging">MERGING</function> ( <parameter>property</parameter> )
+</synopsis>
+   The <parameter>property</parameter> key word specifies what to return:
+
+   <variablelist>
+    <varlistentry>
+     <term><literal>ACTION</literal></term>
+     <listitem>
+      <para>
+       The merge action command executed for the current row.  This will be
+       <literal>'INSERT'</literal>, <literal>'UPDATE'</literal>, or
+       <literal>'DELETE'</literal>.
+      </para>
+     </listitem>
+    </varlistentry>
+
+    <varlistentry>
+     <term><literal>CLAUSE_NUMBER</literal></term>
+     <listitem>
+      <para>
+       The 1-based index of the <literal>WHEN</literal> clause executed for
+       the current row.
+      </para>
+     </listitem>
+    </varlistentry>
+   </variablelist>
+  </para>
+
+  <para>
+   The following example illustrates how this may be used to determine
+   which actions were performed for each row affected by the
+   <command>MERGE</command> command:
+<screen><![CDATA[
+MERGE INTO products p
+  USING stock s ON p.product_id = s.product_id
+  WHEN MATCHED AND s.quantity > 0 THEN
+    UPDATE SET in_stock = true, quantity = s.quantity
+  WHEN MATCHED THEN
+    UPDATE SET in_stock = false, quantity = 0
+  WHEN NOT MATCHED THEN
+    INSERT (product_id, in_stock, quantity)
+      VALUES (s.product_id, true, s.quantity)
+  RETURNING MERGING(ACTION), MERGING(CLAUSE_NUMBER), p.*;
+
+ action | clause_number | product_id | in_stock | quantity
+--------+---------------+------------+----------+----------
+ UPDATE |             1 |       1001 | t        |       50
+ UPDATE |             2 |       1002 | f        |        0
+ INSERT |             3 |       1003 | t        |       10
+]]></screen>
+  </para>
+
+  <para>
+   Note that this function can only be used in the <literal>RETURNING</literal>
+   list of a <command>MERGE</command> command.  It is an error to use it in any
+   other part of a query.
+  </para>
+
+ </sect1>
+
  <sect1 id="functions-subquery">
   <title>Subquery Expressions</title>
 
diff --git a/doc/src/sgml/glossary.sgml b/doc/src/sgml/glossary.sgml
new file mode 100644
index 8c2f114..a81c17a
--- a/doc/src/sgml/glossary.sgml
+++ b/doc/src/sgml/glossary.sgml
@@ -1442,9 +1442,9 @@
      to a <glossterm linkend="glossary-client">client</glossterm> upon the
      completion of an <acronym>SQL</acronym> command, usually a
      <command>SELECT</command> but it can be an
-     <command>INSERT</command>, <command>UPDATE</command>, or
-     <command>DELETE</command> command if the <literal>RETURNING</literal>
-     clause is specified.
+     <command>INSERT</command>, <command>UPDATE</command>,
+     <command>DELETE</command>, or <command>MERGE</command> command if the
+     <literal>RETURNING</literal> clause is specified.
     </para>
     <para>
      The fact that a result set is a relation means that a query can be used
diff --git a/doc/src/sgml/plpgsql.sgml b/doc/src/sgml/plpgsql.sgml
new file mode 100644
index c2b9c6a..406834e
--- a/doc/src/sgml/plpgsql.sgml
+++ b/doc/src/sgml/plpgsql.sgml
@@ -1043,8 +1043,8 @@ INSERT INTO mytable VALUES (1,'one'), (2
     </para>
 
     <para>
-     If the command does return rows (for example <command>SELECT</command>,
-     or <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>
+     If the command does return rows (for example <command>SELECT</command>, or
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
      with <literal>RETURNING</literal>), there are two ways to proceed.
      When the command will return at most one row, or you only care about
      the first row of output, write the command as usual but add
@@ -1172,6 +1172,7 @@ SELECT <replaceable>select_expressions</
 INSERT ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 UPDATE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 DELETE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
+MERGE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 </synopsis>
 
      where <replaceable>target</replaceable> can be a record variable, a row
@@ -1182,8 +1183,8 @@ DELETE ... RETURNING <replaceable>expres
      <literal>INTO</literal> clause) just as described above,
      and the plan is cached in the same way.
      This works for <command>SELECT</command>,
-     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
-     <literal>RETURNING</literal>, and certain utility commands
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
+     with <literal>RETURNING</literal>, and certain utility commands
      that return row sets, such as <command>EXPLAIN</command>.
      Except for the <literal>INTO</literal> clause, the SQL command is the same
      as it would be written outside <application>PL/pgSQL</application>.
@@ -1259,7 +1260,7 @@ END;
     </para>
 
     <para>
-     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
+     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command> with
      <literal>RETURNING</literal>, <application>PL/pgSQL</application> reports
      an error for more than one returned row, even when
      <literal>STRICT</literal> is not specified.  This is because there
diff --git a/doc/src/sgml/queries.sgml b/doc/src/sgml/queries.sgml
new file mode 100644
index 648b283..372cce1
--- a/doc/src/sgml/queries.sgml
+++ b/doc/src/sgml/queries.sgml
@@ -2063,9 +2063,10 @@ SELECT <replaceable>select_list</replace
    Table Expressions or <acronym>CTE</acronym>s, can be thought of as defining
    temporary tables that exist just for one query.  Each auxiliary statement
    in a <literal>WITH</literal> clause can be a <command>SELECT</command>,
-   <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>; and the
+   <command>INSERT</command>, <command>UPDATE</command>, <command>DELETE</command>,
+   or <command>MERGE</command>; and the
    <literal>WITH</literal> clause itself is attached to a primary statement that can
-   be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
+   also be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
    <command>DELETE</command>, or <command>MERGE</command>.
   </para>
 
@@ -2599,8 +2600,8 @@ SELECT * FROM w AS w1 JOIN w AS w2 ON w1
    <title>Data-Modifying Statements in <literal>WITH</literal></title>
 
    <para>
-    You can use most data-modifying statements (<command>INSERT</command>,
-    <command>UPDATE</command>, or <command>DELETE</command>, but not
+    You can use data-modifying statements (<command>INSERT</command>,
+    <command>UPDATE</command>, <command>DELETE</command>, or
     <command>MERGE</command>) in <literal>WITH</literal>.  This
     allows you to perform several different operations in the same query.
     An example is:
diff --git a/doc/src/sgml/ref/copy.sgml b/doc/src/sgml/ref/copy.sgml
new file mode 100644
index 55764fc..6c83e30
--- a/doc/src/sgml/ref/copy.sgml
+++ b/doc/src/sgml/ref/copy.sgml
@@ -130,16 +130,19 @@ COPY { <replaceable class="parameter">ta
       A <link linkend="sql-select"><command>SELECT</command></link>,
       <link linkend="sql-values"><command>VALUES</command></link>,
       <link linkend="sql-insert"><command>INSERT</command></link>,
-      <link linkend="sql-update"><command>UPDATE</command></link>, or
-      <link linkend="sql-delete"><command>DELETE</command></link> command whose results are to be
-      copied.  Note that parentheses are required around the query.
+      <link linkend="sql-update"><command>UPDATE</command></link>,
+      <link linkend="sql-delete"><command>DELETE</command></link>, or
+      <link linkend="sql-merge"><command>MERGE</command></link> command
+      whose results are to be copied.  Note that parentheses are required
+      around the query.
      </para>
      <para>
-      For <command>INSERT</command>, <command>UPDATE</command> and
-      <command>DELETE</command> queries a <literal>RETURNING</literal> clause
-      must be provided, and the target relation must not have a conditional
-      rule, nor an <literal>ALSO</literal> rule, nor an
-      <literal>INSTEAD</literal> rule that expands to multiple statements.
+      For <command>INSERT</command>, <command>UPDATE</command>,
+      <command>DELETE</command>, and <command>MERGE</command> queries a
+      <literal>RETURNING</literal> clause must be provided, and the target
+      relation must not have a conditional rule, nor an
+      <literal>ALSO</literal> rule, nor an <literal>INSTEAD</literal> rule
+      that expands to multiple statements.
      </para>
     </listitem>
    </varlistentry>
diff --git a/doc/src/sgml/ref/merge.sgml b/doc/src/sgml/ref/merge.sgml
new file mode 100644
index 7a8ac40..2f0ad34
--- a/doc/src/sgml/ref/merge.sgml
+++ b/doc/src/sgml/ref/merge.sgml
@@ -25,6 +25,7 @@ PostgreSQL documentation
 MERGE INTO [ ONLY ] <replaceable class="parameter">target_table_name</replaceable> [ * ] [ [ AS ] <replaceable class="parameter">target_alias</replaceable> ]
 USING <replaceable class="parameter">data_source</replaceable> ON <replaceable class="parameter">join_condition</replaceable>
 <replaceable class="parameter">when_clause</replaceable> [...]
+[ RETURNING * | <replaceable class="parameter">output_expression</replaceable> [ [ AS ] <replaceable class="parameter">output_name</replaceable> ] [, ...] ]
 
 <phrase>where <replaceable class="parameter">data_source</replaceable> is:</phrase>
 
@@ -95,6 +96,18 @@ DELETE
   </para>
 
   <para>
+   The optional <literal>RETURNING</literal> clause causes <command>MERGE</command>
+   to compute and return value(s) based on each row inserted, updated, or
+   deleted.  Any expression using the source or target table's columns, or
+   the <xref linkend="function-merging"/> function can be computed.  When an
+   <command>INSERT</command> or <command>UPDATE</command> action is performed,
+   the new values of the target table's columns are used.  When a
+   <command>DELETE</command> is performed, the old values of the target table's
+   columns are used.  The syntax of the <literal>RETURNING</literal> list is
+   identical to that of the output list of <command>SELECT</command>.
+  </para>
+
+  <para>
    There is no separate <literal>MERGE</literal> privilege.
    If you specify an update action, you must have the
    <literal>UPDATE</literal> privilege on the column(s)
@@ -129,6 +142,18 @@ DELETE
 
   <variablelist>
    <varlistentry>
+    <term><replaceable class="parameter">with_query</replaceable></term>
+    <listitem>
+     <para>
+      The <literal>WITH</literal> clause allows you to specify one or more
+      subqueries that can be referenced by name in the <command>MERGE</command>
+      query. See <xref linkend="queries-with"/> and <xref linkend="sql-select"/>
+      for details.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
     <term><replaceable class="parameter">target_table_name</replaceable></term>
     <listitem>
      <para>
@@ -411,13 +436,30 @@ DELETE
    </varlistentry>
 
    <varlistentry>
-    <term><replaceable class="parameter">with_query</replaceable></term>
+    <term><replaceable class="parameter">output_expression</replaceable></term>
     <listitem>
      <para>
-      The <literal>WITH</literal> clause allows you to specify one or more
-      subqueries that can be referenced by name in the <command>MERGE</command>
-      query. See <xref linkend="queries-with"/> and <xref linkend="sql-select"/>
-      for details.
+      An expression to be computed and returned by the <command>MERGE</command>
+      command after each row is merged.  The expression can use any columns of
+      the source or target tables, or the <xref linkend="function-merging"/>
+      function to return additional information about the action executed.
+     </para>
+     <para>
+      Writing <literal>*</literal> will return all columns from the source
+      table, followed by all columns from the target table.  Often this will
+      lead to a lot of duplication, since it is common for the source and
+      target tables to have a lot of the same columns.  This can be avoided by
+      qualifying the <literal>*</literal> with the name or alias of the source
+      or target table.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
+    <term><replaceable class="parameter">output_name</replaceable></term>
+    <listitem>
+     <para>
+      A name to use for a returned column.
      </para>
     </listitem>
    </varlistentry>
@@ -440,6 +482,13 @@ MERGE <replaceable class="parameter">tot
    were changed in any way.
   </para>
 
+  <para>
+   If the <command>MERGE</command> command contains a <literal>RETURNING</literal>
+   clause, the result will be similar to that of a <command>SELECT</command>
+   statement containing the columns and values defined in the
+   <literal>RETURNING</literal> list, computed over the row(s) inserted, updated,
+   or deleted by the command.
+  </para>
  </refsect1>
 
  <refsect1>
@@ -560,13 +609,6 @@ MERGE <replaceable class="parameter">tot
   </para>
 
   <para>
-   There is no <literal>RETURNING</literal> clause with
-   <command>MERGE</command>.  Actions of <command>INSERT</command>,
-   <command>UPDATE</command> and <command>DELETE</command> cannot contain
-   <literal>RETURNING</literal> or <literal>WITH</literal> clauses.
-  </para>
-
-  <para>
    When <command>MERGE</command> is run concurrently with other commands
    that modify the target table, the usual transaction isolation rules
    apply; see <xref linkend="transaction-iso"/> for an explanation
@@ -618,7 +660,8 @@ WHEN NOT MATCHED THEN
   <para>
    Attempt to insert a new stock item along with the quantity of stock. If
    the item already exists, instead update the stock count of the existing
-   item. Don't allow entries that have zero stock.
+   item. Don't allow entries that have zero stock. Return details of all
+   changes made.
 <programlisting>
 MERGE INTO wines w
 USING wine_stock_changes s
@@ -628,7 +671,8 @@ WHEN NOT MATCHED AND s.stock_delta > 0 T
 WHEN MATCHED AND w.stock + s.stock_delta > 0 THEN
   UPDATE SET stock = w.stock + s.stock_delta
 WHEN MATCHED THEN
-  DELETE;
+  DELETE
+RETURNING MERGING(ACTION), w.*;
 </programlisting>
 
    The <literal>wine_stock_changes</literal> table might be, for example, a
@@ -643,8 +687,9 @@ WHEN MATCHED THEN
     This command conforms to the <acronym>SQL</acronym> standard.
   </para>
    <para>
-    The <literal>WITH</literal> clause and <literal>DO NOTHING</literal>
-    action are extensions to the <acronym>SQL</acronym> standard.
+    The <literal>WITH</literal> clause, <literal>DO NOTHING</literal> action,
+    and <literal>RETURNING</literal> clause are extensions to the
+    <acronym>SQL</acronym> standard.
   </para>
  </refsect1>
 </refentry>
diff --git a/doc/src/sgml/ref/select.sgml b/doc/src/sgml/ref/select.sgml
new file mode 100644
index 9917df7..066aed4
--- a/doc/src/sgml/ref/select.sgml
+++ b/doc/src/sgml/ref/select.sgml
@@ -74,7 +74,7 @@ SELECT [ ALL | DISTINCT [ ON ( <replacea
 
 <phrase>and <replaceable class="parameter">with_query</replaceable> is:</phrase>
 
-    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> )
+    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> | <replaceable class="parameter">merge</replaceable> )
         [ SEARCH { BREADTH | DEPTH } FIRST BY <replaceable>column_name</replaceable> [, ...] SET <replaceable>search_seq_col_name</replaceable> ]
         [ CYCLE <replaceable>column_name</replaceable> [, ...] SET <replaceable>cycle_mark_col_name</replaceable> [ TO <replaceable>cycle_mark_value</replaceable> DEFAULT <replaceable>cycle_mark_default</replaceable> ] USING <replaceable>cycle_path_col_name</replaceable> ]
 
@@ -230,10 +230,10 @@ TABLE [ ONLY ] <replaceable class="param
     The subqueries effectively act as temporary tables or views
     for the duration of the primary query.
     Each subquery can be a <command>SELECT</command>, <command>TABLE</command>, <command>VALUES</command>,
-    <command>INSERT</command>, <command>UPDATE</command> or
-    <command>DELETE</command> statement.
+    <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command> statement.
     When writing a data-modifying statement (<command>INSERT</command>,
-    <command>UPDATE</command> or <command>DELETE</command>) in
+    <command>UPDATE</command>, <command>DELETE</command>, or <command>MERGE</command>) in
     <literal>WITH</literal>, it is usual to include a <literal>RETURNING</literal> clause.
     It is the output of <literal>RETURNING</literal>, <emphasis>not</emphasis> the underlying
     table that the statement modifies, that forms the temporary table that is
@@ -2184,7 +2184,8 @@ SELECT 2+2;
 
    <para>
     <productname>PostgreSQL</productname> allows <command>INSERT</command>,
-    <command>UPDATE</command>, and <command>DELETE</command> to be used as <literal>WITH</literal>
+    <command>UPDATE</command>, <command>DELETE</command>, and
+    <command>MERGE</command> to be used as <literal>WITH</literal>
     queries.  This is not found in the SQL standard.
    </para>
   </refsect2>
diff --git a/doc/src/sgml/rowtypes.sgml b/doc/src/sgml/rowtypes.sgml
new file mode 100644
index 4d86f97..bbeac84
--- a/doc/src/sgml/rowtypes.sgml
+++ b/doc/src/sgml/rowtypes.sgml
@@ -348,7 +348,7 @@ SELECT m.* FROM some_table, LATERAL myfu
    column expansion of this kind when it appears at the top level of
    a <link linkend="queries-select-lists"><command>SELECT</command> output
    list</link>, a <link linkend="dml-returning"><literal>RETURNING</literal>
-   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>,
+   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>,
    a <link linkend="queries-values"><literal>VALUES</literal> clause</link>, or
    a <link linkend="sql-syntax-row-constructors">row constructor</link>.
    In all other contexts (including when nested inside one of those
diff --git a/doc/src/sgml/spi.sgml b/doc/src/sgml/spi.sgml
new file mode 100644
index 47f4b5b..bb37786
--- a/doc/src/sgml/spi.sgml
+++ b/doc/src/sgml/spi.sgml
@@ -301,8 +301,9 @@ SPI_execute("INSERT INTO foo SELECT * FR
    is returned in the global variable <varname>SPI_processed</varname>.
    If the return value of the function is <symbol>SPI_OK_SELECT</symbol>,
    <symbol>SPI_OK_INSERT_RETURNING</symbol>,
-   <symbol>SPI_OK_DELETE_RETURNING</symbol>, or
-   <symbol>SPI_OK_UPDATE_RETURNING</symbol>,
+   <symbol>SPI_OK_DELETE_RETURNING</symbol>,
+   <symbol>SPI_OK_UPDATE_RETURNING</symbol>, or
+   <symbol>SPI_OK_MERGE_RETURNING</symbol>,
    then you can use the
    global pointer <literal>SPITupleTable *SPI_tuptable</literal> to
    access the result rows.  Some utility commands (such as
@@ -472,6 +473,15 @@ typedef struct SPITupleTable
       </para>
      </listitem>
     </varlistentry>
+
+    <varlistentry>
+     <term><symbol>SPI_OK_MERGE_RETURNING</symbol></term>
+     <listitem>
+      <para>
+       if a <command>MERGE RETURNING</command> was executed
+      </para>
+     </listitem>
+    </varlistentry>
 
     <varlistentry>
      <term><symbol>SPI_OK_UTILITY</symbol></term>
diff --git a/doc/src/sgml/xfunc.sgml b/doc/src/sgml/xfunc.sgml
new file mode 100644
index 8a79ad0..7d05369
--- a/doc/src/sgml/xfunc.sgml
+++ b/doc/src/sgml/xfunc.sgml
@@ -177,7 +177,8 @@
     statements separated by semicolons.  A semicolon after the last
     statement is optional.  Unless the function is declared to return
     <type>void</type>, the last statement must be a <command>SELECT</command>,
-    or an <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>
+    or an <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command>
     that has a <literal>RETURNING</literal> clause.
    </para>
 
@@ -1236,8 +1237,9 @@ SELECT x, CASE WHEN x &gt; 0 THEN genera
 
     <note>
      <para>
-      If a function's last command is <command>INSERT</command>, <command>UPDATE</command>,
-      or <command>DELETE</command> with <literal>RETURNING</literal>, that command will
+      If a function's last command is <command>INSERT</command>,
+      <command>UPDATE</command>, <command>DELETE</command>, or
+      <command>MERGE</command> with <literal>RETURNING</literal>, that command will
       always be executed to completion, even if the function is not declared
       with <literal>SETOF</literal> or the calling query does not fetch all the
       result rows.  Any extra rows produced by the <literal>RETURNING</literal>
diff --git a/src/backend/commands/copy.c b/src/backend/commands/copy.c
new file mode 100644
index cc0786c..e96067a
--- a/src/backend/commands/copy.c
+++ b/src/backend/commands/copy.c
@@ -283,12 +283,6 @@ DoCopy(ParseState *pstate, const CopyStm
 	{
 		Assert(stmt->query);
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(stmt->query, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in COPY"));
-
 		query = makeNode(RawStmt);
 		query->stmt = stmt->query;
 		query->stmt_location = stmt_location;
diff --git a/src/backend/commands/copyto.c b/src/backend/commands/copyto.c
new file mode 100644
index 20ffc90..583f990
--- a/src/backend/commands/copyto.c
+++ b/src/backend/commands/copyto.c
@@ -510,7 +510,8 @@ BeginCopyTo(ParseState *pstate,
 		{
 			Assert(query->commandType == CMD_INSERT ||
 				   query->commandType == CMD_UPDATE ||
-				   query->commandType == CMD_DELETE);
+				   query->commandType == CMD_DELETE ||
+				   query->commandType == CMD_MERGE);
 
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
diff --git a/src/backend/executor/execExpr.c b/src/backend/executor/execExpr.c
new file mode 100644
index 3181b11..c4c2fb5
--- a/src/backend/executor/execExpr.c
+++ b/src/backend/executor/execExpr.c
@@ -1108,6 +1108,21 @@ ExecInitExprRec(Expr *node, ExprState *s
 				break;
 			}
 
+		case T_MergingFunc:
+			{
+				MergingFunc *mf_node = (MergingFunc *) node;
+
+				if (!state->parent ||
+					!IsA(state->parent, ModifyTableState) ||
+					((ModifyTableState *) state->parent)->operation != CMD_MERGE)
+					elog(ERROR, "MergingFunc found in non-merge plan node");
+
+				scratch.opcode = EEOP_MERGING_FUNC;
+				scratch.d.merging_func.mfop = mf_node->mfop;
+				ExprEvalPushStep(state, &scratch);
+				break;
+			}
+
 		case T_SubscriptingRef:
 			{
 				SubscriptingRef *sbsref = (SubscriptingRef *) node;
diff --git a/src/backend/executor/execExprInterp.c b/src/backend/executor/execExprInterp.c
new file mode 100644
index 3f20f1d..4bcad29
--- a/src/backend/executor/execExprInterp.c
+++ b/src/backend/executor/execExprInterp.c
@@ -485,6 +485,7 @@ ExecInterpExpr(ExprState *state, ExprCon
 		&&CASE_EEOP_AGGREF,
 		&&CASE_EEOP_GROUPING_FUNC,
 		&&CASE_EEOP_WINDOW_FUNC,
+		&&CASE_EEOP_MERGING_FUNC,
 		&&CASE_EEOP_SUBPLAN,
 		&&CASE_EEOP_AGG_STRICT_DESERIALIZE,
 		&&CASE_EEOP_AGG_DESERIALIZE,
@@ -1593,6 +1594,14 @@ ExecInterpExpr(ExprState *state, ExprCon
 			EEO_NEXT();
 		}
 
+		EEO_CASE(EEOP_MERGING_FUNC)
+		{
+			/* too complex/uncommon for an inline implementation */
+			ExecEvalMergingFunc(state, op, econtext);
+
+			EEO_NEXT();
+		}
+
 		EEO_CASE(EEOP_SUBPLAN)
 		{
 			/* too complex for an inline implementation */
@@ -4247,6 +4256,59 @@ ExecEvalGroupingFunc(ExprState *state, E
 }
 
 /*
+ * ExecEvalMergingFunc
+ *
+ * Returns information about the current MERGE action for its RETURNING list.
+ */
+void
+ExecEvalMergingFunc(ExprState *state, ExprEvalStep *op, ExprContext *econtext)
+{
+	MergingFuncOp mfop = op->d.merging_func.mfop;
+	ModifyTableState *mtstate = castNode(ModifyTableState, state->parent);
+	MergeActionState *relaction = mtstate->mt_merge_action;
+
+	if (!relaction)
+		elog(ERROR, "no merge action in progress");
+
+	switch (mfop)
+	{
+		case MERGING_ACTION:
+			/* Return the MERGE action ("INSERT", "UPDATE", or "DELETE") */
+			switch (relaction->mas_action->commandType)
+			{
+				case CMD_INSERT:
+					*op->resvalue = PointerGetDatum(cstring_to_text_with_len("INSERT", 6));
+					*op->resnull = false;
+					break;
+				case CMD_UPDATE:
+					*op->resvalue = PointerGetDatum(cstring_to_text_with_len("UPDATE", 6));
+					*op->resnull = false;
+					break;
+				case CMD_DELETE:
+					*op->resvalue = PointerGetDatum(cstring_to_text_with_len("DELETE", 6));
+					*op->resnull = false;
+					break;
+				case CMD_NOTHING:
+					elog(ERROR, "unexpected merge action: DO NOTHING");
+					break;
+				default:
+					elog(ERROR, "unrecognized commandType: %d",
+						 (int) relaction->mas_action->commandType);
+			}
+			break;
+
+		case MERGING_CLAUSE_NUMBER:
+			/* Return the 1-based index of the MERGE action */
+			*op->resvalue = Int32GetDatum(relaction->mas_action->index);
+			*op->resnull = false;
+			break;
+
+		default:
+			elog(ERROR, "unrecognized MergingFuncOp: %d", (int) mfop);
+	}
+}
+
+/*
  * Hand off evaluation of a subplan to nodeSubplan.c
  */
 void
diff --git a/src/backend/executor/execPartition.c b/src/backend/executor/execPartition.c
new file mode 100644
index 429f0a8..813f76a
--- a/src/backend/executor/execPartition.c
+++ b/src/backend/executor/execPartition.c
@@ -612,8 +612,8 @@ ExecInitPartitionInfo(ModifyTableState *
 	 * Build the RETURNING projection for the partition.  Note that we didn't
 	 * build the returningList for partitions within the planner, but simple
 	 * translation of varattnos will suffice.  This only occurs for the INSERT
-	 * case or in the case of UPDATE tuple routing where we didn't find a
-	 * result rel to reuse.
+	 * case or in the case of UPDATE/MERGE tuple routing where we didn't find
+	 * a result rel to reuse.
 	 */
 	if (node && node->returningLists != NIL)
 	{
@@ -622,12 +622,14 @@ ExecInitPartitionInfo(ModifyTableState *
 		List	   *returningList;
 
 		/* See the comment above for WCO lists. */
-		/* (except no RETURNING support for MERGE yet) */
 		Assert((node->operation == CMD_INSERT &&
 				list_length(node->returningLists) == 1 &&
 				list_length(node->resultRelations) == 1) ||
 			   (node->operation == CMD_UPDATE &&
 				list_length(node->returningLists) ==
+				list_length(node->resultRelations)) ||
+			   (node->operation == CMD_MERGE &&
+				list_length(node->returningLists) ==
 				list_length(node->resultRelations)));
 
 		/*
diff --git a/src/backend/executor/functions.c b/src/backend/executor/functions.c
new file mode 100644
index 0f811fd..7da02b0
--- a/src/backend/executor/functions.c
+++ b/src/backend/executor/functions.c
@@ -1660,8 +1660,8 @@ check_sql_fn_retval(List *queryTreeLists
 
 	/*
 	 * If it's a plain SELECT, it returns whatever the targetlist says.
-	 * Otherwise, if it's INSERT/UPDATE/DELETE with RETURNING, it returns
-	 * that. Otherwise, the function return type must be VOID.
+	 * Otherwise, if it's INSERT/UPDATE/DELETE/MERGE with RETURNING, it
+	 * returns that. Otherwise, the function return type must be VOID.
 	 *
 	 * Note: eventually replace this test with QueryReturnsTuples?	We'd need
 	 * a more general method of determining the output type, though.  Also, it
@@ -1679,7 +1679,8 @@ check_sql_fn_retval(List *queryTreeLists
 	else if (parse &&
 			 (parse->commandType == CMD_INSERT ||
 			  parse->commandType == CMD_UPDATE ||
-			  parse->commandType == CMD_DELETE) &&
+			  parse->commandType == CMD_DELETE ||
+			  parse->commandType == CMD_MERGE) &&
 			 parse->returningList)
 	{
 		tlist = parse->returningList;
@@ -1693,7 +1694,7 @@ check_sql_fn_retval(List *queryTreeLists
 				(errcode(ERRCODE_INVALID_FUNCTION_DEFINITION),
 				 errmsg("return type mismatch in function declared to return %s",
 						format_type_be(rettype)),
-				 errdetail("Function's final statement must be SELECT or INSERT/UPDATE/DELETE RETURNING.")));
+				 errdetail("Function's final statement must be SELECT or INSERT/UPDATE/DELETE/MERGE RETURNING.")));
 		return false;			/* keep compiler quiet */
 	}
 
diff --git a/src/backend/executor/nodeModifyTable.c b/src/backend/executor/nodeModifyTable.c
new file mode 100644
index ff7ec84..e36c66b
--- a/src/backend/executor/nodeModifyTable.c
+++ b/src/backend/executor/nodeModifyTable.c
@@ -36,8 +36,7 @@
  *		RETURNING tuple after completing each row insert, update, or delete.
  *		It must be called again to continue the operation.  Without RETURNING,
  *		we just loop within the node until all the work is done, then
- *		return NULL.  This avoids useless call/return overhead.  (MERGE does
- *		not support RETURNING.)
+ *		return NULL.  This avoids useless call/return overhead.
  */
 
 #include "postgres.h"
@@ -88,9 +87,6 @@ typedef struct ModifyTableContext
 	 */
 	TupleTableSlot *planSlot;
 
-	/* MERGE specific */
-	MergeActionState *relaction;	/* MERGE action in progress */
-
 	/*
 	 * Information about the changes that were made concurrently to a tuple
 	 * being updated or deleted
@@ -153,14 +149,15 @@ static TupleTableSlot *ExecMerge(ModifyT
 								 HeapTuple oldtuple,
 								 bool canSetTag);
 static void ExecInitMerge(ModifyTableState *mtstate, EState *estate);
-static bool ExecMergeMatched(ModifyTableContext *context,
-							 ResultRelInfo *resultRelInfo,
-							 ItemPointer tupleid,
-							 HeapTuple oldtuple,
-							 bool canSetTag);
-static void ExecMergeNotMatched(ModifyTableContext *context,
-								ResultRelInfo *resultRelInfo,
-								bool canSetTag);
+static TupleTableSlot *ExecMergeMatched(ModifyTableContext *context,
+										ResultRelInfo *resultRelInfo,
+										ItemPointer tupleid,
+										HeapTuple oldtuple,
+										bool canSetTag,
+										bool *matched);
+static TupleTableSlot *ExecMergeNotMatched(ModifyTableContext *context,
+										   ResultRelInfo *resultRelInfo,
+										   bool canSetTag);
 
 
 /*
@@ -980,7 +977,7 @@ ExecInsert(ModifyTableContext *context,
 		if (mtstate->operation == CMD_UPDATE)
 			wco_kind = WCO_RLS_UPDATE_CHECK;
 		else if (mtstate->operation == CMD_MERGE)
-			wco_kind = (context->relaction->mas_action->commandType == CMD_UPDATE) ?
+			wco_kind = (mtstate->mt_merge_action->mas_action->commandType == CMD_UPDATE) ?
 				WCO_RLS_UPDATE_CHECK : WCO_RLS_INSERT_CHECK;
 		else
 			wco_kind = WCO_RLS_INSERT_CHECK;
@@ -1834,7 +1831,7 @@ ExecCrossPartitionUpdate(ModifyTableCont
 		 * additional rechecking, and might end up executing a different
 		 * action entirely).
 		 */
-		if (context->relaction != NULL)
+		if (mtstate->operation == CMD_MERGE)
 			return *tmresult == TM_Ok;
 		else if (TupIsNull(epqslot))
 			return true;
@@ -2075,7 +2072,7 @@ lreplace:
 		 * No luck, a retry is needed.  If running MERGE, we do not do so
 		 * here; instead let it handle that on its own rules.
 		 */
-		if (context->relaction != NULL)
+		if (context->mtstate->operation == CMD_MERGE)
 			return result;
 
 		/*
@@ -2716,6 +2713,7 @@ static TupleTableSlot *
 ExecMerge(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 		  ItemPointer tupleid, HeapTuple oldtuple, bool canSetTag)
 {
+	TupleTableSlot *rslot = NULL;
 	bool		matched;
 
 	/*-----
@@ -2764,19 +2762,18 @@ ExecMerge(ModifyTableContext *context, R
 	 */
 	matched = tupleid != NULL || oldtuple != NULL;
 	if (matched)
-		matched = ExecMergeMatched(context, resultRelInfo, tupleid, oldtuple,
-								   canSetTag);
+		rslot = ExecMergeMatched(context, resultRelInfo, tupleid, oldtuple,
+								 canSetTag, &matched);
 
 	/*
-	 * Either we were dealing with a NOT MATCHED tuple or ExecMergeMatched()
-	 * returned "false", indicating the previously MATCHED tuple no longer
-	 * matches.
+	 * Deal with the NOT MATCHED case (either a NOT MATCHED tuple from the
+	 * join, or a previously MATCHED tuple for which ExecMergeMatched() set
+	 * "matched" to false, indicating that it no longer matches).
 	 */
 	if (!matched)
-		ExecMergeNotMatched(context, resultRelInfo, canSetTag);
+		rslot = ExecMergeNotMatched(context, resultRelInfo, canSetTag);
 
-	/* No RETURNING support yet */
-	return NULL;
+	return rslot;
 }
 
 /*
@@ -2788,8 +2785,8 @@ ExecMerge(ModifyTableContext *context, R
  * We start from the first WHEN MATCHED action and check if the WHEN quals
  * pass, if any. If the WHEN quals for the first action do not pass, we
  * check the second, then the third and so on. If we reach to the end, no
- * action is taken and we return true, indicating that no further action is
- * required for this tuple.
+ * action is taken and "matched" is set to true, indicating that no further
+ * action is required for this tuple.
  *
  * If we do find a qualifying action, then we attempt to execute the action.
  *
@@ -2798,16 +2795,18 @@ ExecMerge(ModifyTableContext *context, R
  * with individual actions are evaluated by this routine via ExecQual, while
  * EvalPlanQual checks for the join quals. If EvalPlanQual tells us that the
  * updated tuple still passes the join quals, then we restart from the first
- * action to look for a qualifying action. Otherwise, we return false --
- * meaning that a NOT MATCHED action must now be executed for the current
- * source tuple.
+ * action to look for a qualifying action. Otherwise, "matched" is set to
+ * false -- meaning that a NOT MATCHED action must now be executed for the
+ * current source tuple.
  */
-static bool
+static TupleTableSlot *
 ExecMergeMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
-				 ItemPointer tupleid, HeapTuple oldtuple, bool canSetTag)
+				 ItemPointer tupleid, HeapTuple oldtuple, bool canSetTag,
+				 bool *matched)
 {
 	ModifyTableState *mtstate = context->mtstate;
-	TupleTableSlot *newslot;
+	TupleTableSlot *rslot = NULL;
+	TupleTableSlot *newslot = NULL;
 	EState	   *estate = context->estate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
 	bool		isNull;
@@ -2818,7 +2817,10 @@ ExecMergeMatched(ModifyTableContext *con
 	 * If there are no WHEN MATCHED actions, we are done.
 	 */
 	if (resultRelInfo->ri_matchedMergeAction == NIL)
-		return true;
+	{
+		*matched = true;
+		return NULL;
+	}
 
 	/*
 	 * Make tuple and any needed join variables available to ExecQual and
@@ -2908,12 +2910,15 @@ lmerge_matched:
 				 */
 				newslot = ExecProject(relaction->mas_proj);
 
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				if (!ExecUpdatePrologue(context, resultRelInfo,
 										tupleid, NULL, newslot, &result))
 				{
 					if (result == TM_Ok)
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 					break;		/* concurrent update/delete */
 				}
 
@@ -2923,7 +2928,10 @@ lmerge_matched:
 				{
 					if (!ExecIRUpdateTriggers(estate, resultRelInfo,
 											  oldtuple, newslot))
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 				}
 				else
 				{
@@ -2936,12 +2944,15 @@ lmerge_matched:
 					 * cross-partition update was done, then there's nothing
 					 * else for us to do --- the UPDATE has been turned into a
 					 * DELETE and an INSERT, and we must not perform any of
-					 * the usual post-update tasks.
+					 * the usual post-update tasks.  Also, the RETURNING tuple
+					 * (if any) has been projected, so we can just return
+					 * that.
 					 */
 					if (updateCxt.crossPartUpdate)
 					{
 						mtstate->mt_merge_updated += 1;
-						return true;
+						*matched = true;
+						return context->cpUpdateReturningSlot;
 					}
 				}
 
@@ -2954,12 +2965,15 @@ lmerge_matched:
 				break;
 
 			case CMD_DELETE:
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				if (!ExecDeletePrologue(context, resultRelInfo, tupleid,
 										NULL, NULL, &result))
 				{
 					if (result == TM_Ok)
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 					break;		/* concurrent update/delete */
 				}
 
@@ -2969,7 +2983,10 @@ lmerge_matched:
 				{
 					if (!ExecIRDeleteTriggers(estate, resultRelInfo,
 											  oldtuple))
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 				}
 				else
 					result = ExecDeleteAct(context, resultRelInfo, tupleid,
@@ -3027,7 +3044,8 @@ lmerge_matched:
 				 * If the tuple was already deleted, return to let caller
 				 * handle it under NOT MATCHED clauses.
 				 */
-				return false;
+				*matched = false;
+				return NULL;
 
 			case TM_Updated:
 				{
@@ -3073,13 +3091,19 @@ lmerge_matched:
 							 * NOT MATCHED actions.
 							 */
 							if (TupIsNull(epqslot))
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							(void) ExecGetJunkAttribute(epqslot,
 														resultRelInfo->ri_RowIdAttNo,
 														&isNull);
 							if (isNull)
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							/*
 							 * When a tuple was updated and migrated to
@@ -3114,7 +3138,8 @@ lmerge_matched:
 							 * tuple already deleted; tell caller to run NOT
 							 * MATCHED actions
 							 */
-							return false;
+							*matched = false;
+							return NULL;
 
 						case TM_SelfModified:
 
@@ -3134,13 +3159,14 @@ lmerge_matched:
 										(errcode(ERRCODE_TRIGGERED_DATA_CHANGE_VIOLATION),
 										 errmsg("tuple to be updated or deleted was already modified by an operation triggered by the current command"),
 										 errhint("Consider using an AFTER trigger instead of a BEFORE trigger to propagate changes to other rows.")));
-							return false;
+							*matched = false;
+							return NULL;
 
 						default:
 							/* see table_tuple_lock call in ExecDelete() */
 							elog(ERROR, "unexpected table_tuple_lock status: %u",
 								 result);
-							return false;
+							return NULL;
 					}
 				}
 
@@ -3152,6 +3178,31 @@ lmerge_matched:
 				break;
 		}
 
+		/* Process RETURNING if present */
+		if (resultRelInfo->ri_projectReturning)
+		{
+			switch (commandType)
+			{
+				case CMD_UPDATE:
+					rslot = ExecProcessReturning(resultRelInfo, newslot,
+												 context->planSlot);
+					break;
+
+				case CMD_DELETE:
+					rslot = ExecProcessReturning(resultRelInfo,
+												 resultRelInfo->ri_oldTupleSlot,
+												 context->planSlot);
+					break;
+
+				case CMD_NOTHING:
+					break;
+
+				default:
+					elog(ERROR, "unrecognized commandType: %d",
+						 (int) commandType);
+			}
+		}
+
 		/*
 		 * We've activated one of the WHEN clauses, so we don't search
 		 * further. This is required behaviour, not an optimization.
@@ -3162,19 +3213,22 @@ lmerge_matched:
 	/*
 	 * Successfully executed an action or no qualifying action was found.
 	 */
-	return true;
+	*matched = true;
+
+	return rslot;
 }
 
 /*
  * Execute the first qualifying NOT MATCHED action.
  */
-static void
+static TupleTableSlot *
 ExecMergeNotMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 					bool canSetTag)
 {
 	ModifyTableState *mtstate = context->mtstate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
 	List	   *actionStates = NIL;
+	TupleTableSlot *rslot = NULL;
 	ListCell   *l;
 
 	/*
@@ -3224,10 +3278,10 @@ ExecMergeNotMatched(ModifyTableContext *
 				 * so we don't need to map the tuple here.
 				 */
 				newslot = ExecProject(action->mas_proj);
-				context->relaction = action;
+				mtstate->mt_merge_action = action;
 
-				(void) ExecInsert(context, mtstate->rootResultRelInfo, newslot,
-								  canSetTag, NULL, NULL);
+				rslot = ExecInsert(context, mtstate->rootResultRelInfo,
+								   newslot, canSetTag, NULL, NULL);
 				mtstate->mt_merge_inserted += 1;
 				break;
 			case CMD_NOTHING:
@@ -3243,6 +3297,8 @@ ExecMergeNotMatched(ModifyTableContext *
 		 */
 		break;
 	}
+
+	return rslot;
 }
 
 /*
@@ -3705,9 +3761,17 @@ ExecModifyTable(PlanState *pstate)
 				{
 					EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-					ExecMerge(&context, node->resultRelInfo, NULL, NULL,
-							  node->canSetTag);
-					continue;	/* no RETURNING support yet */
+					slot = ExecMerge(&context, node->resultRelInfo,
+									 NULL, NULL, node->canSetTag);
+
+					/*
+					 * If we got a RETURNING result, return it to the caller.
+					 * We'll continue the work on next call.
+					 */
+					if (slot)
+						return slot;
+
+					continue;	/* continue with the next tuple */
 				}
 
 				elog(ERROR, "tableoid is NULL");
@@ -3784,9 +3848,17 @@ ExecModifyTable(PlanState *pstate)
 					{
 						EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-						ExecMerge(&context, node->resultRelInfo, NULL, NULL,
-								  node->canSetTag);
-						continue;	/* no RETURNING support yet */
+						slot = ExecMerge(&context, node->resultRelInfo,
+										 NULL, NULL, node->canSetTag);
+
+						/*
+						 * If we got a RETURNING result, return it to the
+						 * caller.  We'll continue the work on next call.
+						 */
+						if (slot)
+							return slot;
+
+						continue;	/* continue with the next tuple */
 					}
 
 					elog(ERROR, "ctid is NULL");
@@ -3833,9 +3905,17 @@ ExecModifyTable(PlanState *pstate)
 					{
 						EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-						ExecMerge(&context, node->resultRelInfo, NULL, NULL,
-								  node->canSetTag);
-						continue;	/* no RETURNING support yet */
+						slot = ExecMerge(&context, node->resultRelInfo,
+										 NULL, NULL, node->canSetTag);
+
+						/*
+						 * If we got a RETURNING result, return it to the
+						 * caller.  We'll continue the work on next call.
+						 */
+						if (slot)
+							return slot;
+
+						continue;	/* continue with the next tuple */
 					}
 
 					elog(ERROR, "wholerow is NULL");
@@ -3897,7 +3977,6 @@ ExecModifyTable(PlanState *pstate)
 				}
 				slot = ExecGetUpdateNewTuple(resultRelInfo, context.planSlot,
 											 oldSlot);
-				context.relaction = NULL;
 
 				/* Now apply the update. */
 				slot = ExecUpdate(&context, resultRelInfo, tupleid, oldtuple,
diff --git a/src/backend/executor/spi.c b/src/backend/executor/spi.c
new file mode 100644
index d71484c..abb8b0d
--- a/src/backend/executor/spi.c
+++ b/src/backend/executor/spi.c
@@ -2033,6 +2033,8 @@ SPI_result_code_string(int code)
 			return "SPI_OK_TD_REGISTER";
 		case SPI_OK_MERGE:
 			return "SPI_OK_MERGE";
+		case SPI_OK_MERGE_RETURNING:
+			return "SPI_OK_MERGE_RETURNING";
 	}
 	/* Unrecognized code ... return something useful ... */
 	sprintf(buf, "Unrecognized SPI code %d", code);
@@ -2886,7 +2888,10 @@ _SPI_pquery(QueryDesc *queryDesc, bool f
 				res = SPI_OK_UPDATE;
 			break;
 		case CMD_MERGE:
-			res = SPI_OK_MERGE;
+			if (queryDesc->plannedstmt->hasReturning)
+				res = SPI_OK_MERGE_RETURNING;
+			else
+				res = SPI_OK_MERGE;
 			break;
 		default:
 			return SPI_ERROR_OPUNKNOWN;
diff --git a/src/backend/jit/llvm/llvmjit_expr.c b/src/backend/jit/llvm/llvmjit_expr.c
new file mode 100644
index 0c44842..a5cd6fb
--- a/src/backend/jit/llvm/llvmjit_expr.c
+++ b/src/backend/jit/llvm/llvmjit_expr.c
@@ -1986,6 +1986,12 @@ llvm_compile_expr(ExprState *state)
 					break;
 				}
 
+			case EEOP_MERGING_FUNC:
+				build_EvalXFunc(b, mod, "ExecEvalMergingFunc",
+								v_state, op, v_econtext);
+				LLVMBuildBr(b, opblocks[opno + 1]);
+				break;
+
 			case EEOP_SUBPLAN:
 				build_EvalXFunc(b, mod, "ExecEvalSubPlan",
 								v_state, op, v_econtext);
diff --git a/src/backend/jit/llvm/llvmjit_types.c b/src/backend/jit/llvm/llvmjit_types.c
new file mode 100644
index 47c9daf..912d9ad
--- a/src/backend/jit/llvm/llvmjit_types.c
+++ b/src/backend/jit/llvm/llvmjit_types.c
@@ -155,6 +155,7 @@ void	   *referenced_functions[] =
 	ExecEvalFuncExprFusage,
 	ExecEvalFuncExprStrictFusage,
 	ExecEvalGroupingFunc,
+	ExecEvalMergingFunc,
 	ExecEvalMinMax,
 	ExecEvalNextValueExpr,
 	ExecEvalParamExec,
diff --git a/src/backend/nodes/nodeFuncs.c b/src/backend/nodes/nodeFuncs.c
new file mode 100644
index e1a5bc7..2468063
--- a/src/backend/nodes/nodeFuncs.c
+++ b/src/backend/nodes/nodeFuncs.c
@@ -67,6 +67,9 @@ exprType(const Node *expr)
 		case T_WindowFunc:
 			type = ((const WindowFunc *) expr)->wintype;
 			break;
+		case T_MergingFunc:
+			type = ((const MergingFunc *) expr)->mftype;
+			break;
 		case T_SubscriptingRef:
 			type = ((const SubscriptingRef *) expr)->refrestype;
 			break;
@@ -810,6 +813,9 @@ exprCollation(const Node *expr)
 		case T_WindowFunc:
 			coll = ((const WindowFunc *) expr)->wincollid;
 			break;
+		case T_MergingFunc:
+			coll = ((const MergingFunc *) expr)->mfcollid;
+			break;
 		case T_SubscriptingRef:
 			coll = ((const SubscriptingRef *) expr)->refcollid;
 			break;
@@ -1085,6 +1091,9 @@ exprSetCollation(Node *expr, Oid collati
 		case T_WindowFunc:
 			((WindowFunc *) expr)->wincollid = collation;
 			break;
+		case T_MergingFunc:
+			((MergingFunc *) expr)->mfcollid = collation;
+			break;
 		case T_SubscriptingRef:
 			((SubscriptingRef *) expr)->refcollid = collation;
 			break;
@@ -1343,6 +1352,9 @@ exprLocation(const Node *expr)
 			/* function name should always be the first thing */
 			loc = ((const WindowFunc *) expr)->location;
 			break;
+		case T_MergingFunc:
+			loc = ((const MergingFunc *) expr)->location;
+			break;
 		case T_SubscriptingRef:
 			/* just use container argument's location */
 			loc = exprLocation((Node *) ((const SubscriptingRef *) expr)->refexpr);
@@ -2035,6 +2047,7 @@ expression_tree_walker_impl(Node *node,
 		case T_RangeTblRef:
 		case T_SortGroupClause:
 		case T_CTESearchClause:
+		case T_MergingFunc:
 			/* primitive node types with no expression subnodes */
 			break;
 		case T_WithCheckOption:
@@ -2869,6 +2882,7 @@ expression_tree_mutator_impl(Node *node,
 		case T_RangeTblRef:
 		case T_SortGroupClause:
 		case T_CTESearchClause:
+		case T_MergingFunc:
 			return (Node *) copyObject(node);
 		case T_WithCheckOption:
 			{
@@ -3833,6 +3847,7 @@ raw_expression_tree_walker_impl(Node *no
 		case T_ParamRef:
 		case T_A_Const:
 		case T_A_Star:
+		case T_MergingFunc:
 			/* primitive node types with no subnodes */
 			break;
 		case T_Alias:
@@ -4053,6 +4068,8 @@ raw_expression_tree_walker_impl(Node *no
 					return true;
 				if (WALK(stmt->mergeWhenClauses))
 					return true;
+				if (WALK(stmt->returningList))
+					return true;
 				if (WALK(stmt->withClause))
 					return true;
 			}
diff --git a/src/backend/optimizer/plan/subselect.c b/src/backend/optimizer/plan/subselect.c
new file mode 100644
index 47e1472..b6a83c2
--- a/src/backend/optimizer/plan/subselect.c
+++ b/src/backend/optimizer/plan/subselect.c
@@ -1855,7 +1855,8 @@ convert_EXISTS_to_ANY(PlannerInfo *root,
 /*
  * Replace correlation vars (uplevel vars) with Params.
  *
- * Uplevel PlaceHolderVars and aggregates are replaced, too.
+ * Uplevel PlaceHolderVars, aggregates, GROUPING() expressions, and MERGING()
+ * expressions are replaced, too.
  *
  * Note: it is critical that this runs immediately after SS_process_sublinks.
  * Since we do not recurse into the arguments of uplevel PHVs and aggregates,
@@ -1909,6 +1910,11 @@ replace_correlation_vars_mutator(Node *n
 		if (((GroupingFunc *) node)->agglevelsup > 0)
 			return (Node *) replace_outer_grouping(root, (GroupingFunc *) node);
 	}
+	if (IsA(node, MergingFunc))
+	{
+		if (root->parse->commandType != CMD_MERGE)
+			return (Node *) replace_outer_merging(root, (MergingFunc *) node);
+	}
 	return expression_tree_mutator(node,
 								   replace_correlation_vars_mutator,
 								   (void *) root);
diff --git a/src/backend/optimizer/util/paramassign.c b/src/backend/optimizer/util/paramassign.c
new file mode 100644
index a58da7c..6ea6db2
--- a/src/backend/optimizer/util/paramassign.c
+++ b/src/backend/optimizer/util/paramassign.c
@@ -308,6 +308,56 @@ replace_outer_grouping(PlannerInfo *root
 }
 
 /*
+ * Generate a Param node to replace the given MergingFunc expression which is
+ * expected to be in the RETURNING list of an upper-level MERGE query.  Record
+ * the need for the MergingFunc in the proper upper-level root->plan_params.
+ */
+Param *
+replace_outer_merging(PlannerInfo *root, MergingFunc *mf)
+{
+	Param	   *retval;
+	PlannerParamItem *pitem;
+	Oid			ptype = exprType((Node *) mf);
+
+	Assert(root->parse->commandType != CMD_MERGE);
+
+	/*
+	 * The parser should have ensured that the MergingFunc is in the RETURNING
+	 * list of an upper-level MERGE query, so find that query.
+	 */
+	do
+	{
+		root = root->parent_root;
+		if (root == NULL)
+			elog(ERROR, "MergingFunc found outside MERGE");
+	} while (root->parse->commandType != CMD_MERGE);
+
+	/*
+	 * It does not seem worthwhile to try to de-duplicate references to outer
+	 * MergingFunc expressions.  Just make a new slot every time.
+	 */
+	mf = copyObject(mf);
+
+	pitem = makeNode(PlannerParamItem);
+	pitem->item = (Node *) mf;
+	pitem->paramId = list_length(root->glob->paramExecTypes);
+	root->glob->paramExecTypes = lappend_oid(root->glob->paramExecTypes,
+											 ptype);
+
+	root->plan_params = lappend(root->plan_params, pitem);
+
+	retval = makeNode(Param);
+	retval->paramkind = PARAM_EXEC;
+	retval->paramid = pitem->paramId;
+	retval->paramtype = ptype;
+	retval->paramtypmod = -1;
+	retval->paramcollid = InvalidOid;
+	retval->location = mf->location;
+
+	return retval;
+}
+
+/*
  * Generate a Param node to replace the given Var,
  * which is expected to come from some upper NestLoop plan node.
  * Record the need for the Var in root->curOuterParams.
diff --git a/src/backend/parser/analyze.c b/src/backend/parser/analyze.c
new file mode 100644
index dbdf6bf..44893d8
--- a/src/backend/parser/analyze.c
+++ b/src/backend/parser/analyze.c
@@ -74,7 +74,6 @@ static void determineRecursiveColTypes(P
 									   Node *larg, List *nrtargetlist);
 static Query *transformReturnStmt(ParseState *pstate, ReturnStmt *stmt);
 static Query *transformUpdateStmt(ParseState *pstate, UpdateStmt *stmt);
-static List *transformReturningList(ParseState *pstate, List *returningList);
 static Query *transformPLAssignStmt(ParseState *pstate,
 									PLAssignStmt *stmt);
 static Query *transformDeclareCursorStmt(ParseState *pstate,
@@ -553,7 +552,8 @@ transformDeleteStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -980,7 +980,8 @@ transformInsertStmt(ParseState *pstate,
 	/* Process RETURNING, if any. */
 	if (stmt->returningList)
 		qry->returningList = transformReturningList(pstate,
-													stmt->returningList);
+													stmt->returningList,
+													EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -2445,7 +2446,8 @@ transformUpdateStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/*
 	 * Now we are done with SELECT-like processing, and can get on with
@@ -2542,10 +2544,11 @@ transformUpdateTargetList(ParseState *ps
 
 /*
  * transformReturningList -
- *	handle a RETURNING clause in INSERT/UPDATE/DELETE
+ *	handle a RETURNING clause in INSERT/UPDATE/DELETE/MERGE
  */
-static List *
-transformReturningList(ParseState *pstate, List *returningList)
+List *
+transformReturningList(ParseState *pstate, List *returningList,
+					   ParseExprKind exprKind)
 {
 	List	   *rlist;
 	int			save_next_resno;
@@ -2562,7 +2565,7 @@ transformReturningList(ParseState *pstat
 	pstate->p_next_resno = 1;
 
 	/* transform RETURNING identically to a SELECT targetlist */
-	rlist = transformTargetList(pstate, returningList, EXPR_KIND_RETURNING);
+	rlist = transformTargetList(pstate, returningList, exprKind);
 
 	/*
 	 * Complain if the nonempty tlist expanded to nothing (which is possible
diff --git a/src/backend/parser/gram.y b/src/backend/parser/gram.y
new file mode 100644
index 130f7fc..04fee0d
--- a/src/backend/parser/gram.y
+++ b/src/backend/parser/gram.y
@@ -694,7 +694,7 @@ static Node *makeRecursiveViewSelect(cha
 	BOOLEAN_P BOTH BREADTH BY
 
 	CACHE CALL CALLED CASCADE CASCADED CASE CAST CATALOG_P CHAIN CHAR_P
-	CHARACTER CHARACTERISTICS CHECK CHECKPOINT CLASS CLOSE
+	CHARACTER CHARACTERISTICS CHECK CHECKPOINT CLASS CLAUSE_NUMBER CLOSE
 	CLUSTER COALESCE COLLATE COLLATION COLUMN COLUMNS COMMENT COMMENTS COMMIT
 	COMMITTED COMPRESSION CONCURRENTLY CONFIGURATION CONFLICT
 	CONNECTION CONSTRAINT CONSTRAINTS CONTENT_P CONTINUE_P CONVERSION_P COPY
@@ -732,7 +732,7 @@ static Node *makeRecursiveViewSelect(cha
 	LEADING LEAKPROOF LEAST LEFT LEVEL LIKE LIMIT LISTEN LOAD LOCAL
 	LOCALTIME LOCALTIMESTAMP LOCATION LOCK_P LOCKED LOGGED
 
-	MAPPING MATCH MATCHED MATERIALIZED MAXVALUE MERGE METHOD
+	MAPPING MATCH MATCHED MATERIALIZED MAXVALUE MERGE MERGING METHOD
 	MINUTE_P MINVALUE MODE MONTH_P MOVE
 
 	NAME_P NAMES NATIONAL NATURAL NCHAR NEW NEXT NFC NFD NFKC NFKD NO NONE
@@ -12368,6 +12368,7 @@ MergeStmt:
 			USING table_ref
 			ON a_expr
 			merge_when_list
+			returning_clause
 				{
 					MergeStmt  *m = makeNode(MergeStmt);
 
@@ -12376,6 +12377,7 @@ MergeStmt:
 					m->sourceRelation = $6;
 					m->joinCondition = $8;
 					m->mergeWhenClauses = $9;
+					m->returningList = $10;
 
 					$$ = (Node *) m;
 				}
@@ -15789,6 +15791,24 @@ func_expr_common_subexpr:
 					n->location = @1;
 					$$ = (Node *) n;
 				}
+			| MERGING '(' ACTION ')'
+				{
+					MergingFunc *m = makeNode(MergingFunc);
+
+					m->mfop = MERGING_ACTION;
+					m->mftype = TEXTOID;
+					m->location = @1;
+					$$ = (Node *) m;
+				}
+			| MERGING '(' CLAUSE_NUMBER ')'
+				{
+					MergingFunc *m = makeNode(MergingFunc);
+
+					m->mfop = MERGING_CLAUSE_NUMBER;
+					m->mftype = INT4OID;
+					m->location = @1;
+					$$ = (Node *) m;
+				}
 			;
 
 
@@ -17167,6 +17187,7 @@ unreserved_keyword:
 			| CHARACTERISTICS
 			| CHECKPOINT
 			| CLASS
+			| CLAUSE_NUMBER
 			| CLOSE
 			| CLUSTER
 			| COLUMNS
@@ -17486,6 +17507,7 @@ col_name_keyword:
 			| JSON_SCALAR
 			| JSON_SERIALIZE
 			| LEAST
+			| MERGING
 			| NATIONAL
 			| NCHAR
 			| NONE
@@ -17704,6 +17726,7 @@ bare_label_keyword:
 			| CHECK
 			| CHECKPOINT
 			| CLASS
+			| CLAUSE_NUMBER
 			| CLOSE
 			| CLUSTER
 			| COALESCE
@@ -17875,6 +17898,7 @@ bare_label_keyword:
 			| MATERIALIZED
 			| MAXVALUE
 			| MERGE
+			| MERGING
 			| METHOD
 			| MINVALUE
 			| MODE
diff --git a/src/backend/parser/parse_agg.c b/src/backend/parser/parse_agg.c
new file mode 100644
index 9d151a8..bee7d83
--- a/src/backend/parser/parse_agg.c
+++ b/src/backend/parser/parse_agg.c
@@ -468,6 +468,7 @@ check_agglevels_and_constraints(ParseSta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
@@ -915,6 +916,7 @@ transformWindowFuncCall(ParseState *psta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_cte.c b/src/backend/parser/parse_cte.c
new file mode 100644
index 3c88c9a..6826d4f
--- a/src/backend/parser/parse_cte.c
+++ b/src/backend/parser/parse_cte.c
@@ -126,13 +126,6 @@ transformWithClause(ParseState *pstate,
 		CommonTableExpr *cte = (CommonTableExpr *) lfirst(lc);
 		ListCell   *rest;
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(cte->ctequery, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in WITH query"),
-					parser_errposition(pstate, cte->location));
-
 		for_each_cell(rest, withClause->ctes, lnext(withClause->ctes, lc))
 		{
 			CommonTableExpr *cte2 = (CommonTableExpr *) lfirst(rest);
@@ -153,7 +146,8 @@ transformWithClause(ParseState *pstate,
 			/* must be a data-modifying statement */
 			Assert(IsA(cte->ctequery, InsertStmt) ||
 				   IsA(cte->ctequery, UpdateStmt) ||
-				   IsA(cte->ctequery, DeleteStmt));
+				   IsA(cte->ctequery, DeleteStmt) ||
+				   IsA(cte->ctequery, MergeStmt));
 
 			pstate->p_hasModifyingCTE = true;
 		}
diff --git a/src/backend/parser/parse_expr.c b/src/backend/parser/parse_expr.c
new file mode 100644
index 9300c7b..ab7e564
--- a/src/backend/parser/parse_expr.c
+++ b/src/backend/parser/parse_expr.c
@@ -54,6 +54,7 @@ static Node *transformAExprDistinct(Pars
 static Node *transformAExprNullIf(ParseState *pstate, A_Expr *a);
 static Node *transformAExprIn(ParseState *pstate, A_Expr *a);
 static Node *transformAExprBetween(ParseState *pstate, A_Expr *a);
+static Node *transformMergingFunc(ParseState *pstate, MergingFunc *f);
 static Node *transformBoolExpr(ParseState *pstate, BoolExpr *a);
 static Node *transformFuncCall(ParseState *pstate, FuncCall *fn);
 static Node *transformMultiAssignRef(ParseState *pstate, MultiAssignRef *maref);
@@ -227,6 +228,10 @@ transformExprRecurse(ParseState *pstate,
 			result = transformGroupingFunc(pstate, (GroupingFunc *) expr);
 			break;
 
+		case T_MergingFunc:
+			result = transformMergingFunc(pstate, (MergingFunc *) expr);
+			break;
+
 		case T_NamedArgExpr:
 			{
 				NamedArgExpr *na = (NamedArgExpr *) expr;
@@ -541,6 +546,7 @@ transformColumnRef(ParseState *pstate, C
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CHECK_CONSTRAINT:
@@ -1354,6 +1360,32 @@ transformAExprBetween(ParseState *pstate
 }
 
 static Node *
+transformMergingFunc(ParseState *pstate, MergingFunc *f)
+{
+	/*
+	 * All we need to do is check that we're in the RETURNING list of a MERGE
+	 * command.  If so, we just return the node as-is.
+	 */
+	if (pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+	{
+		ParseState *parent_pstate = pstate->parentParseState;
+
+		while (parent_pstate &&
+			   parent_pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+			parent_pstate = parent_pstate->parentParseState;
+
+		if (!parent_pstate ||
+			parent_pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+			ereport(ERROR,
+					errcode(ERRCODE_SYNTAX_ERROR),
+					errmsg("MERGING() can only be used in the RETURNING list of a MERGE command"),
+					parser_errposition(pstate, f->location));
+	}
+
+	return (Node *) f;
+}
+
+static Node *
 transformBoolExpr(ParseState *pstate, BoolExpr *a)
 {
 	List	   *args = NIL;
@@ -1767,6 +1799,7 @@ transformSubLink(ParseState *pstate, Sub
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CYCLE_MARK:
@@ -3115,6 +3148,7 @@ ParseExprKindName(ParseExprKind exprKind
 		case EXPR_KIND_OFFSET:
 			return "OFFSET";
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			return "RETURNING";
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
diff --git a/src/backend/parser/parse_func.c b/src/backend/parser/parse_func.c
new file mode 100644
index fdb3e6d..0cbc950
--- a/src/backend/parser/parse_func.c
+++ b/src/backend/parser/parse_func.c
@@ -2599,6 +2599,7 @@ check_srf_call_placement(ParseState *pst
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_merge.c b/src/backend/parser/parse_merge.c
new file mode 100644
index 4356d61..c568a59
--- a/src/backend/parser/parse_merge.c
+++ b/src/backend/parser/parse_merge.c
@@ -237,6 +237,10 @@ transformMergeStmt(ParseState *pstate, M
 	 */
 	qry->jointree = makeFromExpr(pstate->p_joinlist, joinExpr);
 
+	/* Transform the RETURNING list, if any */
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_MERGE_RETURNING);
+
 	/*
 	 * We now have a good query shape, so now look at the WHEN conditions and
 	 * action targetlists.
@@ -258,6 +262,7 @@ transformMergeStmt(ParseState *pstate, M
 		MergeAction *action;
 
 		action = makeNode(MergeAction);
+		action->index = foreach_current_index(l) + 1;
 		action->commandType = mergeWhenClause->commandType;
 		action->matched = mergeWhenClause->matched;
 
@@ -394,9 +399,6 @@ transformMergeStmt(ParseState *pstate, M
 
 	qry->mergeActionList = mergeActionList;
 
-	/* RETURNING could potentially be added in the future, but not in SQL std */
-	qry->returningList = NULL;
-
 	qry->hasTargetSRFs = false;
 	qry->hasSubLinks = pstate->p_hasSubLinks;
 
diff --git a/src/backend/parser/parse_relation.c b/src/backend/parser/parse_relation.c
new file mode 100644
index 34a0ec5..60b7e4f
--- a/src/backend/parser/parse_relation.c
+++ b/src/backend/parser/parse_relation.c
@@ -2346,9 +2346,10 @@ addRangeTableEntryForCTE(ParseState *pst
 		cte->cterefcount++;
 
 	/*
-	 * We throw error if the CTE is INSERT/UPDATE/DELETE without RETURNING.
-	 * This won't get checked in case of a self-reference, but that's OK
-	 * because data-modifying CTEs aren't allowed to be recursive anyhow.
+	 * We throw error if the CTE is INSERT/UPDATE/DELETE/MERGE without
+	 * RETURNING.  This won't get checked in case of a self-reference, but
+	 * that's OK because data-modifying CTEs aren't allowed to be recursive
+	 * anyhow.
 	 */
 	if (IsA(cte->ctequery, Query))
 	{
diff --git a/src/backend/parser/parse_target.c b/src/backend/parser/parse_target.c
new file mode 100644
index 0cd904f..eab78f4
--- a/src/backend/parser/parse_target.c
+++ b/src/backend/parser/parse_target.c
@@ -1807,6 +1807,20 @@ FigureColnameInternal(Node *node, char *
 			/* make GROUPING() act like a regular function */
 			*name = "grouping";
 			return 2;
+		case T_MergingFunc:
+			switch (((MergingFunc *) node)->mfop)
+			{
+				case MERGING_ACTION:
+					*name = "action";
+					return 2;
+				case MERGING_CLAUSE_NUMBER:
+					*name = "clause_number";
+					return 2;
+				default:
+					elog(ERROR, "unrecognized MergingFuncOp: %d",
+						 (int) ((MergingFunc *) node)->mfop);
+			}
+			break;
 		case T_SubLink:
 			switch (((SubLink *) node)->subLinkType)
 			{
diff --git a/src/backend/rewrite/rewriteHandler.c b/src/backend/rewrite/rewriteHandler.c
new file mode 100644
index b8839b5..b8e883f
--- a/src/backend/rewrite/rewriteHandler.c
+++ b/src/backend/rewrite/rewriteHandler.c
@@ -3834,9 +3834,9 @@ RewriteQuery(Query *parsetree, List *rew
 	ListCell   *lc1;
 
 	/*
-	 * First, recursively process any insert/update/delete statements in WITH
-	 * clauses.  (We have to do this first because the WITH clauses may get
-	 * copied into rule actions below.)
+	 * First, recursively process any insert/update/delete/merge statements in
+	 * WITH clauses.  (We have to do this first because the WITH clauses may
+	 * get copied into rule actions below.)
 	 */
 	foreach(lc1, parsetree->cteList)
 	{
@@ -3861,7 +3861,8 @@ RewriteQuery(Query *parsetree, List *rew
 			if (!(ctequery->commandType == CMD_SELECT ||
 				  ctequery->commandType == CMD_UPDATE ||
 				  ctequery->commandType == CMD_INSERT ||
-				  ctequery->commandType == CMD_DELETE))
+				  ctequery->commandType == CMD_DELETE ||
+				  ctequery->commandType == CMD_MERGE))
 			{
 				/*
 				 * Currently it could only be NOTIFY; this error message will
diff --git a/src/backend/rewrite/rowsecurity.c b/src/backend/rewrite/rowsecurity.c
new file mode 100644
index 94268a2..65ce579
--- a/src/backend/rewrite/rowsecurity.c
+++ b/src/backend/rewrite/rowsecurity.c
@@ -395,10 +395,10 @@ get_row_security_policies(Query *root, R
 	 * on the final action we take.
 	 *
 	 * We already fetched the SELECT policies above, to check existing rows,
-	 * but we must also check that new rows created by UPDATE actions are
-	 * visible, if SELECT rights are required for this relation. We don't do
-	 * this for INSERT actions, since an INSERT command would only do this
-	 * check if it had a RETURNING list, and MERGE does not support RETURNING.
+	 * but we must also check that new rows created by INSERT/UPDATE actions
+	 * are visible, if SELECT rights are required. For INSERT actions, we only
+	 * do this if RETURNING is specified, to be consistent with a plain INSERT
+	 * command, which can only require SELECT rights when RETURNING is used.
 	 *
 	 * We don't push the UPDATE/DELETE USING quals to the RTE because we don't
 	 * really want to apply them while scanning the relation since we don't
@@ -420,6 +420,8 @@ get_row_security_policies(Query *root, R
 		List	   *merge_delete_restrictive_policies;
 		List	   *merge_insert_permissive_policies;
 		List	   *merge_insert_restrictive_policies;
+		List	   *merge_select_permissive_policies = NIL;
+		List	   *merge_select_restrictive_policies = NIL;
 
 		/*
 		 * Fetch the UPDATE policies and set them up to execute on the
@@ -457,9 +459,6 @@ get_row_security_policies(Query *root, R
 		 */
 		if (perminfo->requiredPerms & ACL_SELECT)
 		{
-			List	   *merge_select_permissive_policies;
-			List	   *merge_select_restrictive_policies;
-
 			get_policies_for_relation(rel, CMD_SELECT, user_id,
 									  &merge_select_permissive_policies,
 									  &merge_select_restrictive_policies);
@@ -508,6 +507,21 @@ get_row_security_policies(Query *root, R
 							   withCheckOptions,
 							   hasSubLinks,
 							   false);
+
+		/*
+		 * Add ALL/SELECT policies as WCO_RLS_INSERT_CHECK WCOs, to ensure
+		 * that the inserted row is visible when executing an INSERT action,
+		 * if RETURNING is specified and SELECT rights are required for this
+		 * relation.
+		 */
+		if (perminfo->requiredPerms & ACL_SELECT && root->returningList)
+			add_with_check_options(rel, rt_index,
+								   WCO_RLS_INSERT_CHECK,
+								   merge_select_permissive_policies,
+								   merge_select_restrictive_policies,
+								   withCheckOptions,
+								   hasSubLinks,
+								   true);
 	}
 
 	table_close(rel, NoLock);
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
new file mode 100644
index 8de821f..f84a943
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -2145,11 +2145,10 @@ QueryReturnsTuples(Query *parsetree)
 		case CMD_SELECT:
 			/* returns tuples */
 			return true;
-		case CMD_MERGE:
-			return false;
 		case CMD_INSERT:
 		case CMD_UPDATE:
 		case CMD_DELETE:
+		case CMD_MERGE:
 			/* the forms with RETURNING return tuples */
 			if (parsetree->returningList)
 				return true;
diff --git a/src/backend/utils/adt/ruleutils.c b/src/backend/utils/adt/ruleutils.c
new file mode 100644
index a928a8c..ff999d3
--- a/src/backend/utils/adt/ruleutils.c
+++ b/src/backend/utils/adt/ruleutils.c
@@ -7197,8 +7197,13 @@ get_merge_query_def(Query *query, depars
 			appendStringInfoString(buf, "DO NOTHING");
 	}
 
-	/* No RETURNING support in MERGE yet */
-	Assert(query->returningList == NIL);
+	/* Add RETURNING if present */
+	if (query->returningList)
+	{
+		appendContextKeyword(context, " RETURNING",
+							 -PRETTYINDENT_STD, PRETTYINDENT_STD, 1);
+		get_target_list(query->returningList, context, NULL, colNamesVisible);
+	}
 }
 
 
@@ -8298,6 +8303,7 @@ isSimpleNode(Node *node, Node *parentNod
 		case T_Aggref:
 		case T_GroupingFunc:
 		case T_WindowFunc:
+		case T_MergingFunc:
 		case T_FuncExpr:
 		case T_JsonConstructorExpr:
 			/* function-like: name(..) or name[..] */
@@ -8652,6 +8658,27 @@ get_rule_expr(Node *node, deparse_contex
 			get_windowfunc_expr((WindowFunc *) node, context);
 			break;
 
+		case T_MergingFunc:
+			{
+				MergingFunc *mexpr = (MergingFunc *) node;
+
+				appendStringInfoString(buf, "MERGING(");
+				switch (mexpr->mfop)
+				{
+					case MERGING_ACTION:
+						appendStringInfoString(buf, "ACTION");
+						break;
+					case MERGING_CLAUSE_NUMBER:
+						appendStringInfoString(buf, "CLAUSE_NUMBER");
+						break;
+					default:
+						elog(ERROR, "unrecognized MergingFuncOp: %d",
+							 (int) mexpr->mfop);
+				}
+				appendStringInfoChar(buf, ')');
+			}
+			break;
+
 		case T_SubscriptingRef:
 			{
 				SubscriptingRef *sbsref = (SubscriptingRef *) node;
diff --git a/src/bin/psql/common.c b/src/bin/psql/common.c
new file mode 100644
index 76e01b0..2830bde
--- a/src/bin/psql/common.c
+++ b/src/bin/psql/common.c
@@ -983,13 +983,17 @@ PrintQueryResult(PGresult *result, bool
 			else
 				success = true;
 
-			/* if it's INSERT/UPDATE/DELETE RETURNING, also print status */
+			/*
+			 * If it's INSERT/UPDATE/DELETE/MERGE RETURNING, also print
+			 * status.
+			 */
 			if (last || pset.show_all_results)
 			{
 				cmdstatus = PQcmdStatus(result);
 				if (strncmp(cmdstatus, "INSERT", 6) == 0 ||
 					strncmp(cmdstatus, "UPDATE", 6) == 0 ||
-					strncmp(cmdstatus, "DELETE", 6) == 0)
+					strncmp(cmdstatus, "DELETE", 6) == 0 ||
+					strncmp(cmdstatus, "MERGE", 5) == 0)
 					PrintQueryStatus(result, printStatusFout);
 			}
 
diff --git a/src/include/executor/execExpr.h b/src/include/executor/execExpr.h
new file mode 100644
index a28ddcd..abb0fb4
--- a/src/include/executor/execExpr.h
+++ b/src/include/executor/execExpr.h
@@ -243,6 +243,7 @@ typedef enum ExprEvalOp
 	EEOP_AGGREF,
 	EEOP_GROUPING_FUNC,
 	EEOP_WINDOW_FUNC,
+	EEOP_MERGING_FUNC,
 	EEOP_SUBPLAN,
 
 	/* aggregation related nodes */
@@ -626,6 +627,12 @@ typedef struct ExprEvalStep
 			WindowFuncExprState *wfstate;
 		}			window_func;
 
+		/* for EEOP_MERGING_FUNC */
+		struct
+		{
+			MergingFuncOp mfop; /* the MERGING() operation to perform */
+		}			merging_func;
+
 		/* for EEOP_SUBPLAN */
 		struct
 		{
@@ -810,6 +817,8 @@ extern void ExecEvalJsonConstructor(Expr
 									ExprContext *econtext);
 extern void ExecEvalJsonIsPredicate(ExprState *state, ExprEvalStep *op);
 extern void ExecEvalGroupingFunc(ExprState *state, ExprEvalStep *op);
+extern void ExecEvalMergingFunc(ExprState *state, ExprEvalStep *op,
+								ExprContext *econtext);
 extern void ExecEvalSubPlan(ExprState *state, ExprEvalStep *op,
 							ExprContext *econtext);
 extern void ExecEvalWholeRowVar(ExprState *state, ExprEvalStep *op,
diff --git a/src/include/executor/spi.h b/src/include/executor/spi.h
new file mode 100644
index c13e1f6..48b8773
--- a/src/include/executor/spi.h
+++ b/src/include/executor/spi.h
@@ -97,6 +97,7 @@ typedef struct _SPI_plan *SPIPlanPtr;
 #define SPI_OK_REL_UNREGISTER	16
 #define SPI_OK_TD_REGISTER		17
 #define SPI_OK_MERGE			18
+#define SPI_OK_MERGE_RETURNING	19
 
 #define SPI_OPT_NONATOMIC		(1 << 0)
 
diff --git a/src/include/nodes/execnodes.h b/src/include/nodes/execnodes.h
new file mode 100644
index 444a5f0..57fdfca
--- a/src/include/nodes/execnodes.h
+++ b/src/include/nodes/execnodes.h
@@ -1325,6 +1325,9 @@ typedef struct ModifyTableState
 	/* Flags showing which subcommands are present INS/UPD/DEL/DO NOTHING */
 	int			mt_merge_subcommands;
 
+	/* For MERGE, the action currently being executed */
+	MergeActionState *mt_merge_action;
+
 	/* tuple counters for MERGE */
 	double		mt_merge_inserted;
 	double		mt_merge_updated;
diff --git a/src/include/nodes/parsenodes.h b/src/include/nodes/parsenodes.h
new file mode 100644
index d60e148..21d8d64
--- a/src/include/nodes/parsenodes.h
+++ b/src/include/nodes/parsenodes.h
@@ -1935,6 +1935,7 @@ typedef struct MergeStmt
 	Node	   *sourceRelation; /* source relation */
 	Node	   *joinCondition;	/* join condition between source and target */
 	List	   *mergeWhenClauses;	/* list of MergeWhenClause(es) */
+	List	   *returningList;	/* list of expressions to return */
 	WithClause *withClause;		/* WITH clause */
 } MergeStmt;
 
diff --git a/src/include/nodes/primnodes.h b/src/include/nodes/primnodes.h
new file mode 100644
index 4a15460..97806f5
--- a/src/include/nodes/primnodes.h
+++ b/src/include/nodes/primnodes.h
@@ -572,6 +572,39 @@ typedef struct WindowFunc
 } WindowFunc;
 
 /*
+ * MergingFunc
+ *
+ * A MergingFunc is a MERGING(...) expression that can only appear in the
+ * RETURNING list of a MERGE command.  It returns information about the
+ * currently executing merge action.  Possible operations are:
+ *
+ *	MERGING_ACTION:
+ *		Return the command string of the current merge action ("INSERT",
+ *		"UPDATE" or "DELETE").
+ *
+ *	MERGING_CLAUSE_NUMBER:
+ *		Return the 1-based index of the current merge WHEN clause.
+ */
+typedef enum MergingFuncOp
+{
+	MERGING_ACTION,
+	MERGING_CLAUSE_NUMBER,
+} MergingFuncOp;
+
+typedef struct MergingFunc
+{
+	Expr		xpr;
+	/* operation to perform */
+	MergingFuncOp mfop;
+	/* type Oid of result */
+	Oid			mftype pg_node_attr(query_jumble_ignore);
+	/* OID of collation, or InvalidOid if none */
+	Oid			mfcollid pg_node_attr(query_jumble_ignore);
+	/* token location, or -1 if unknown */
+	int			location;
+} MergingFunc;
+
+/*
  * SubscriptingRef: describes a subscripting operation over a container
  * (array, etc).
  *
@@ -1735,6 +1768,7 @@ typedef struct BooleanTest
 typedef struct MergeAction
 {
 	NodeTag		type;
+	int			index;			/* 1-based index of the clause */
 	bool		matched;		/* true=MATCHED, false=NOT MATCHED */
 	CmdType		commandType;	/* INSERT/UPDATE/DELETE/DO NOTHING */
 	/* OVERRIDING clause */
diff --git a/src/include/optimizer/paramassign.h b/src/include/optimizer/paramassign.h
new file mode 100644
index 501f728..1e95c8f
--- a/src/include/optimizer/paramassign.h
+++ b/src/include/optimizer/paramassign.h
@@ -20,6 +20,7 @@ extern Param *replace_outer_placeholderv
 										   PlaceHolderVar *phv);
 extern Param *replace_outer_agg(PlannerInfo *root, Aggref *agg);
 extern Param *replace_outer_grouping(PlannerInfo *root, GroupingFunc *grp);
+extern Param *replace_outer_merging(PlannerInfo *root, MergingFunc *mf);
 extern Param *replace_nestloop_param_var(PlannerInfo *root, Var *var);
 extern Param *replace_nestloop_param_placeholdervar(PlannerInfo *root,
 													PlaceHolderVar *phv);
diff --git a/src/include/parser/analyze.h b/src/include/parser/analyze.h
new file mode 100644
index 5ecdd10..28b66fc
--- a/src/include/parser/analyze.h
+++ b/src/include/parser/analyze.h
@@ -44,6 +44,8 @@ extern List *transformInsertRow(ParseSta
 								bool strip_indirection);
 extern List *transformUpdateTargetList(ParseState *pstate,
 									   List *origTlist);
+extern List *transformReturningList(ParseState *pstate, List *returningList,
+									ParseExprKind exprKind);
 extern Query *transformTopLevelStmt(ParseState *pstate, RawStmt *parseTree);
 extern Query *transformStmt(ParseState *pstate, Node *parseTree);
 
diff --git a/src/include/parser/kwlist.h b/src/include/parser/kwlist.h
new file mode 100644
index 2331aca..9c5d6ea
--- a/src/include/parser/kwlist.h
+++ b/src/include/parser/kwlist.h
@@ -80,6 +80,7 @@ PG_KEYWORD("characteristics", CHARACTERI
 PG_KEYWORD("check", CHECK, RESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("checkpoint", CHECKPOINT, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("class", CLASS, UNRESERVED_KEYWORD, BARE_LABEL)
+PG_KEYWORD("clause_number", CLAUSE_NUMBER, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("close", CLOSE, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("cluster", CLUSTER, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("coalesce", COALESCE, COL_NAME_KEYWORD, BARE_LABEL)
@@ -266,6 +267,7 @@ PG_KEYWORD("matched", MATCHED, UNRESERVE
 PG_KEYWORD("materialized", MATERIALIZED, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("maxvalue", MAXVALUE, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("merge", MERGE, UNRESERVED_KEYWORD, BARE_LABEL)
+PG_KEYWORD("merging", MERGING, COL_NAME_KEYWORD, BARE_LABEL)
 PG_KEYWORD("method", METHOD, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("minute", MINUTE_P, UNRESERVED_KEYWORD, AS_LABEL)
 PG_KEYWORD("minvalue", MINVALUE, UNRESERVED_KEYWORD, BARE_LABEL)
diff --git a/src/include/parser/parse_node.h b/src/include/parser/parse_node.h
new file mode 100644
index 99d6515..5b781d8
--- a/src/include/parser/parse_node.h
+++ b/src/include/parser/parse_node.h
@@ -61,7 +61,8 @@ typedef enum ParseExprKind
 	EXPR_KIND_DISTINCT_ON,		/* DISTINCT ON */
 	EXPR_KIND_LIMIT,			/* LIMIT */
 	EXPR_KIND_OFFSET,			/* OFFSET */
-	EXPR_KIND_RETURNING,		/* RETURNING */
+	EXPR_KIND_RETURNING,		/* RETURNING in INSERT/UPDATE/DELETE */
+	EXPR_KIND_MERGE_RETURNING,	/* RETURNING in MERGE */
 	EXPR_KIND_VALUES,			/* VALUES */
 	EXPR_KIND_VALUES_SINGLE,	/* single-row VALUES (in INSERT only) */
 	EXPR_KIND_CHECK_CONSTRAINT, /* CHECK constraint for a table */
diff --git a/src/pl/plpgsql/src/pl_exec.c b/src/pl/plpgsql/src/pl_exec.c
new file mode 100644
index 6d16913..51de094
--- a/src/pl/plpgsql/src/pl_exec.c
+++ b/src/pl/plpgsql/src/pl_exec.c
@@ -4267,9 +4267,9 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 	/*
 	 * If we have INTO, then we only need one row back ... but if we have INTO
 	 * STRICT or extra check too_many_rows, ask for two rows, so that we can
-	 * verify the statement returns only one.  INSERT/UPDATE/DELETE are always
-	 * treated strictly. Without INTO, just run the statement to completion
-	 * (tcount = 0).
+	 * verify the statement returns only one.  INSERT/UPDATE/DELETE/MERGE are
+	 * always treated strictly. Without INTO, just run the statement to
+	 * completion (tcount = 0).
 	 *
 	 * We could just ask for two rows always when using INTO, but there are
 	 * some cases where demanding the extra row costs significant time, eg by
@@ -4307,10 +4307,11 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 			Assert(stmt->mod_stmt);
 			exec_set_found(estate, (SPI_processed != 0));
 			break;
@@ -4489,10 +4490,11 @@ exec_stmt_dynexecute(PLpgSQL_execstate *
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 		case SPI_OK_UTILITY:
 		case SPI_OK_REWRITTEN:
 			break;
diff --git a/src/pl/tcl/pltcl.c b/src/pl/tcl/pltcl.c
new file mode 100644
index 6187e15..18d14a2
--- a/src/pl/tcl/pltcl.c
+++ b/src/pl/tcl/pltcl.c
@@ -2456,6 +2456,7 @@ pltcl_process_SPI_result(Tcl_Interp *int
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
+		case SPI_OK_MERGE_RETURNING:
 
 			/*
 			 * Process the tuples we got
diff --git a/src/test/regress/expected/merge.out b/src/test/regress/expected/merge.out
new file mode 100644
index e3ebf46..778e7c8
--- a/src/test/regress/expected/merge.out
+++ b/src/test/regress/expected/merge.out
@@ -124,20 +124,20 @@ ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
 ERROR:  name "target" specified more than once
 DETAIL:  The name is used both as MERGE target table and data source.
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
-ERROR:  MERGE not supported in WITH query
-LINE 1: WITH foo AS (
-             ^
--- used in COPY
+ERROR:  WITH query "foo" does not have a RETURNING clause
+LINE 4: ) SELECT * FROM foo;
+                        ^
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) TO stdout;
-ERROR:  MERGE not supported in COPY
+ERROR:  COPY query must have a RETURNING clause
 -- unsupported relation types
 -- materialized view
 CREATE MATERIALIZED VIEW mv AS SELECT * FROM target;
@@ -1255,7 +1255,7 @@ BEGIN;
 MERGE INTO sq_target
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1270,7 +1270,7 @@ INSERT INTO sq_source (sid, balance, del
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1294,7 +1294,7 @@ WITH targq AS (
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1303,20 +1303,191 @@ WHEN MATCHED AND tid < 2 THEN
 ROLLBACK;
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
+CREATE TABLE merge_actions(action text, abbrev text);
+INSERT INTO merge_actions VALUES ('INSERT', 'ins'), ('UPDATE', 'upd'), ('DELETE', 'del');
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
-ERROR:  syntax error at or near "RETURNING"
-LINE 10: RETURNING *;
-         ^
+    DELETE
+RETURNING MERGING(CLAUSE_NUMBER) AS clause,
+          (SELECT abbrev FROM merge_actions
+            WHERE action = MERGING(ACTION)) AS action,
+          t.*,
+          CASE MERGING(ACTION)
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ clause | action | tid | balance |     description     
+--------+--------+-----+---------+---------------------
+      3 | del    |   1 |     100 | Removed (1,100)
+      1 | upd    |   2 |     220 | Added 20 to balance
+      2 | ins    |   4 |      40 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- error when using MERGING() outside MERGE
+SELECT MERGING(ACTION) FROM sq_target;
+ERROR:  MERGING() can only be used in the RETURNING list of a MERGE command
+LINE 1: SELECT MERGING(ACTION) FROM sq_target;
+               ^
+UPDATE sq_target SET balance = balance + 1 RETURNING MERGING(CLAUSE_NUMBER);
+ERROR:  MERGING() can only be used in the RETURNING list of a MERGE command
+LINE 1: ...ATE sq_target SET balance = balance + 1 RETURNING MERGING(CL...
+                                                             ^
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(CLAUSE_NUMBER), MERGING(ACTION), t.*,
+              CASE MERGING(ACTION)
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING action, MERGING(ACTION) AS log_action, l.*
+)
+SELECT * FROM m2;
+ action | log_action | tid |     last_change     
+--------+------------+-----+---------------------
+ DELETE | UPDATE     |   1 | Removed (1,100)
+ UPDATE | INSERT     |   2 | Added 20 to balance
+ INSERT | INSERT     |   4 | Inserted (4,40)
+(3 rows)
+
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ tid |     last_change     
+-----+---------------------
+   1 | Removed (1,100)
+   2 | Added 20 to balance
+   4 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.*
+) TO stdout;
+DELETE	1	100
+UPDATE	2	220
+INSERT	4	40
+ROLLBACK;
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ action | tid | new_balance 
+--------+-----+-------------
+ DELETE |   1 |         100
+ UPDATE |   3 |         320
+ INSERT |   4 |         110
+(3 rows)
+
+ROLLBACK;
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ action | tid | balance 
+--------+-----+---------
+ DELETE |   1 |     100
+ UPDATE |   2 |     220
+ INSERT |   4 |      40
+(3 rows)
+
+ROLLBACK;
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ r_action | r_tid | r_balance 
+----------+-------+-----------
+ DELETE   |     1 |       100
+ UPDATE   |     3 |       320
+ INSERT   |     4 |       110
+(3 rows)
+
 ROLLBACK;
 -- EXPLAIN
 CREATE TABLE ex_mtarget (a int, b int)
@@ -1513,7 +1684,7 @@ SELECT * FROM sq_target WHERE tid = 1;
 (1 row)
 
 ROLLBACK;
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 NOTICE:  drop cascades to view v
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1639,6 +1810,32 @@ SELECT * FROM pa_target ORDER BY tid;
 (14 rows)
 
 ROLLBACK;
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING MERGING(ACTION), MERGING(CLAUSE_NUMBER), t.*;
+ action | clause_number | tid | balance |           val            
+--------+---------------+-----+---------+--------------------------
+ UPDATE |             1 |   2 |     110 | initial updated by merge
+(1 row)
+
+SELECT * FROM pa_target ORDER BY tid;
+ tid | balance |           val            
+-----+---------+--------------------------
+   2 |     110 | initial updated by merge
+   3 |     300 | initial
+   5 |     500 | initial
+   7 |     700 | initial
+   9 |     900 | initial
+  11 |    1100 | initial
+  13 |    1300 | initial
+(7 rows)
+
+ROLLBACK;
 DROP TABLE pa_target CASCADE;
 -- The target table is partitioned in the same way, but this time by attaching
 -- partitions which have columns in different order, dropped columns etc.
@@ -1888,7 +2085,21 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING MERGING(ACTION), MERGING(CLAUSE_NUMBER), t.*;
+ action | clause_number |          logts           | tid | balance |           val            
+--------+---------------+--------------------------+-----+---------+--------------------------
+ UPDATE |             1 | Tue Jan 31 00:00:00 2017 |   1 |     110 | initial updated by merge
+ UPDATE |             1 | Tue Feb 28 00:00:00 2017 |   2 |     220 | initial updated by merge
+ INSERT |             2 | Sun Jan 15 00:00:00 2017 |   3 |      30 | inserted by merge
+ UPDATE |             1 | Tue Jan 31 00:00:00 2017 |   4 |     440 | initial updated by merge
+ UPDATE |             1 | Tue Feb 28 00:00:00 2017 |   5 |     550 | initial updated by merge
+ INSERT |             2 | Sun Jan 15 00:00:00 2017 |   6 |      60 | inserted by merge
+ UPDATE |             1 | Tue Jan 31 00:00:00 2017 |   7 |     770 | initial updated by merge
+ UPDATE |             1 | Tue Feb 28 00:00:00 2017 |   8 |     880 | initial updated by merge
+ INSERT |             2 | Sun Jan 15 00:00:00 2017 |   9 |      90 | inserted by merge
+(9 rows)
+
 SELECT * FROM pa_target ORDER BY tid;
           logts           | tid | balance |           val            
 --------------------------+-----+---------+--------------------------
diff --git a/src/test/regress/expected/rowsecurity.out b/src/test/regress/expected/rowsecurity.out
new file mode 100644
index 6988128..3fd4fab
--- a/src/test/regress/expected/rowsecurity.out
+++ b/src/test/regress/expected/rowsecurity.out
@@ -2302,6 +2302,35 @@ WHEN MATCHED THEN
 	UPDATE SET dnotes = dnotes || ' notes added by merge8 '
 WHEN NOT MATCHED THEN
 	INSERT VALUES (13, 44, 1, 'regress_rls_bob', 'new manga');
+SELECT * FROM document WHERE did = 13;
+ did | cid | dlevel | dauthor | dtitle | dnotes 
+-----+-----+--------+---------+--------+--------
+(0 rows)
+
+-- but not OK if RETURNING is used
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge9 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 44, 1, 'regress_rls_bob', 'new manga')
+RETURNING *;
+ERROR:  new row violates row-level security policy for table "document"
+-- but OK if new row is visible
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge10 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 11, 1, 'regress_rls_bob', 'new novel')
+RETURNING *;
+ sdid | did | cid | dlevel |     dauthor     |  dtitle   | dnotes 
+------+-----+-----+--------+-----------------+-----------+--------
+   14 |  14 |  11 |      1 | regress_rls_bob | new novel | 
+(1 row)
+
 RESET SESSION AUTHORIZATION;
 -- drop the restrictive SELECT policy so that we can look at the
 -- final state of the table
@@ -2325,7 +2354,8 @@ SELECT * FROM document;
   12 |  11 |      1 | regress_rls_bob   | another novel                    | 
    1 |  11 |      1 | regress_rls_bob   | my first novel                   |  notes added by merge2  notes added by merge3  notes added by merge4  notes added by merge7 
   13 |  44 |      1 | regress_rls_bob   | new manga                        | 
-(15 rows)
+  14 |  11 |      1 | regress_rls_bob   | new novel                        | 
+(16 rows)
 
 --
 -- ROLE/GROUP
diff --git a/src/test/regress/expected/rules.out b/src/test/regress/expected/rules.out
new file mode 100644
index 0cd2c64..31beacc
--- a/src/test/regress/expected/rules.out
+++ b/src/test/regress/expected/rules.out
@@ -3638,7 +3638,8 @@ MERGE INTO rule_merge1 t USING (SELECT 1
 -- test deparsing
 CREATE TABLE sf_target(id int, data text, filling int[]);
 CREATE FUNCTION merge_sf_test()
- RETURNS void
+ RETURNS TABLE(action text, clause_number int,
+               a int, b text, id int, data text, filling int[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -3675,11 +3676,13 @@ WHEN NOT MATCHED
    VALUES (s.a, s.b, DEFAULT)
 WHEN NOT MATCHED
    THEN INSERT (filling[1], id)
-   VALUES (s.a, s.a);
+   VALUES (s.a, s.a)
+RETURNING
+   merging(action), merging(clause_number) AS clause_number, *;
 END;
 \sf merge_sf_test
 CREATE OR REPLACE FUNCTION public.merge_sf_test()
- RETURNS void
+ RETURNS TABLE(action text, clause_number integer, a integer, b text, id integer, data text, filling integer[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -3716,7 +3719,14 @@ BEGIN ATOMIC
       VALUES (s.a, s.b, DEFAULT)
     WHEN NOT MATCHED
      THEN INSERT (filling[1], id)
-      VALUES (s.a, s.a);
+      VALUES (s.a, s.a)
+   RETURNING MERGING(ACTION) AS action,
+     MERGING(CLAUSE_NUMBER) AS clause_number,
+     s.a,
+     s.b,
+     t.id,
+     t.data,
+     t.filling;
 END
 DROP FUNCTION merge_sf_test;
 DROP TABLE sf_target;
diff --git a/src/test/regress/expected/updatable_views.out b/src/test/regress/expected/updatable_views.out
new file mode 100644
index 794cf9c..9d1e68c
--- a/src/test/regress/expected/updatable_views.out
+++ b/src/test/regress/expected/updatable_views.out
@@ -479,7 +479,15 @@ MERGE INTO rw_view1 t
                 (2, 'ROW 2'), (3, 'ROW 3')) AS v(a,b) ON t.a = v.a
   WHEN MATCHED AND t.a <= 1 THEN UPDATE SET b = v.b
   WHEN MATCHED THEN DELETE
-  WHEN NOT MATCHED AND a > 0 THEN INSERT (a) VALUES (v.a);
+  WHEN NOT MATCHED AND a > 0 THEN INSERT (a) VALUES (v.a)
+  RETURNING MERGING(ACTION), v.*, t.*;
+ action | a |   b   | a |      b      
+--------+---+-------+---+-------------
+ UPDATE | 1 | ROW 1 | 1 | ROW 1
+ DELETE | 3 | ROW 3 | 3 | Row 3
+ INSERT | 2 | ROW 2 | 2 | Unspecified
+(3 rows)
+
 SELECT * FROM base_tbl ORDER BY a;
  a  |      b      
 ----+-------------
@@ -631,7 +639,15 @@ MERGE INTO rw_view2 t
   USING (VALUES (3, 'R3'), (4, 'R4'), (5, 'R5')) AS v(a,b) ON aaa = v.a
   WHEN MATCHED AND aaa = 3 THEN DELETE
   WHEN MATCHED THEN UPDATE SET bbb = v.b
-  WHEN NOT MATCHED THEN INSERT (aaa) VALUES (v.a);
+  WHEN NOT MATCHED THEN INSERT (aaa) VALUES (v.a)
+  RETURNING MERGING(ACTION), v.*, t.*;
+ action | a | b  | aaa |     bbb     
+--------+---+----+-----+-------------
+ DELETE | 3 | R3 |   3 | Row 3
+ UPDATE | 4 | R4 |   4 | R4
+ INSERT | 5 | R5 |   5 | Unspecified
+(3 rows)
+
 SELECT * FROM rw_view2 ORDER BY aaa;
  aaa |     bbb     
 -----+-------------
@@ -1071,7 +1087,15 @@ MERGE INTO rw_view2 t
   USING (SELECT x, 'R'||x FROM generate_series(0,3) x) AS s(a,b) ON t.a = s.a
   WHEN MATCHED AND t.a <= 1 THEN DELETE
   WHEN MATCHED THEN UPDATE SET b = s.b
-  WHEN NOT MATCHED AND s.a > 0 THEN INSERT VALUES (s.a, s.b);
+  WHEN NOT MATCHED AND s.a > 0 THEN INSERT VALUES (s.a, s.b)
+  RETURNING MERGING(ACTION), s.*, t.*;
+ action | a | b  | a |   b   
+--------+---+----+---+-------
+ DELETE | 1 | R1 | 1 | Row 1
+ UPDATE | 2 | R2 | 2 | R2
+ INSERT | 3 | R3 | 3 | R3
+(3 rows)
+
 SELECT * FROM base_tbl ORDER BY a;
  a  |   b    
 ----+--------
diff --git a/src/test/regress/expected/with.out b/src/test/regress/expected/with.out
new file mode 100644
index 7d796ea..4666873
--- a/src/test/regress/expected/with.out
+++ b/src/test/regress/expected/with.out
@@ -3493,6 +3493,16 @@ SELECT * FROM t;
 ERROR:  WITH query "t" does not have a RETURNING clause
 LINE 4: SELECT * FROM t;
                       ^
+-- RETURNING tries to return its own output
+WITH RECURSIVE t(action, a) AS (
+	MERGE INTO y USING (VALUES (11)) v(a) ON y.a = v.a
+		WHEN NOT MATCHED THEN INSERT VALUES (v.a)
+		RETURNING MERGING(ACTION), (SELECT a FROM t)
+)
+SELECT * FROM t;
+ERROR:  recursive query "t" must not contain data-modifying statements
+LINE 1: WITH RECURSIVE t(action, a) AS (
+                       ^
 -- data-modifying WITH allowed only at the top level
 SELECT * FROM (
 	WITH t AS (UPDATE y SET a=a+1 RETURNING *)
diff --git a/src/test/regress/sql/merge.sql b/src/test/regress/sql/merge.sql
new file mode 100644
index 79a1868..bcab004
--- a/src/test/regress/sql/merge.sql
+++ b/src/test/regress/sql/merge.sql
@@ -88,12 +88,12 @@ MERGE INTO target
 USING target
 ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
--- used in COPY
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
@@ -817,7 +817,7 @@ BEGIN;
 MERGE INTO sq_target
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -830,7 +830,7 @@ INSERT INTO sq_source (sid, balance, del
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -848,7 +848,7 @@ WITH targq AS (
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -858,17 +858,146 @@ ROLLBACK;
 
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
+CREATE TABLE merge_actions(action text, abbrev text);
+INSERT INTO merge_actions VALUES ('INSERT', 'ins'), ('UPDATE', 'upd'), ('DELETE', 'del');
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
+    DELETE
+RETURNING MERGING(CLAUSE_NUMBER) AS clause,
+          (SELECT abbrev FROM merge_actions
+            WHERE action = MERGING(ACTION)) AS action,
+          t.*,
+          CASE MERGING(ACTION)
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ROLLBACK;
+
+-- error when using MERGING() outside MERGE
+SELECT MERGING(ACTION) FROM sq_target;
+UPDATE sq_target SET balance = balance + 1 RETURNING MERGING(CLAUSE_NUMBER);
+
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(CLAUSE_NUMBER), MERGING(ACTION), t.*,
+              CASE MERGING(ACTION)
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING action, MERGING(ACTION) AS log_action, l.*
+)
+SELECT * FROM m2;
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ROLLBACK;
+
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.*
+) TO stdout;
+ROLLBACK;
+
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ROLLBACK;
+
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ROLLBACK;
+
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING MERGING(ACTION), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
 ROLLBACK;
 
 -- EXPLAIN
@@ -967,7 +1096,7 @@ WHEN MATCHED THEN
 SELECT * FROM sq_target WHERE tid = 1;
 ROLLBACK;
 
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1034,6 +1163,17 @@ SELECT merge_func();
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING MERGING(ACTION), MERGING(CLAUSE_NUMBER), t.*;
+SELECT * FROM pa_target ORDER BY tid;
+ROLLBACK;
+
 DROP TABLE pa_target CASCADE;
 
 -- The target table is partitioned in the same way, but this time by attaching
@@ -1210,7 +1350,8 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING MERGING(ACTION), MERGING(CLAUSE_NUMBER), t.*;
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
diff --git a/src/test/regress/sql/rowsecurity.sql b/src/test/regress/sql/rowsecurity.sql
new file mode 100644
index dec7340..1d5ed0a
--- a/src/test/regress/sql/rowsecurity.sql
+++ b/src/test/regress/sql/rowsecurity.sql
@@ -992,6 +992,27 @@ WHEN MATCHED THEN
 	UPDATE SET dnotes = dnotes || ' notes added by merge8 '
 WHEN NOT MATCHED THEN
 	INSERT VALUES (13, 44, 1, 'regress_rls_bob', 'new manga');
+SELECT * FROM document WHERE did = 13;
+
+-- but not OK if RETURNING is used
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge9 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 44, 1, 'regress_rls_bob', 'new manga')
+RETURNING *;
+
+-- but OK if new row is visible
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge10 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 11, 1, 'regress_rls_bob', 'new novel')
+RETURNING *;
 
 RESET SESSION AUTHORIZATION;
 -- drop the restrictive SELECT policy so that we can look at the
diff --git a/src/test/regress/sql/rules.sql b/src/test/regress/sql/rules.sql
new file mode 100644
index 6924012..efd3287
--- a/src/test/regress/sql/rules.sql
+++ b/src/test/regress/sql/rules.sql
@@ -1294,7 +1294,8 @@ MERGE INTO rule_merge1 t USING (SELECT 1
 CREATE TABLE sf_target(id int, data text, filling int[]);
 
 CREATE FUNCTION merge_sf_test()
- RETURNS void
+ RETURNS TABLE(action text, clause_number int,
+               a int, b text, id int, data text, filling int[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -1331,7 +1332,9 @@ WHEN NOT MATCHED
    VALUES (s.a, s.b, DEFAULT)
 WHEN NOT MATCHED
    THEN INSERT (filling[1], id)
-   VALUES (s.a, s.a);
+   VALUES (s.a, s.a)
+RETURNING
+   merging(action), merging(clause_number) AS clause_number, *;
 END;
 
 \sf merge_sf_test
diff --git a/src/test/regress/sql/updatable_views.sql b/src/test/regress/sql/updatable_views.sql
new file mode 100644
index ae11e46..650b315
--- a/src/test/regress/sql/updatable_views.sql
+++ b/src/test/regress/sql/updatable_views.sql
@@ -175,7 +175,8 @@ MERGE INTO rw_view1 t
                 (2, 'ROW 2'), (3, 'ROW 3')) AS v(a,b) ON t.a = v.a
   WHEN MATCHED AND t.a <= 1 THEN UPDATE SET b = v.b
   WHEN MATCHED THEN DELETE
-  WHEN NOT MATCHED AND a > 0 THEN INSERT (a) VALUES (v.a);
+  WHEN NOT MATCHED AND a > 0 THEN INSERT (a) VALUES (v.a)
+  RETURNING MERGING(ACTION), v.*, t.*;
 SELECT * FROM base_tbl ORDER BY a;
 
 EXPLAIN (costs off) UPDATE rw_view1 SET a=6 WHERE a=5;
@@ -246,7 +247,8 @@ MERGE INTO rw_view2 t
   USING (VALUES (3, 'R3'), (4, 'R4'), (5, 'R5')) AS v(a,b) ON aaa = v.a
   WHEN MATCHED AND aaa = 3 THEN DELETE
   WHEN MATCHED THEN UPDATE SET bbb = v.b
-  WHEN NOT MATCHED THEN INSERT (aaa) VALUES (v.a);
+  WHEN NOT MATCHED THEN INSERT (aaa) VALUES (v.a)
+  RETURNING MERGING(ACTION), v.*, t.*;
 SELECT * FROM rw_view2 ORDER BY aaa;
 
 EXPLAIN (costs off) UPDATE rw_view2 SET aaa=5 WHERE aaa=4;
@@ -458,7 +460,8 @@ MERGE INTO rw_view2 t
   USING (SELECT x, 'R'||x FROM generate_series(0,3) x) AS s(a,b) ON t.a = s.a
   WHEN MATCHED AND t.a <= 1 THEN DELETE
   WHEN MATCHED THEN UPDATE SET b = s.b
-  WHEN NOT MATCHED AND s.a > 0 THEN INSERT VALUES (s.a, s.b);
+  WHEN NOT MATCHED AND s.a > 0 THEN INSERT VALUES (s.a, s.b)
+  RETURNING MERGING(ACTION), s.*, t.*;
 SELECT * FROM base_tbl ORDER BY a;
 
 EXPLAIN (costs off) UPDATE rw_view2 SET a=3 WHERE a=2;
diff --git a/src/test/regress/sql/with.sql b/src/test/regress/sql/with.sql
new file mode 100644
index f8a213e..1d82d4a
--- a/src/test/regress/sql/with.sql
+++ b/src/test/regress/sql/with.sql
@@ -1591,6 +1591,14 @@ WITH t AS (
 )
 SELECT * FROM t;
 
+-- RETURNING tries to return its own output
+WITH RECURSIVE t(action, a) AS (
+	MERGE INTO y USING (VALUES (11)) v(a) ON y.a = v.a
+		WHEN NOT MATCHED THEN INSERT VALUES (v.a)
+		RETURNING MERGING(ACTION), (SELECT a FROM t)
+)
+SELECT * FROM t;
+
 -- data-modifying WITH allowed only at the top level
 SELECT * FROM (
 	WITH t AS (UPDATE y SET a=a+1 RETURNING *)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
new file mode 100644
index fc8b15d..9dc7e24
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -1577,6 +1577,8 @@ MergePath
 MergeScanSelCache
 MergeStmt
 MergeWhenClause
+MergingFunc
+MergingFuncOp
 MetaCommand
 MinMaxAggInfo
 MinMaxAggPath

diff --git a/doc/src/sgml/dml.sgml b/doc/src/sgml/dml.sgml
new file mode 100644
index cbbc5e2..3d95bdb
--- a/doc/src/sgml/dml.sgml
+++ b/doc/src/sgml/dml.sgml
@@ -283,10 +283,15 @@ DELETE FROM products;
    <secondary>RETURNING</secondary>
   </indexterm>
 
+  <indexterm zone="dml-returning">
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
   <para>
    Sometimes it is useful to obtain data from modified rows while they are
    being manipulated.  The <command>INSERT</command>, <command>UPDATE</command>,
-   and <command>DELETE</command> commands all have an
+   <command>DELETE</command>, and <command>MERGE</command> commands all have an
    optional <literal>RETURNING</literal> clause that supports this.  Use
    of <literal>RETURNING</literal> avoids performing an extra database query to
    collect the data, and is especially valuable when it would otherwise be
@@ -339,6 +344,21 @@ DELETE FROM products
 </programlisting>
   </para>
 
+  <para>
+   In a <command>MERGE</command>, the data available to <literal>RETURNING</literal> is
+   the content of the source row plus the content of the inserted, updated, or
+   deleted target row.  Since it is quite common for the source and target to
+   have many of the same columns, specifying <literal>RETURNING *</literal>
+   can lead to a lot of duplicated columns, so it is often more useful to
+   qualify it so as to return just the source or target row.  For example:
+<programlisting>
+MERGE INTO products p USING new_products n ON p.product_no = n.product_no
+  WHEN NOT MATCHED THEN INSERT VALUES (n.product_no, n.name, n.price)
+  WHEN MATCHED THEN UPDATE SET name = n.name, price = n.price
+  RETURNING p.*;
+</programlisting>
+  </para>
+
   <para>
    If there are triggers (<xref linkend="triggers"/>) on the target table,
    the data available to <literal>RETURNING</literal> is the row as modified by
diff --git a/doc/src/sgml/func.sgml b/doc/src/sgml/func.sgml
new file mode 100644
index 0bb7aeb..0e2b888
--- a/doc/src/sgml/func.sgml
+++ b/doc/src/sgml/func.sgml
@@ -22421,6 +22421,85 @@ SELECT count(*) FROM sometable;
 
  </sect1>
 
+ <sect1 id="functions-merge-support">
+  <title>Merge Support Functions</title>
+
+  <indexterm>
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
+  <para>
+   <productname>PostgreSQL</productname> includes one merge support function
+   that may be used in the <literal>RETURNING</literal> list of a
+   <xref linkend="sql-merge"/> command to identify the action taken for each
+   row.
+  </para>
+
+  <table id="functions-merge-support-table">
+   <title>Merge Support Functions</title>
+
+   <tgroup cols="1">
+    <thead>
+     <row>
+      <entry role="func_table_entry"><para role="func_signature">
+       Function
+      </para>
+      <para>
+       Description
+      </para></entry>
+     </row>
+    </thead>
+
+    <tbody>
+     <row>
+      <entry role="func_table_entry"><para role="func_signature">
+       <indexterm>
+        <primary>merge_action</primary>
+       </indexterm>
+       <function id="merge_action">merge_action</function> ( )
+       <returnvalue>text</returnvalue>
+      </para>
+      <para>
+       Returns the merge action command executed for the current row.  This
+       will be <literal>'INSERT'</literal>, <literal>'UPDATE'</literal>, or
+       <literal>'DELETE'</literal>.
+      </para></entry>
+     </row>
+    </tbody>
+   </tgroup>
+  </table>
+
+  <para>
+   Example:
+<screen><![CDATA[
+MERGE INTO products p
+  USING stock s ON p.product_id = s.product_id
+  WHEN MATCHED AND s.quantity > 0 THEN
+    UPDATE SET in_stock = true, quantity = s.quantity
+  WHEN MATCHED THEN
+    UPDATE SET in_stock = false, quantity = 0
+  WHEN NOT MATCHED THEN
+    INSERT (product_id, in_stock, quantity)
+      VALUES (s.product_id, true, s.quantity)
+  RETURNING merge_action(), p.*;
+
+ merge_action | product_id | in_stock | quantity
+--------------+------------+----------+----------
+ UPDATE       |       1001 | t        |       50
+ UPDATE       |       1002 | f        |        0
+ INSERT       |       1003 | t        |       10
+]]></screen>
+  </para>
+
+  <para>
+   Note that this function can only be used in the <literal>RETURNING</literal>
+   list of a <command>MERGE</command> command.  It is an error to use it in any
+   other part of a query.
+  </para>
+
+ </sect1>
+
  <sect1 id="functions-subquery">
   <title>Subquery Expressions</title>
 
diff --git a/doc/src/sgml/glossary.sgml b/doc/src/sgml/glossary.sgml
new file mode 100644
index 8c2f114..a81c17a
--- a/doc/src/sgml/glossary.sgml
+++ b/doc/src/sgml/glossary.sgml
@@ -1442,9 +1442,9 @@
      to a <glossterm linkend="glossary-client">client</glossterm> upon the
      completion of an <acronym>SQL</acronym> command, usually a
      <command>SELECT</command> but it can be an
-     <command>INSERT</command>, <command>UPDATE</command>, or
-     <command>DELETE</command> command if the <literal>RETURNING</literal>
-     clause is specified.
+     <command>INSERT</command>, <command>UPDATE</command>,
+     <command>DELETE</command>, or <command>MERGE</command> command if the
+     <literal>RETURNING</literal> clause is specified.
     </para>
     <para>
      The fact that a result set is a relation means that a query can be used
diff --git a/doc/src/sgml/plpgsql.sgml b/doc/src/sgml/plpgsql.sgml
new file mode 100644
index c2b9c6a..406834e
--- a/doc/src/sgml/plpgsql.sgml
+++ b/doc/src/sgml/plpgsql.sgml
@@ -1043,8 +1043,8 @@ INSERT INTO mytable VALUES (1,'one'), (2
     </para>
 
     <para>
-     If the command does return rows (for example <command>SELECT</command>,
-     or <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>
+     If the command does return rows (for example <command>SELECT</command>, or
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
      with <literal>RETURNING</literal>), there are two ways to proceed.
      When the command will return at most one row, or you only care about
      the first row of output, write the command as usual but add
@@ -1172,6 +1172,7 @@ SELECT <replaceable>select_expressions</
 INSERT ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 UPDATE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 DELETE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
+MERGE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 </synopsis>
 
      where <replaceable>target</replaceable> can be a record variable, a row
@@ -1182,8 +1183,8 @@ DELETE ... RETURNING <replaceable>expres
      <literal>INTO</literal> clause) just as described above,
      and the plan is cached in the same way.
      This works for <command>SELECT</command>,
-     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
-     <literal>RETURNING</literal>, and certain utility commands
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
+     with <literal>RETURNING</literal>, and certain utility commands
      that return row sets, such as <command>EXPLAIN</command>.
      Except for the <literal>INTO</literal> clause, the SQL command is the same
      as it would be written outside <application>PL/pgSQL</application>.
@@ -1259,7 +1260,7 @@ END;
     </para>
 
     <para>
-     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
+     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command> with
      <literal>RETURNING</literal>, <application>PL/pgSQL</application> reports
      an error for more than one returned row, even when
      <literal>STRICT</literal> is not specified.  This is because there
diff --git a/doc/src/sgml/queries.sgml b/doc/src/sgml/queries.sgml
new file mode 100644
index 648b283..372cce1
--- a/doc/src/sgml/queries.sgml
+++ b/doc/src/sgml/queries.sgml
@@ -2063,9 +2063,10 @@ SELECT <replaceable>select_list</replace
    Table Expressions or <acronym>CTE</acronym>s, can be thought of as defining
    temporary tables that exist just for one query.  Each auxiliary statement
    in a <literal>WITH</literal> clause can be a <command>SELECT</command>,
-   <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>; and the
+   <command>INSERT</command>, <command>UPDATE</command>, <command>DELETE</command>,
+   or <command>MERGE</command>; and the
    <literal>WITH</literal> clause itself is attached to a primary statement that can
-   be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
+   also be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
    <command>DELETE</command>, or <command>MERGE</command>.
   </para>
 
@@ -2599,8 +2600,8 @@ SELECT * FROM w AS w1 JOIN w AS w2 ON w1
    <title>Data-Modifying Statements in <literal>WITH</literal></title>
 
    <para>
-    You can use most data-modifying statements (<command>INSERT</command>,
-    <command>UPDATE</command>, or <command>DELETE</command>, but not
+    You can use data-modifying statements (<command>INSERT</command>,
+    <command>UPDATE</command>, <command>DELETE</command>, or
     <command>MERGE</command>) in <literal>WITH</literal>.  This
     allows you to perform several different operations in the same query.
     An example is:
diff --git a/doc/src/sgml/ref/copy.sgml b/doc/src/sgml/ref/copy.sgml
new file mode 100644
index 55764fc..6c83e30
--- a/doc/src/sgml/ref/copy.sgml
+++ b/doc/src/sgml/ref/copy.sgml
@@ -130,16 +130,19 @@ COPY { <replaceable class="parameter">ta
       A <link linkend="sql-select"><command>SELECT</command></link>,
       <link linkend="sql-values"><command>VALUES</command></link>,
       <link linkend="sql-insert"><command>INSERT</command></link>,
-      <link linkend="sql-update"><command>UPDATE</command></link>, or
-      <link linkend="sql-delete"><command>DELETE</command></link> command whose results are to be
-      copied.  Note that parentheses are required around the query.
+      <link linkend="sql-update"><command>UPDATE</command></link>,
+      <link linkend="sql-delete"><command>DELETE</command></link>, or
+      <link linkend="sql-merge"><command>MERGE</command></link> command
+      whose results are to be copied.  Note that parentheses are required
+      around the query.
      </para>
      <para>
-      For <command>INSERT</command>, <command>UPDATE</command> and
-      <command>DELETE</command> queries a <literal>RETURNING</literal> clause
-      must be provided, and the target relation must not have a conditional
-      rule, nor an <literal>ALSO</literal> rule, nor an
-      <literal>INSTEAD</literal> rule that expands to multiple statements.
+      For <command>INSERT</command>, <command>UPDATE</command>,
+      <command>DELETE</command>, and <command>MERGE</command> queries a
+      <literal>RETURNING</literal> clause must be provided, and the target
+      relation must not have a conditional rule, nor an
+      <literal>ALSO</literal> rule, nor an <literal>INSTEAD</literal> rule
+      that expands to multiple statements.
      </para>
     </listitem>
    </varlistentry>
diff --git a/doc/src/sgml/ref/merge.sgml b/doc/src/sgml/ref/merge.sgml
new file mode 100644
index 7a8ac40..3d8c792
--- a/doc/src/sgml/ref/merge.sgml
+++ b/doc/src/sgml/ref/merge.sgml
@@ -25,6 +25,7 @@ PostgreSQL documentation
 MERGE INTO [ ONLY ] <replaceable class="parameter">target_table_name</replaceable> [ * ] [ [ AS ] <replaceable class="parameter">target_alias</replaceable> ]
 USING <replaceable class="parameter">data_source</replaceable> ON <replaceable class="parameter">join_condition</replaceable>
 <replaceable class="parameter">when_clause</replaceable> [...]
+[ RETURNING * | <replaceable class="parameter">output_expression</replaceable> [ [ AS ] <replaceable class="parameter">output_name</replaceable> ] [, ...] ]
 
 <phrase>where <replaceable class="parameter">data_source</replaceable> is:</phrase>
 
@@ -95,6 +96,18 @@ DELETE
   </para>
 
   <para>
+   The optional <literal>RETURNING</literal> clause causes <command>MERGE</command>
+   to compute and return value(s) based on each row inserted, updated, or
+   deleted.  Any expression using the source or target table's columns, or
+   the <xref linkend="merge_action"/> function can be computed.  When an
+   <command>INSERT</command> or <command>UPDATE</command> action is performed,
+   the new values of the target table's columns are used.  When a
+   <command>DELETE</command> is performed, the old values of the target table's
+   columns are used.  The syntax of the <literal>RETURNING</literal> list is
+   identical to that of the output list of <command>SELECT</command>.
+  </para>
+
+  <para>
    There is no separate <literal>MERGE</literal> privilege.
    If you specify an update action, you must have the
    <literal>UPDATE</literal> privilege on the column(s)
@@ -129,6 +142,18 @@ DELETE
 
   <variablelist>
    <varlistentry>
+    <term><replaceable class="parameter">with_query</replaceable></term>
+    <listitem>
+     <para>
+      The <literal>WITH</literal> clause allows you to specify one or more
+      subqueries that can be referenced by name in the <command>MERGE</command>
+      query. See <xref linkend="queries-with"/> and <xref linkend="sql-select"/>
+      for details.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
     <term><replaceable class="parameter">target_table_name</replaceable></term>
     <listitem>
      <para>
@@ -411,13 +436,30 @@ DELETE
    </varlistentry>
 
    <varlistentry>
-    <term><replaceable class="parameter">with_query</replaceable></term>
+    <term><replaceable class="parameter">output_expression</replaceable></term>
     <listitem>
      <para>
-      The <literal>WITH</literal> clause allows you to specify one or more
-      subqueries that can be referenced by name in the <command>MERGE</command>
-      query. See <xref linkend="queries-with"/> and <xref linkend="sql-select"/>
-      for details.
+      An expression to be computed and returned by the <command>MERGE</command>
+      command after each row is merged.  The expression can use any columns of
+      the source or target tables, or the <xref linkend="merge_action"/>
+      function to return additional information about the action executed.
+     </para>
+     <para>
+      Writing <literal>*</literal> will return all columns from the source
+      table, followed by all columns from the target table.  Often this will
+      lead to a lot of duplication, since it is common for the source and
+      target tables to have a lot of the same columns.  This can be avoided by
+      qualifying the <literal>*</literal> with the name or alias of the source
+      or target table.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
+    <term><replaceable class="parameter">output_name</replaceable></term>
+    <listitem>
+     <para>
+      A name to use for a returned column.
      </para>
     </listitem>
    </varlistentry>
@@ -440,6 +482,13 @@ MERGE <replaceable class="parameter">tot
    were changed in any way.
   </para>
 
+  <para>
+   If the <command>MERGE</command> command contains a <literal>RETURNING</literal>
+   clause, the result will be similar to that of a <command>SELECT</command>
+   statement containing the columns and values defined in the
+   <literal>RETURNING</literal> list, computed over the row(s) inserted, updated,
+   or deleted by the command.
+  </para>
  </refsect1>
 
  <refsect1>
@@ -560,13 +609,6 @@ MERGE <replaceable class="parameter">tot
   </para>
 
   <para>
-   There is no <literal>RETURNING</literal> clause with
-   <command>MERGE</command>.  Actions of <command>INSERT</command>,
-   <command>UPDATE</command> and <command>DELETE</command> cannot contain
-   <literal>RETURNING</literal> or <literal>WITH</literal> clauses.
-  </para>
-
-  <para>
    When <command>MERGE</command> is run concurrently with other commands
    that modify the target table, the usual transaction isolation rules
    apply; see <xref linkend="transaction-iso"/> for an explanation
@@ -618,7 +660,8 @@ WHEN NOT MATCHED THEN
   <para>
    Attempt to insert a new stock item along with the quantity of stock. If
    the item already exists, instead update the stock count of the existing
-   item. Don't allow entries that have zero stock.
+   item. Don't allow entries that have zero stock. Return details of all
+   changes made.
 <programlisting>
 MERGE INTO wines w
 USING wine_stock_changes s
@@ -628,7 +671,8 @@ WHEN NOT MATCHED AND s.stock_delta > 0 T
 WHEN MATCHED AND w.stock + s.stock_delta > 0 THEN
   UPDATE SET stock = w.stock + s.stock_delta
 WHEN MATCHED THEN
-  DELETE;
+  DELETE
+RETURNING merge_action(), w.*;
 </programlisting>
 
    The <literal>wine_stock_changes</literal> table might be, for example, a
@@ -643,8 +687,9 @@ WHEN MATCHED THEN
     This command conforms to the <acronym>SQL</acronym> standard.
   </para>
    <para>
-    The <literal>WITH</literal> clause and <literal>DO NOTHING</literal>
-    action are extensions to the <acronym>SQL</acronym> standard.
+    The <literal>WITH</literal> clause, <literal>DO NOTHING</literal> action,
+    and <literal>RETURNING</literal> clause are extensions to the
+    <acronym>SQL</acronym> standard.
   </para>
  </refsect1>
 </refentry>
diff --git a/doc/src/sgml/ref/select.sgml b/doc/src/sgml/ref/select.sgml
new file mode 100644
index 9917df7..066aed4
--- a/doc/src/sgml/ref/select.sgml
+++ b/doc/src/sgml/ref/select.sgml
@@ -74,7 +74,7 @@ SELECT [ ALL | DISTINCT [ ON ( <replacea
 
 <phrase>and <replaceable class="parameter">with_query</replaceable> is:</phrase>
 
-    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> )
+    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> | <replaceable class="parameter">merge</replaceable> )
         [ SEARCH { BREADTH | DEPTH } FIRST BY <replaceable>column_name</replaceable> [, ...] SET <replaceable>search_seq_col_name</replaceable> ]
         [ CYCLE <replaceable>column_name</replaceable> [, ...] SET <replaceable>cycle_mark_col_name</replaceable> [ TO <replaceable>cycle_mark_value</replaceable> DEFAULT <replaceable>cycle_mark_default</replaceable> ] USING <replaceable>cycle_path_col_name</replaceable> ]
 
@@ -230,10 +230,10 @@ TABLE [ ONLY ] <replaceable class="param
     The subqueries effectively act as temporary tables or views
     for the duration of the primary query.
     Each subquery can be a <command>SELECT</command>, <command>TABLE</command>, <command>VALUES</command>,
-    <command>INSERT</command>, <command>UPDATE</command> or
-    <command>DELETE</command> statement.
+    <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command> statement.
     When writing a data-modifying statement (<command>INSERT</command>,
-    <command>UPDATE</command> or <command>DELETE</command>) in
+    <command>UPDATE</command>, <command>DELETE</command>, or <command>MERGE</command>) in
     <literal>WITH</literal>, it is usual to include a <literal>RETURNING</literal> clause.
     It is the output of <literal>RETURNING</literal>, <emphasis>not</emphasis> the underlying
     table that the statement modifies, that forms the temporary table that is
@@ -2184,7 +2184,8 @@ SELECT 2+2;
 
    <para>
     <productname>PostgreSQL</productname> allows <command>INSERT</command>,
-    <command>UPDATE</command>, and <command>DELETE</command> to be used as <literal>WITH</literal>
+    <command>UPDATE</command>, <command>DELETE</command>, and
+    <command>MERGE</command> to be used as <literal>WITH</literal>
     queries.  This is not found in the SQL standard.
    </para>
   </refsect2>
diff --git a/doc/src/sgml/rowtypes.sgml b/doc/src/sgml/rowtypes.sgml
new file mode 100644
index 4d86f97..bbeac84
--- a/doc/src/sgml/rowtypes.sgml
+++ b/doc/src/sgml/rowtypes.sgml
@@ -348,7 +348,7 @@ SELECT m.* FROM some_table, LATERAL myfu
    column expansion of this kind when it appears at the top level of
    a <link linkend="queries-select-lists"><command>SELECT</command> output
    list</link>, a <link linkend="dml-returning"><literal>RETURNING</literal>
-   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>,
+   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>,
    a <link linkend="queries-values"><literal>VALUES</literal> clause</link>, or
    a <link linkend="sql-syntax-row-constructors">row constructor</link>.
    In all other contexts (including when nested inside one of those
diff --git a/doc/src/sgml/spi.sgml b/doc/src/sgml/spi.sgml
new file mode 100644
index 47f4b5b..bb37786
--- a/doc/src/sgml/spi.sgml
+++ b/doc/src/sgml/spi.sgml
@@ -301,8 +301,9 @@ SPI_execute("INSERT INTO foo SELECT * FR
    is returned in the global variable <varname>SPI_processed</varname>.
    If the return value of the function is <symbol>SPI_OK_SELECT</symbol>,
    <symbol>SPI_OK_INSERT_RETURNING</symbol>,
-   <symbol>SPI_OK_DELETE_RETURNING</symbol>, or
-   <symbol>SPI_OK_UPDATE_RETURNING</symbol>,
+   <symbol>SPI_OK_DELETE_RETURNING</symbol>,
+   <symbol>SPI_OK_UPDATE_RETURNING</symbol>, or
+   <symbol>SPI_OK_MERGE_RETURNING</symbol>,
    then you can use the
    global pointer <literal>SPITupleTable *SPI_tuptable</literal> to
    access the result rows.  Some utility commands (such as
@@ -472,6 +473,15 @@ typedef struct SPITupleTable
       </para>
      </listitem>
     </varlistentry>
+
+    <varlistentry>
+     <term><symbol>SPI_OK_MERGE_RETURNING</symbol></term>
+     <listitem>
+      <para>
+       if a <command>MERGE RETURNING</command> was executed
+      </para>
+     </listitem>
+    </varlistentry>
 
     <varlistentry>
      <term><symbol>SPI_OK_UTILITY</symbol></term>
diff --git a/doc/src/sgml/xfunc.sgml b/doc/src/sgml/xfunc.sgml
new file mode 100644
index 8a79ad0..7d05369
--- a/doc/src/sgml/xfunc.sgml
+++ b/doc/src/sgml/xfunc.sgml
@@ -177,7 +177,8 @@
     statements separated by semicolons.  A semicolon after the last
     statement is optional.  Unless the function is declared to return
     <type>void</type>, the last statement must be a <command>SELECT</command>,
-    or an <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>
+    or an <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command>
     that has a <literal>RETURNING</literal> clause.
    </para>
 
@@ -1236,8 +1237,9 @@ SELECT x, CASE WHEN x &gt; 0 THEN genera
 
     <note>
      <para>
-      If a function's last command is <command>INSERT</command>, <command>UPDATE</command>,
-      or <command>DELETE</command> with <literal>RETURNING</literal>, that command will
+      If a function's last command is <command>INSERT</command>,
+      <command>UPDATE</command>, <command>DELETE</command>, or
+      <command>MERGE</command> with <literal>RETURNING</literal>, that command will
       always be executed to completion, even if the function is not declared
       with <literal>SETOF</literal> or the calling query does not fetch all the
       result rows.  Any extra rows produced by the <literal>RETURNING</literal>
diff --git a/src/backend/commands/copy.c b/src/backend/commands/copy.c
new file mode 100644
index 056b673..28cf8b0
--- a/src/backend/commands/copy.c
+++ b/src/backend/commands/copy.c
@@ -281,12 +281,6 @@ DoCopy(ParseState *pstate, const CopyStm
 	{
 		Assert(stmt->query);
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(stmt->query, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in COPY"));
-
 		query = makeNode(RawStmt);
 		query->stmt = stmt->query;
 		query->stmt_location = stmt_location;
diff --git a/src/backend/commands/copyto.c b/src/backend/commands/copyto.c
new file mode 100644
index a6962e0..ae8b2e3
--- a/src/backend/commands/copyto.c
+++ b/src/backend/commands/copyto.c
@@ -503,7 +503,8 @@ BeginCopyTo(ParseState *pstate,
 		{
 			Assert(query->commandType == CMD_INSERT ||
 				   query->commandType == CMD_UPDATE ||
-				   query->commandType == CMD_DELETE);
+				   query->commandType == CMD_DELETE ||
+				   query->commandType == CMD_MERGE);
 
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
diff --git a/src/backend/executor/execExpr.c b/src/backend/executor/execExpr.c
new file mode 100644
index ffd3ca4..728c8d5
--- a/src/backend/executor/execExpr.c
+++ b/src/backend/executor/execExpr.c
@@ -1107,6 +1107,19 @@ ExecInitExprRec(Expr *node, ExprState *s
 				break;
 			}
 
+		case T_MergeSupportFunc:
+			{
+				/* must be in a MERGE, else something messed up */
+				if (!state->parent ||
+					!IsA(state->parent, ModifyTableState) ||
+					((ModifyTableState *) state->parent)->operation != CMD_MERGE)
+					elog(ERROR, "MergeSupportFunc found in non-merge plan node");
+
+				scratch.opcode = EEOP_MERGE_SUPPORT_FUNC;
+				ExprEvalPushStep(state, &scratch);
+				break;
+			}
+
 		case T_SubscriptingRef:
 			{
 				SubscriptingRef *sbsref = (SubscriptingRef *) node;
diff --git a/src/backend/executor/execExprInterp.c b/src/backend/executor/execExprInterp.c
new file mode 100644
index 7c1f51e..a25ab75
--- a/src/backend/executor/execExprInterp.c
+++ b/src/backend/executor/execExprInterp.c
@@ -484,6 +484,7 @@ ExecInterpExpr(ExprState *state, ExprCon
 		&&CASE_EEOP_AGGREF,
 		&&CASE_EEOP_GROUPING_FUNC,
 		&&CASE_EEOP_WINDOW_FUNC,
+		&&CASE_EEOP_MERGE_SUPPORT_FUNC,
 		&&CASE_EEOP_SUBPLAN,
 		&&CASE_EEOP_AGG_STRICT_DESERIALIZE,
 		&&CASE_EEOP_AGG_DESERIALIZE,
@@ -1592,6 +1593,14 @@ ExecInterpExpr(ExprState *state, ExprCon
 			EEO_NEXT();
 		}
 
+		EEO_CASE(EEOP_MERGE_SUPPORT_FUNC)
+		{
+			/* too complex/uncommon for an inline implementation */
+			ExecEvalMergeSupportFunc(state, op, econtext);
+
+			EEO_NEXT();
+		}
+
 		EEO_CASE(EEOP_SUBPLAN)
 		{
 			/* too complex for an inline implementation */
@@ -4246,6 +4255,45 @@ ExecEvalGroupingFunc(ExprState *state, E
 }
 
 /*
+ * ExecEvalMergeSupportFunc
+ *
+ * Returns information about the current MERGE action for its RETURNING list.
+ */
+void
+ExecEvalMergeSupportFunc(ExprState *state, ExprEvalStep *op,
+						 ExprContext *econtext)
+{
+	ModifyTableState *mtstate = castNode(ModifyTableState, state->parent);
+	MergeActionState *relaction = mtstate->mt_merge_action;
+
+	if (!relaction)
+		elog(ERROR, "no merge action in progress");
+
+	/* Return the MERGE action ("INSERT", "UPDATE", or "DELETE") */
+	switch (relaction->mas_action->commandType)
+	{
+		case CMD_INSERT:
+			*op->resvalue = PointerGetDatum(cstring_to_text_with_len("INSERT", 6));
+			*op->resnull = false;
+			break;
+		case CMD_UPDATE:
+			*op->resvalue = PointerGetDatum(cstring_to_text_with_len("UPDATE", 6));
+			*op->resnull = false;
+			break;
+		case CMD_DELETE:
+			*op->resvalue = PointerGetDatum(cstring_to_text_with_len("DELETE", 6));
+			*op->resnull = false;
+			break;
+		case CMD_NOTHING:
+			elog(ERROR, "unexpected merge action: DO NOTHING");
+			break;
+		default:
+			elog(ERROR, "unrecognized commandType: %d",
+				 (int) relaction->mas_action->commandType);
+	}
+}
+
+/*
  * Hand off evaluation of a subplan to nodeSubplan.c
  */
 void
diff --git a/src/backend/executor/execPartition.c b/src/backend/executor/execPartition.c
new file mode 100644
index 8ca512d..64fcb01
--- a/src/backend/executor/execPartition.c
+++ b/src/backend/executor/execPartition.c
@@ -609,8 +609,8 @@ ExecInitPartitionInfo(ModifyTableState *
 	 * Build the RETURNING projection for the partition.  Note that we didn't
 	 * build the returningList for partitions within the planner, but simple
 	 * translation of varattnos will suffice.  This only occurs for the INSERT
-	 * case or in the case of UPDATE tuple routing where we didn't find a
-	 * result rel to reuse.
+	 * case or in the case of UPDATE/MERGE tuple routing where we didn't find
+	 * a result rel to reuse.
 	 */
 	if (node && node->returningLists != NIL)
 	{
@@ -619,12 +619,14 @@ ExecInitPartitionInfo(ModifyTableState *
 		List	   *returningList;
 
 		/* See the comment above for WCO lists. */
-		/* (except no RETURNING support for MERGE yet) */
 		Assert((node->operation == CMD_INSERT &&
 				list_length(node->returningLists) == 1 &&
 				list_length(node->resultRelations) == 1) ||
 			   (node->operation == CMD_UPDATE &&
 				list_length(node->returningLists) ==
+				list_length(node->resultRelations)) ||
+			   (node->operation == CMD_MERGE &&
+				list_length(node->returningLists) ==
 				list_length(node->resultRelations)));
 
 		/*
diff --git a/src/backend/executor/functions.c b/src/backend/executor/functions.c
new file mode 100644
index a4b6e1e..f6ccdf0
--- a/src/backend/executor/functions.c
+++ b/src/backend/executor/functions.c
@@ -1660,8 +1660,8 @@ check_sql_fn_retval(List *queryTreeLists
 
 	/*
 	 * If it's a plain SELECT, it returns whatever the targetlist says.
-	 * Otherwise, if it's INSERT/UPDATE/DELETE with RETURNING, it returns
-	 * that. Otherwise, the function return type must be VOID.
+	 * Otherwise, if it's INSERT/UPDATE/DELETE/MERGE with RETURNING, it
+	 * returns that. Otherwise, the function return type must be VOID.
 	 *
 	 * Note: eventually replace this test with QueryReturnsTuples?	We'd need
 	 * a more general method of determining the output type, though.  Also, it
@@ -1679,7 +1679,8 @@ check_sql_fn_retval(List *queryTreeLists
 	else if (parse &&
 			 (parse->commandType == CMD_INSERT ||
 			  parse->commandType == CMD_UPDATE ||
-			  parse->commandType == CMD_DELETE) &&
+			  parse->commandType == CMD_DELETE ||
+			  parse->commandType == CMD_MERGE) &&
 			 parse->returningList)
 	{
 		tlist = parse->returningList;
@@ -1693,7 +1694,7 @@ check_sql_fn_retval(List *queryTreeLists
 				(errcode(ERRCODE_INVALID_FUNCTION_DEFINITION),
 				 errmsg("return type mismatch in function declared to return %s",
 						format_type_be(rettype)),
-				 errdetail("Function's final statement must be SELECT or INSERT/UPDATE/DELETE RETURNING.")));
+				 errdetail("Function's final statement must be SELECT or INSERT/UPDATE/DELETE/MERGE RETURNING.")));
 		return false;			/* keep compiler quiet */
 	}
 
diff --git a/src/backend/executor/nodeModifyTable.c b/src/backend/executor/nodeModifyTable.c
new file mode 100644
index 9351fbc..b8a75c5
--- a/src/backend/executor/nodeModifyTable.c
+++ b/src/backend/executor/nodeModifyTable.c
@@ -36,8 +36,7 @@
  *		RETURNING tuple after completing each row insert, update, or delete.
  *		It must be called again to continue the operation.  Without RETURNING,
  *		we just loop within the node until all the work is done, then
- *		return NULL.  This avoids useless call/return overhead.  (MERGE does
- *		not support RETURNING.)
+ *		return NULL.  This avoids useless call/return overhead.
  */
 
 #include "postgres.h"
@@ -85,9 +84,6 @@ typedef struct ModifyTableContext
 	 */
 	TupleTableSlot *planSlot;
 
-	/* MERGE specific */
-	MergeActionState *relaction;	/* MERGE action in progress */
-
 	/*
 	 * Information about the changes that were made concurrently to a tuple
 	 * being updated or deleted
@@ -150,14 +146,15 @@ static TupleTableSlot *ExecMerge(ModifyT
 								 HeapTuple oldtuple,
 								 bool canSetTag);
 static void ExecInitMerge(ModifyTableState *mtstate, EState *estate);
-static bool ExecMergeMatched(ModifyTableContext *context,
-							 ResultRelInfo *resultRelInfo,
-							 ItemPointer tupleid,
-							 HeapTuple oldtuple,
-							 bool canSetTag);
-static void ExecMergeNotMatched(ModifyTableContext *context,
-								ResultRelInfo *resultRelInfo,
-								bool canSetTag);
+static TupleTableSlot *ExecMergeMatched(ModifyTableContext *context,
+										ResultRelInfo *resultRelInfo,
+										ItemPointer tupleid,
+										HeapTuple oldtuple,
+										bool canSetTag,
+										bool *matched);
+static TupleTableSlot *ExecMergeNotMatched(ModifyTableContext *context,
+										   ResultRelInfo *resultRelInfo,
+										   bool canSetTag);
 
 
 /*
@@ -977,7 +974,7 @@ ExecInsert(ModifyTableContext *context,
 		if (mtstate->operation == CMD_UPDATE)
 			wco_kind = WCO_RLS_UPDATE_CHECK;
 		else if (mtstate->operation == CMD_MERGE)
-			wco_kind = (context->relaction->mas_action->commandType == CMD_UPDATE) ?
+			wco_kind = (mtstate->mt_merge_action->mas_action->commandType == CMD_UPDATE) ?
 				WCO_RLS_UPDATE_CHECK : WCO_RLS_INSERT_CHECK;
 		else
 			wco_kind = WCO_RLS_INSERT_CHECK;
@@ -1831,7 +1828,7 @@ ExecCrossPartitionUpdate(ModifyTableCont
 		 * additional rechecking, and might end up executing a different
 		 * action entirely).
 		 */
-		if (context->relaction != NULL)
+		if (mtstate->operation == CMD_MERGE)
 			return *tmresult == TM_Ok;
 		else if (TupIsNull(epqslot))
 			return true;
@@ -2072,7 +2069,7 @@ lreplace:
 		 * No luck, a retry is needed.  If running MERGE, we do not do so
 		 * here; instead let it handle that on its own rules.
 		 */
-		if (context->relaction != NULL)
+		if (context->mtstate->operation == CMD_MERGE)
 			return result;
 
 		/*
@@ -2713,6 +2710,7 @@ static TupleTableSlot *
 ExecMerge(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 		  ItemPointer tupleid, HeapTuple oldtuple, bool canSetTag)
 {
+	TupleTableSlot *rslot = NULL;
 	bool		matched;
 
 	/*-----
@@ -2761,19 +2759,18 @@ ExecMerge(ModifyTableContext *context, R
 	 */
 	matched = tupleid != NULL || oldtuple != NULL;
 	if (matched)
-		matched = ExecMergeMatched(context, resultRelInfo, tupleid, oldtuple,
-								   canSetTag);
+		rslot = ExecMergeMatched(context, resultRelInfo, tupleid, oldtuple,
+								 canSetTag, &matched);
 
 	/*
-	 * Either we were dealing with a NOT MATCHED tuple or ExecMergeMatched()
-	 * returned "false", indicating the previously MATCHED tuple no longer
-	 * matches.
+	 * Deal with the NOT MATCHED case (either a NOT MATCHED tuple from the
+	 * join, or a previously MATCHED tuple for which ExecMergeMatched() set
+	 * "matched" to false, indicating that it no longer matches).
 	 */
 	if (!matched)
-		ExecMergeNotMatched(context, resultRelInfo, canSetTag);
+		rslot = ExecMergeNotMatched(context, resultRelInfo, canSetTag);
 
-	/* No RETURNING support yet */
-	return NULL;
+	return rslot;
 }
 
 /*
@@ -2785,8 +2782,8 @@ ExecMerge(ModifyTableContext *context, R
  * We start from the first WHEN MATCHED action and check if the WHEN quals
  * pass, if any. If the WHEN quals for the first action do not pass, we
  * check the second, then the third and so on. If we reach to the end, no
- * action is taken and we return true, indicating that no further action is
- * required for this tuple.
+ * action is taken and "matched" is set to true, indicating that no further
+ * action is required for this tuple.
  *
  * If we do find a qualifying action, then we attempt to execute the action.
  *
@@ -2795,16 +2792,18 @@ ExecMerge(ModifyTableContext *context, R
  * with individual actions are evaluated by this routine via ExecQual, while
  * EvalPlanQual checks for the join quals. If EvalPlanQual tells us that the
  * updated tuple still passes the join quals, then we restart from the first
- * action to look for a qualifying action. Otherwise, we return false --
- * meaning that a NOT MATCHED action must now be executed for the current
- * source tuple.
+ * action to look for a qualifying action. Otherwise, "matched" is set to
+ * false -- meaning that a NOT MATCHED action must now be executed for the
+ * current source tuple.
  */
-static bool
+static TupleTableSlot *
 ExecMergeMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
-				 ItemPointer tupleid, HeapTuple oldtuple, bool canSetTag)
+				 ItemPointer tupleid, HeapTuple oldtuple, bool canSetTag,
+				 bool *matched)
 {
 	ModifyTableState *mtstate = context->mtstate;
-	TupleTableSlot *newslot;
+	TupleTableSlot *rslot = NULL;
+	TupleTableSlot *newslot = NULL;
 	EState	   *estate = context->estate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
 	bool		isNull;
@@ -2815,7 +2814,10 @@ ExecMergeMatched(ModifyTableContext *con
 	 * If there are no WHEN MATCHED actions, we are done.
 	 */
 	if (resultRelInfo->ri_matchedMergeAction == NIL)
-		return true;
+	{
+		*matched = true;
+		return NULL;
+	}
 
 	/*
 	 * Make tuple and any needed join variables available to ExecQual and
@@ -2905,12 +2907,15 @@ lmerge_matched:
 				 */
 				newslot = ExecProject(relaction->mas_proj);
 
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				if (!ExecUpdatePrologue(context, resultRelInfo,
 										tupleid, NULL, newslot, &result))
 				{
 					if (result == TM_Ok)
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 					break;		/* concurrent update/delete */
 				}
 
@@ -2920,7 +2925,10 @@ lmerge_matched:
 				{
 					if (!ExecIRUpdateTriggers(estate, resultRelInfo,
 											  oldtuple, newslot))
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 				}
 				else
 				{
@@ -2933,12 +2941,15 @@ lmerge_matched:
 					 * cross-partition update was done, then there's nothing
 					 * else for us to do --- the UPDATE has been turned into a
 					 * DELETE and an INSERT, and we must not perform any of
-					 * the usual post-update tasks.
+					 * the usual post-update tasks.  Also, the RETURNING tuple
+					 * (if any) has been projected, so we can just return
+					 * that.
 					 */
 					if (updateCxt.crossPartUpdate)
 					{
 						mtstate->mt_merge_updated += 1;
-						return true;
+						*matched = true;
+						return context->cpUpdateReturningSlot;
 					}
 				}
 
@@ -2951,12 +2962,15 @@ lmerge_matched:
 				break;
 
 			case CMD_DELETE:
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				if (!ExecDeletePrologue(context, resultRelInfo, tupleid,
 										NULL, NULL, &result))
 				{
 					if (result == TM_Ok)
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 					break;		/* concurrent update/delete */
 				}
 
@@ -2966,7 +2980,10 @@ lmerge_matched:
 				{
 					if (!ExecIRDeleteTriggers(estate, resultRelInfo,
 											  oldtuple))
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 				}
 				else
 					result = ExecDeleteAct(context, resultRelInfo, tupleid,
@@ -3046,7 +3063,8 @@ lmerge_matched:
 				 * If the tuple was already deleted, return to let caller
 				 * handle it under NOT MATCHED clauses.
 				 */
-				return false;
+				*matched = false;
+				return NULL;
 
 			case TM_Updated:
 				{
@@ -3092,13 +3110,19 @@ lmerge_matched:
 							 * NOT MATCHED actions.
 							 */
 							if (TupIsNull(epqslot))
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							(void) ExecGetJunkAttribute(epqslot,
 														resultRelInfo->ri_RowIdAttNo,
 														&isNull);
 							if (isNull)
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							/*
 							 * When a tuple was updated and migrated to
@@ -3133,7 +3157,8 @@ lmerge_matched:
 							 * tuple already deleted; tell caller to run NOT
 							 * MATCHED actions
 							 */
-							return false;
+							*matched = false;
+							return NULL;
 
 						case TM_SelfModified:
 
@@ -3161,13 +3186,13 @@ lmerge_matched:
 
 							/* This shouldn't happen */
 							elog(ERROR, "attempted to update or delete invisible tuple");
-							return false;
+							return NULL;
 
 						default:
 							/* see table_tuple_lock call in ExecDelete() */
 							elog(ERROR, "unexpected table_tuple_lock status: %u",
 								 result);
-							return false;
+							return NULL;
 					}
 				}
 
@@ -3179,6 +3204,31 @@ lmerge_matched:
 				break;
 		}
 
+		/* Process RETURNING if present */
+		if (resultRelInfo->ri_projectReturning)
+		{
+			switch (commandType)
+			{
+				case CMD_UPDATE:
+					rslot = ExecProcessReturning(resultRelInfo, newslot,
+												 context->planSlot);
+					break;
+
+				case CMD_DELETE:
+					rslot = ExecProcessReturning(resultRelInfo,
+												 resultRelInfo->ri_oldTupleSlot,
+												 context->planSlot);
+					break;
+
+				case CMD_NOTHING:
+					break;
+
+				default:
+					elog(ERROR, "unrecognized commandType: %d",
+						 (int) commandType);
+			}
+		}
+
 		/*
 		 * We've activated one of the WHEN clauses, so we don't search
 		 * further. This is required behaviour, not an optimization.
@@ -3189,19 +3239,22 @@ lmerge_matched:
 	/*
 	 * Successfully executed an action or no qualifying action was found.
 	 */
-	return true;
+	*matched = true;
+
+	return rslot;
 }
 
 /*
  * Execute the first qualifying NOT MATCHED action.
  */
-static void
+static TupleTableSlot *
 ExecMergeNotMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 					bool canSetTag)
 {
 	ModifyTableState *mtstate = context->mtstate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
 	List	   *actionStates = NIL;
+	TupleTableSlot *rslot = NULL;
 	ListCell   *l;
 
 	/*
@@ -3251,10 +3304,10 @@ ExecMergeNotMatched(ModifyTableContext *
 				 * so we don't need to map the tuple here.
 				 */
 				newslot = ExecProject(action->mas_proj);
-				context->relaction = action;
+				mtstate->mt_merge_action = action;
 
-				(void) ExecInsert(context, mtstate->rootResultRelInfo, newslot,
-								  canSetTag, NULL, NULL);
+				rslot = ExecInsert(context, mtstate->rootResultRelInfo,
+								   newslot, canSetTag, NULL, NULL);
 				mtstate->mt_merge_inserted += 1;
 				break;
 			case CMD_NOTHING:
@@ -3270,6 +3323,8 @@ ExecMergeNotMatched(ModifyTableContext *
 		 */
 		break;
 	}
+
+	return rslot;
 }
 
 /*
@@ -3732,9 +3787,17 @@ ExecModifyTable(PlanState *pstate)
 				{
 					EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-					ExecMerge(&context, node->resultRelInfo, NULL, NULL,
-							  node->canSetTag);
-					continue;	/* no RETURNING support yet */
+					slot = ExecMerge(&context, node->resultRelInfo,
+									 NULL, NULL, node->canSetTag);
+
+					/*
+					 * If we got a RETURNING result, return it to the caller.
+					 * We'll continue the work on next call.
+					 */
+					if (slot)
+						return slot;
+
+					continue;	/* continue with the next tuple */
 				}
 
 				elog(ERROR, "tableoid is NULL");
@@ -3811,9 +3874,17 @@ ExecModifyTable(PlanState *pstate)
 					{
 						EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-						ExecMerge(&context, node->resultRelInfo, NULL, NULL,
-								  node->canSetTag);
-						continue;	/* no RETURNING support yet */
+						slot = ExecMerge(&context, node->resultRelInfo,
+										 NULL, NULL, node->canSetTag);
+
+						/*
+						 * If we got a RETURNING result, return it to the
+						 * caller.  We'll continue the work on next call.
+						 */
+						if (slot)
+							return slot;
+
+						continue;	/* continue with the next tuple */
 					}
 
 					elog(ERROR, "ctid is NULL");
@@ -3860,9 +3931,17 @@ ExecModifyTable(PlanState *pstate)
 					{
 						EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-						ExecMerge(&context, node->resultRelInfo, NULL, NULL,
-								  node->canSetTag);
-						continue;	/* no RETURNING support yet */
+						slot = ExecMerge(&context, node->resultRelInfo,
+										 NULL, NULL, node->canSetTag);
+
+						/*
+						 * If we got a RETURNING result, return it to the
+						 * caller.  We'll continue the work on next call.
+						 */
+						if (slot)
+							return slot;
+
+						continue;	/* continue with the next tuple */
 					}
 
 					elog(ERROR, "wholerow is NULL");
@@ -3924,7 +4003,6 @@ ExecModifyTable(PlanState *pstate)
 				}
 				slot = ExecGetUpdateNewTuple(resultRelInfo, context.planSlot,
 											 oldSlot);
-				context.relaction = NULL;
 
 				/* Now apply the update. */
 				slot = ExecUpdate(&context, resultRelInfo, tupleid, oldtuple,
diff --git a/src/backend/executor/spi.c b/src/backend/executor/spi.c
new file mode 100644
index 8585771..a97a7e3
--- a/src/backend/executor/spi.c
+++ b/src/backend/executor/spi.c
@@ -2032,6 +2032,8 @@ SPI_result_code_string(int code)
 			return "SPI_OK_TD_REGISTER";
 		case SPI_OK_MERGE:
 			return "SPI_OK_MERGE";
+		case SPI_OK_MERGE_RETURNING:
+			return "SPI_OK_MERGE_RETURNING";
 	}
 	/* Unrecognized code ... return something useful ... */
 	sprintf(buf, "Unrecognized SPI code %d", code);
@@ -2885,7 +2887,10 @@ _SPI_pquery(QueryDesc *queryDesc, bool f
 				res = SPI_OK_UPDATE;
 			break;
 		case CMD_MERGE:
-			res = SPI_OK_MERGE;
+			if (queryDesc->plannedstmt->hasReturning)
+				res = SPI_OK_MERGE_RETURNING;
+			else
+				res = SPI_OK_MERGE;
 			break;
 		default:
 			return SPI_ERROR_OPUNKNOWN;
diff --git a/src/backend/jit/llvm/llvmjit_expr.c b/src/backend/jit/llvm/llvmjit_expr.c
new file mode 100644
index 0c44842..2a7d84f
--- a/src/backend/jit/llvm/llvmjit_expr.c
+++ b/src/backend/jit/llvm/llvmjit_expr.c
@@ -1986,6 +1986,12 @@ llvm_compile_expr(ExprState *state)
 					break;
 				}
 
+			case EEOP_MERGE_SUPPORT_FUNC:
+				build_EvalXFunc(b, mod, "ExecEvalMergeSupportFunc",
+								v_state, op, v_econtext);
+				LLVMBuildBr(b, opblocks[opno + 1]);
+				break;
+
 			case EEOP_SUBPLAN:
 				build_EvalXFunc(b, mod, "ExecEvalSubPlan",
 								v_state, op, v_econtext);
diff --git a/src/backend/jit/llvm/llvmjit_types.c b/src/backend/jit/llvm/llvmjit_types.c
new file mode 100644
index 47c9daf..7d7aeee
--- a/src/backend/jit/llvm/llvmjit_types.c
+++ b/src/backend/jit/llvm/llvmjit_types.c
@@ -155,6 +155,7 @@ void	   *referenced_functions[] =
 	ExecEvalFuncExprFusage,
 	ExecEvalFuncExprStrictFusage,
 	ExecEvalGroupingFunc,
+	ExecEvalMergeSupportFunc,
 	ExecEvalMinMax,
 	ExecEvalNextValueExpr,
 	ExecEvalParamExec,
diff --git a/src/backend/nodes/nodeFuncs.c b/src/backend/nodes/nodeFuncs.c
new file mode 100644
index 6ba8e73..5b70280
--- a/src/backend/nodes/nodeFuncs.c
+++ b/src/backend/nodes/nodeFuncs.c
@@ -66,6 +66,9 @@ exprType(const Node *expr)
 		case T_WindowFunc:
 			type = ((const WindowFunc *) expr)->wintype;
 			break;
+		case T_MergeSupportFunc:
+			type = ((const MergeSupportFunc *) expr)->msftype;
+			break;
 		case T_SubscriptingRef:
 			type = ((const SubscriptingRef *) expr)->refrestype;
 			break;
@@ -809,6 +812,9 @@ exprCollation(const Node *expr)
 		case T_WindowFunc:
 			coll = ((const WindowFunc *) expr)->wincollid;
 			break;
+		case T_MergeSupportFunc:
+			coll = ((const MergeSupportFunc *) expr)->msfcollid;
+			break;
 		case T_SubscriptingRef:
 			coll = ((const SubscriptingRef *) expr)->refcollid;
 			break;
@@ -1084,6 +1090,9 @@ exprSetCollation(Node *expr, Oid collati
 		case T_WindowFunc:
 			((WindowFunc *) expr)->wincollid = collation;
 			break;
+		case T_MergeSupportFunc:
+			((MergeSupportFunc *) expr)->msfcollid = collation;
+			break;
 		case T_SubscriptingRef:
 			((SubscriptingRef *) expr)->refcollid = collation;
 			break;
@@ -1342,6 +1351,9 @@ exprLocation(const Node *expr)
 			/* function name should always be the first thing */
 			loc = ((const WindowFunc *) expr)->location;
 			break;
+		case T_MergeSupportFunc:
+			loc = ((const MergeSupportFunc *) expr)->location;
+			break;
 		case T_SubscriptingRef:
 			/* just use container argument's location */
 			loc = exprLocation((Node *) ((const SubscriptingRef *) expr)->refexpr);
@@ -2034,6 +2046,7 @@ expression_tree_walker_impl(Node *node,
 		case T_RangeTblRef:
 		case T_SortGroupClause:
 		case T_CTESearchClause:
+		case T_MergeSupportFunc:
 			/* primitive node types with no expression subnodes */
 			break;
 		case T_WithCheckOption:
@@ -2868,6 +2881,7 @@ expression_tree_mutator_impl(Node *node,
 		case T_RangeTblRef:
 		case T_SortGroupClause:
 		case T_CTESearchClause:
+		case T_MergeSupportFunc:
 			return (Node *) copyObject(node);
 		case T_WithCheckOption:
 			{
@@ -3832,6 +3846,7 @@ raw_expression_tree_walker_impl(Node *no
 		case T_ParamRef:
 		case T_A_Const:
 		case T_A_Star:
+		case T_MergeSupportFunc:
 			/* primitive node types with no subnodes */
 			break;
 		case T_Alias:
@@ -4052,6 +4067,8 @@ raw_expression_tree_walker_impl(Node *no
 					return true;
 				if (WALK(stmt->mergeWhenClauses))
 					return true;
+				if (WALK(stmt->returningList))
+					return true;
 				if (WALK(stmt->withClause))
 					return true;
 			}
diff --git a/src/backend/optimizer/plan/subselect.c b/src/backend/optimizer/plan/subselect.c
new file mode 100644
index 47e1472..b9e0c96
--- a/src/backend/optimizer/plan/subselect.c
+++ b/src/backend/optimizer/plan/subselect.c
@@ -1855,7 +1855,8 @@ convert_EXISTS_to_ANY(PlannerInfo *root,
 /*
  * Replace correlation vars (uplevel vars) with Params.
  *
- * Uplevel PlaceHolderVars and aggregates are replaced, too.
+ * Uplevel PlaceHolderVars, aggregates, GROUPING() expressions, and
+ * MergeSupportFuncs are replaced, too.
  *
  * Note: it is critical that this runs immediately after SS_process_sublinks.
  * Since we do not recurse into the arguments of uplevel PHVs and aggregates,
@@ -1909,6 +1910,12 @@ replace_correlation_vars_mutator(Node *n
 		if (((GroupingFunc *) node)->agglevelsup > 0)
 			return (Node *) replace_outer_grouping(root, (GroupingFunc *) node);
 	}
+	if (IsA(node, MergeSupportFunc))
+	{
+		if (root->parse->commandType != CMD_MERGE)
+			return (Node *) replace_outer_merge_support(root,
+														(MergeSupportFunc *) node);
+	}
 	return expression_tree_mutator(node,
 								   replace_correlation_vars_mutator,
 								   (void *) root);
diff --git a/src/backend/optimizer/util/paramassign.c b/src/backend/optimizer/util/paramassign.c
new file mode 100644
index a58da7c..f461fed
--- a/src/backend/optimizer/util/paramassign.c
+++ b/src/backend/optimizer/util/paramassign.c
@@ -308,6 +308,57 @@ replace_outer_grouping(PlannerInfo *root
 }
 
 /*
+ * Generate a Param node to replace the given MergeSupportFunc expression
+ * which is expected to be in the RETURNING list of an upper-level MERGE
+ * query.  Record the need for the MergeSupportFunc in the proper upper-level
+ * root->plan_params.
+ */
+Param *
+replace_outer_merge_support(PlannerInfo *root, MergeSupportFunc *msf)
+{
+	Param	   *retval;
+	PlannerParamItem *pitem;
+	Oid			ptype = exprType((Node *) msf);
+
+	Assert(root->parse->commandType != CMD_MERGE);
+
+	/*
+	 * The parser should have ensured that the MergeSupportFunc is in the
+	 * RETURNING list of an upper-level MERGE query, so find that query.
+	 */
+	do
+	{
+		root = root->parent_root;
+		if (root == NULL)
+			elog(ERROR, "MergeSupportFunc found outside MERGE");
+	} while (root->parse->commandType != CMD_MERGE);
+
+	/*
+	 * It does not seem worthwhile to try to de-duplicate references to outer
+	 * MergeSupportFunc expressions.  Just make a new slot every time.
+	 */
+	msf = copyObject(msf);
+
+	pitem = makeNode(PlannerParamItem);
+	pitem->item = (Node *) msf;
+	pitem->paramId = list_length(root->glob->paramExecTypes);
+	root->glob->paramExecTypes = lappend_oid(root->glob->paramExecTypes,
+											 ptype);
+
+	root->plan_params = lappend(root->plan_params, pitem);
+
+	retval = makeNode(Param);
+	retval->paramkind = PARAM_EXEC;
+	retval->paramid = pitem->paramId;
+	retval->paramtype = ptype;
+	retval->paramtypmod = -1;
+	retval->paramcollid = InvalidOid;
+	retval->location = msf->location;
+
+	return retval;
+}
+
+/*
  * Generate a Param node to replace the given Var,
  * which is expected to come from some upper NestLoop plan node.
  * Record the need for the Var in root->curOuterParams.
diff --git a/src/backend/parser/analyze.c b/src/backend/parser/analyze.c
new file mode 100644
index 2255314..7bd947b
--- a/src/backend/parser/analyze.c
+++ b/src/backend/parser/analyze.c
@@ -72,7 +72,6 @@ static void determineRecursiveColTypes(P
 									   Node *larg, List *nrtargetlist);
 static Query *transformReturnStmt(ParseState *pstate, ReturnStmt *stmt);
 static Query *transformUpdateStmt(ParseState *pstate, UpdateStmt *stmt);
-static List *transformReturningList(ParseState *pstate, List *returningList);
 static Query *transformPLAssignStmt(ParseState *pstate,
 									PLAssignStmt *stmt);
 static Query *transformDeclareCursorStmt(ParseState *pstate,
@@ -551,7 +550,8 @@ transformDeleteStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -978,7 +978,8 @@ transformInsertStmt(ParseState *pstate,
 	/* Process RETURNING, if any. */
 	if (stmt->returningList)
 		qry->returningList = transformReturningList(pstate,
-													stmt->returningList);
+													stmt->returningList,
+													EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -2443,7 +2444,8 @@ transformUpdateStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/*
 	 * Now we are done with SELECT-like processing, and can get on with
@@ -2540,10 +2542,11 @@ transformUpdateTargetList(ParseState *ps
 
 /*
  * transformReturningList -
- *	handle a RETURNING clause in INSERT/UPDATE/DELETE
+ *	handle a RETURNING clause in INSERT/UPDATE/DELETE/MERGE
  */
-static List *
-transformReturningList(ParseState *pstate, List *returningList)
+List *
+transformReturningList(ParseState *pstate, List *returningList,
+					   ParseExprKind exprKind)
 {
 	List	   *rlist;
 	int			save_next_resno;
@@ -2560,7 +2563,7 @@ transformReturningList(ParseState *pstat
 	pstate->p_next_resno = 1;
 
 	/* transform RETURNING identically to a SELECT targetlist */
-	rlist = transformTargetList(pstate, returningList, EXPR_KIND_RETURNING);
+	rlist = transformTargetList(pstate, returningList, exprKind);
 
 	/*
 	 * Complain if the nonempty tlist expanded to nothing (which is possible
diff --git a/src/backend/parser/gram.y b/src/backend/parser/gram.y
new file mode 100644
index c6e2f67..6419b32
--- a/src/backend/parser/gram.y
+++ b/src/backend/parser/gram.y
@@ -733,7 +733,7 @@ static Node *makeRecursiveViewSelect(cha
 	LEADING LEAKPROOF LEAST LEFT LEVEL LIKE LIMIT LISTEN LOAD LOCAL
 	LOCALTIME LOCALTIMESTAMP LOCATION LOCK_P LOCKED LOGGED
 
-	MAPPING MATCH MATCHED MATERIALIZED MAXVALUE MERGE METHOD
+	MAPPING MATCH MATCHED MATERIALIZED MAXVALUE MERGE MERGE_ACTION METHOD
 	MINUTE_P MINVALUE MODE MONTH_P MOVE
 
 	NAME_P NAMES NATIONAL NATURAL NCHAR NEW NEXT NFC NFD NFKC NFKD NO NONE
@@ -12374,6 +12374,7 @@ MergeStmt:
 			USING table_ref
 			ON a_expr
 			merge_when_list
+			returning_clause
 				{
 					MergeStmt  *m = makeNode(MergeStmt);
 
@@ -12382,6 +12383,7 @@ MergeStmt:
 					m->sourceRelation = $6;
 					m->joinCondition = $8;
 					m->mergeWhenClauses = $9;
+					m->returningList = $10;
 
 					$$ = (Node *) m;
 				}
@@ -15795,6 +15797,14 @@ func_expr_common_subexpr:
 					n->location = @1;
 					$$ = (Node *) n;
 				}
+			| MERGE_ACTION '(' ')'
+				{
+					MergeSupportFunc *m = makeNode(MergeSupportFunc);
+
+					m->msftype = TEXTOID;
+					m->location = @1;
+					$$ = (Node *) m;
+				}
 			;
 
 
@@ -17492,6 +17502,7 @@ col_name_keyword:
 			| JSON_SCALAR
 			| JSON_SERIALIZE
 			| LEAST
+			| MERGE_ACTION
 			| NATIONAL
 			| NCHAR
 			| NONE
@@ -17881,6 +17892,7 @@ bare_label_keyword:
 			| MATERIALIZED
 			| MAXVALUE
 			| MERGE
+			| MERGE_ACTION
 			| METHOD
 			| MINVALUE
 			| MODE
diff --git a/src/backend/parser/parse_agg.c b/src/backend/parser/parse_agg.c
new file mode 100644
index 9d151a8..bee7d83
--- a/src/backend/parser/parse_agg.c
+++ b/src/backend/parser/parse_agg.c
@@ -468,6 +468,7 @@ check_agglevels_and_constraints(ParseSta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
@@ -915,6 +916,7 @@ transformWindowFuncCall(ParseState *psta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_cte.c b/src/backend/parser/parse_cte.c
new file mode 100644
index 3c88c9a..6826d4f
--- a/src/backend/parser/parse_cte.c
+++ b/src/backend/parser/parse_cte.c
@@ -126,13 +126,6 @@ transformWithClause(ParseState *pstate,
 		CommonTableExpr *cte = (CommonTableExpr *) lfirst(lc);
 		ListCell   *rest;
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(cte->ctequery, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in WITH query"),
-					parser_errposition(pstate, cte->location));
-
 		for_each_cell(rest, withClause->ctes, lnext(withClause->ctes, lc))
 		{
 			CommonTableExpr *cte2 = (CommonTableExpr *) lfirst(rest);
@@ -153,7 +146,8 @@ transformWithClause(ParseState *pstate,
 			/* must be a data-modifying statement */
 			Assert(IsA(cte->ctequery, InsertStmt) ||
 				   IsA(cte->ctequery, UpdateStmt) ||
-				   IsA(cte->ctequery, DeleteStmt));
+				   IsA(cte->ctequery, DeleteStmt) ||
+				   IsA(cte->ctequery, MergeStmt));
 
 			pstate->p_hasModifyingCTE = true;
 		}
diff --git a/src/backend/parser/parse_expr.c b/src/backend/parser/parse_expr.c
new file mode 100644
index 9300c7b..fe56101
--- a/src/backend/parser/parse_expr.c
+++ b/src/backend/parser/parse_expr.c
@@ -54,6 +54,7 @@ static Node *transformAExprDistinct(Pars
 static Node *transformAExprNullIf(ParseState *pstate, A_Expr *a);
 static Node *transformAExprIn(ParseState *pstate, A_Expr *a);
 static Node *transformAExprBetween(ParseState *pstate, A_Expr *a);
+static Node *transformMergeSupportFunc(ParseState *pstate, MergeSupportFunc *f);
 static Node *transformBoolExpr(ParseState *pstate, BoolExpr *a);
 static Node *transformFuncCall(ParseState *pstate, FuncCall *fn);
 static Node *transformMultiAssignRef(ParseState *pstate, MultiAssignRef *maref);
@@ -227,6 +228,11 @@ transformExprRecurse(ParseState *pstate,
 			result = transformGroupingFunc(pstate, (GroupingFunc *) expr);
 			break;
 
+		case T_MergeSupportFunc:
+			result = transformMergeSupportFunc(pstate,
+											   (MergeSupportFunc *) expr);
+			break;
+
 		case T_NamedArgExpr:
 			{
 				NamedArgExpr *na = (NamedArgExpr *) expr;
@@ -541,6 +547,7 @@ transformColumnRef(ParseState *pstate, C
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CHECK_CONSTRAINT:
@@ -1354,6 +1361,32 @@ transformAExprBetween(ParseState *pstate
 }
 
 static Node *
+transformMergeSupportFunc(ParseState *pstate, MergeSupportFunc *f)
+{
+	/*
+	 * All we need to do is check that we're in the RETURNING list of a MERGE
+	 * command.  If so, we just return the node as-is.
+	 */
+	if (pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+	{
+		ParseState *parent_pstate = pstate->parentParseState;
+
+		while (parent_pstate &&
+			   parent_pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+			parent_pstate = parent_pstate->parentParseState;
+
+		if (!parent_pstate ||
+			parent_pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+			ereport(ERROR,
+					errcode(ERRCODE_SYNTAX_ERROR),
+					errmsg("MERGE_ACTION() can only be used in the RETURNING list of a MERGE command"),
+					parser_errposition(pstate, f->location));
+	}
+
+	return (Node *) f;
+}
+
+static Node *
 transformBoolExpr(ParseState *pstate, BoolExpr *a)
 {
 	List	   *args = NIL;
@@ -1767,6 +1800,7 @@ transformSubLink(ParseState *pstate, Sub
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CYCLE_MARK:
@@ -3115,6 +3149,7 @@ ParseExprKindName(ParseExprKind exprKind
 		case EXPR_KIND_OFFSET:
 			return "OFFSET";
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			return "RETURNING";
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
diff --git a/src/backend/parser/parse_func.c b/src/backend/parser/parse_func.c
new file mode 100644
index fdb3e6d..0cbc950
--- a/src/backend/parser/parse_func.c
+++ b/src/backend/parser/parse_func.c
@@ -2599,6 +2599,7 @@ check_srf_call_placement(ParseState *pst
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_merge.c b/src/backend/parser/parse_merge.c
new file mode 100644
index 6b99f55..efb739d
--- a/src/backend/parser/parse_merge.c
+++ b/src/backend/parser/parse_merge.c
@@ -234,6 +234,10 @@ transformMergeStmt(ParseState *pstate, M
 	 */
 	qry->jointree = makeFromExpr(pstate->p_joinlist, joinExpr);
 
+	/* Transform the RETURNING list, if any */
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_MERGE_RETURNING);
+
 	/*
 	 * We now have a good query shape, so now look at the WHEN conditions and
 	 * action targetlists.
@@ -391,9 +395,6 @@ transformMergeStmt(ParseState *pstate, M
 
 	qry->mergeActionList = mergeActionList;
 
-	/* RETURNING could potentially be added in the future, but not in SQL std */
-	qry->returningList = NULL;
-
 	qry->hasTargetSRFs = false;
 	qry->hasSubLinks = pstate->p_hasSubLinks;
 
diff --git a/src/backend/parser/parse_relation.c b/src/backend/parser/parse_relation.c
new file mode 100644
index 6f5d9e2..427b732
--- a/src/backend/parser/parse_relation.c
+++ b/src/backend/parser/parse_relation.c
@@ -2341,9 +2341,10 @@ addRangeTableEntryForCTE(ParseState *pst
 		cte->cterefcount++;
 
 	/*
-	 * We throw error if the CTE is INSERT/UPDATE/DELETE without RETURNING.
-	 * This won't get checked in case of a self-reference, but that's OK
-	 * because data-modifying CTEs aren't allowed to be recursive anyhow.
+	 * We throw error if the CTE is INSERT/UPDATE/DELETE/MERGE without
+	 * RETURNING.  This won't get checked in case of a self-reference, but
+	 * that's OK because data-modifying CTEs aren't allowed to be recursive
+	 * anyhow.
 	 */
 	if (IsA(cte->ctequery, Query))
 	{
diff --git a/src/backend/parser/parse_target.c b/src/backend/parser/parse_target.c
new file mode 100644
index f10fc42..8923be1
--- a/src/backend/parser/parse_target.c
+++ b/src/backend/parser/parse_target.c
@@ -1806,6 +1806,10 @@ FigureColnameInternal(Node *node, char *
 			/* make GROUPING() act like a regular function */
 			*name = "grouping";
 			return 2;
+		case T_MergeSupportFunc:
+			/* make MERGE_ACTION() act like a regular function */
+			*name = "merge_action";
+			return 2;
 		case T_SubLink:
 			switch (((SubLink *) node)->subLinkType)
 			{
diff --git a/src/backend/rewrite/rewriteHandler.c b/src/backend/rewrite/rewriteHandler.c
new file mode 100644
index 89187d9..c3fa282
--- a/src/backend/rewrite/rewriteHandler.c
+++ b/src/backend/rewrite/rewriteHandler.c
@@ -3833,9 +3833,9 @@ RewriteQuery(Query *parsetree, List *rew
 	ListCell   *lc1;
 
 	/*
-	 * First, recursively process any insert/update/delete statements in WITH
-	 * clauses.  (We have to do this first because the WITH clauses may get
-	 * copied into rule actions below.)
+	 * First, recursively process any insert/update/delete/merge statements in
+	 * WITH clauses.  (We have to do this first because the WITH clauses may
+	 * get copied into rule actions below.)
 	 */
 	foreach(lc1, parsetree->cteList)
 	{
@@ -3860,7 +3860,8 @@ RewriteQuery(Query *parsetree, List *rew
 			if (!(ctequery->commandType == CMD_SELECT ||
 				  ctequery->commandType == CMD_UPDATE ||
 				  ctequery->commandType == CMD_INSERT ||
-				  ctequery->commandType == CMD_DELETE))
+				  ctequery->commandType == CMD_DELETE ||
+				  ctequery->commandType == CMD_MERGE))
 			{
 				/*
 				 * Currently it could only be NOTIFY; this error message will
diff --git a/src/backend/rewrite/rowsecurity.c b/src/backend/rewrite/rowsecurity.c
new file mode 100644
index 450af27..59fd305
--- a/src/backend/rewrite/rowsecurity.c
+++ b/src/backend/rewrite/rowsecurity.c
@@ -384,10 +384,10 @@ get_row_security_policies(Query *root, R
 	 * on the final action we take.
 	 *
 	 * We already fetched the SELECT policies above, to check existing rows,
-	 * but we must also check that new rows created by UPDATE actions are
-	 * visible, if SELECT rights are required for this relation. We don't do
-	 * this for INSERT actions, since an INSERT command would only do this
-	 * check if it had a RETURNING list, and MERGE does not support RETURNING.
+	 * but we must also check that new rows created by INSERT/UPDATE actions
+	 * are visible, if SELECT rights are required. For INSERT actions, we only
+	 * do this if RETURNING is specified, to be consistent with a plain INSERT
+	 * command, which can only require SELECT rights when RETURNING is used.
 	 *
 	 * We don't push the UPDATE/DELETE USING quals to the RTE because we don't
 	 * really want to apply them while scanning the relation since we don't
@@ -409,6 +409,8 @@ get_row_security_policies(Query *root, R
 		List	   *merge_delete_restrictive_policies;
 		List	   *merge_insert_permissive_policies;
 		List	   *merge_insert_restrictive_policies;
+		List	   *merge_select_permissive_policies = NIL;
+		List	   *merge_select_restrictive_policies = NIL;
 
 		/*
 		 * Fetch the UPDATE policies and set them up to execute on the
@@ -446,9 +448,6 @@ get_row_security_policies(Query *root, R
 		 */
 		if (perminfo->requiredPerms & ACL_SELECT)
 		{
-			List	   *merge_select_permissive_policies;
-			List	   *merge_select_restrictive_policies;
-
 			get_policies_for_relation(rel, CMD_SELECT, user_id,
 									  &merge_select_permissive_policies,
 									  &merge_select_restrictive_policies);
@@ -497,6 +496,21 @@ get_row_security_policies(Query *root, R
 							   withCheckOptions,
 							   hasSubLinks,
 							   false);
+
+		/*
+		 * Add ALL/SELECT policies as WCO_RLS_INSERT_CHECK WCOs, to ensure
+		 * that the inserted row is visible when executing an INSERT action,
+		 * if RETURNING is specified and SELECT rights are required for this
+		 * relation.
+		 */
+		if (perminfo->requiredPerms & ACL_SELECT && root->returningList)
+			add_with_check_options(rel, rt_index,
+								   WCO_RLS_INSERT_CHECK,
+								   merge_select_permissive_policies,
+								   merge_select_restrictive_policies,
+								   withCheckOptions,
+								   hasSubLinks,
+								   true);
 	}
 
 	table_close(rel, NoLock);
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
new file mode 100644
index 83f86a4..fa66b80
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -2138,11 +2138,10 @@ QueryReturnsTuples(Query *parsetree)
 		case CMD_SELECT:
 			/* returns tuples */
 			return true;
-		case CMD_MERGE:
-			return false;
 		case CMD_INSERT:
 		case CMD_UPDATE:
 		case CMD_DELETE:
+		case CMD_MERGE:
 			/* the forms with RETURNING return tuples */
 			if (parsetree->returningList)
 				return true;
diff --git a/src/backend/utils/adt/ruleutils.c b/src/backend/utils/adt/ruleutils.c
new file mode 100644
index 2a1ee69..e2b46b6
--- a/src/backend/utils/adt/ruleutils.c
+++ b/src/backend/utils/adt/ruleutils.c
@@ -7196,8 +7196,13 @@ get_merge_query_def(Query *query, depars
 			appendStringInfoString(buf, "DO NOTHING");
 	}
 
-	/* No RETURNING support in MERGE yet */
-	Assert(query->returningList == NIL);
+	/* Add RETURNING if present */
+	if (query->returningList)
+	{
+		appendContextKeyword(context, " RETURNING",
+							 -PRETTYINDENT_STD, PRETTYINDENT_STD, 1);
+		get_target_list(query->returningList, context, NULL, colNamesVisible);
+	}
 }
 
 
@@ -8297,6 +8302,7 @@ isSimpleNode(Node *node, Node *parentNod
 		case T_Aggref:
 		case T_GroupingFunc:
 		case T_WindowFunc:
+		case T_MergeSupportFunc:
 		case T_FuncExpr:
 		case T_JsonConstructorExpr:
 			/* function-like: name(..) or name[..] */
@@ -8651,6 +8657,10 @@ get_rule_expr(Node *node, deparse_contex
 			get_windowfunc_expr((WindowFunc *) node, context);
 			break;
 
+		case T_MergeSupportFunc:
+			appendStringInfoString(buf, "MERGE_ACTION()");
+			break;
+
 		case T_SubscriptingRef:
 			{
 				SubscriptingRef *sbsref = (SubscriptingRef *) node;
diff --git a/src/bin/psql/common.c b/src/bin/psql/common.c
new file mode 100644
index 76e01b0..2830bde
--- a/src/bin/psql/common.c
+++ b/src/bin/psql/common.c
@@ -983,13 +983,17 @@ PrintQueryResult(PGresult *result, bool
 			else
 				success = true;
 
-			/* if it's INSERT/UPDATE/DELETE RETURNING, also print status */
+			/*
+			 * If it's INSERT/UPDATE/DELETE/MERGE RETURNING, also print
+			 * status.
+			 */
 			if (last || pset.show_all_results)
 			{
 				cmdstatus = PQcmdStatus(result);
 				if (strncmp(cmdstatus, "INSERT", 6) == 0 ||
 					strncmp(cmdstatus, "UPDATE", 6) == 0 ||
-					strncmp(cmdstatus, "DELETE", 6) == 0)
+					strncmp(cmdstatus, "DELETE", 6) == 0 ||
+					strncmp(cmdstatus, "MERGE", 5) == 0)
 					PrintQueryStatus(result, printStatusFout);
 			}
 
diff --git a/src/include/executor/execExpr.h b/src/include/executor/execExpr.h
new file mode 100644
index a28ddcd..8953d76
--- a/src/include/executor/execExpr.h
+++ b/src/include/executor/execExpr.h
@@ -243,6 +243,7 @@ typedef enum ExprEvalOp
 	EEOP_AGGREF,
 	EEOP_GROUPING_FUNC,
 	EEOP_WINDOW_FUNC,
+	EEOP_MERGE_SUPPORT_FUNC,
 	EEOP_SUBPLAN,
 
 	/* aggregation related nodes */
@@ -810,6 +811,8 @@ extern void ExecEvalJsonConstructor(Expr
 									ExprContext *econtext);
 extern void ExecEvalJsonIsPredicate(ExprState *state, ExprEvalStep *op);
 extern void ExecEvalGroupingFunc(ExprState *state, ExprEvalStep *op);
+extern void ExecEvalMergeSupportFunc(ExprState *state, ExprEvalStep *op,
+									 ExprContext *econtext);
 extern void ExecEvalSubPlan(ExprState *state, ExprEvalStep *op,
 							ExprContext *econtext);
 extern void ExecEvalWholeRowVar(ExprState *state, ExprEvalStep *op,
diff --git a/src/include/executor/spi.h b/src/include/executor/spi.h
new file mode 100644
index c13e1f6..48b8773
--- a/src/include/executor/spi.h
+++ b/src/include/executor/spi.h
@@ -97,6 +97,7 @@ typedef struct _SPI_plan *SPIPlanPtr;
 #define SPI_OK_REL_UNREGISTER	16
 #define SPI_OK_TD_REGISTER		17
 #define SPI_OK_MERGE			18
+#define SPI_OK_MERGE_RETURNING	19
 
 #define SPI_OPT_NONATOMIC		(1 << 0)
 
diff --git a/src/include/nodes/execnodes.h b/src/include/nodes/execnodes.h
new file mode 100644
index 444a5f0..57fdfca
--- a/src/include/nodes/execnodes.h
+++ b/src/include/nodes/execnodes.h
@@ -1325,6 +1325,9 @@ typedef struct ModifyTableState
 	/* Flags showing which subcommands are present INS/UPD/DEL/DO NOTHING */
 	int			mt_merge_subcommands;
 
+	/* For MERGE, the action currently being executed */
+	MergeActionState *mt_merge_action;
+
 	/* tuple counters for MERGE */
 	double		mt_merge_inserted;
 	double		mt_merge_updated;
diff --git a/src/include/nodes/parsenodes.h b/src/include/nodes/parsenodes.h
new file mode 100644
index 2380821..2bae226
--- a/src/include/nodes/parsenodes.h
+++ b/src/include/nodes/parsenodes.h
@@ -1938,6 +1938,7 @@ typedef struct MergeStmt
 	Node	   *sourceRelation; /* source relation */
 	Node	   *joinCondition;	/* join condition between source and target */
 	List	   *mergeWhenClauses;	/* list of MergeWhenClause(es) */
+	List	   *returningList;	/* list of expressions to return */
 	WithClause *withClause;		/* WITH clause */
 } MergeStmt;
 
diff --git a/src/include/nodes/primnodes.h b/src/include/nodes/primnodes.h
new file mode 100644
index 4a15460..8df8884
--- a/src/include/nodes/primnodes.h
+++ b/src/include/nodes/primnodes.h
@@ -572,6 +572,27 @@ typedef struct WindowFunc
 } WindowFunc;
 
 /*
+ * MergeSupportFunc
+ *
+ * A MergeSupportFunc is a merge support function expression that can only
+ * appear in the RETURNING list of a MERGE command.  It returns information
+ * about the currently executing merge action.
+ *
+ * Currently, the only supported function is MERGE_ACTION(), which returns the
+ * command executed ("INSERT", "UPDATE", or "DELETE").
+ */
+typedef struct MergeSupportFunc
+{
+	Expr		xpr;
+	/* type Oid of result */
+	Oid			msftype;
+	/* OID of collation, or InvalidOid if none */
+	Oid			msfcollid;
+	/* token location, or -1 if unknown */
+	int			location;
+} MergeSupportFunc;
+
+/*
  * SubscriptingRef: describes a subscripting operation over a container
  * (array, etc).
  *
diff --git a/src/include/optimizer/paramassign.h b/src/include/optimizer/paramassign.h
new file mode 100644
index 501f728..4026b74
--- a/src/include/optimizer/paramassign.h
+++ b/src/include/optimizer/paramassign.h
@@ -20,6 +20,8 @@ extern Param *replace_outer_placeholderv
 										   PlaceHolderVar *phv);
 extern Param *replace_outer_agg(PlannerInfo *root, Aggref *agg);
 extern Param *replace_outer_grouping(PlannerInfo *root, GroupingFunc *grp);
+extern Param *replace_outer_merge_support(PlannerInfo *root,
+										  MergeSupportFunc *msf);
 extern Param *replace_nestloop_param_var(PlannerInfo *root, Var *var);
 extern Param *replace_nestloop_param_placeholdervar(PlannerInfo *root,
 													PlaceHolderVar *phv);
diff --git a/src/include/parser/analyze.h b/src/include/parser/analyze.h
new file mode 100644
index 5ecdd10..28b66fc
--- a/src/include/parser/analyze.h
+++ b/src/include/parser/analyze.h
@@ -44,6 +44,8 @@ extern List *transformInsertRow(ParseSta
 								bool strip_indirection);
 extern List *transformUpdateTargetList(ParseState *pstate,
 									   List *origTlist);
+extern List *transformReturningList(ParseState *pstate, List *returningList,
+									ParseExprKind exprKind);
 extern Query *transformTopLevelStmt(ParseState *pstate, RawStmt *parseTree);
 extern Query *transformStmt(ParseState *pstate, Node *parseTree);
 
diff --git a/src/include/parser/kwlist.h b/src/include/parser/kwlist.h
new file mode 100644
index 2331aca..0993534
--- a/src/include/parser/kwlist.h
+++ b/src/include/parser/kwlist.h
@@ -266,6 +266,7 @@ PG_KEYWORD("matched", MATCHED, UNRESERVE
 PG_KEYWORD("materialized", MATERIALIZED, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("maxvalue", MAXVALUE, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("merge", MERGE, UNRESERVED_KEYWORD, BARE_LABEL)
+PG_KEYWORD("merge_action", MERGE_ACTION, COL_NAME_KEYWORD, BARE_LABEL)
 PG_KEYWORD("method", METHOD, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("minute", MINUTE_P, UNRESERVED_KEYWORD, AS_LABEL)
 PG_KEYWORD("minvalue", MINVALUE, UNRESERVED_KEYWORD, BARE_LABEL)
diff --git a/src/include/parser/parse_node.h b/src/include/parser/parse_node.h
new file mode 100644
index 99d6515..5b781d8
--- a/src/include/parser/parse_node.h
+++ b/src/include/parser/parse_node.h
@@ -61,7 +61,8 @@ typedef enum ParseExprKind
 	EXPR_KIND_DISTINCT_ON,		/* DISTINCT ON */
 	EXPR_KIND_LIMIT,			/* LIMIT */
 	EXPR_KIND_OFFSET,			/* OFFSET */
-	EXPR_KIND_RETURNING,		/* RETURNING */
+	EXPR_KIND_RETURNING,		/* RETURNING in INSERT/UPDATE/DELETE */
+	EXPR_KIND_MERGE_RETURNING,	/* RETURNING in MERGE */
 	EXPR_KIND_VALUES,			/* VALUES */
 	EXPR_KIND_VALUES_SINGLE,	/* single-row VALUES (in INSERT only) */
 	EXPR_KIND_CHECK_CONSTRAINT, /* CHECK constraint for a table */
diff --git a/src/pl/plpgsql/src/pl_exec.c b/src/pl/plpgsql/src/pl_exec.c
new file mode 100644
index ed51694..6947575
--- a/src/pl/plpgsql/src/pl_exec.c
+++ b/src/pl/plpgsql/src/pl_exec.c
@@ -4267,9 +4267,9 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 	/*
 	 * If we have INTO, then we only need one row back ... but if we have INTO
 	 * STRICT or extra check too_many_rows, ask for two rows, so that we can
-	 * verify the statement returns only one.  INSERT/UPDATE/DELETE are always
-	 * treated strictly. Without INTO, just run the statement to completion
-	 * (tcount = 0).
+	 * verify the statement returns only one.  INSERT/UPDATE/DELETE/MERGE are
+	 * always treated strictly. Without INTO, just run the statement to
+	 * completion (tcount = 0).
 	 *
 	 * We could just ask for two rows always when using INTO, but there are
 	 * some cases where demanding the extra row costs significant time, eg by
@@ -4307,10 +4307,11 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 			Assert(stmt->mod_stmt);
 			exec_set_found(estate, (SPI_processed != 0));
 			break;
@@ -4489,10 +4490,11 @@ exec_stmt_dynexecute(PLpgSQL_execstate *
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 		case SPI_OK_UTILITY:
 		case SPI_OK_REWRITTEN:
 			break;
diff --git a/src/pl/tcl/pltcl.c b/src/pl/tcl/pltcl.c
new file mode 100644
index 6187e15..18d14a2
--- a/src/pl/tcl/pltcl.c
+++ b/src/pl/tcl/pltcl.c
@@ -2456,6 +2456,7 @@ pltcl_process_SPI_result(Tcl_Interp *int
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
+		case SPI_OK_MERGE_RETURNING:
 
 			/*
 			 * Process the tuples we got
diff --git a/src/test/regress/expected/merge.out b/src/test/regress/expected/merge.out
new file mode 100644
index e3ebf46..28b39b9
--- a/src/test/regress/expected/merge.out
+++ b/src/test/regress/expected/merge.out
@@ -124,20 +124,20 @@ ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
 ERROR:  name "target" specified more than once
 DETAIL:  The name is used both as MERGE target table and data source.
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
-ERROR:  MERGE not supported in WITH query
-LINE 1: WITH foo AS (
-             ^
--- used in COPY
+ERROR:  WITH query "foo" does not have a RETURNING clause
+LINE 4: ) SELECT * FROM foo;
+                        ^
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) TO stdout;
-ERROR:  MERGE not supported in COPY
+ERROR:  COPY query must have a RETURNING clause
 -- unsupported relation types
 -- materialized view
 CREATE MATERIALIZED VIEW mv AS SELECT * FROM target;
@@ -1255,7 +1255,7 @@ BEGIN;
 MERGE INTO sq_target
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1270,7 +1270,7 @@ INSERT INTO sq_source (sid, balance, del
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1294,7 +1294,7 @@ WITH targq AS (
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1303,20 +1303,190 @@ WHEN MATCHED AND tid < 2 THEN
 ROLLBACK;
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
+CREATE TABLE merge_actions(action text, abbrev text);
+INSERT INTO merge_actions VALUES ('INSERT', 'ins'), ('UPDATE', 'upd'), ('DELETE', 'del');
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
-ERROR:  syntax error at or near "RETURNING"
-LINE 10: RETURNING *;
-         ^
+    DELETE
+RETURNING (SELECT abbrev FROM merge_actions
+            WHERE action = merge_action()) AS action,
+          t.*,
+          CASE merge_action()
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ action | tid | balance |     description     
+--------+-----+---------+---------------------
+ del    |   1 |     100 | Removed (1,100)
+ upd    |   2 |     220 | Added 20 to balance
+ ins    |   4 |      40 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- error when using merge_action() outside MERGE
+SELECT merge_action() FROM sq_target;
+ERROR:  MERGE_ACTION() can only be used in the RETURNING list of a MERGE command
+LINE 1: SELECT merge_action() FROM sq_target;
+               ^
+UPDATE sq_target SET balance = balance + 1 RETURNING merge_action();
+ERROR:  MERGE_ACTION() can only be used in the RETURNING list of a MERGE command
+LINE 1: ...ATE sq_target SET balance = balance + 1 RETURNING merge_acti...
+                                                             ^
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING merge_action() AS action, t.*,
+              CASE merge_action()
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING action, merge_action() AS log_action, l.*
+)
+SELECT * FROM m2;
+ action | log_action | tid |     last_change     
+--------+------------+-----+---------------------
+ DELETE | UPDATE     |   1 | Removed (1,100)
+ UPDATE | INSERT     |   2 | Added 20 to balance
+ INSERT | INSERT     |   4 | Inserted (4,40)
+(3 rows)
+
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ tid |     last_change     
+-----+---------------------
+   1 | Removed (1,100)
+   2 | Added 20 to balance
+   4 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING merge_action(), t.*
+) TO stdout;
+DELETE	1	100
+UPDATE	2	220
+INSERT	4	40
+ROLLBACK;
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING merge_action(), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ action | tid | new_balance 
+--------+-----+-------------
+ DELETE |   1 |         100
+ UPDATE |   3 |         320
+ INSERT |   4 |         110
+(3 rows)
+
+ROLLBACK;
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING merge_action(), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ action | tid | balance 
+--------+-----+---------
+ DELETE |   1 |     100
+ UPDATE |   2 |     220
+ INSERT |   4 |      40
+(3 rows)
+
+ROLLBACK;
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING merge_action(), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ r_action | r_tid | r_balance 
+----------+-------+-----------
+ DELETE   |     1 |       100
+ UPDATE   |     3 |       320
+ INSERT   |     4 |       110
+(3 rows)
+
 ROLLBACK;
 -- EXPLAIN
 CREATE TABLE ex_mtarget (a int, b int)
@@ -1513,7 +1683,7 @@ SELECT * FROM sq_target WHERE tid = 1;
 (1 row)
 
 ROLLBACK;
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 NOTICE:  drop cascades to view v
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1639,6 +1809,32 @@ SELECT * FROM pa_target ORDER BY tid;
 (14 rows)
 
 ROLLBACK;
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING merge_action(), t.*;
+ merge_action | tid | balance |           val            
+--------------+-----+---------+--------------------------
+ UPDATE       |   2 |     110 | initial updated by merge
+(1 row)
+
+SELECT * FROM pa_target ORDER BY tid;
+ tid | balance |           val            
+-----+---------+--------------------------
+   2 |     110 | initial updated by merge
+   3 |     300 | initial
+   5 |     500 | initial
+   7 |     700 | initial
+   9 |     900 | initial
+  11 |    1100 | initial
+  13 |    1300 | initial
+(7 rows)
+
+ROLLBACK;
 DROP TABLE pa_target CASCADE;
 -- The target table is partitioned in the same way, but this time by attaching
 -- partitions which have columns in different order, dropped columns etc.
@@ -1888,7 +2084,21 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING merge_action(), t.*;
+ merge_action |          logts           | tid | balance |           val            
+--------------+--------------------------+-----+---------+--------------------------
+ UPDATE       | Tue Jan 31 00:00:00 2017 |   1 |     110 | initial updated by merge
+ UPDATE       | Tue Feb 28 00:00:00 2017 |   2 |     220 | initial updated by merge
+ INSERT       | Sun Jan 15 00:00:00 2017 |   3 |      30 | inserted by merge
+ UPDATE       | Tue Jan 31 00:00:00 2017 |   4 |     440 | initial updated by merge
+ UPDATE       | Tue Feb 28 00:00:00 2017 |   5 |     550 | initial updated by merge
+ INSERT       | Sun Jan 15 00:00:00 2017 |   6 |      60 | inserted by merge
+ UPDATE       | Tue Jan 31 00:00:00 2017 |   7 |     770 | initial updated by merge
+ UPDATE       | Tue Feb 28 00:00:00 2017 |   8 |     880 | initial updated by merge
+ INSERT       | Sun Jan 15 00:00:00 2017 |   9 |      90 | inserted by merge
+(9 rows)
+
 SELECT * FROM pa_target ORDER BY tid;
           logts           | tid | balance |           val            
 --------------------------+-----+---------+--------------------------
diff --git a/src/test/regress/expected/rowsecurity.out b/src/test/regress/expected/rowsecurity.out
new file mode 100644
index 6988128..3fd4fab
--- a/src/test/regress/expected/rowsecurity.out
+++ b/src/test/regress/expected/rowsecurity.out
@@ -2302,6 +2302,35 @@ WHEN MATCHED THEN
 	UPDATE SET dnotes = dnotes || ' notes added by merge8 '
 WHEN NOT MATCHED THEN
 	INSERT VALUES (13, 44, 1, 'regress_rls_bob', 'new manga');
+SELECT * FROM document WHERE did = 13;
+ did | cid | dlevel | dauthor | dtitle | dnotes 
+-----+-----+--------+---------+--------+--------
+(0 rows)
+
+-- but not OK if RETURNING is used
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge9 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 44, 1, 'regress_rls_bob', 'new manga')
+RETURNING *;
+ERROR:  new row violates row-level security policy for table "document"
+-- but OK if new row is visible
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge10 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 11, 1, 'regress_rls_bob', 'new novel')
+RETURNING *;
+ sdid | did | cid | dlevel |     dauthor     |  dtitle   | dnotes 
+------+-----+-----+--------+-----------------+-----------+--------
+   14 |  14 |  11 |      1 | regress_rls_bob | new novel | 
+(1 row)
+
 RESET SESSION AUTHORIZATION;
 -- drop the restrictive SELECT policy so that we can look at the
 -- final state of the table
@@ -2325,7 +2354,8 @@ SELECT * FROM document;
   12 |  11 |      1 | regress_rls_bob   | another novel                    | 
    1 |  11 |      1 | regress_rls_bob   | my first novel                   |  notes added by merge2  notes added by merge3  notes added by merge4  notes added by merge7 
   13 |  44 |      1 | regress_rls_bob   | new manga                        | 
-(15 rows)
+  14 |  11 |      1 | regress_rls_bob   | new novel                        | 
+(16 rows)
 
 --
 -- ROLE/GROUP
diff --git a/src/test/regress/expected/rules.out b/src/test/regress/expected/rules.out
new file mode 100644
index 0cd2c64..84e359f
--- a/src/test/regress/expected/rules.out
+++ b/src/test/regress/expected/rules.out
@@ -3638,7 +3638,7 @@ MERGE INTO rule_merge1 t USING (SELECT 1
 -- test deparsing
 CREATE TABLE sf_target(id int, data text, filling int[]);
 CREATE FUNCTION merge_sf_test()
- RETURNS void
+ RETURNS TABLE(action text, a int, b text, id int, data text, filling int[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -3675,11 +3675,13 @@ WHEN NOT MATCHED
    VALUES (s.a, s.b, DEFAULT)
 WHEN NOT MATCHED
    THEN INSERT (filling[1], id)
-   VALUES (s.a, s.a);
+   VALUES (s.a, s.a)
+RETURNING
+   merge_action() AS action, *;
 END;
 \sf merge_sf_test
 CREATE OR REPLACE FUNCTION public.merge_sf_test()
- RETURNS void
+ RETURNS TABLE(action text, a integer, b text, id integer, data text, filling integer[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -3716,7 +3718,13 @@ BEGIN ATOMIC
       VALUES (s.a, s.b, DEFAULT)
     WHEN NOT MATCHED
      THEN INSERT (filling[1], id)
-      VALUES (s.a, s.a);
+      VALUES (s.a, s.a)
+   RETURNING MERGE_ACTION() AS action,
+     s.a,
+     s.b,
+     t.id,
+     t.data,
+     t.filling;
 END
 DROP FUNCTION merge_sf_test;
 DROP TABLE sf_target;
diff --git a/src/test/regress/expected/updatable_views.out b/src/test/regress/expected/updatable_views.out
new file mode 100644
index 794cf9c..1062c34
--- a/src/test/regress/expected/updatable_views.out
+++ b/src/test/regress/expected/updatable_views.out
@@ -479,7 +479,15 @@ MERGE INTO rw_view1 t
                 (2, 'ROW 2'), (3, 'ROW 3')) AS v(a,b) ON t.a = v.a
   WHEN MATCHED AND t.a <= 1 THEN UPDATE SET b = v.b
   WHEN MATCHED THEN DELETE
-  WHEN NOT MATCHED AND a > 0 THEN INSERT (a) VALUES (v.a);
+  WHEN NOT MATCHED AND a > 0 THEN INSERT (a) VALUES (v.a)
+  RETURNING merge_action(), v.*, t.*;
+ merge_action | a |   b   | a |      b      
+--------------+---+-------+---+-------------
+ UPDATE       | 1 | ROW 1 | 1 | ROW 1
+ DELETE       | 3 | ROW 3 | 3 | Row 3
+ INSERT       | 2 | ROW 2 | 2 | Unspecified
+(3 rows)
+
 SELECT * FROM base_tbl ORDER BY a;
  a  |      b      
 ----+-------------
@@ -631,7 +639,15 @@ MERGE INTO rw_view2 t
   USING (VALUES (3, 'R3'), (4, 'R4'), (5, 'R5')) AS v(a,b) ON aaa = v.a
   WHEN MATCHED AND aaa = 3 THEN DELETE
   WHEN MATCHED THEN UPDATE SET bbb = v.b
-  WHEN NOT MATCHED THEN INSERT (aaa) VALUES (v.a);
+  WHEN NOT MATCHED THEN INSERT (aaa) VALUES (v.a)
+  RETURNING merge_action(), v.*, t.*;
+ merge_action | a | b  | aaa |     bbb     
+--------------+---+----+-----+-------------
+ DELETE       | 3 | R3 |   3 | Row 3
+ UPDATE       | 4 | R4 |   4 | R4
+ INSERT       | 5 | R5 |   5 | Unspecified
+(3 rows)
+
 SELECT * FROM rw_view2 ORDER BY aaa;
  aaa |     bbb     
 -----+-------------
@@ -1071,7 +1087,15 @@ MERGE INTO rw_view2 t
   USING (SELECT x, 'R'||x FROM generate_series(0,3) x) AS s(a,b) ON t.a = s.a
   WHEN MATCHED AND t.a <= 1 THEN DELETE
   WHEN MATCHED THEN UPDATE SET b = s.b
-  WHEN NOT MATCHED AND s.a > 0 THEN INSERT VALUES (s.a, s.b);
+  WHEN NOT MATCHED AND s.a > 0 THEN INSERT VALUES (s.a, s.b)
+  RETURNING merge_action(), s.*, t.*;
+ merge_action | a | b  | a |   b   
+--------------+---+----+---+-------
+ DELETE       | 1 | R1 | 1 | Row 1
+ UPDATE       | 2 | R2 | 2 | R2
+ INSERT       | 3 | R3 | 3 | R3
+(3 rows)
+
 SELECT * FROM base_tbl ORDER BY a;
  a  |   b    
 ----+--------
diff --git a/src/test/regress/expected/with.out b/src/test/regress/expected/with.out
new file mode 100644
index 7d796ea..6de347b
--- a/src/test/regress/expected/with.out
+++ b/src/test/regress/expected/with.out
@@ -3493,6 +3493,16 @@ SELECT * FROM t;
 ERROR:  WITH query "t" does not have a RETURNING clause
 LINE 4: SELECT * FROM t;
                       ^
+-- RETURNING tries to return its own output
+WITH RECURSIVE t(action, a) AS (
+	MERGE INTO y USING (VALUES (11)) v(a) ON y.a = v.a
+		WHEN NOT MATCHED THEN INSERT VALUES (v.a)
+		RETURNING merge_action(), (SELECT a FROM t)
+)
+SELECT * FROM t;
+ERROR:  recursive query "t" must not contain data-modifying statements
+LINE 1: WITH RECURSIVE t(action, a) AS (
+                       ^
 -- data-modifying WITH allowed only at the top level
 SELECT * FROM (
 	WITH t AS (UPDATE y SET a=a+1 RETURNING *)
diff --git a/src/test/regress/sql/merge.sql b/src/test/regress/sql/merge.sql
new file mode 100644
index 79a1868..101f2e9
--- a/src/test/regress/sql/merge.sql
+++ b/src/test/regress/sql/merge.sql
@@ -88,12 +88,12 @@ MERGE INTO target
 USING target
 ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
--- used in COPY
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
@@ -817,7 +817,7 @@ BEGIN;
 MERGE INTO sq_target
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -830,7 +830,7 @@ INSERT INTO sq_source (sid, balance, del
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -848,7 +848,7 @@ WITH targq AS (
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -858,17 +858,145 @@ ROLLBACK;
 
 -- RETURNING
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
+CREATE TABLE merge_actions(action text, abbrev text);
+INSERT INTO merge_actions VALUES ('INSERT', 'ins'), ('UPDATE', 'upd'), ('DELETE', 'del');
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
+    DELETE
+RETURNING (SELECT abbrev FROM merge_actions
+            WHERE action = merge_action()) AS action,
+          t.*,
+          CASE merge_action()
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ROLLBACK;
+
+-- error when using merge_action() outside MERGE
+SELECT merge_action() FROM sq_target;
+UPDATE sq_target SET balance = balance + 1 RETURNING merge_action();
+
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING merge_action() AS action, t.*,
+              CASE merge_action()
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING action, merge_action() AS log_action, l.*
+)
+SELECT * FROM m2;
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ROLLBACK;
+
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING merge_action(), t.*
+) TO stdout;
+ROLLBACK;
+
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING merge_action(), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ROLLBACK;
+
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING merge_action(), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ROLLBACK;
+
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING merge_action(), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
 ROLLBACK;
 
 -- EXPLAIN
@@ -967,7 +1095,7 @@ WHEN MATCHED THEN
 SELECT * FROM sq_target WHERE tid = 1;
 ROLLBACK;
 
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1034,6 +1162,17 @@ SELECT merge_func();
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING merge_action(), t.*;
+SELECT * FROM pa_target ORDER BY tid;
+ROLLBACK;
+
 DROP TABLE pa_target CASCADE;
 
 -- The target table is partitioned in the same way, but this time by attaching
@@ -1210,7 +1349,8 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING merge_action(), t.*;
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
diff --git a/src/test/regress/sql/rowsecurity.sql b/src/test/regress/sql/rowsecurity.sql
new file mode 100644
index dec7340..1d5ed0a
--- a/src/test/regress/sql/rowsecurity.sql
+++ b/src/test/regress/sql/rowsecurity.sql
@@ -992,6 +992,27 @@ WHEN MATCHED THEN
 	UPDATE SET dnotes = dnotes || ' notes added by merge8 '
 WHEN NOT MATCHED THEN
 	INSERT VALUES (13, 44, 1, 'regress_rls_bob', 'new manga');
+SELECT * FROM document WHERE did = 13;
+
+-- but not OK if RETURNING is used
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge9 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 44, 1, 'regress_rls_bob', 'new manga')
+RETURNING *;
+
+-- but OK if new row is visible
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge10 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 11, 1, 'regress_rls_bob', 'new novel')
+RETURNING *;
 
 RESET SESSION AUTHORIZATION;
 -- drop the restrictive SELECT policy so that we can look at the
diff --git a/src/test/regress/sql/rules.sql b/src/test/regress/sql/rules.sql
new file mode 100644
index 6924012..27340ba
--- a/src/test/regress/sql/rules.sql
+++ b/src/test/regress/sql/rules.sql
@@ -1294,7 +1294,7 @@ MERGE INTO rule_merge1 t USING (SELECT 1
 CREATE TABLE sf_target(id int, data text, filling int[]);
 
 CREATE FUNCTION merge_sf_test()
- RETURNS void
+ RETURNS TABLE(action text, a int, b text, id int, data text, filling int[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -1331,7 +1331,9 @@ WHEN NOT MATCHED
    VALUES (s.a, s.b, DEFAULT)
 WHEN NOT MATCHED
    THEN INSERT (filling[1], id)
-   VALUES (s.a, s.a);
+   VALUES (s.a, s.a)
+RETURNING
+   merge_action() AS action, *;
 END;
 
 \sf merge_sf_test
diff --git a/src/test/regress/sql/updatable_views.sql b/src/test/regress/sql/updatable_views.sql
new file mode 100644
index ae11e46..afdf331
--- a/src/test/regress/sql/updatable_views.sql
+++ b/src/test/regress/sql/updatable_views.sql
@@ -175,7 +175,8 @@ MERGE INTO rw_view1 t
                 (2, 'ROW 2'), (3, 'ROW 3')) AS v(a,b) ON t.a = v.a
   WHEN MATCHED AND t.a <= 1 THEN UPDATE SET b = v.b
   WHEN MATCHED THEN DELETE
-  WHEN NOT MATCHED AND a > 0 THEN INSERT (a) VALUES (v.a);
+  WHEN NOT MATCHED AND a > 0 THEN INSERT (a) VALUES (v.a)
+  RETURNING merge_action(), v.*, t.*;
 SELECT * FROM base_tbl ORDER BY a;
 
 EXPLAIN (costs off) UPDATE rw_view1 SET a=6 WHERE a=5;
@@ -246,7 +247,8 @@ MERGE INTO rw_view2 t
   USING (VALUES (3, 'R3'), (4, 'R4'), (5, 'R5')) AS v(a,b) ON aaa = v.a
   WHEN MATCHED AND aaa = 3 THEN DELETE
   WHEN MATCHED THEN UPDATE SET bbb = v.b
-  WHEN NOT MATCHED THEN INSERT (aaa) VALUES (v.a);
+  WHEN NOT MATCHED THEN INSERT (aaa) VALUES (v.a)
+  RETURNING merge_action(), v.*, t.*;
 SELECT * FROM rw_view2 ORDER BY aaa;
 
 EXPLAIN (costs off) UPDATE rw_view2 SET aaa=5 WHERE aaa=4;
@@ -458,7 +460,8 @@ MERGE INTO rw_view2 t
   USING (SELECT x, 'R'||x FROM generate_series(0,3) x) AS s(a,b) ON t.a = s.a
   WHEN MATCHED AND t.a <= 1 THEN DELETE
   WHEN MATCHED THEN UPDATE SET b = s.b
-  WHEN NOT MATCHED AND s.a > 0 THEN INSERT VALUES (s.a, s.b);
+  WHEN NOT MATCHED AND s.a > 0 THEN INSERT VALUES (s.a, s.b)
+  RETURNING merge_action(), s.*, t.*;
 SELECT * FROM base_tbl ORDER BY a;
 
 EXPLAIN (costs off) UPDATE rw_view2 SET a=3 WHERE a=2;
diff --git a/src/test/regress/sql/with.sql b/src/test/regress/sql/with.sql
new file mode 100644
index f8a213e..037bc0a
--- a/src/test/regress/sql/with.sql
+++ b/src/test/regress/sql/with.sql
@@ -1591,6 +1591,14 @@ WITH t AS (
 )
 SELECT * FROM t;
 
+-- RETURNING tries to return its own output
+WITH RECURSIVE t(action, a) AS (
+	MERGE INTO y USING (VALUES (11)) v(a) ON y.a = v.a
+		WHEN NOT MATCHED THEN INSERT VALUES (v.a)
+		RETURNING merge_action(), (SELECT a FROM t)
+)
+SELECT * FROM t;
+
 -- data-modifying WITH allowed only at the top level
 SELECT * FROM (
 	WITH t AS (UPDATE y SET a=a+1 RETURNING *)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
new file mode 100644
index d3a7f75..4e3aa38
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -1576,6 +1576,7 @@ MergeJoinState
 MergePath
 MergeScanSelCache
 MergeStmt
+MergeSupportFunc
 MergeWhenClause
 MetaCommand
 MinMaxAggInfo

diff --git a/doc/src/sgml/dml.sgml b/doc/src/sgml/dml.sgml
new file mode 100644
index cbbc5e2..3d95bdb
--- a/doc/src/sgml/dml.sgml
+++ b/doc/src/sgml/dml.sgml
@@ -283,10 +283,15 @@ DELETE FROM products;
    <secondary>RETURNING</secondary>
   </indexterm>
 
+  <indexterm zone="dml-returning">
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
   <para>
    Sometimes it is useful to obtain data from modified rows while they are
    being manipulated.  The <command>INSERT</command>, <command>UPDATE</command>,
-   and <command>DELETE</command> commands all have an
+   <command>DELETE</command>, and <command>MERGE</command> commands all have an
    optional <literal>RETURNING</literal> clause that supports this.  Use
    of <literal>RETURNING</literal> avoids performing an extra database query to
    collect the data, and is especially valuable when it would otherwise be
@@ -339,6 +344,21 @@ DELETE FROM products
 </programlisting>
   </para>
 
+  <para>
+   In a <command>MERGE</command>, the data available to <literal>RETURNING</literal> is
+   the content of the source row plus the content of the inserted, updated, or
+   deleted target row.  Since it is quite common for the source and target to
+   have many of the same columns, specifying <literal>RETURNING *</literal>
+   can lead to a lot of duplicated columns, so it is often more useful to
+   qualify it so as to return just the source or target row.  For example:
+<programlisting>
+MERGE INTO products p USING new_products n ON p.product_no = n.product_no
+  WHEN NOT MATCHED THEN INSERT VALUES (n.product_no, n.name, n.price)
+  WHEN MATCHED THEN UPDATE SET name = n.name, price = n.price
+  RETURNING p.*;
+</programlisting>
+  </para>
+
   <para>
    If there are triggers (<xref linkend="triggers"/>) on the target table,
    the data available to <literal>RETURNING</literal> is the row as modified by
diff --git a/doc/src/sgml/func.sgml b/doc/src/sgml/func.sgml
new file mode 100644
index 0bb7aeb..0e2b888
--- a/doc/src/sgml/func.sgml
+++ b/doc/src/sgml/func.sgml
@@ -22421,6 +22421,85 @@ SELECT count(*) FROM sometable;
 
  </sect1>
 
+ <sect1 id="functions-merge-support">
+  <title>Merge Support Functions</title>
+
+  <indexterm>
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
+  <para>
+   <productname>PostgreSQL</productname> includes one merge support function
+   that may be used in the <literal>RETURNING</literal> list of a
+   <xref linkend="sql-merge"/> command to identify the action taken for each
+   row.
+  </para>
+
+  <table id="functions-merge-support-table">
+   <title>Merge Support Functions</title>
+
+   <tgroup cols="1">
+    <thead>
+     <row>
+      <entry role="func_table_entry"><para role="func_signature">
+       Function
+      </para>
+      <para>
+       Description
+      </para></entry>
+     </row>
+    </thead>
+
+    <tbody>
+     <row>
+      <entry role="func_table_entry"><para role="func_signature">
+       <indexterm>
+        <primary>merge_action</primary>
+       </indexterm>
+       <function id="merge_action">merge_action</function> ( )
+       <returnvalue>text</returnvalue>
+      </para>
+      <para>
+       Returns the merge action command executed for the current row.  This
+       will be <literal>'INSERT'</literal>, <literal>'UPDATE'</literal>, or
+       <literal>'DELETE'</literal>.
+      </para></entry>
+     </row>
+    </tbody>
+   </tgroup>
+  </table>
+
+  <para>
+   Example:
+<screen><![CDATA[
+MERGE INTO products p
+  USING stock s ON p.product_id = s.product_id
+  WHEN MATCHED AND s.quantity > 0 THEN
+    UPDATE SET in_stock = true, quantity = s.quantity
+  WHEN MATCHED THEN
+    UPDATE SET in_stock = false, quantity = 0
+  WHEN NOT MATCHED THEN
+    INSERT (product_id, in_stock, quantity)
+      VALUES (s.product_id, true, s.quantity)
+  RETURNING merge_action(), p.*;
+
+ merge_action | product_id | in_stock | quantity
+--------------+------------+----------+----------
+ UPDATE       |       1001 | t        |       50
+ UPDATE       |       1002 | f        |        0
+ INSERT       |       1003 | t        |       10
+]]></screen>
+  </para>
+
+  <para>
+   Note that this function can only be used in the <literal>RETURNING</literal>
+   list of a <command>MERGE</command> command.  It is an error to use it in any
+   other part of a query.
+  </para>
+
+ </sect1>
+
  <sect1 id="functions-subquery">
   <title>Subquery Expressions</title>
 
diff --git a/doc/src/sgml/glossary.sgml b/doc/src/sgml/glossary.sgml
new file mode 100644
index 8c2f114..a81c17a
--- a/doc/src/sgml/glossary.sgml
+++ b/doc/src/sgml/glossary.sgml
@@ -1442,9 +1442,9 @@
      to a <glossterm linkend="glossary-client">client</glossterm> upon the
      completion of an <acronym>SQL</acronym> command, usually a
      <command>SELECT</command> but it can be an
-     <command>INSERT</command>, <command>UPDATE</command>, or
-     <command>DELETE</command> command if the <literal>RETURNING</literal>
-     clause is specified.
+     <command>INSERT</command>, <command>UPDATE</command>,
+     <command>DELETE</command>, or <command>MERGE</command> command if the
+     <literal>RETURNING</literal> clause is specified.
     </para>
     <para>
      The fact that a result set is a relation means that a query can be used
diff --git a/doc/src/sgml/plpgsql.sgml b/doc/src/sgml/plpgsql.sgml
new file mode 100644
index c2b9c6a..406834e
--- a/doc/src/sgml/plpgsql.sgml
+++ b/doc/src/sgml/plpgsql.sgml
@@ -1043,8 +1043,8 @@ INSERT INTO mytable VALUES (1,'one'), (2
     </para>
 
     <para>
-     If the command does return rows (for example <command>SELECT</command>,
-     or <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>
+     If the command does return rows (for example <command>SELECT</command>, or
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
      with <literal>RETURNING</literal>), there are two ways to proceed.
      When the command will return at most one row, or you only care about
      the first row of output, write the command as usual but add
@@ -1172,6 +1172,7 @@ SELECT <replaceable>select_expressions</
 INSERT ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 UPDATE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 DELETE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
+MERGE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 </synopsis>
 
      where <replaceable>target</replaceable> can be a record variable, a row
@@ -1182,8 +1183,8 @@ DELETE ... RETURNING <replaceable>expres
      <literal>INTO</literal> clause) just as described above,
      and the plan is cached in the same way.
      This works for <command>SELECT</command>,
-     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
-     <literal>RETURNING</literal>, and certain utility commands
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
+     with <literal>RETURNING</literal>, and certain utility commands
      that return row sets, such as <command>EXPLAIN</command>.
      Except for the <literal>INTO</literal> clause, the SQL command is the same
      as it would be written outside <application>PL/pgSQL</application>.
@@ -1259,7 +1260,7 @@ END;
     </para>
 
     <para>
-     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
+     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command> with
      <literal>RETURNING</literal>, <application>PL/pgSQL</application> reports
      an error for more than one returned row, even when
      <literal>STRICT</literal> is not specified.  This is because there
diff --git a/doc/src/sgml/queries.sgml b/doc/src/sgml/queries.sgml
new file mode 100644
index 648b283..372cce1
--- a/doc/src/sgml/queries.sgml
+++ b/doc/src/sgml/queries.sgml
@@ -2063,9 +2063,10 @@ SELECT <replaceable>select_list</replace
    Table Expressions or <acronym>CTE</acronym>s, can be thought of as defining
    temporary tables that exist just for one query.  Each auxiliary statement
    in a <literal>WITH</literal> clause can be a <command>SELECT</command>,
-   <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>; and the
+   <command>INSERT</command>, <command>UPDATE</command>, <command>DELETE</command>,
+   or <command>MERGE</command>; and the
    <literal>WITH</literal> clause itself is attached to a primary statement that can
-   be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
+   also be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
    <command>DELETE</command>, or <command>MERGE</command>.
   </para>
 
@@ -2599,8 +2600,8 @@ SELECT * FROM w AS w1 JOIN w AS w2 ON w1
    <title>Data-Modifying Statements in <literal>WITH</literal></title>
 
    <para>
-    You can use most data-modifying statements (<command>INSERT</command>,
-    <command>UPDATE</command>, or <command>DELETE</command>, but not
+    You can use data-modifying statements (<command>INSERT</command>,
+    <command>UPDATE</command>, <command>DELETE</command>, or
     <command>MERGE</command>) in <literal>WITH</literal>.  This
     allows you to perform several different operations in the same query.
     An example is:
diff --git a/doc/src/sgml/ref/copy.sgml b/doc/src/sgml/ref/copy.sgml
new file mode 100644
index 55764fc..6c83e30
--- a/doc/src/sgml/ref/copy.sgml
+++ b/doc/src/sgml/ref/copy.sgml
@@ -130,16 +130,19 @@ COPY { <replaceable class="parameter">ta
       A <link linkend="sql-select"><command>SELECT</command></link>,
       <link linkend="sql-values"><command>VALUES</command></link>,
       <link linkend="sql-insert"><command>INSERT</command></link>,
-      <link linkend="sql-update"><command>UPDATE</command></link>, or
-      <link linkend="sql-delete"><command>DELETE</command></link> command whose results are to be
-      copied.  Note that parentheses are required around the query.
+      <link linkend="sql-update"><command>UPDATE</command></link>,
+      <link linkend="sql-delete"><command>DELETE</command></link>, or
+      <link linkend="sql-merge"><command>MERGE</command></link> command
+      whose results are to be copied.  Note that parentheses are required
+      around the query.
      </para>
      <para>
-      For <command>INSERT</command>, <command>UPDATE</command> and
-      <command>DELETE</command> queries a <literal>RETURNING</literal> clause
-      must be provided, and the target relation must not have a conditional
-      rule, nor an <literal>ALSO</literal> rule, nor an
-      <literal>INSTEAD</literal> rule that expands to multiple statements.
+      For <command>INSERT</command>, <command>UPDATE</command>,
+      <command>DELETE</command>, and <command>MERGE</command> queries a
+      <literal>RETURNING</literal> clause must be provided, and the target
+      relation must not have a conditional rule, nor an
+      <literal>ALSO</literal> rule, nor an <literal>INSTEAD</literal> rule
+      that expands to multiple statements.
      </para>
     </listitem>
    </varlistentry>
diff --git a/doc/src/sgml/ref/merge.sgml b/doc/src/sgml/ref/merge.sgml
new file mode 100644
index e745fbd..71feb66
--- a/doc/src/sgml/ref/merge.sgml
+++ b/doc/src/sgml/ref/merge.sgml
@@ -25,6 +25,7 @@ PostgreSQL documentation
 MERGE INTO [ ONLY ] <replaceable class="parameter">target_table_name</replaceable> [ * ] [ [ AS ] <replaceable class="parameter">target_alias</replaceable> ]
 USING <replaceable class="parameter">data_source</replaceable> ON <replaceable class="parameter">join_condition</replaceable>
 <replaceable class="parameter">when_clause</replaceable> [...]
+[ RETURNING * | <replaceable class="parameter">output_expression</replaceable> [ [ AS ] <replaceable class="parameter">output_name</replaceable> ] [, ...] ]
 
 <phrase>where <replaceable class="parameter">data_source</replaceable> is:</phrase>
 
@@ -97,6 +98,18 @@ DELETE
   </para>
 
   <para>
+   The optional <literal>RETURNING</literal> clause causes <command>MERGE</command>
+   to compute and return value(s) based on each row inserted, updated, or
+   deleted.  Any expression using the source or target table's columns, or
+   the <xref linkend="merge_action"/> function can be computed.  When an
+   <command>INSERT</command> or <command>UPDATE</command> action is performed,
+   the new values of the target table's columns are used.  When a
+   <command>DELETE</command> is performed, the old values of the target table's
+   columns are used.  The syntax of the <literal>RETURNING</literal> list is
+   identical to that of the output list of <command>SELECT</command>.
+  </para>
+
+  <para>
    There is no separate <literal>MERGE</literal> privilege.
    If you specify an update action, you must have the
    <literal>UPDATE</literal> privilege on the column(s)
@@ -442,6 +455,36 @@ DELETE
     </listitem>
    </varlistentry>
 
+   <varlistentry>
+    <term><replaceable class="parameter">output_expression</replaceable></term>
+    <listitem>
+     <para>
+      An expression to be computed and returned by the <command>MERGE</command>
+      command after each row is changed (whether inserted, updated, or deleted).
+      The expression can use any columns of the source or target tables, or the
+      <xref linkend="merge_action"/> function to return additional information
+      about the action executed.
+     </para>
+     <para>
+      Writing <literal>*</literal> will return all columns from the source
+      table, followed by all columns from the target table.  Often this will
+      lead to a lot of duplication, since it is common for the source and
+      target tables to have a lot of the same columns.  This can be avoided by
+      qualifying the <literal>*</literal> with the name or alias of the source
+      or target table.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
+    <term><replaceable class="parameter">output_name</replaceable></term>
+    <listitem>
+     <para>
+      A name to use for a returned column.
+     </para>
+    </listitem>
+   </varlistentry>
+
   </variablelist>
  </refsect1>
 
@@ -460,6 +503,13 @@ MERGE <replaceable class="parameter">tot
    were changed in any way.
   </para>
 
+  <para>
+   If the <command>MERGE</command> command contains a <literal>RETURNING</literal>
+   clause, the result will be similar to that of a <command>SELECT</command>
+   statement containing the columns and values defined in the
+   <literal>RETURNING</literal> list, computed over the row(s) inserted, updated,
+   or deleted by the command.
+  </para>
  </refsect1>
 
  <refsect1>
@@ -580,13 +630,6 @@ MERGE <replaceable class="parameter">tot
   </para>
 
   <para>
-   There is no <literal>RETURNING</literal> clause with
-   <command>MERGE</command>.  Actions of <command>INSERT</command>,
-   <command>UPDATE</command> and <command>DELETE</command> cannot contain
-   <literal>RETURNING</literal> or <literal>WITH</literal> clauses.
-  </para>
-
-  <para>
    When <command>MERGE</command> is run concurrently with other commands
    that modify the target table, the usual transaction isolation rules
    apply; see <xref linkend="transaction-iso"/> for an explanation
@@ -638,7 +681,8 @@ WHEN NOT MATCHED THEN
   <para>
    Attempt to insert a new stock item along with the quantity of stock. If
    the item already exists, instead update the stock count of the existing
-   item. Don't allow entries that have zero stock.
+   item. Don't allow entries that have zero stock. Return details of all
+   changes made.
 <programlisting>
 MERGE INTO wines w
 USING wine_stock_changes s
@@ -648,7 +692,8 @@ WHEN NOT MATCHED AND s.stock_delta > 0 T
 WHEN MATCHED AND w.stock + s.stock_delta > 0 THEN
   UPDATE SET stock = w.stock + s.stock_delta
 WHEN MATCHED THEN
-  DELETE;
+  DELETE
+RETURNING merge_action(), w.*;
 </programlisting>
 
    The <literal>wine_stock_changes</literal> table might be, for example, a
@@ -663,8 +708,9 @@ WHEN MATCHED THEN
     This command conforms to the <acronym>SQL</acronym> standard.
   </para>
    <para>
-    The <literal>WITH</literal> clause and <literal>DO NOTHING</literal>
-    action are extensions to the <acronym>SQL</acronym> standard.
+    The <literal>WITH</literal> clause, <literal>DO NOTHING</literal> action,
+    and <literal>RETURNING</literal> clause are extensions to the
+    <acronym>SQL</acronym> standard.
   </para>
  </refsect1>
 </refentry>
diff --git a/doc/src/sgml/ref/select.sgml b/doc/src/sgml/ref/select.sgml
new file mode 100644
index 9917df7..066aed4
--- a/doc/src/sgml/ref/select.sgml
+++ b/doc/src/sgml/ref/select.sgml
@@ -74,7 +74,7 @@ SELECT [ ALL | DISTINCT [ ON ( <replacea
 
 <phrase>and <replaceable class="parameter">with_query</replaceable> is:</phrase>
 
-    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> )
+    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> | <replaceable class="parameter">merge</replaceable> )
         [ SEARCH { BREADTH | DEPTH } FIRST BY <replaceable>column_name</replaceable> [, ...] SET <replaceable>search_seq_col_name</replaceable> ]
         [ CYCLE <replaceable>column_name</replaceable> [, ...] SET <replaceable>cycle_mark_col_name</replaceable> [ TO <replaceable>cycle_mark_value</replaceable> DEFAULT <replaceable>cycle_mark_default</replaceable> ] USING <replaceable>cycle_path_col_name</replaceable> ]
 
@@ -230,10 +230,10 @@ TABLE [ ONLY ] <replaceable class="param
     The subqueries effectively act as temporary tables or views
     for the duration of the primary query.
     Each subquery can be a <command>SELECT</command>, <command>TABLE</command>, <command>VALUES</command>,
-    <command>INSERT</command>, <command>UPDATE</command> or
-    <command>DELETE</command> statement.
+    <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command> statement.
     When writing a data-modifying statement (<command>INSERT</command>,
-    <command>UPDATE</command> or <command>DELETE</command>) in
+    <command>UPDATE</command>, <command>DELETE</command>, or <command>MERGE</command>) in
     <literal>WITH</literal>, it is usual to include a <literal>RETURNING</literal> clause.
     It is the output of <literal>RETURNING</literal>, <emphasis>not</emphasis> the underlying
     table that the statement modifies, that forms the temporary table that is
@@ -2184,7 +2184,8 @@ SELECT 2+2;
 
    <para>
     <productname>PostgreSQL</productname> allows <command>INSERT</command>,
-    <command>UPDATE</command>, and <command>DELETE</command> to be used as <literal>WITH</literal>
+    <command>UPDATE</command>, <command>DELETE</command>, and
+    <command>MERGE</command> to be used as <literal>WITH</literal>
     queries.  This is not found in the SQL standard.
    </para>
   </refsect2>
diff --git a/doc/src/sgml/rowtypes.sgml b/doc/src/sgml/rowtypes.sgml
new file mode 100644
index 4d86f97..bbeac84
--- a/doc/src/sgml/rowtypes.sgml
+++ b/doc/src/sgml/rowtypes.sgml
@@ -348,7 +348,7 @@ SELECT m.* FROM some_table, LATERAL myfu
    column expansion of this kind when it appears at the top level of
    a <link linkend="queries-select-lists"><command>SELECT</command> output
    list</link>, a <link linkend="dml-returning"><literal>RETURNING</literal>
-   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>,
+   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>,
    a <link linkend="queries-values"><literal>VALUES</literal> clause</link>, or
    a <link linkend="sql-syntax-row-constructors">row constructor</link>.
    In all other contexts (including when nested inside one of those
diff --git a/doc/src/sgml/spi.sgml b/doc/src/sgml/spi.sgml
new file mode 100644
index 47f4b5b..bb37786
--- a/doc/src/sgml/spi.sgml
+++ b/doc/src/sgml/spi.sgml
@@ -301,8 +301,9 @@ SPI_execute("INSERT INTO foo SELECT * FR
    is returned in the global variable <varname>SPI_processed</varname>.
    If the return value of the function is <symbol>SPI_OK_SELECT</symbol>,
    <symbol>SPI_OK_INSERT_RETURNING</symbol>,
-   <symbol>SPI_OK_DELETE_RETURNING</symbol>, or
-   <symbol>SPI_OK_UPDATE_RETURNING</symbol>,
+   <symbol>SPI_OK_DELETE_RETURNING</symbol>,
+   <symbol>SPI_OK_UPDATE_RETURNING</symbol>, or
+   <symbol>SPI_OK_MERGE_RETURNING</symbol>,
    then you can use the
    global pointer <literal>SPITupleTable *SPI_tuptable</literal> to
    access the result rows.  Some utility commands (such as
@@ -472,6 +473,15 @@ typedef struct SPITupleTable
       </para>
      </listitem>
     </varlistentry>
+
+    <varlistentry>
+     <term><symbol>SPI_OK_MERGE_RETURNING</symbol></term>
+     <listitem>
+      <para>
+       if a <command>MERGE RETURNING</command> was executed
+      </para>
+     </listitem>
+    </varlistentry>
 
     <varlistentry>
      <term><symbol>SPI_OK_UTILITY</symbol></term>
diff --git a/doc/src/sgml/xfunc.sgml b/doc/src/sgml/xfunc.sgml
new file mode 100644
index 8a79ad0..7d05369
--- a/doc/src/sgml/xfunc.sgml
+++ b/doc/src/sgml/xfunc.sgml
@@ -177,7 +177,8 @@
     statements separated by semicolons.  A semicolon after the last
     statement is optional.  Unless the function is declared to return
     <type>void</type>, the last statement must be a <command>SELECT</command>,
-    or an <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>
+    or an <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command>
     that has a <literal>RETURNING</literal> clause.
    </para>
 
@@ -1236,8 +1237,9 @@ SELECT x, CASE WHEN x &gt; 0 THEN genera
 
     <note>
      <para>
-      If a function's last command is <command>INSERT</command>, <command>UPDATE</command>,
-      or <command>DELETE</command> with <literal>RETURNING</literal>, that command will
+      If a function's last command is <command>INSERT</command>,
+      <command>UPDATE</command>, <command>DELETE</command>, or
+      <command>MERGE</command> with <literal>RETURNING</literal>, that command will
       always be executed to completion, even if the function is not declared
       with <literal>SETOF</literal> or the calling query does not fetch all the
       result rows.  Any extra rows produced by the <literal>RETURNING</literal>
diff --git a/src/backend/commands/copy.c b/src/backend/commands/copy.c
new file mode 100644
index 056b673..28cf8b0
--- a/src/backend/commands/copy.c
+++ b/src/backend/commands/copy.c
@@ -281,12 +281,6 @@ DoCopy(ParseState *pstate, const CopyStm
 	{
 		Assert(stmt->query);
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(stmt->query, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in COPY"));
-
 		query = makeNode(RawStmt);
 		query->stmt = stmt->query;
 		query->stmt_location = stmt_location;
diff --git a/src/backend/commands/copyto.c b/src/backend/commands/copyto.c
new file mode 100644
index a6962e0..ae8b2e3
--- a/src/backend/commands/copyto.c
+++ b/src/backend/commands/copyto.c
@@ -503,7 +503,8 @@ BeginCopyTo(ParseState *pstate,
 		{
 			Assert(query->commandType == CMD_INSERT ||
 				   query->commandType == CMD_UPDATE ||
-				   query->commandType == CMD_DELETE);
+				   query->commandType == CMD_DELETE ||
+				   query->commandType == CMD_MERGE);
 
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
diff --git a/src/backend/executor/execExpr.c b/src/backend/executor/execExpr.c
new file mode 100644
index ffd3ca4..728c8d5
--- a/src/backend/executor/execExpr.c
+++ b/src/backend/executor/execExpr.c
@@ -1107,6 +1107,19 @@ ExecInitExprRec(Expr *node, ExprState *s
 				break;
 			}
 
+		case T_MergeSupportFunc:
+			{
+				/* must be in a MERGE, else something messed up */
+				if (!state->parent ||
+					!IsA(state->parent, ModifyTableState) ||
+					((ModifyTableState *) state->parent)->operation != CMD_MERGE)
+					elog(ERROR, "MergeSupportFunc found in non-merge plan node");
+
+				scratch.opcode = EEOP_MERGE_SUPPORT_FUNC;
+				ExprEvalPushStep(state, &scratch);
+				break;
+			}
+
 		case T_SubscriptingRef:
 			{
 				SubscriptingRef *sbsref = (SubscriptingRef *) node;
diff --git a/src/backend/executor/execExprInterp.c b/src/backend/executor/execExprInterp.c
new file mode 100644
index 7c1f51e..a25ab75
--- a/src/backend/executor/execExprInterp.c
+++ b/src/backend/executor/execExprInterp.c
@@ -484,6 +484,7 @@ ExecInterpExpr(ExprState *state, ExprCon
 		&&CASE_EEOP_AGGREF,
 		&&CASE_EEOP_GROUPING_FUNC,
 		&&CASE_EEOP_WINDOW_FUNC,
+		&&CASE_EEOP_MERGE_SUPPORT_FUNC,
 		&&CASE_EEOP_SUBPLAN,
 		&&CASE_EEOP_AGG_STRICT_DESERIALIZE,
 		&&CASE_EEOP_AGG_DESERIALIZE,
@@ -1592,6 +1593,14 @@ ExecInterpExpr(ExprState *state, ExprCon
 			EEO_NEXT();
 		}
 
+		EEO_CASE(EEOP_MERGE_SUPPORT_FUNC)
+		{
+			/* too complex/uncommon for an inline implementation */
+			ExecEvalMergeSupportFunc(state, op, econtext);
+
+			EEO_NEXT();
+		}
+
 		EEO_CASE(EEOP_SUBPLAN)
 		{
 			/* too complex for an inline implementation */
@@ -4246,6 +4255,45 @@ ExecEvalGroupingFunc(ExprState *state, E
 }
 
 /*
+ * ExecEvalMergeSupportFunc
+ *
+ * Returns information about the current MERGE action for its RETURNING list.
+ */
+void
+ExecEvalMergeSupportFunc(ExprState *state, ExprEvalStep *op,
+						 ExprContext *econtext)
+{
+	ModifyTableState *mtstate = castNode(ModifyTableState, state->parent);
+	MergeActionState *relaction = mtstate->mt_merge_action;
+
+	if (!relaction)
+		elog(ERROR, "no merge action in progress");
+
+	/* Return the MERGE action ("INSERT", "UPDATE", or "DELETE") */
+	switch (relaction->mas_action->commandType)
+	{
+		case CMD_INSERT:
+			*op->resvalue = PointerGetDatum(cstring_to_text_with_len("INSERT", 6));
+			*op->resnull = false;
+			break;
+		case CMD_UPDATE:
+			*op->resvalue = PointerGetDatum(cstring_to_text_with_len("UPDATE", 6));
+			*op->resnull = false;
+			break;
+		case CMD_DELETE:
+			*op->resvalue = PointerGetDatum(cstring_to_text_with_len("DELETE", 6));
+			*op->resnull = false;
+			break;
+		case CMD_NOTHING:
+			elog(ERROR, "unexpected merge action: DO NOTHING");
+			break;
+		default:
+			elog(ERROR, "unrecognized commandType: %d",
+				 (int) relaction->mas_action->commandType);
+	}
+}
+
+/*
  * Hand off evaluation of a subplan to nodeSubplan.c
  */
 void
diff --git a/src/backend/executor/execPartition.c b/src/backend/executor/execPartition.c
new file mode 100644
index 8ca512d..64fcb01
--- a/src/backend/executor/execPartition.c
+++ b/src/backend/executor/execPartition.c
@@ -609,8 +609,8 @@ ExecInitPartitionInfo(ModifyTableState *
 	 * Build the RETURNING projection for the partition.  Note that we didn't
 	 * build the returningList for partitions within the planner, but simple
 	 * translation of varattnos will suffice.  This only occurs for the INSERT
-	 * case or in the case of UPDATE tuple routing where we didn't find a
-	 * result rel to reuse.
+	 * case or in the case of UPDATE/MERGE tuple routing where we didn't find
+	 * a result rel to reuse.
 	 */
 	if (node && node->returningLists != NIL)
 	{
@@ -619,12 +619,14 @@ ExecInitPartitionInfo(ModifyTableState *
 		List	   *returningList;
 
 		/* See the comment above for WCO lists. */
-		/* (except no RETURNING support for MERGE yet) */
 		Assert((node->operation == CMD_INSERT &&
 				list_length(node->returningLists) == 1 &&
 				list_length(node->resultRelations) == 1) ||
 			   (node->operation == CMD_UPDATE &&
 				list_length(node->returningLists) ==
+				list_length(node->resultRelations)) ||
+			   (node->operation == CMD_MERGE &&
+				list_length(node->returningLists) ==
 				list_length(node->resultRelations)));
 
 		/*
diff --git a/src/backend/executor/functions.c b/src/backend/executor/functions.c
new file mode 100644
index 6e926ef..539cd0a
--- a/src/backend/executor/functions.c
+++ b/src/backend/executor/functions.c
@@ -1662,8 +1662,8 @@ check_sql_fn_retval(List *queryTreeLists
 
 	/*
 	 * If it's a plain SELECT, it returns whatever the targetlist says.
-	 * Otherwise, if it's INSERT/UPDATE/DELETE with RETURNING, it returns
-	 * that. Otherwise, the function return type must be VOID.
+	 * Otherwise, if it's INSERT/UPDATE/DELETE/MERGE with RETURNING, it
+	 * returns that. Otherwise, the function return type must be VOID.
 	 *
 	 * Note: eventually replace this test with QueryReturnsTuples?	We'd need
 	 * a more general method of determining the output type, though.  Also, it
@@ -1681,7 +1681,8 @@ check_sql_fn_retval(List *queryTreeLists
 	else if (parse &&
 			 (parse->commandType == CMD_INSERT ||
 			  parse->commandType == CMD_UPDATE ||
-			  parse->commandType == CMD_DELETE) &&
+			  parse->commandType == CMD_DELETE ||
+			  parse->commandType == CMD_MERGE) &&
 			 parse->returningList)
 	{
 		tlist = parse->returningList;
@@ -1695,7 +1696,7 @@ check_sql_fn_retval(List *queryTreeLists
 				(errcode(ERRCODE_INVALID_FUNCTION_DEFINITION),
 				 errmsg("return type mismatch in function declared to return %s",
 						format_type_be(rettype)),
-				 errdetail("Function's final statement must be SELECT or INSERT/UPDATE/DELETE RETURNING.")));
+				 errdetail("Function's final statement must be SELECT or INSERT/UPDATE/DELETE/MERGE RETURNING.")));
 		return false;			/* keep compiler quiet */
 	}
 
diff --git a/src/backend/executor/nodeModifyTable.c b/src/backend/executor/nodeModifyTable.c
new file mode 100644
index 9351fbc..b8a75c5
--- a/src/backend/executor/nodeModifyTable.c
+++ b/src/backend/executor/nodeModifyTable.c
@@ -36,8 +36,7 @@
  *		RETURNING tuple after completing each row insert, update, or delete.
  *		It must be called again to continue the operation.  Without RETURNING,
  *		we just loop within the node until all the work is done, then
- *		return NULL.  This avoids useless call/return overhead.  (MERGE does
- *		not support RETURNING.)
+ *		return NULL.  This avoids useless call/return overhead.
  */
 
 #include "postgres.h"
@@ -85,9 +84,6 @@ typedef struct ModifyTableContext
 	 */
 	TupleTableSlot *planSlot;
 
-	/* MERGE specific */
-	MergeActionState *relaction;	/* MERGE action in progress */
-
 	/*
 	 * Information about the changes that were made concurrently to a tuple
 	 * being updated or deleted
@@ -150,14 +146,15 @@ static TupleTableSlot *ExecMerge(ModifyT
 								 HeapTuple oldtuple,
 								 bool canSetTag);
 static void ExecInitMerge(ModifyTableState *mtstate, EState *estate);
-static bool ExecMergeMatched(ModifyTableContext *context,
-							 ResultRelInfo *resultRelInfo,
-							 ItemPointer tupleid,
-							 HeapTuple oldtuple,
-							 bool canSetTag);
-static void ExecMergeNotMatched(ModifyTableContext *context,
-								ResultRelInfo *resultRelInfo,
-								bool canSetTag);
+static TupleTableSlot *ExecMergeMatched(ModifyTableContext *context,
+										ResultRelInfo *resultRelInfo,
+										ItemPointer tupleid,
+										HeapTuple oldtuple,
+										bool canSetTag,
+										bool *matched);
+static TupleTableSlot *ExecMergeNotMatched(ModifyTableContext *context,
+										   ResultRelInfo *resultRelInfo,
+										   bool canSetTag);
 
 
 /*
@@ -977,7 +974,7 @@ ExecInsert(ModifyTableContext *context,
 		if (mtstate->operation == CMD_UPDATE)
 			wco_kind = WCO_RLS_UPDATE_CHECK;
 		else if (mtstate->operation == CMD_MERGE)
-			wco_kind = (context->relaction->mas_action->commandType == CMD_UPDATE) ?
+			wco_kind = (mtstate->mt_merge_action->mas_action->commandType == CMD_UPDATE) ?
 				WCO_RLS_UPDATE_CHECK : WCO_RLS_INSERT_CHECK;
 		else
 			wco_kind = WCO_RLS_INSERT_CHECK;
@@ -1831,7 +1828,7 @@ ExecCrossPartitionUpdate(ModifyTableCont
 		 * additional rechecking, and might end up executing a different
 		 * action entirely).
 		 */
-		if (context->relaction != NULL)
+		if (mtstate->operation == CMD_MERGE)
 			return *tmresult == TM_Ok;
 		else if (TupIsNull(epqslot))
 			return true;
@@ -2072,7 +2069,7 @@ lreplace:
 		 * No luck, a retry is needed.  If running MERGE, we do not do so
 		 * here; instead let it handle that on its own rules.
 		 */
-		if (context->relaction != NULL)
+		if (context->mtstate->operation == CMD_MERGE)
 			return result;
 
 		/*
@@ -2713,6 +2710,7 @@ static TupleTableSlot *
 ExecMerge(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 		  ItemPointer tupleid, HeapTuple oldtuple, bool canSetTag)
 {
+	TupleTableSlot *rslot = NULL;
 	bool		matched;
 
 	/*-----
@@ -2761,19 +2759,18 @@ ExecMerge(ModifyTableContext *context, R
 	 */
 	matched = tupleid != NULL || oldtuple != NULL;
 	if (matched)
-		matched = ExecMergeMatched(context, resultRelInfo, tupleid, oldtuple,
-								   canSetTag);
+		rslot = ExecMergeMatched(context, resultRelInfo, tupleid, oldtuple,
+								 canSetTag, &matched);
 
 	/*
-	 * Either we were dealing with a NOT MATCHED tuple or ExecMergeMatched()
-	 * returned "false", indicating the previously MATCHED tuple no longer
-	 * matches.
+	 * Deal with the NOT MATCHED case (either a NOT MATCHED tuple from the
+	 * join, or a previously MATCHED tuple for which ExecMergeMatched() set
+	 * "matched" to false, indicating that it no longer matches).
 	 */
 	if (!matched)
-		ExecMergeNotMatched(context, resultRelInfo, canSetTag);
+		rslot = ExecMergeNotMatched(context, resultRelInfo, canSetTag);
 
-	/* No RETURNING support yet */
-	return NULL;
+	return rslot;
 }
 
 /*
@@ -2785,8 +2782,8 @@ ExecMerge(ModifyTableContext *context, R
  * We start from the first WHEN MATCHED action and check if the WHEN quals
  * pass, if any. If the WHEN quals for the first action do not pass, we
  * check the second, then the third and so on. If we reach to the end, no
- * action is taken and we return true, indicating that no further action is
- * required for this tuple.
+ * action is taken and "matched" is set to true, indicating that no further
+ * action is required for this tuple.
  *
  * If we do find a qualifying action, then we attempt to execute the action.
  *
@@ -2795,16 +2792,18 @@ ExecMerge(ModifyTableContext *context, R
  * with individual actions are evaluated by this routine via ExecQual, while
  * EvalPlanQual checks for the join quals. If EvalPlanQual tells us that the
  * updated tuple still passes the join quals, then we restart from the first
- * action to look for a qualifying action. Otherwise, we return false --
- * meaning that a NOT MATCHED action must now be executed for the current
- * source tuple.
+ * action to look for a qualifying action. Otherwise, "matched" is set to
+ * false -- meaning that a NOT MATCHED action must now be executed for the
+ * current source tuple.
  */
-static bool
+static TupleTableSlot *
 ExecMergeMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
-				 ItemPointer tupleid, HeapTuple oldtuple, bool canSetTag)
+				 ItemPointer tupleid, HeapTuple oldtuple, bool canSetTag,
+				 bool *matched)
 {
 	ModifyTableState *mtstate = context->mtstate;
-	TupleTableSlot *newslot;
+	TupleTableSlot *rslot = NULL;
+	TupleTableSlot *newslot = NULL;
 	EState	   *estate = context->estate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
 	bool		isNull;
@@ -2815,7 +2814,10 @@ ExecMergeMatched(ModifyTableContext *con
 	 * If there are no WHEN MATCHED actions, we are done.
 	 */
 	if (resultRelInfo->ri_matchedMergeAction == NIL)
-		return true;
+	{
+		*matched = true;
+		return NULL;
+	}
 
 	/*
 	 * Make tuple and any needed join variables available to ExecQual and
@@ -2905,12 +2907,15 @@ lmerge_matched:
 				 */
 				newslot = ExecProject(relaction->mas_proj);
 
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				if (!ExecUpdatePrologue(context, resultRelInfo,
 										tupleid, NULL, newslot, &result))
 				{
 					if (result == TM_Ok)
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 					break;		/* concurrent update/delete */
 				}
 
@@ -2920,7 +2925,10 @@ lmerge_matched:
 				{
 					if (!ExecIRUpdateTriggers(estate, resultRelInfo,
 											  oldtuple, newslot))
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 				}
 				else
 				{
@@ -2933,12 +2941,15 @@ lmerge_matched:
 					 * cross-partition update was done, then there's nothing
 					 * else for us to do --- the UPDATE has been turned into a
 					 * DELETE and an INSERT, and we must not perform any of
-					 * the usual post-update tasks.
+					 * the usual post-update tasks.  Also, the RETURNING tuple
+					 * (if any) has been projected, so we can just return
+					 * that.
 					 */
 					if (updateCxt.crossPartUpdate)
 					{
 						mtstate->mt_merge_updated += 1;
-						return true;
+						*matched = true;
+						return context->cpUpdateReturningSlot;
 					}
 				}
 
@@ -2951,12 +2962,15 @@ lmerge_matched:
 				break;
 
 			case CMD_DELETE:
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				if (!ExecDeletePrologue(context, resultRelInfo, tupleid,
 										NULL, NULL, &result))
 				{
 					if (result == TM_Ok)
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 					break;		/* concurrent update/delete */
 				}
 
@@ -2966,7 +2980,10 @@ lmerge_matched:
 				{
 					if (!ExecIRDeleteTriggers(estate, resultRelInfo,
 											  oldtuple))
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 				}
 				else
 					result = ExecDeleteAct(context, resultRelInfo, tupleid,
@@ -3046,7 +3063,8 @@ lmerge_matched:
 				 * If the tuple was already deleted, return to let caller
 				 * handle it under NOT MATCHED clauses.
 				 */
-				return false;
+				*matched = false;
+				return NULL;
 
 			case TM_Updated:
 				{
@@ -3092,13 +3110,19 @@ lmerge_matched:
 							 * NOT MATCHED actions.
 							 */
 							if (TupIsNull(epqslot))
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							(void) ExecGetJunkAttribute(epqslot,
 														resultRelInfo->ri_RowIdAttNo,
 														&isNull);
 							if (isNull)
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							/*
 							 * When a tuple was updated and migrated to
@@ -3133,7 +3157,8 @@ lmerge_matched:
 							 * tuple already deleted; tell caller to run NOT
 							 * MATCHED actions
 							 */
-							return false;
+							*matched = false;
+							return NULL;
 
 						case TM_SelfModified:
 
@@ -3161,13 +3186,13 @@ lmerge_matched:
 
 							/* This shouldn't happen */
 							elog(ERROR, "attempted to update or delete invisible tuple");
-							return false;
+							return NULL;
 
 						default:
 							/* see table_tuple_lock call in ExecDelete() */
 							elog(ERROR, "unexpected table_tuple_lock status: %u",
 								 result);
-							return false;
+							return NULL;
 					}
 				}
 
@@ -3179,6 +3204,31 @@ lmerge_matched:
 				break;
 		}
 
+		/* Process RETURNING if present */
+		if (resultRelInfo->ri_projectReturning)
+		{
+			switch (commandType)
+			{
+				case CMD_UPDATE:
+					rslot = ExecProcessReturning(resultRelInfo, newslot,
+												 context->planSlot);
+					break;
+
+				case CMD_DELETE:
+					rslot = ExecProcessReturning(resultRelInfo,
+												 resultRelInfo->ri_oldTupleSlot,
+												 context->planSlot);
+					break;
+
+				case CMD_NOTHING:
+					break;
+
+				default:
+					elog(ERROR, "unrecognized commandType: %d",
+						 (int) commandType);
+			}
+		}
+
 		/*
 		 * We've activated one of the WHEN clauses, so we don't search
 		 * further. This is required behaviour, not an optimization.
@@ -3189,19 +3239,22 @@ lmerge_matched:
 	/*
 	 * Successfully executed an action or no qualifying action was found.
 	 */
-	return true;
+	*matched = true;
+
+	return rslot;
 }
 
 /*
  * Execute the first qualifying NOT MATCHED action.
  */
-static void
+static TupleTableSlot *
 ExecMergeNotMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 					bool canSetTag)
 {
 	ModifyTableState *mtstate = context->mtstate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
 	List	   *actionStates = NIL;
+	TupleTableSlot *rslot = NULL;
 	ListCell   *l;
 
 	/*
@@ -3251,10 +3304,10 @@ ExecMergeNotMatched(ModifyTableContext *
 				 * so we don't need to map the tuple here.
 				 */
 				newslot = ExecProject(action->mas_proj);
-				context->relaction = action;
+				mtstate->mt_merge_action = action;
 
-				(void) ExecInsert(context, mtstate->rootResultRelInfo, newslot,
-								  canSetTag, NULL, NULL);
+				rslot = ExecInsert(context, mtstate->rootResultRelInfo,
+								   newslot, canSetTag, NULL, NULL);
 				mtstate->mt_merge_inserted += 1;
 				break;
 			case CMD_NOTHING:
@@ -3270,6 +3323,8 @@ ExecMergeNotMatched(ModifyTableContext *
 		 */
 		break;
 	}
+
+	return rslot;
 }
 
 /*
@@ -3732,9 +3787,17 @@ ExecModifyTable(PlanState *pstate)
 				{
 					EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-					ExecMerge(&context, node->resultRelInfo, NULL, NULL,
-							  node->canSetTag);
-					continue;	/* no RETURNING support yet */
+					slot = ExecMerge(&context, node->resultRelInfo,
+									 NULL, NULL, node->canSetTag);
+
+					/*
+					 * If we got a RETURNING result, return it to the caller.
+					 * We'll continue the work on next call.
+					 */
+					if (slot)
+						return slot;
+
+					continue;	/* continue with the next tuple */
 				}
 
 				elog(ERROR, "tableoid is NULL");
@@ -3811,9 +3874,17 @@ ExecModifyTable(PlanState *pstate)
 					{
 						EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-						ExecMerge(&context, node->resultRelInfo, NULL, NULL,
-								  node->canSetTag);
-						continue;	/* no RETURNING support yet */
+						slot = ExecMerge(&context, node->resultRelInfo,
+										 NULL, NULL, node->canSetTag);
+
+						/*
+						 * If we got a RETURNING result, return it to the
+						 * caller.  We'll continue the work on next call.
+						 */
+						if (slot)
+							return slot;
+
+						continue;	/* continue with the next tuple */
 					}
 
 					elog(ERROR, "ctid is NULL");
@@ -3860,9 +3931,17 @@ ExecModifyTable(PlanState *pstate)
 					{
 						EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-						ExecMerge(&context, node->resultRelInfo, NULL, NULL,
-								  node->canSetTag);
-						continue;	/* no RETURNING support yet */
+						slot = ExecMerge(&context, node->resultRelInfo,
+										 NULL, NULL, node->canSetTag);
+
+						/*
+						 * If we got a RETURNING result, return it to the
+						 * caller.  We'll continue the work on next call.
+						 */
+						if (slot)
+							return slot;
+
+						continue;	/* continue with the next tuple */
 					}
 
 					elog(ERROR, "wholerow is NULL");
@@ -3924,7 +4003,6 @@ ExecModifyTable(PlanState *pstate)
 				}
 				slot = ExecGetUpdateNewTuple(resultRelInfo, context.planSlot,
 											 oldSlot);
-				context.relaction = NULL;
 
 				/* Now apply the update. */
 				slot = ExecUpdate(&context, resultRelInfo, tupleid, oldtuple,
diff --git a/src/backend/executor/spi.c b/src/backend/executor/spi.c
new file mode 100644
index 8585771..a97a7e3
--- a/src/backend/executor/spi.c
+++ b/src/backend/executor/spi.c
@@ -2032,6 +2032,8 @@ SPI_result_code_string(int code)
 			return "SPI_OK_TD_REGISTER";
 		case SPI_OK_MERGE:
 			return "SPI_OK_MERGE";
+		case SPI_OK_MERGE_RETURNING:
+			return "SPI_OK_MERGE_RETURNING";
 	}
 	/* Unrecognized code ... return something useful ... */
 	sprintf(buf, "Unrecognized SPI code %d", code);
@@ -2885,7 +2887,10 @@ _SPI_pquery(QueryDesc *queryDesc, bool f
 				res = SPI_OK_UPDATE;
 			break;
 		case CMD_MERGE:
-			res = SPI_OK_MERGE;
+			if (queryDesc->plannedstmt->hasReturning)
+				res = SPI_OK_MERGE_RETURNING;
+			else
+				res = SPI_OK_MERGE;
 			break;
 		default:
 			return SPI_ERROR_OPUNKNOWN;
diff --git a/src/backend/jit/llvm/llvmjit_expr.c b/src/backend/jit/llvm/llvmjit_expr.c
new file mode 100644
index 0c44842..2a7d84f
--- a/src/backend/jit/llvm/llvmjit_expr.c
+++ b/src/backend/jit/llvm/llvmjit_expr.c
@@ -1986,6 +1986,12 @@ llvm_compile_expr(ExprState *state)
 					break;
 				}
 
+			case EEOP_MERGE_SUPPORT_FUNC:
+				build_EvalXFunc(b, mod, "ExecEvalMergeSupportFunc",
+								v_state, op, v_econtext);
+				LLVMBuildBr(b, opblocks[opno + 1]);
+				break;
+
 			case EEOP_SUBPLAN:
 				build_EvalXFunc(b, mod, "ExecEvalSubPlan",
 								v_state, op, v_econtext);
diff --git a/src/backend/jit/llvm/llvmjit_types.c b/src/backend/jit/llvm/llvmjit_types.c
new file mode 100644
index 47c9daf..7d7aeee
--- a/src/backend/jit/llvm/llvmjit_types.c
+++ b/src/backend/jit/llvm/llvmjit_types.c
@@ -155,6 +155,7 @@ void	   *referenced_functions[] =
 	ExecEvalFuncExprFusage,
 	ExecEvalFuncExprStrictFusage,
 	ExecEvalGroupingFunc,
+	ExecEvalMergeSupportFunc,
 	ExecEvalMinMax,
 	ExecEvalNextValueExpr,
 	ExecEvalParamExec,
diff --git a/src/backend/nodes/nodeFuncs.c b/src/backend/nodes/nodeFuncs.c
new file mode 100644
index 6ba8e73..5b70280
--- a/src/backend/nodes/nodeFuncs.c
+++ b/src/backend/nodes/nodeFuncs.c
@@ -66,6 +66,9 @@ exprType(const Node *expr)
 		case T_WindowFunc:
 			type = ((const WindowFunc *) expr)->wintype;
 			break;
+		case T_MergeSupportFunc:
+			type = ((const MergeSupportFunc *) expr)->msftype;
+			break;
 		case T_SubscriptingRef:
 			type = ((const SubscriptingRef *) expr)->refrestype;
 			break;
@@ -809,6 +812,9 @@ exprCollation(const Node *expr)
 		case T_WindowFunc:
 			coll = ((const WindowFunc *) expr)->wincollid;
 			break;
+		case T_MergeSupportFunc:
+			coll = ((const MergeSupportFunc *) expr)->msfcollid;
+			break;
 		case T_SubscriptingRef:
 			coll = ((const SubscriptingRef *) expr)->refcollid;
 			break;
@@ -1084,6 +1090,9 @@ exprSetCollation(Node *expr, Oid collati
 		case T_WindowFunc:
 			((WindowFunc *) expr)->wincollid = collation;
 			break;
+		case T_MergeSupportFunc:
+			((MergeSupportFunc *) expr)->msfcollid = collation;
+			break;
 		case T_SubscriptingRef:
 			((SubscriptingRef *) expr)->refcollid = collation;
 			break;
@@ -1342,6 +1351,9 @@ exprLocation(const Node *expr)
 			/* function name should always be the first thing */
 			loc = ((const WindowFunc *) expr)->location;
 			break;
+		case T_MergeSupportFunc:
+			loc = ((const MergeSupportFunc *) expr)->location;
+			break;
 		case T_SubscriptingRef:
 			/* just use container argument's location */
 			loc = exprLocation((Node *) ((const SubscriptingRef *) expr)->refexpr);
@@ -2034,6 +2046,7 @@ expression_tree_walker_impl(Node *node,
 		case T_RangeTblRef:
 		case T_SortGroupClause:
 		case T_CTESearchClause:
+		case T_MergeSupportFunc:
 			/* primitive node types with no expression subnodes */
 			break;
 		case T_WithCheckOption:
@@ -2868,6 +2881,7 @@ expression_tree_mutator_impl(Node *node,
 		case T_RangeTblRef:
 		case T_SortGroupClause:
 		case T_CTESearchClause:
+		case T_MergeSupportFunc:
 			return (Node *) copyObject(node);
 		case T_WithCheckOption:
 			{
@@ -3832,6 +3846,7 @@ raw_expression_tree_walker_impl(Node *no
 		case T_ParamRef:
 		case T_A_Const:
 		case T_A_Star:
+		case T_MergeSupportFunc:
 			/* primitive node types with no subnodes */
 			break;
 		case T_Alias:
@@ -4052,6 +4067,8 @@ raw_expression_tree_walker_impl(Node *no
 					return true;
 				if (WALK(stmt->mergeWhenClauses))
 					return true;
+				if (WALK(stmt->returningList))
+					return true;
 				if (WALK(stmt->withClause))
 					return true;
 			}
diff --git a/src/backend/optimizer/plan/subselect.c b/src/backend/optimizer/plan/subselect.c
new file mode 100644
index 47e1472..b9e0c96
--- a/src/backend/optimizer/plan/subselect.c
+++ b/src/backend/optimizer/plan/subselect.c
@@ -1855,7 +1855,8 @@ convert_EXISTS_to_ANY(PlannerInfo *root,
 /*
  * Replace correlation vars (uplevel vars) with Params.
  *
- * Uplevel PlaceHolderVars and aggregates are replaced, too.
+ * Uplevel PlaceHolderVars, aggregates, GROUPING() expressions, and
+ * MergeSupportFuncs are replaced, too.
  *
  * Note: it is critical that this runs immediately after SS_process_sublinks.
  * Since we do not recurse into the arguments of uplevel PHVs and aggregates,
@@ -1909,6 +1910,12 @@ replace_correlation_vars_mutator(Node *n
 		if (((GroupingFunc *) node)->agglevelsup > 0)
 			return (Node *) replace_outer_grouping(root, (GroupingFunc *) node);
 	}
+	if (IsA(node, MergeSupportFunc))
+	{
+		if (root->parse->commandType != CMD_MERGE)
+			return (Node *) replace_outer_merge_support(root,
+														(MergeSupportFunc *) node);
+	}
 	return expression_tree_mutator(node,
 								   replace_correlation_vars_mutator,
 								   (void *) root);
diff --git a/src/backend/optimizer/util/paramassign.c b/src/backend/optimizer/util/paramassign.c
new file mode 100644
index a58da7c..f461fed
--- a/src/backend/optimizer/util/paramassign.c
+++ b/src/backend/optimizer/util/paramassign.c
@@ -308,6 +308,57 @@ replace_outer_grouping(PlannerInfo *root
 }
 
 /*
+ * Generate a Param node to replace the given MergeSupportFunc expression
+ * which is expected to be in the RETURNING list of an upper-level MERGE
+ * query.  Record the need for the MergeSupportFunc in the proper upper-level
+ * root->plan_params.
+ */
+Param *
+replace_outer_merge_support(PlannerInfo *root, MergeSupportFunc *msf)
+{
+	Param	   *retval;
+	PlannerParamItem *pitem;
+	Oid			ptype = exprType((Node *) msf);
+
+	Assert(root->parse->commandType != CMD_MERGE);
+
+	/*
+	 * The parser should have ensured that the MergeSupportFunc is in the
+	 * RETURNING list of an upper-level MERGE query, so find that query.
+	 */
+	do
+	{
+		root = root->parent_root;
+		if (root == NULL)
+			elog(ERROR, "MergeSupportFunc found outside MERGE");
+	} while (root->parse->commandType != CMD_MERGE);
+
+	/*
+	 * It does not seem worthwhile to try to de-duplicate references to outer
+	 * MergeSupportFunc expressions.  Just make a new slot every time.
+	 */
+	msf = copyObject(msf);
+
+	pitem = makeNode(PlannerParamItem);
+	pitem->item = (Node *) msf;
+	pitem->paramId = list_length(root->glob->paramExecTypes);
+	root->glob->paramExecTypes = lappend_oid(root->glob->paramExecTypes,
+											 ptype);
+
+	root->plan_params = lappend(root->plan_params, pitem);
+
+	retval = makeNode(Param);
+	retval->paramkind = PARAM_EXEC;
+	retval->paramid = pitem->paramId;
+	retval->paramtype = ptype;
+	retval->paramtypmod = -1;
+	retval->paramcollid = InvalidOid;
+	retval->location = msf->location;
+
+	return retval;
+}
+
+/*
  * Generate a Param node to replace the given Var,
  * which is expected to come from some upper NestLoop plan node.
  * Record the need for the Var in root->curOuterParams.
diff --git a/src/backend/parser/analyze.c b/src/backend/parser/analyze.c
new file mode 100644
index 2255314..7bd947b
--- a/src/backend/parser/analyze.c
+++ b/src/backend/parser/analyze.c
@@ -72,7 +72,6 @@ static void determineRecursiveColTypes(P
 									   Node *larg, List *nrtargetlist);
 static Query *transformReturnStmt(ParseState *pstate, ReturnStmt *stmt);
 static Query *transformUpdateStmt(ParseState *pstate, UpdateStmt *stmt);
-static List *transformReturningList(ParseState *pstate, List *returningList);
 static Query *transformPLAssignStmt(ParseState *pstate,
 									PLAssignStmt *stmt);
 static Query *transformDeclareCursorStmt(ParseState *pstate,
@@ -551,7 +550,8 @@ transformDeleteStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -978,7 +978,8 @@ transformInsertStmt(ParseState *pstate,
 	/* Process RETURNING, if any. */
 	if (stmt->returningList)
 		qry->returningList = transformReturningList(pstate,
-													stmt->returningList);
+													stmt->returningList,
+													EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -2443,7 +2444,8 @@ transformUpdateStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/*
 	 * Now we are done with SELECT-like processing, and can get on with
@@ -2540,10 +2542,11 @@ transformUpdateTargetList(ParseState *ps
 
 /*
  * transformReturningList -
- *	handle a RETURNING clause in INSERT/UPDATE/DELETE
+ *	handle a RETURNING clause in INSERT/UPDATE/DELETE/MERGE
  */
-static List *
-transformReturningList(ParseState *pstate, List *returningList)
+List *
+transformReturningList(ParseState *pstate, List *returningList,
+					   ParseExprKind exprKind)
 {
 	List	   *rlist;
 	int			save_next_resno;
@@ -2560,7 +2563,7 @@ transformReturningList(ParseState *pstat
 	pstate->p_next_resno = 1;
 
 	/* transform RETURNING identically to a SELECT targetlist */
-	rlist = transformTargetList(pstate, returningList, EXPR_KIND_RETURNING);
+	rlist = transformTargetList(pstate, returningList, exprKind);
 
 	/*
 	 * Complain if the nonempty tlist expanded to nothing (which is possible
diff --git a/src/backend/parser/gram.y b/src/backend/parser/gram.y
new file mode 100644
index c6e2f67..6419b32
--- a/src/backend/parser/gram.y
+++ b/src/backend/parser/gram.y
@@ -733,7 +733,7 @@ static Node *makeRecursiveViewSelect(cha
 	LEADING LEAKPROOF LEAST LEFT LEVEL LIKE LIMIT LISTEN LOAD LOCAL
 	LOCALTIME LOCALTIMESTAMP LOCATION LOCK_P LOCKED LOGGED
 
-	MAPPING MATCH MATCHED MATERIALIZED MAXVALUE MERGE METHOD
+	MAPPING MATCH MATCHED MATERIALIZED MAXVALUE MERGE MERGE_ACTION METHOD
 	MINUTE_P MINVALUE MODE MONTH_P MOVE
 
 	NAME_P NAMES NATIONAL NATURAL NCHAR NEW NEXT NFC NFD NFKC NFKD NO NONE
@@ -12374,6 +12374,7 @@ MergeStmt:
 			USING table_ref
 			ON a_expr
 			merge_when_list
+			returning_clause
 				{
 					MergeStmt  *m = makeNode(MergeStmt);
 
@@ -12382,6 +12383,7 @@ MergeStmt:
 					m->sourceRelation = $6;
 					m->joinCondition = $8;
 					m->mergeWhenClauses = $9;
+					m->returningList = $10;
 
 					$$ = (Node *) m;
 				}
@@ -15795,6 +15797,14 @@ func_expr_common_subexpr:
 					n->location = @1;
 					$$ = (Node *) n;
 				}
+			| MERGE_ACTION '(' ')'
+				{
+					MergeSupportFunc *m = makeNode(MergeSupportFunc);
+
+					m->msftype = TEXTOID;
+					m->location = @1;
+					$$ = (Node *) m;
+				}
 			;
 
 
@@ -17492,6 +17502,7 @@ col_name_keyword:
 			| JSON_SCALAR
 			| JSON_SERIALIZE
 			| LEAST
+			| MERGE_ACTION
 			| NATIONAL
 			| NCHAR
 			| NONE
@@ -17881,6 +17892,7 @@ bare_label_keyword:
 			| MATERIALIZED
 			| MAXVALUE
 			| MERGE
+			| MERGE_ACTION
 			| METHOD
 			| MINVALUE
 			| MODE
diff --git a/src/backend/parser/parse_agg.c b/src/backend/parser/parse_agg.c
new file mode 100644
index 9d151a8..bee7d83
--- a/src/backend/parser/parse_agg.c
+++ b/src/backend/parser/parse_agg.c
@@ -468,6 +468,7 @@ check_agglevels_and_constraints(ParseSta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
@@ -915,6 +916,7 @@ transformWindowFuncCall(ParseState *psta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_cte.c b/src/backend/parser/parse_cte.c
new file mode 100644
index 3c88c9a..6826d4f
--- a/src/backend/parser/parse_cte.c
+++ b/src/backend/parser/parse_cte.c
@@ -126,13 +126,6 @@ transformWithClause(ParseState *pstate,
 		CommonTableExpr *cte = (CommonTableExpr *) lfirst(lc);
 		ListCell   *rest;
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(cte->ctequery, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in WITH query"),
-					parser_errposition(pstate, cte->location));
-
 		for_each_cell(rest, withClause->ctes, lnext(withClause->ctes, lc))
 		{
 			CommonTableExpr *cte2 = (CommonTableExpr *) lfirst(rest);
@@ -153,7 +146,8 @@ transformWithClause(ParseState *pstate,
 			/* must be a data-modifying statement */
 			Assert(IsA(cte->ctequery, InsertStmt) ||
 				   IsA(cte->ctequery, UpdateStmt) ||
-				   IsA(cte->ctequery, DeleteStmt));
+				   IsA(cte->ctequery, DeleteStmt) ||
+				   IsA(cte->ctequery, MergeStmt));
 
 			pstate->p_hasModifyingCTE = true;
 		}
diff --git a/src/backend/parser/parse_expr.c b/src/backend/parser/parse_expr.c
new file mode 100644
index 9300c7b..fe56101
--- a/src/backend/parser/parse_expr.c
+++ b/src/backend/parser/parse_expr.c
@@ -54,6 +54,7 @@ static Node *transformAExprDistinct(Pars
 static Node *transformAExprNullIf(ParseState *pstate, A_Expr *a);
 static Node *transformAExprIn(ParseState *pstate, A_Expr *a);
 static Node *transformAExprBetween(ParseState *pstate, A_Expr *a);
+static Node *transformMergeSupportFunc(ParseState *pstate, MergeSupportFunc *f);
 static Node *transformBoolExpr(ParseState *pstate, BoolExpr *a);
 static Node *transformFuncCall(ParseState *pstate, FuncCall *fn);
 static Node *transformMultiAssignRef(ParseState *pstate, MultiAssignRef *maref);
@@ -227,6 +228,11 @@ transformExprRecurse(ParseState *pstate,
 			result = transformGroupingFunc(pstate, (GroupingFunc *) expr);
 			break;
 
+		case T_MergeSupportFunc:
+			result = transformMergeSupportFunc(pstate,
+											   (MergeSupportFunc *) expr);
+			break;
+
 		case T_NamedArgExpr:
 			{
 				NamedArgExpr *na = (NamedArgExpr *) expr;
@@ -541,6 +547,7 @@ transformColumnRef(ParseState *pstate, C
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CHECK_CONSTRAINT:
@@ -1354,6 +1361,32 @@ transformAExprBetween(ParseState *pstate
 }
 
 static Node *
+transformMergeSupportFunc(ParseState *pstate, MergeSupportFunc *f)
+{
+	/*
+	 * All we need to do is check that we're in the RETURNING list of a MERGE
+	 * command.  If so, we just return the node as-is.
+	 */
+	if (pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+	{
+		ParseState *parent_pstate = pstate->parentParseState;
+
+		while (parent_pstate &&
+			   parent_pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+			parent_pstate = parent_pstate->parentParseState;
+
+		if (!parent_pstate ||
+			parent_pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+			ereport(ERROR,
+					errcode(ERRCODE_SYNTAX_ERROR),
+					errmsg("MERGE_ACTION() can only be used in the RETURNING list of a MERGE command"),
+					parser_errposition(pstate, f->location));
+	}
+
+	return (Node *) f;
+}
+
+static Node *
 transformBoolExpr(ParseState *pstate, BoolExpr *a)
 {
 	List	   *args = NIL;
@@ -1767,6 +1800,7 @@ transformSubLink(ParseState *pstate, Sub
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CYCLE_MARK:
@@ -3115,6 +3149,7 @@ ParseExprKindName(ParseExprKind exprKind
 		case EXPR_KIND_OFFSET:
 			return "OFFSET";
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			return "RETURNING";
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
diff --git a/src/backend/parser/parse_func.c b/src/backend/parser/parse_func.c
new file mode 100644
index fdb3e6d..0cbc950
--- a/src/backend/parser/parse_func.c
+++ b/src/backend/parser/parse_func.c
@@ -2599,6 +2599,7 @@ check_srf_call_placement(ParseState *pst
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_merge.c b/src/backend/parser/parse_merge.c
new file mode 100644
index 6b99f55..efb739d
--- a/src/backend/parser/parse_merge.c
+++ b/src/backend/parser/parse_merge.c
@@ -234,6 +234,10 @@ transformMergeStmt(ParseState *pstate, M
 	 */
 	qry->jointree = makeFromExpr(pstate->p_joinlist, joinExpr);
 
+	/* Transform the RETURNING list, if any */
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_MERGE_RETURNING);
+
 	/*
 	 * We now have a good query shape, so now look at the WHEN conditions and
 	 * action targetlists.
@@ -391,9 +395,6 @@ transformMergeStmt(ParseState *pstate, M
 
 	qry->mergeActionList = mergeActionList;
 
-	/* RETURNING could potentially be added in the future, but not in SQL std */
-	qry->returningList = NULL;
-
 	qry->hasTargetSRFs = false;
 	qry->hasSubLinks = pstate->p_hasSubLinks;
 
diff --git a/src/backend/parser/parse_relation.c b/src/backend/parser/parse_relation.c
new file mode 100644
index 6f5d9e2..427b732
--- a/src/backend/parser/parse_relation.c
+++ b/src/backend/parser/parse_relation.c
@@ -2341,9 +2341,10 @@ addRangeTableEntryForCTE(ParseState *pst
 		cte->cterefcount++;
 
 	/*
-	 * We throw error if the CTE is INSERT/UPDATE/DELETE without RETURNING.
-	 * This won't get checked in case of a self-reference, but that's OK
-	 * because data-modifying CTEs aren't allowed to be recursive anyhow.
+	 * We throw error if the CTE is INSERT/UPDATE/DELETE/MERGE without
+	 * RETURNING.  This won't get checked in case of a self-reference, but
+	 * that's OK because data-modifying CTEs aren't allowed to be recursive
+	 * anyhow.
 	 */
 	if (IsA(cte->ctequery, Query))
 	{
diff --git a/src/backend/parser/parse_target.c b/src/backend/parser/parse_target.c
new file mode 100644
index f10fc42..8923be1
--- a/src/backend/parser/parse_target.c
+++ b/src/backend/parser/parse_target.c
@@ -1806,6 +1806,10 @@ FigureColnameInternal(Node *node, char *
 			/* make GROUPING() act like a regular function */
 			*name = "grouping";
 			return 2;
+		case T_MergeSupportFunc:
+			/* make MERGE_ACTION() act like a regular function */
+			*name = "merge_action";
+			return 2;
 		case T_SubLink:
 			switch (((SubLink *) node)->subLinkType)
 			{
diff --git a/src/backend/rewrite/rewriteHandler.c b/src/backend/rewrite/rewriteHandler.c
new file mode 100644
index 89187d9..c3fa282
--- a/src/backend/rewrite/rewriteHandler.c
+++ b/src/backend/rewrite/rewriteHandler.c
@@ -3833,9 +3833,9 @@ RewriteQuery(Query *parsetree, List *rew
 	ListCell   *lc1;
 
 	/*
-	 * First, recursively process any insert/update/delete statements in WITH
-	 * clauses.  (We have to do this first because the WITH clauses may get
-	 * copied into rule actions below.)
+	 * First, recursively process any insert/update/delete/merge statements in
+	 * WITH clauses.  (We have to do this first because the WITH clauses may
+	 * get copied into rule actions below.)
 	 */
 	foreach(lc1, parsetree->cteList)
 	{
@@ -3860,7 +3860,8 @@ RewriteQuery(Query *parsetree, List *rew
 			if (!(ctequery->commandType == CMD_SELECT ||
 				  ctequery->commandType == CMD_UPDATE ||
 				  ctequery->commandType == CMD_INSERT ||
-				  ctequery->commandType == CMD_DELETE))
+				  ctequery->commandType == CMD_DELETE ||
+				  ctequery->commandType == CMD_MERGE))
 			{
 				/*
 				 * Currently it could only be NOTIFY; this error message will
diff --git a/src/backend/rewrite/rowsecurity.c b/src/backend/rewrite/rowsecurity.c
new file mode 100644
index 450af27..59fd305
--- a/src/backend/rewrite/rowsecurity.c
+++ b/src/backend/rewrite/rowsecurity.c
@@ -384,10 +384,10 @@ get_row_security_policies(Query *root, R
 	 * on the final action we take.
 	 *
 	 * We already fetched the SELECT policies above, to check existing rows,
-	 * but we must also check that new rows created by UPDATE actions are
-	 * visible, if SELECT rights are required for this relation. We don't do
-	 * this for INSERT actions, since an INSERT command would only do this
-	 * check if it had a RETURNING list, and MERGE does not support RETURNING.
+	 * but we must also check that new rows created by INSERT/UPDATE actions
+	 * are visible, if SELECT rights are required. For INSERT actions, we only
+	 * do this if RETURNING is specified, to be consistent with a plain INSERT
+	 * command, which can only require SELECT rights when RETURNING is used.
 	 *
 	 * We don't push the UPDATE/DELETE USING quals to the RTE because we don't
 	 * really want to apply them while scanning the relation since we don't
@@ -409,6 +409,8 @@ get_row_security_policies(Query *root, R
 		List	   *merge_delete_restrictive_policies;
 		List	   *merge_insert_permissive_policies;
 		List	   *merge_insert_restrictive_policies;
+		List	   *merge_select_permissive_policies = NIL;
+		List	   *merge_select_restrictive_policies = NIL;
 
 		/*
 		 * Fetch the UPDATE policies and set them up to execute on the
@@ -446,9 +448,6 @@ get_row_security_policies(Query *root, R
 		 */
 		if (perminfo->requiredPerms & ACL_SELECT)
 		{
-			List	   *merge_select_permissive_policies;
-			List	   *merge_select_restrictive_policies;
-
 			get_policies_for_relation(rel, CMD_SELECT, user_id,
 									  &merge_select_permissive_policies,
 									  &merge_select_restrictive_policies);
@@ -497,6 +496,21 @@ get_row_security_policies(Query *root, R
 							   withCheckOptions,
 							   hasSubLinks,
 							   false);
+
+		/*
+		 * Add ALL/SELECT policies as WCO_RLS_INSERT_CHECK WCOs, to ensure
+		 * that the inserted row is visible when executing an INSERT action,
+		 * if RETURNING is specified and SELECT rights are required for this
+		 * relation.
+		 */
+		if (perminfo->requiredPerms & ACL_SELECT && root->returningList)
+			add_with_check_options(rel, rt_index,
+								   WCO_RLS_INSERT_CHECK,
+								   merge_select_permissive_policies,
+								   merge_select_restrictive_policies,
+								   withCheckOptions,
+								   hasSubLinks,
+								   true);
 	}
 
 	table_close(rel, NoLock);
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
new file mode 100644
index 83f86a4..fa66b80
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -2138,11 +2138,10 @@ QueryReturnsTuples(Query *parsetree)
 		case CMD_SELECT:
 			/* returns tuples */
 			return true;
-		case CMD_MERGE:
-			return false;
 		case CMD_INSERT:
 		case CMD_UPDATE:
 		case CMD_DELETE:
+		case CMD_MERGE:
 			/* the forms with RETURNING return tuples */
 			if (parsetree->returningList)
 				return true;
diff --git a/src/backend/utils/adt/ruleutils.c b/src/backend/utils/adt/ruleutils.c
new file mode 100644
index 2a1ee69..e2b46b6
--- a/src/backend/utils/adt/ruleutils.c
+++ b/src/backend/utils/adt/ruleutils.c
@@ -7196,8 +7196,13 @@ get_merge_query_def(Query *query, depars
 			appendStringInfoString(buf, "DO NOTHING");
 	}
 
-	/* No RETURNING support in MERGE yet */
-	Assert(query->returningList == NIL);
+	/* Add RETURNING if present */
+	if (query->returningList)
+	{
+		appendContextKeyword(context, " RETURNING",
+							 -PRETTYINDENT_STD, PRETTYINDENT_STD, 1);
+		get_target_list(query->returningList, context, NULL, colNamesVisible);
+	}
 }
 
 
@@ -8297,6 +8302,7 @@ isSimpleNode(Node *node, Node *parentNod
 		case T_Aggref:
 		case T_GroupingFunc:
 		case T_WindowFunc:
+		case T_MergeSupportFunc:
 		case T_FuncExpr:
 		case T_JsonConstructorExpr:
 			/* function-like: name(..) or name[..] */
@@ -8651,6 +8657,10 @@ get_rule_expr(Node *node, deparse_contex
 			get_windowfunc_expr((WindowFunc *) node, context);
 			break;
 
+		case T_MergeSupportFunc:
+			appendStringInfoString(buf, "MERGE_ACTION()");
+			break;
+
 		case T_SubscriptingRef:
 			{
 				SubscriptingRef *sbsref = (SubscriptingRef *) node;
diff --git a/src/bin/psql/common.c b/src/bin/psql/common.c
new file mode 100644
index 76e01b0..2830bde
--- a/src/bin/psql/common.c
+++ b/src/bin/psql/common.c
@@ -983,13 +983,17 @@ PrintQueryResult(PGresult *result, bool
 			else
 				success = true;
 
-			/* if it's INSERT/UPDATE/DELETE RETURNING, also print status */
+			/*
+			 * If it's INSERT/UPDATE/DELETE/MERGE RETURNING, also print
+			 * status.
+			 */
 			if (last || pset.show_all_results)
 			{
 				cmdstatus = PQcmdStatus(result);
 				if (strncmp(cmdstatus, "INSERT", 6) == 0 ||
 					strncmp(cmdstatus, "UPDATE", 6) == 0 ||
-					strncmp(cmdstatus, "DELETE", 6) == 0)
+					strncmp(cmdstatus, "DELETE", 6) == 0 ||
+					strncmp(cmdstatus, "MERGE", 5) == 0)
 					PrintQueryStatus(result, printStatusFout);
 			}
 
diff --git a/src/include/executor/execExpr.h b/src/include/executor/execExpr.h
new file mode 100644
index a28ddcd..8953d76
--- a/src/include/executor/execExpr.h
+++ b/src/include/executor/execExpr.h
@@ -243,6 +243,7 @@ typedef enum ExprEvalOp
 	EEOP_AGGREF,
 	EEOP_GROUPING_FUNC,
 	EEOP_WINDOW_FUNC,
+	EEOP_MERGE_SUPPORT_FUNC,
 	EEOP_SUBPLAN,
 
 	/* aggregation related nodes */
@@ -810,6 +811,8 @@ extern void ExecEvalJsonConstructor(Expr
 									ExprContext *econtext);
 extern void ExecEvalJsonIsPredicate(ExprState *state, ExprEvalStep *op);
 extern void ExecEvalGroupingFunc(ExprState *state, ExprEvalStep *op);
+extern void ExecEvalMergeSupportFunc(ExprState *state, ExprEvalStep *op,
+									 ExprContext *econtext);
 extern void ExecEvalSubPlan(ExprState *state, ExprEvalStep *op,
 							ExprContext *econtext);
 extern void ExecEvalWholeRowVar(ExprState *state, ExprEvalStep *op,
diff --git a/src/include/executor/spi.h b/src/include/executor/spi.h
new file mode 100644
index c13e1f6..48b8773
--- a/src/include/executor/spi.h
+++ b/src/include/executor/spi.h
@@ -97,6 +97,7 @@ typedef struct _SPI_plan *SPIPlanPtr;
 #define SPI_OK_REL_UNREGISTER	16
 #define SPI_OK_TD_REGISTER		17
 #define SPI_OK_MERGE			18
+#define SPI_OK_MERGE_RETURNING	19
 
 #define SPI_OPT_NONATOMIC		(1 << 0)
 
diff --git a/src/include/nodes/execnodes.h b/src/include/nodes/execnodes.h
new file mode 100644
index 444a5f0..57fdfca
--- a/src/include/nodes/execnodes.h
+++ b/src/include/nodes/execnodes.h
@@ -1325,6 +1325,9 @@ typedef struct ModifyTableState
 	/* Flags showing which subcommands are present INS/UPD/DEL/DO NOTHING */
 	int			mt_merge_subcommands;
 
+	/* For MERGE, the action currently being executed */
+	MergeActionState *mt_merge_action;
+
 	/* tuple counters for MERGE */
 	double		mt_merge_inserted;
 	double		mt_merge_updated;
diff --git a/src/include/nodes/parsenodes.h b/src/include/nodes/parsenodes.h
new file mode 100644
index 2380821..2bae226
--- a/src/include/nodes/parsenodes.h
+++ b/src/include/nodes/parsenodes.h
@@ -1938,6 +1938,7 @@ typedef struct MergeStmt
 	Node	   *sourceRelation; /* source relation */
 	Node	   *joinCondition;	/* join condition between source and target */
 	List	   *mergeWhenClauses;	/* list of MergeWhenClause(es) */
+	List	   *returningList;	/* list of expressions to return */
 	WithClause *withClause;		/* WITH clause */
 } MergeStmt;
 
diff --git a/src/include/nodes/primnodes.h b/src/include/nodes/primnodes.h
new file mode 100644
index 4a15460..8df8884
--- a/src/include/nodes/primnodes.h
+++ b/src/include/nodes/primnodes.h
@@ -572,6 +572,27 @@ typedef struct WindowFunc
 } WindowFunc;
 
 /*
+ * MergeSupportFunc
+ *
+ * A MergeSupportFunc is a merge support function expression that can only
+ * appear in the RETURNING list of a MERGE command.  It returns information
+ * about the currently executing merge action.
+ *
+ * Currently, the only supported function is MERGE_ACTION(), which returns the
+ * command executed ("INSERT", "UPDATE", or "DELETE").
+ */
+typedef struct MergeSupportFunc
+{
+	Expr		xpr;
+	/* type Oid of result */
+	Oid			msftype;
+	/* OID of collation, or InvalidOid if none */
+	Oid			msfcollid;
+	/* token location, or -1 if unknown */
+	int			location;
+} MergeSupportFunc;
+
+/*
  * SubscriptingRef: describes a subscripting operation over a container
  * (array, etc).
  *
diff --git a/src/include/optimizer/paramassign.h b/src/include/optimizer/paramassign.h
new file mode 100644
index 501f728..4026b74
--- a/src/include/optimizer/paramassign.h
+++ b/src/include/optimizer/paramassign.h
@@ -20,6 +20,8 @@ extern Param *replace_outer_placeholderv
 										   PlaceHolderVar *phv);
 extern Param *replace_outer_agg(PlannerInfo *root, Aggref *agg);
 extern Param *replace_outer_grouping(PlannerInfo *root, GroupingFunc *grp);
+extern Param *replace_outer_merge_support(PlannerInfo *root,
+										  MergeSupportFunc *msf);
 extern Param *replace_nestloop_param_var(PlannerInfo *root, Var *var);
 extern Param *replace_nestloop_param_placeholdervar(PlannerInfo *root,
 													PlaceHolderVar *phv);
diff --git a/src/include/parser/analyze.h b/src/include/parser/analyze.h
new file mode 100644
index 5ecdd10..28b66fc
--- a/src/include/parser/analyze.h
+++ b/src/include/parser/analyze.h
@@ -44,6 +44,8 @@ extern List *transformInsertRow(ParseSta
 								bool strip_indirection);
 extern List *transformUpdateTargetList(ParseState *pstate,
 									   List *origTlist);
+extern List *transformReturningList(ParseState *pstate, List *returningList,
+									ParseExprKind exprKind);
 extern Query *transformTopLevelStmt(ParseState *pstate, RawStmt *parseTree);
 extern Query *transformStmt(ParseState *pstate, Node *parseTree);
 
diff --git a/src/include/parser/kwlist.h b/src/include/parser/kwlist.h
new file mode 100644
index 2331aca..0993534
--- a/src/include/parser/kwlist.h
+++ b/src/include/parser/kwlist.h
@@ -266,6 +266,7 @@ PG_KEYWORD("matched", MATCHED, UNRESERVE
 PG_KEYWORD("materialized", MATERIALIZED, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("maxvalue", MAXVALUE, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("merge", MERGE, UNRESERVED_KEYWORD, BARE_LABEL)
+PG_KEYWORD("merge_action", MERGE_ACTION, COL_NAME_KEYWORD, BARE_LABEL)
 PG_KEYWORD("method", METHOD, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("minute", MINUTE_P, UNRESERVED_KEYWORD, AS_LABEL)
 PG_KEYWORD("minvalue", MINVALUE, UNRESERVED_KEYWORD, BARE_LABEL)
diff --git a/src/include/parser/parse_node.h b/src/include/parser/parse_node.h
new file mode 100644
index 99d6515..5b781d8
--- a/src/include/parser/parse_node.h
+++ b/src/include/parser/parse_node.h
@@ -61,7 +61,8 @@ typedef enum ParseExprKind
 	EXPR_KIND_DISTINCT_ON,		/* DISTINCT ON */
 	EXPR_KIND_LIMIT,			/* LIMIT */
 	EXPR_KIND_OFFSET,			/* OFFSET */
-	EXPR_KIND_RETURNING,		/* RETURNING */
+	EXPR_KIND_RETURNING,		/* RETURNING in INSERT/UPDATE/DELETE */
+	EXPR_KIND_MERGE_RETURNING,	/* RETURNING in MERGE */
 	EXPR_KIND_VALUES,			/* VALUES */
 	EXPR_KIND_VALUES_SINGLE,	/* single-row VALUES (in INSERT only) */
 	EXPR_KIND_CHECK_CONSTRAINT, /* CHECK constraint for a table */
diff --git a/src/pl/plpgsql/src/pl_exec.c b/src/pl/plpgsql/src/pl_exec.c
new file mode 100644
index ed51694..6947575
--- a/src/pl/plpgsql/src/pl_exec.c
+++ b/src/pl/plpgsql/src/pl_exec.c
@@ -4267,9 +4267,9 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 	/*
 	 * If we have INTO, then we only need one row back ... but if we have INTO
 	 * STRICT or extra check too_many_rows, ask for two rows, so that we can
-	 * verify the statement returns only one.  INSERT/UPDATE/DELETE are always
-	 * treated strictly. Without INTO, just run the statement to completion
-	 * (tcount = 0).
+	 * verify the statement returns only one.  INSERT/UPDATE/DELETE/MERGE are
+	 * always treated strictly. Without INTO, just run the statement to
+	 * completion (tcount = 0).
 	 *
 	 * We could just ask for two rows always when using INTO, but there are
 	 * some cases where demanding the extra row costs significant time, eg by
@@ -4307,10 +4307,11 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 			Assert(stmt->mod_stmt);
 			exec_set_found(estate, (SPI_processed != 0));
 			break;
@@ -4489,10 +4490,11 @@ exec_stmt_dynexecute(PLpgSQL_execstate *
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 		case SPI_OK_UTILITY:
 		case SPI_OK_REWRITTEN:
 			break;
diff --git a/src/pl/tcl/pltcl.c b/src/pl/tcl/pltcl.c
new file mode 100644
index 6187e15..18d14a2
--- a/src/pl/tcl/pltcl.c
+++ b/src/pl/tcl/pltcl.c
@@ -2456,6 +2456,7 @@ pltcl_process_SPI_result(Tcl_Interp *int
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
+		case SPI_OK_MERGE_RETURNING:
 
 			/*
 			 * Process the tuples we got
diff --git a/src/test/regress/expected/merge.out b/src/test/regress/expected/merge.out
new file mode 100644
index e3ebf46..f9f6c2a
--- a/src/test/regress/expected/merge.out
+++ b/src/test/regress/expected/merge.out
@@ -124,20 +124,20 @@ ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
 ERROR:  name "target" specified more than once
 DETAIL:  The name is used both as MERGE target table and data source.
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
-ERROR:  MERGE not supported in WITH query
-LINE 1: WITH foo AS (
-             ^
--- used in COPY
+ERROR:  WITH query "foo" does not have a RETURNING clause
+LINE 4: ) SELECT * FROM foo;
+                        ^
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) TO stdout;
-ERROR:  MERGE not supported in COPY
+ERROR:  COPY query must have a RETURNING clause
 -- unsupported relation types
 -- materialized view
 CREATE MATERIALIZED VIEW mv AS SELECT * FROM target;
@@ -1255,7 +1255,7 @@ BEGIN;
 MERGE INTO sq_target
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1270,7 +1270,7 @@ INSERT INTO sq_source (sid, balance, del
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1294,7 +1294,7 @@ WITH targq AS (
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1302,21 +1302,207 @@ WHEN MATCHED AND tid < 2 THEN
 	DELETE;
 ROLLBACK;
 -- RETURNING
+SELECT * FROM sq_source ORDER BY sid;
+ delta | sid | balance 
+-------+-----+---------
+    10 |   1 |       0
+    20 |   2 |       0
+    40 |   4 |       0
+(3 rows)
+
+SELECT * FROM sq_target ORDER BY tid;
+ tid | balance 
+-----+---------
+   1 |     100
+   2 |     200
+   3 |     300
+(3 rows)
+
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
+CREATE TABLE merge_actions(action text, abbrev text);
+INSERT INTO merge_actions VALUES ('INSERT', 'ins'), ('UPDATE', 'upd'), ('DELETE', 'del');
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
-ERROR:  syntax error at or near "RETURNING"
-LINE 10: RETURNING *;
-         ^
+    DELETE
+RETURNING (SELECT abbrev FROM merge_actions
+            WHERE action = merge_action()) AS action,
+          t.*,
+          CASE merge_action()
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ action | tid | balance |     description     
+--------+-----+---------+---------------------
+ del    |   1 |     100 | Removed (1,100)
+ upd    |   2 |     220 | Added 20 to balance
+ ins    |   4 |      40 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- error when using merge_action() outside MERGE
+SELECT merge_action() FROM sq_target;
+ERROR:  MERGE_ACTION() can only be used in the RETURNING list of a MERGE command
+LINE 1: SELECT merge_action() FROM sq_target;
+               ^
+UPDATE sq_target SET balance = balance + 1 RETURNING merge_action();
+ERROR:  MERGE_ACTION() can only be used in the RETURNING list of a MERGE command
+LINE 1: ...ATE sq_target SET balance = balance + 1 RETURNING merge_acti...
+                                                             ^
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING merge_action() AS action, t.*,
+              CASE merge_action()
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING action, merge_action() AS log_action, l.*
+)
+SELECT * FROM m2;
+ action | log_action | tid |     last_change     
+--------+------------+-----+---------------------
+ DELETE | UPDATE     |   1 | Removed (1,100)
+ UPDATE | INSERT     |   2 | Added 20 to balance
+ INSERT | INSERT     |   4 | Inserted (4,40)
+(3 rows)
+
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ tid |     last_change     
+-----+---------------------
+   1 | Removed (1,100)
+   2 | Added 20 to balance
+   4 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING merge_action(), t.*
+) TO stdout;
+DELETE	1	100
+UPDATE	2	220
+INSERT	4	40
+ROLLBACK;
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING merge_action(), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ action | tid | new_balance 
+--------+-----+-------------
+ DELETE |   1 |         100
+ UPDATE |   3 |         320
+ INSERT |   4 |         110
+(3 rows)
+
+ROLLBACK;
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING merge_action(), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ action | tid | balance 
+--------+-----+---------
+ DELETE |   1 |     100
+ UPDATE |   2 |     220
+ INSERT |   4 |      40
+(3 rows)
+
+ROLLBACK;
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING merge_action(), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ r_action | r_tid | r_balance 
+----------+-------+-----------
+ DELETE   |     1 |       100
+ UPDATE   |     3 |       320
+ INSERT   |     4 |       110
+(3 rows)
+
 ROLLBACK;
 -- EXPLAIN
 CREATE TABLE ex_mtarget (a int, b int)
@@ -1513,7 +1699,7 @@ SELECT * FROM sq_target WHERE tid = 1;
 (1 row)
 
 ROLLBACK;
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 NOTICE:  drop cascades to view v
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1639,6 +1825,32 @@ SELECT * FROM pa_target ORDER BY tid;
 (14 rows)
 
 ROLLBACK;
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING merge_action(), t.*;
+ merge_action | tid | balance |           val            
+--------------+-----+---------+--------------------------
+ UPDATE       |   2 |     110 | initial updated by merge
+(1 row)
+
+SELECT * FROM pa_target ORDER BY tid;
+ tid | balance |           val            
+-----+---------+--------------------------
+   2 |     110 | initial updated by merge
+   3 |     300 | initial
+   5 |     500 | initial
+   7 |     700 | initial
+   9 |     900 | initial
+  11 |    1100 | initial
+  13 |    1300 | initial
+(7 rows)
+
+ROLLBACK;
 DROP TABLE pa_target CASCADE;
 -- The target table is partitioned in the same way, but this time by attaching
 -- partitions which have columns in different order, dropped columns etc.
@@ -1888,7 +2100,21 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING merge_action(), t.*;
+ merge_action |          logts           | tid | balance |           val            
+--------------+--------------------------+-----+---------+--------------------------
+ UPDATE       | Tue Jan 31 00:00:00 2017 |   1 |     110 | initial updated by merge
+ UPDATE       | Tue Feb 28 00:00:00 2017 |   2 |     220 | initial updated by merge
+ INSERT       | Sun Jan 15 00:00:00 2017 |   3 |      30 | inserted by merge
+ UPDATE       | Tue Jan 31 00:00:00 2017 |   4 |     440 | initial updated by merge
+ UPDATE       | Tue Feb 28 00:00:00 2017 |   5 |     550 | initial updated by merge
+ INSERT       | Sun Jan 15 00:00:00 2017 |   6 |      60 | inserted by merge
+ UPDATE       | Tue Jan 31 00:00:00 2017 |   7 |     770 | initial updated by merge
+ UPDATE       | Tue Feb 28 00:00:00 2017 |   8 |     880 | initial updated by merge
+ INSERT       | Sun Jan 15 00:00:00 2017 |   9 |      90 | inserted by merge
+(9 rows)
+
 SELECT * FROM pa_target ORDER BY tid;
           logts           | tid | balance |           val            
 --------------------------+-----+---------+--------------------------
diff --git a/src/test/regress/expected/rowsecurity.out b/src/test/regress/expected/rowsecurity.out
new file mode 100644
index 6988128..3fd4fab
--- a/src/test/regress/expected/rowsecurity.out
+++ b/src/test/regress/expected/rowsecurity.out
@@ -2302,6 +2302,35 @@ WHEN MATCHED THEN
 	UPDATE SET dnotes = dnotes || ' notes added by merge8 '
 WHEN NOT MATCHED THEN
 	INSERT VALUES (13, 44, 1, 'regress_rls_bob', 'new manga');
+SELECT * FROM document WHERE did = 13;
+ did | cid | dlevel | dauthor | dtitle | dnotes 
+-----+-----+--------+---------+--------+--------
+(0 rows)
+
+-- but not OK if RETURNING is used
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge9 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 44, 1, 'regress_rls_bob', 'new manga')
+RETURNING *;
+ERROR:  new row violates row-level security policy for table "document"
+-- but OK if new row is visible
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge10 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 11, 1, 'regress_rls_bob', 'new novel')
+RETURNING *;
+ sdid | did | cid | dlevel |     dauthor     |  dtitle   | dnotes 
+------+-----+-----+--------+-----------------+-----------+--------
+   14 |  14 |  11 |      1 | regress_rls_bob | new novel | 
+(1 row)
+
 RESET SESSION AUTHORIZATION;
 -- drop the restrictive SELECT policy so that we can look at the
 -- final state of the table
@@ -2325,7 +2354,8 @@ SELECT * FROM document;
   12 |  11 |      1 | regress_rls_bob   | another novel                    | 
    1 |  11 |      1 | regress_rls_bob   | my first novel                   |  notes added by merge2  notes added by merge3  notes added by merge4  notes added by merge7 
   13 |  44 |      1 | regress_rls_bob   | new manga                        | 
-(15 rows)
+  14 |  11 |      1 | regress_rls_bob   | new novel                        | 
+(16 rows)
 
 --
 -- ROLE/GROUP
diff --git a/src/test/regress/expected/rules.out b/src/test/regress/expected/rules.out
new file mode 100644
index 0cd2c64..84e359f
--- a/src/test/regress/expected/rules.out
+++ b/src/test/regress/expected/rules.out
@@ -3638,7 +3638,7 @@ MERGE INTO rule_merge1 t USING (SELECT 1
 -- test deparsing
 CREATE TABLE sf_target(id int, data text, filling int[]);
 CREATE FUNCTION merge_sf_test()
- RETURNS void
+ RETURNS TABLE(action text, a int, b text, id int, data text, filling int[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -3675,11 +3675,13 @@ WHEN NOT MATCHED
    VALUES (s.a, s.b, DEFAULT)
 WHEN NOT MATCHED
    THEN INSERT (filling[1], id)
-   VALUES (s.a, s.a);
+   VALUES (s.a, s.a)
+RETURNING
+   merge_action() AS action, *;
 END;
 \sf merge_sf_test
 CREATE OR REPLACE FUNCTION public.merge_sf_test()
- RETURNS void
+ RETURNS TABLE(action text, a integer, b text, id integer, data text, filling integer[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -3716,7 +3718,13 @@ BEGIN ATOMIC
       VALUES (s.a, s.b, DEFAULT)
     WHEN NOT MATCHED
      THEN INSERT (filling[1], id)
-      VALUES (s.a, s.a);
+      VALUES (s.a, s.a)
+   RETURNING MERGE_ACTION() AS action,
+     s.a,
+     s.b,
+     t.id,
+     t.data,
+     t.filling;
 END
 DROP FUNCTION merge_sf_test;
 DROP TABLE sf_target;
diff --git a/src/test/regress/expected/updatable_views.out b/src/test/regress/expected/updatable_views.out
new file mode 100644
index 794cf9c..1062c34
--- a/src/test/regress/expected/updatable_views.out
+++ b/src/test/regress/expected/updatable_views.out
@@ -479,7 +479,15 @@ MERGE INTO rw_view1 t
                 (2, 'ROW 2'), (3, 'ROW 3')) AS v(a,b) ON t.a = v.a
   WHEN MATCHED AND t.a <= 1 THEN UPDATE SET b = v.b
   WHEN MATCHED THEN DELETE
-  WHEN NOT MATCHED AND a > 0 THEN INSERT (a) VALUES (v.a);
+  WHEN NOT MATCHED AND a > 0 THEN INSERT (a) VALUES (v.a)
+  RETURNING merge_action(), v.*, t.*;
+ merge_action | a |   b   | a |      b      
+--------------+---+-------+---+-------------
+ UPDATE       | 1 | ROW 1 | 1 | ROW 1
+ DELETE       | 3 | ROW 3 | 3 | Row 3
+ INSERT       | 2 | ROW 2 | 2 | Unspecified
+(3 rows)
+
 SELECT * FROM base_tbl ORDER BY a;
  a  |      b      
 ----+-------------
@@ -631,7 +639,15 @@ MERGE INTO rw_view2 t
   USING (VALUES (3, 'R3'), (4, 'R4'), (5, 'R5')) AS v(a,b) ON aaa = v.a
   WHEN MATCHED AND aaa = 3 THEN DELETE
   WHEN MATCHED THEN UPDATE SET bbb = v.b
-  WHEN NOT MATCHED THEN INSERT (aaa) VALUES (v.a);
+  WHEN NOT MATCHED THEN INSERT (aaa) VALUES (v.a)
+  RETURNING merge_action(), v.*, t.*;
+ merge_action | a | b  | aaa |     bbb     
+--------------+---+----+-----+-------------
+ DELETE       | 3 | R3 |   3 | Row 3
+ UPDATE       | 4 | R4 |   4 | R4
+ INSERT       | 5 | R5 |   5 | Unspecified
+(3 rows)
+
 SELECT * FROM rw_view2 ORDER BY aaa;
  aaa |     bbb     
 -----+-------------
@@ -1071,7 +1087,15 @@ MERGE INTO rw_view2 t
   USING (SELECT x, 'R'||x FROM generate_series(0,3) x) AS s(a,b) ON t.a = s.a
   WHEN MATCHED AND t.a <= 1 THEN DELETE
   WHEN MATCHED THEN UPDATE SET b = s.b
-  WHEN NOT MATCHED AND s.a > 0 THEN INSERT VALUES (s.a, s.b);
+  WHEN NOT MATCHED AND s.a > 0 THEN INSERT VALUES (s.a, s.b)
+  RETURNING merge_action(), s.*, t.*;
+ merge_action | a | b  | a |   b   
+--------------+---+----+---+-------
+ DELETE       | 1 | R1 | 1 | Row 1
+ UPDATE       | 2 | R2 | 2 | R2
+ INSERT       | 3 | R3 | 3 | R3
+(3 rows)
+
 SELECT * FROM base_tbl ORDER BY a;
  a  |   b    
 ----+--------
diff --git a/src/test/regress/expected/with.out b/src/test/regress/expected/with.out
new file mode 100644
index 7d796ea..6de347b
--- a/src/test/regress/expected/with.out
+++ b/src/test/regress/expected/with.out
@@ -3493,6 +3493,16 @@ SELECT * FROM t;
 ERROR:  WITH query "t" does not have a RETURNING clause
 LINE 4: SELECT * FROM t;
                       ^
+-- RETURNING tries to return its own output
+WITH RECURSIVE t(action, a) AS (
+	MERGE INTO y USING (VALUES (11)) v(a) ON y.a = v.a
+		WHEN NOT MATCHED THEN INSERT VALUES (v.a)
+		RETURNING merge_action(), (SELECT a FROM t)
+)
+SELECT * FROM t;
+ERROR:  recursive query "t" must not contain data-modifying statements
+LINE 1: WITH RECURSIVE t(action, a) AS (
+                       ^
 -- data-modifying WITH allowed only at the top level
 SELECT * FROM (
 	WITH t AS (UPDATE y SET a=a+1 RETURNING *)
diff --git a/src/test/regress/sql/merge.sql b/src/test/regress/sql/merge.sql
new file mode 100644
index 79a1868..dfcc243
--- a/src/test/regress/sql/merge.sql
+++ b/src/test/regress/sql/merge.sql
@@ -88,12 +88,12 @@ MERGE INTO target
 USING target
 ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
--- used in COPY
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
@@ -817,7 +817,7 @@ BEGIN;
 MERGE INTO sq_target
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -830,7 +830,7 @@ INSERT INTO sq_source (sid, balance, del
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -848,7 +848,7 @@ WITH targq AS (
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -857,18 +857,149 @@ WHEN MATCHED AND tid < 2 THEN
 ROLLBACK;
 
 -- RETURNING
+SELECT * FROM sq_source ORDER BY sid;
+SELECT * FROM sq_target ORDER BY tid;
+
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
+CREATE TABLE merge_actions(action text, abbrev text);
+INSERT INTO merge_actions VALUES ('INSERT', 'ins'), ('UPDATE', 'upd'), ('DELETE', 'del');
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
+    DELETE
+RETURNING (SELECT abbrev FROM merge_actions
+            WHERE action = merge_action()) AS action,
+          t.*,
+          CASE merge_action()
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ROLLBACK;
+
+-- error when using merge_action() outside MERGE
+SELECT merge_action() FROM sq_target;
+UPDATE sq_target SET balance = balance + 1 RETURNING merge_action();
+
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING merge_action() AS action, t.*,
+              CASE merge_action()
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING action, merge_action() AS log_action, l.*
+)
+SELECT * FROM m2;
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ROLLBACK;
+
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING merge_action(), t.*
+) TO stdout;
+ROLLBACK;
+
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING merge_action(), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ROLLBACK;
+
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING merge_action(), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ROLLBACK;
+
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING merge_action(), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
 ROLLBACK;
 
 -- EXPLAIN
@@ -967,7 +1098,7 @@ WHEN MATCHED THEN
 SELECT * FROM sq_target WHERE tid = 1;
 ROLLBACK;
 
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1034,6 +1165,17 @@ SELECT merge_func();
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING merge_action(), t.*;
+SELECT * FROM pa_target ORDER BY tid;
+ROLLBACK;
+
 DROP TABLE pa_target CASCADE;
 
 -- The target table is partitioned in the same way, but this time by attaching
@@ -1210,7 +1352,8 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING merge_action(), t.*;
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
diff --git a/src/test/regress/sql/rowsecurity.sql b/src/test/regress/sql/rowsecurity.sql
new file mode 100644
index dec7340..1d5ed0a
--- a/src/test/regress/sql/rowsecurity.sql
+++ b/src/test/regress/sql/rowsecurity.sql
@@ -992,6 +992,27 @@ WHEN MATCHED THEN
 	UPDATE SET dnotes = dnotes || ' notes added by merge8 '
 WHEN NOT MATCHED THEN
 	INSERT VALUES (13, 44, 1, 'regress_rls_bob', 'new manga');
+SELECT * FROM document WHERE did = 13;
+
+-- but not OK if RETURNING is used
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge9 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 44, 1, 'regress_rls_bob', 'new manga')
+RETURNING *;
+
+-- but OK if new row is visible
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge10 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 11, 1, 'regress_rls_bob', 'new novel')
+RETURNING *;
 
 RESET SESSION AUTHORIZATION;
 -- drop the restrictive SELECT policy so that we can look at the
diff --git a/src/test/regress/sql/rules.sql b/src/test/regress/sql/rules.sql
new file mode 100644
index 6924012..27340ba
--- a/src/test/regress/sql/rules.sql
+++ b/src/test/regress/sql/rules.sql
@@ -1294,7 +1294,7 @@ MERGE INTO rule_merge1 t USING (SELECT 1
 CREATE TABLE sf_target(id int, data text, filling int[]);
 
 CREATE FUNCTION merge_sf_test()
- RETURNS void
+ RETURNS TABLE(action text, a int, b text, id int, data text, filling int[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -1331,7 +1331,9 @@ WHEN NOT MATCHED
    VALUES (s.a, s.b, DEFAULT)
 WHEN NOT MATCHED
    THEN INSERT (filling[1], id)
-   VALUES (s.a, s.a);
+   VALUES (s.a, s.a)
+RETURNING
+   merge_action() AS action, *;
 END;
 
 \sf merge_sf_test
diff --git a/src/test/regress/sql/updatable_views.sql b/src/test/regress/sql/updatable_views.sql
new file mode 100644
index ae11e46..afdf331
--- a/src/test/regress/sql/updatable_views.sql
+++ b/src/test/regress/sql/updatable_views.sql
@@ -175,7 +175,8 @@ MERGE INTO rw_view1 t
                 (2, 'ROW 2'), (3, 'ROW 3')) AS v(a,b) ON t.a = v.a
   WHEN MATCHED AND t.a <= 1 THEN UPDATE SET b = v.b
   WHEN MATCHED THEN DELETE
-  WHEN NOT MATCHED AND a > 0 THEN INSERT (a) VALUES (v.a);
+  WHEN NOT MATCHED AND a > 0 THEN INSERT (a) VALUES (v.a)
+  RETURNING merge_action(), v.*, t.*;
 SELECT * FROM base_tbl ORDER BY a;
 
 EXPLAIN (costs off) UPDATE rw_view1 SET a=6 WHERE a=5;
@@ -246,7 +247,8 @@ MERGE INTO rw_view2 t
   USING (VALUES (3, 'R3'), (4, 'R4'), (5, 'R5')) AS v(a,b) ON aaa = v.a
   WHEN MATCHED AND aaa = 3 THEN DELETE
   WHEN MATCHED THEN UPDATE SET bbb = v.b
-  WHEN NOT MATCHED THEN INSERT (aaa) VALUES (v.a);
+  WHEN NOT MATCHED THEN INSERT (aaa) VALUES (v.a)
+  RETURNING merge_action(), v.*, t.*;
 SELECT * FROM rw_view2 ORDER BY aaa;
 
 EXPLAIN (costs off) UPDATE rw_view2 SET aaa=5 WHERE aaa=4;
@@ -458,7 +460,8 @@ MERGE INTO rw_view2 t
   USING (SELECT x, 'R'||x FROM generate_series(0,3) x) AS s(a,b) ON t.a = s.a
   WHEN MATCHED AND t.a <= 1 THEN DELETE
   WHEN MATCHED THEN UPDATE SET b = s.b
-  WHEN NOT MATCHED AND s.a > 0 THEN INSERT VALUES (s.a, s.b);
+  WHEN NOT MATCHED AND s.a > 0 THEN INSERT VALUES (s.a, s.b)
+  RETURNING merge_action(), s.*, t.*;
 SELECT * FROM base_tbl ORDER BY a;
 
 EXPLAIN (costs off) UPDATE rw_view2 SET a=3 WHERE a=2;
diff --git a/src/test/regress/sql/with.sql b/src/test/regress/sql/with.sql
new file mode 100644
index f8a213e..037bc0a
--- a/src/test/regress/sql/with.sql
+++ b/src/test/regress/sql/with.sql
@@ -1591,6 +1591,14 @@ WITH t AS (
 )
 SELECT * FROM t;
 
+-- RETURNING tries to return its own output
+WITH RECURSIVE t(action, a) AS (
+	MERGE INTO y USING (VALUES (11)) v(a) ON y.a = v.a
+		WHEN NOT MATCHED THEN INSERT VALUES (v.a)
+		RETURNING merge_action(), (SELECT a FROM t)
+)
+SELECT * FROM t;
+
 -- data-modifying WITH allowed only at the top level
 SELECT * FROM (
 	WITH t AS (UPDATE y SET a=a+1 RETURNING *)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
new file mode 100644
index aa7a25b..7a102e8
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -1577,6 +1577,7 @@ MergeJoinState
 MergePath
 MergeScanSelCache
 MergeStmt
+MergeSupportFunc
 MergeWhenClause
 MetaCommand
 MinMaxAggInfo

diff --git a/doc/src/sgml/dml.sgml b/doc/src/sgml/dml.sgml
new file mode 100644
index cbbc5e2..3d95bdb
--- a/doc/src/sgml/dml.sgml
+++ b/doc/src/sgml/dml.sgml
@@ -283,10 +283,15 @@ DELETE FROM products;
    <secondary>RETURNING</secondary>
   </indexterm>
 
+  <indexterm zone="dml-returning">
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
   <para>
    Sometimes it is useful to obtain data from modified rows while they are
    being manipulated.  The <command>INSERT</command>, <command>UPDATE</command>,
-   and <command>DELETE</command> commands all have an
+   <command>DELETE</command>, and <command>MERGE</command> commands all have an
    optional <literal>RETURNING</literal> clause that supports this.  Use
    of <literal>RETURNING</literal> avoids performing an extra database query to
    collect the data, and is especially valuable when it would otherwise be
@@ -339,6 +344,21 @@ DELETE FROM products
 </programlisting>
   </para>
 
+  <para>
+   In a <command>MERGE</command>, the data available to <literal>RETURNING</literal> is
+   the content of the source row plus the content of the inserted, updated, or
+   deleted target row.  Since it is quite common for the source and target to
+   have many of the same columns, specifying <literal>RETURNING *</literal>
+   can lead to a lot of duplicated columns, so it is often more useful to
+   qualify it so as to return just the source or target row.  For example:
+<programlisting>
+MERGE INTO products p USING new_products n ON p.product_no = n.product_no
+  WHEN NOT MATCHED THEN INSERT VALUES (n.product_no, n.name, n.price)
+  WHEN MATCHED THEN UPDATE SET name = n.name, price = n.price
+  RETURNING p.*;
+</programlisting>
+  </para>
+
   <para>
    If there are triggers (<xref linkend="triggers"/>) on the target table,
    the data available to <literal>RETURNING</literal> is the row as modified by
diff --git a/doc/src/sgml/func.sgml b/doc/src/sgml/func.sgml
new file mode 100644
index 72c5175..ebb984a
--- a/doc/src/sgml/func.sgml
+++ b/doc/src/sgml/func.sgml
@@ -22421,6 +22421,85 @@ SELECT count(*) FROM sometable;
 
  </sect1>
 
+ <sect1 id="functions-merge-support">
+  <title>Merge Support Functions</title>
+
+  <indexterm>
+   <primary>MERGE</primary>
+   <secondary>RETURNING</secondary>
+  </indexterm>
+
+  <para>
+   <productname>PostgreSQL</productname> includes one merge support function
+   that may be used in the <literal>RETURNING</literal> list of a
+   <xref linkend="sql-merge"/> command to identify the action taken for each
+   row.
+  </para>
+
+  <table id="functions-merge-support-table">
+   <title>Merge Support Functions</title>
+
+   <tgroup cols="1">
+    <thead>
+     <row>
+      <entry role="func_table_entry"><para role="func_signature">
+       Function
+      </para>
+      <para>
+       Description
+      </para></entry>
+     </row>
+    </thead>
+
+    <tbody>
+     <row>
+      <entry role="func_table_entry"><para role="func_signature">
+       <indexterm>
+        <primary>merge_action</primary>
+       </indexterm>
+       <function id="merge_action">merge_action</function> ( )
+       <returnvalue>text</returnvalue>
+      </para>
+      <para>
+       Returns the merge action command executed for the current row.  This
+       will be <literal>'INSERT'</literal>, <literal>'UPDATE'</literal>, or
+       <literal>'DELETE'</literal>.
+      </para></entry>
+     </row>
+    </tbody>
+   </tgroup>
+  </table>
+
+  <para>
+   Example:
+<screen><![CDATA[
+MERGE INTO products p
+  USING stock s ON p.product_id = s.product_id
+  WHEN MATCHED AND s.quantity > 0 THEN
+    UPDATE SET in_stock = true, quantity = s.quantity
+  WHEN MATCHED THEN
+    UPDATE SET in_stock = false, quantity = 0
+  WHEN NOT MATCHED THEN
+    INSERT (product_id, in_stock, quantity)
+      VALUES (s.product_id, true, s.quantity)
+  RETURNING merge_action(), p.*;
+
+ merge_action | product_id | in_stock | quantity
+--------------+------------+----------+----------
+ UPDATE       |       1001 | t        |       50
+ UPDATE       |       1002 | f        |        0
+ INSERT       |       1003 | t        |       10
+]]></screen>
+  </para>
+
+  <para>
+   Note that this function can only be used in the <literal>RETURNING</literal>
+   list of a <command>MERGE</command> command.  It is an error to use it in any
+   other part of a query.
+  </para>
+
+ </sect1>
+
  <sect1 id="functions-subquery">
   <title>Subquery Expressions</title>
 
diff --git a/doc/src/sgml/glossary.sgml b/doc/src/sgml/glossary.sgml
new file mode 100644
index 8c2f114..a81c17a
--- a/doc/src/sgml/glossary.sgml
+++ b/doc/src/sgml/glossary.sgml
@@ -1442,9 +1442,9 @@
      to a <glossterm linkend="glossary-client">client</glossterm> upon the
      completion of an <acronym>SQL</acronym> command, usually a
      <command>SELECT</command> but it can be an
-     <command>INSERT</command>, <command>UPDATE</command>, or
-     <command>DELETE</command> command if the <literal>RETURNING</literal>
-     clause is specified.
+     <command>INSERT</command>, <command>UPDATE</command>,
+     <command>DELETE</command>, or <command>MERGE</command> command if the
+     <literal>RETURNING</literal> clause is specified.
     </para>
     <para>
      The fact that a result set is a relation means that a query can be used
diff --git a/doc/src/sgml/plpgsql.sgml b/doc/src/sgml/plpgsql.sgml
new file mode 100644
index c2b9c6a..6f880b7
--- a/doc/src/sgml/plpgsql.sgml
+++ b/doc/src/sgml/plpgsql.sgml
@@ -1043,8 +1043,8 @@ INSERT INTO mytable VALUES (1,'one'), (2
     </para>
 
     <para>
-     If the command does return rows (for example <command>SELECT</command>,
-     or <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>
+     If the command does return rows (for example <command>SELECT</command>, or
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
      with <literal>RETURNING</literal>), there are two ways to proceed.
      When the command will return at most one row, or you only care about
      the first row of output, write the command as usual but add
@@ -1172,6 +1172,7 @@ SELECT <replaceable>select_expressions</
 INSERT ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 UPDATE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 DELETE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
+MERGE ... RETURNING <replaceable>expressions</replaceable> INTO <optional>STRICT</optional> <replaceable>target</replaceable>;
 </synopsis>
 
      where <replaceable>target</replaceable> can be a record variable, a row
@@ -1182,8 +1183,8 @@ DELETE ... RETURNING <replaceable>expres
      <literal>INTO</literal> clause) just as described above,
      and the plan is cached in the same way.
      This works for <command>SELECT</command>,
-     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
-     <literal>RETURNING</literal>, and certain utility commands
+     <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>
+     with <literal>RETURNING</literal>, and certain utility commands
      that return row sets, such as <command>EXPLAIN</command>.
      Except for the <literal>INTO</literal> clause, the SQL command is the same
      as it would be written outside <application>PL/pgSQL</application>.
@@ -1259,7 +1260,7 @@ END;
     </para>
 
     <para>
-     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command> with
+     For <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command> with
      <literal>RETURNING</literal>, <application>PL/pgSQL</application> reports
      an error for more than one returned row, even when
      <literal>STRICT</literal> is not specified.  This is because there
@@ -2657,8 +2658,9 @@ $$ LANGUAGE plpgsql;
      The <replaceable>query</replaceable> used in this type of <literal>FOR</literal>
      statement can be any SQL command that returns rows to the caller:
      <command>SELECT</command> is the most common case,
-     but you can also use <command>INSERT</command>, <command>UPDATE</command>, or
-     <command>DELETE</command> with a <literal>RETURNING</literal> clause.  Some utility
+     but you can also use <command>INSERT</command>, <command>UPDATE</command>,
+     <command>DELETE</command>, or <command>MERGE</command> with a
+     <literal>RETURNING</literal> clause.  Some utility
      commands such as <command>EXPLAIN</command> will work too.
     </para>
 
diff --git a/doc/src/sgml/queries.sgml b/doc/src/sgml/queries.sgml
new file mode 100644
index 648b283..372cce1
--- a/doc/src/sgml/queries.sgml
+++ b/doc/src/sgml/queries.sgml
@@ -2063,9 +2063,10 @@ SELECT <replaceable>select_list</replace
    Table Expressions or <acronym>CTE</acronym>s, can be thought of as defining
    temporary tables that exist just for one query.  Each auxiliary statement
    in a <literal>WITH</literal> clause can be a <command>SELECT</command>,
-   <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>; and the
+   <command>INSERT</command>, <command>UPDATE</command>, <command>DELETE</command>,
+   or <command>MERGE</command>; and the
    <literal>WITH</literal> clause itself is attached to a primary statement that can
-   be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
+   also be a <command>SELECT</command>, <command>INSERT</command>, <command>UPDATE</command>,
    <command>DELETE</command>, or <command>MERGE</command>.
   </para>
 
@@ -2599,8 +2600,8 @@ SELECT * FROM w AS w1 JOIN w AS w2 ON w1
    <title>Data-Modifying Statements in <literal>WITH</literal></title>
 
    <para>
-    You can use most data-modifying statements (<command>INSERT</command>,
-    <command>UPDATE</command>, or <command>DELETE</command>, but not
+    You can use data-modifying statements (<command>INSERT</command>,
+    <command>UPDATE</command>, <command>DELETE</command>, or
     <command>MERGE</command>) in <literal>WITH</literal>.  This
     allows you to perform several different operations in the same query.
     An example is:
diff --git a/doc/src/sgml/ref/copy.sgml b/doc/src/sgml/ref/copy.sgml
new file mode 100644
index 55764fc..6c83e30
--- a/doc/src/sgml/ref/copy.sgml
+++ b/doc/src/sgml/ref/copy.sgml
@@ -130,16 +130,19 @@ COPY { <replaceable class="parameter">ta
       A <link linkend="sql-select"><command>SELECT</command></link>,
       <link linkend="sql-values"><command>VALUES</command></link>,
       <link linkend="sql-insert"><command>INSERT</command></link>,
-      <link linkend="sql-update"><command>UPDATE</command></link>, or
-      <link linkend="sql-delete"><command>DELETE</command></link> command whose results are to be
-      copied.  Note that parentheses are required around the query.
+      <link linkend="sql-update"><command>UPDATE</command></link>,
+      <link linkend="sql-delete"><command>DELETE</command></link>, or
+      <link linkend="sql-merge"><command>MERGE</command></link> command
+      whose results are to be copied.  Note that parentheses are required
+      around the query.
      </para>
      <para>
-      For <command>INSERT</command>, <command>UPDATE</command> and
-      <command>DELETE</command> queries a <literal>RETURNING</literal> clause
-      must be provided, and the target relation must not have a conditional
-      rule, nor an <literal>ALSO</literal> rule, nor an
-      <literal>INSTEAD</literal> rule that expands to multiple statements.
+      For <command>INSERT</command>, <command>UPDATE</command>,
+      <command>DELETE</command>, and <command>MERGE</command> queries a
+      <literal>RETURNING</literal> clause must be provided, and the target
+      relation must not have a conditional rule, nor an
+      <literal>ALSO</literal> rule, nor an <literal>INSTEAD</literal> rule
+      that expands to multiple statements.
      </para>
     </listitem>
    </varlistentry>
diff --git a/doc/src/sgml/ref/merge.sgml b/doc/src/sgml/ref/merge.sgml
new file mode 100644
index e745fbd..71feb66
--- a/doc/src/sgml/ref/merge.sgml
+++ b/doc/src/sgml/ref/merge.sgml
@@ -25,6 +25,7 @@ PostgreSQL documentation
 MERGE INTO [ ONLY ] <replaceable class="parameter">target_table_name</replaceable> [ * ] [ [ AS ] <replaceable class="parameter">target_alias</replaceable> ]
 USING <replaceable class="parameter">data_source</replaceable> ON <replaceable class="parameter">join_condition</replaceable>
 <replaceable class="parameter">when_clause</replaceable> [...]
+[ RETURNING * | <replaceable class="parameter">output_expression</replaceable> [ [ AS ] <replaceable class="parameter">output_name</replaceable> ] [, ...] ]
 
 <phrase>where <replaceable class="parameter">data_source</replaceable> is:</phrase>
 
@@ -97,6 +98,18 @@ DELETE
   </para>
 
   <para>
+   The optional <literal>RETURNING</literal> clause causes <command>MERGE</command>
+   to compute and return value(s) based on each row inserted, updated, or
+   deleted.  Any expression using the source or target table's columns, or
+   the <xref linkend="merge_action"/> function can be computed.  When an
+   <command>INSERT</command> or <command>UPDATE</command> action is performed,
+   the new values of the target table's columns are used.  When a
+   <command>DELETE</command> is performed, the old values of the target table's
+   columns are used.  The syntax of the <literal>RETURNING</literal> list is
+   identical to that of the output list of <command>SELECT</command>.
+  </para>
+
+  <para>
    There is no separate <literal>MERGE</literal> privilege.
    If you specify an update action, you must have the
    <literal>UPDATE</literal> privilege on the column(s)
@@ -442,6 +455,36 @@ DELETE
     </listitem>
    </varlistentry>
 
+   <varlistentry>
+    <term><replaceable class="parameter">output_expression</replaceable></term>
+    <listitem>
+     <para>
+      An expression to be computed and returned by the <command>MERGE</command>
+      command after each row is changed (whether inserted, updated, or deleted).
+      The expression can use any columns of the source or target tables, or the
+      <xref linkend="merge_action"/> function to return additional information
+      about the action executed.
+     </para>
+     <para>
+      Writing <literal>*</literal> will return all columns from the source
+      table, followed by all columns from the target table.  Often this will
+      lead to a lot of duplication, since it is common for the source and
+      target tables to have a lot of the same columns.  This can be avoided by
+      qualifying the <literal>*</literal> with the name or alias of the source
+      or target table.
+     </para>
+    </listitem>
+   </varlistentry>
+
+   <varlistentry>
+    <term><replaceable class="parameter">output_name</replaceable></term>
+    <listitem>
+     <para>
+      A name to use for a returned column.
+     </para>
+    </listitem>
+   </varlistentry>
+
   </variablelist>
  </refsect1>
 
@@ -460,6 +503,13 @@ MERGE <replaceable class="parameter">tot
    were changed in any way.
   </para>
 
+  <para>
+   If the <command>MERGE</command> command contains a <literal>RETURNING</literal>
+   clause, the result will be similar to that of a <command>SELECT</command>
+   statement containing the columns and values defined in the
+   <literal>RETURNING</literal> list, computed over the row(s) inserted, updated,
+   or deleted by the command.
+  </para>
  </refsect1>
 
  <refsect1>
@@ -580,13 +630,6 @@ MERGE <replaceable class="parameter">tot
   </para>
 
   <para>
-   There is no <literal>RETURNING</literal> clause with
-   <command>MERGE</command>.  Actions of <command>INSERT</command>,
-   <command>UPDATE</command> and <command>DELETE</command> cannot contain
-   <literal>RETURNING</literal> or <literal>WITH</literal> clauses.
-  </para>
-
-  <para>
    When <command>MERGE</command> is run concurrently with other commands
    that modify the target table, the usual transaction isolation rules
    apply; see <xref linkend="transaction-iso"/> for an explanation
@@ -638,7 +681,8 @@ WHEN NOT MATCHED THEN
   <para>
    Attempt to insert a new stock item along with the quantity of stock. If
    the item already exists, instead update the stock count of the existing
-   item. Don't allow entries that have zero stock.
+   item. Don't allow entries that have zero stock. Return details of all
+   changes made.
 <programlisting>
 MERGE INTO wines w
 USING wine_stock_changes s
@@ -648,7 +692,8 @@ WHEN NOT MATCHED AND s.stock_delta > 0 T
 WHEN MATCHED AND w.stock + s.stock_delta > 0 THEN
   UPDATE SET stock = w.stock + s.stock_delta
 WHEN MATCHED THEN
-  DELETE;
+  DELETE
+RETURNING merge_action(), w.*;
 </programlisting>
 
    The <literal>wine_stock_changes</literal> table might be, for example, a
@@ -663,8 +708,9 @@ WHEN MATCHED THEN
     This command conforms to the <acronym>SQL</acronym> standard.
   </para>
    <para>
-    The <literal>WITH</literal> clause and <literal>DO NOTHING</literal>
-    action are extensions to the <acronym>SQL</acronym> standard.
+    The <literal>WITH</literal> clause, <literal>DO NOTHING</literal> action,
+    and <literal>RETURNING</literal> clause are extensions to the
+    <acronym>SQL</acronym> standard.
   </para>
  </refsect1>
 </refentry>
diff --git a/doc/src/sgml/ref/select.sgml b/doc/src/sgml/ref/select.sgml
new file mode 100644
index 9917df7..066aed4
--- a/doc/src/sgml/ref/select.sgml
+++ b/doc/src/sgml/ref/select.sgml
@@ -74,7 +74,7 @@ SELECT [ ALL | DISTINCT [ ON ( <replacea
 
 <phrase>and <replaceable class="parameter">with_query</replaceable> is:</phrase>
 
-    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> )
+    <replaceable class="parameter">with_query_name</replaceable> [ ( <replaceable class="parameter">column_name</replaceable> [, ...] ) ] AS [ [ NOT ] MATERIALIZED ] ( <replaceable class="parameter">select</replaceable> | <replaceable class="parameter">values</replaceable> | <replaceable class="parameter">insert</replaceable> | <replaceable class="parameter">update</replaceable> | <replaceable class="parameter">delete</replaceable> | <replaceable class="parameter">merge</replaceable> )
         [ SEARCH { BREADTH | DEPTH } FIRST BY <replaceable>column_name</replaceable> [, ...] SET <replaceable>search_seq_col_name</replaceable> ]
         [ CYCLE <replaceable>column_name</replaceable> [, ...] SET <replaceable>cycle_mark_col_name</replaceable> [ TO <replaceable>cycle_mark_value</replaceable> DEFAULT <replaceable>cycle_mark_default</replaceable> ] USING <replaceable>cycle_path_col_name</replaceable> ]
 
@@ -230,10 +230,10 @@ TABLE [ ONLY ] <replaceable class="param
     The subqueries effectively act as temporary tables or views
     for the duration of the primary query.
     Each subquery can be a <command>SELECT</command>, <command>TABLE</command>, <command>VALUES</command>,
-    <command>INSERT</command>, <command>UPDATE</command> or
-    <command>DELETE</command> statement.
+    <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command> statement.
     When writing a data-modifying statement (<command>INSERT</command>,
-    <command>UPDATE</command> or <command>DELETE</command>) in
+    <command>UPDATE</command>, <command>DELETE</command>, or <command>MERGE</command>) in
     <literal>WITH</literal>, it is usual to include a <literal>RETURNING</literal> clause.
     It is the output of <literal>RETURNING</literal>, <emphasis>not</emphasis> the underlying
     table that the statement modifies, that forms the temporary table that is
@@ -2184,7 +2184,8 @@ SELECT 2+2;
 
    <para>
     <productname>PostgreSQL</productname> allows <command>INSERT</command>,
-    <command>UPDATE</command>, and <command>DELETE</command> to be used as <literal>WITH</literal>
+    <command>UPDATE</command>, <command>DELETE</command>, and
+    <command>MERGE</command> to be used as <literal>WITH</literal>
     queries.  This is not found in the SQL standard.
    </para>
   </refsect2>
diff --git a/doc/src/sgml/rowtypes.sgml b/doc/src/sgml/rowtypes.sgml
new file mode 100644
index 4d86f97..bbeac84
--- a/doc/src/sgml/rowtypes.sgml
+++ b/doc/src/sgml/rowtypes.sgml
@@ -348,7 +348,7 @@ SELECT m.* FROM some_table, LATERAL myfu
    column expansion of this kind when it appears at the top level of
    a <link linkend="queries-select-lists"><command>SELECT</command> output
    list</link>, a <link linkend="dml-returning"><literal>RETURNING</literal>
-   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>,
+   list</link> in <command>INSERT</command>/<command>UPDATE</command>/<command>DELETE</command>/<command>MERGE</command>,
    a <link linkend="queries-values"><literal>VALUES</literal> clause</link>, or
    a <link linkend="sql-syntax-row-constructors">row constructor</link>.
    In all other contexts (including when nested inside one of those
diff --git a/doc/src/sgml/spi.sgml b/doc/src/sgml/spi.sgml
new file mode 100644
index 47f4b5b..bb37786
--- a/doc/src/sgml/spi.sgml
+++ b/doc/src/sgml/spi.sgml
@@ -301,8 +301,9 @@ SPI_execute("INSERT INTO foo SELECT * FR
    is returned in the global variable <varname>SPI_processed</varname>.
    If the return value of the function is <symbol>SPI_OK_SELECT</symbol>,
    <symbol>SPI_OK_INSERT_RETURNING</symbol>,
-   <symbol>SPI_OK_DELETE_RETURNING</symbol>, or
-   <symbol>SPI_OK_UPDATE_RETURNING</symbol>,
+   <symbol>SPI_OK_DELETE_RETURNING</symbol>,
+   <symbol>SPI_OK_UPDATE_RETURNING</symbol>, or
+   <symbol>SPI_OK_MERGE_RETURNING</symbol>,
    then you can use the
    global pointer <literal>SPITupleTable *SPI_tuptable</literal> to
    access the result rows.  Some utility commands (such as
@@ -472,6 +473,15 @@ typedef struct SPITupleTable
       </para>
      </listitem>
     </varlistentry>
+
+    <varlistentry>
+     <term><symbol>SPI_OK_MERGE_RETURNING</symbol></term>
+     <listitem>
+      <para>
+       if a <command>MERGE RETURNING</command> was executed
+      </para>
+     </listitem>
+    </varlistentry>
 
     <varlistentry>
      <term><symbol>SPI_OK_UTILITY</symbol></term>
diff --git a/doc/src/sgml/xfunc.sgml b/doc/src/sgml/xfunc.sgml
new file mode 100644
index 8a79ad0..7d05369
--- a/doc/src/sgml/xfunc.sgml
+++ b/doc/src/sgml/xfunc.sgml
@@ -177,7 +177,8 @@
     statements separated by semicolons.  A semicolon after the last
     statement is optional.  Unless the function is declared to return
     <type>void</type>, the last statement must be a <command>SELECT</command>,
-    or an <command>INSERT</command>, <command>UPDATE</command>, or <command>DELETE</command>
+    or an <command>INSERT</command>, <command>UPDATE</command>,
+    <command>DELETE</command>, or <command>MERGE</command>
     that has a <literal>RETURNING</literal> clause.
    </para>
 
@@ -1236,8 +1237,9 @@ SELECT x, CASE WHEN x &gt; 0 THEN genera
 
     <note>
      <para>
-      If a function's last command is <command>INSERT</command>, <command>UPDATE</command>,
-      or <command>DELETE</command> with <literal>RETURNING</literal>, that command will
+      If a function's last command is <command>INSERT</command>,
+      <command>UPDATE</command>, <command>DELETE</command>, or
+      <command>MERGE</command> with <literal>RETURNING</literal>, that command will
       always be executed to completion, even if the function is not declared
       with <literal>SETOF</literal> or the calling query does not fetch all the
       result rows.  Any extra rows produced by the <literal>RETURNING</literal>
diff --git a/src/backend/commands/copy.c b/src/backend/commands/copy.c
new file mode 100644
index 056b673..28cf8b0
--- a/src/backend/commands/copy.c
+++ b/src/backend/commands/copy.c
@@ -281,12 +281,6 @@ DoCopy(ParseState *pstate, const CopyStm
 	{
 		Assert(stmt->query);
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(stmt->query, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in COPY"));
-
 		query = makeNode(RawStmt);
 		query->stmt = stmt->query;
 		query->stmt_location = stmt_location;
diff --git a/src/backend/commands/copyto.c b/src/backend/commands/copyto.c
new file mode 100644
index a6962e0..ae8b2e3
--- a/src/backend/commands/copyto.c
+++ b/src/backend/commands/copyto.c
@@ -503,7 +503,8 @@ BeginCopyTo(ParseState *pstate,
 		{
 			Assert(query->commandType == CMD_INSERT ||
 				   query->commandType == CMD_UPDATE ||
-				   query->commandType == CMD_DELETE);
+				   query->commandType == CMD_DELETE ||
+				   query->commandType == CMD_MERGE);
 
 			ereport(ERROR,
 					(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
diff --git a/src/backend/executor/execExpr.c b/src/backend/executor/execExpr.c
new file mode 100644
index ffd3ca4..728c8d5
--- a/src/backend/executor/execExpr.c
+++ b/src/backend/executor/execExpr.c
@@ -1107,6 +1107,19 @@ ExecInitExprRec(Expr *node, ExprState *s
 				break;
 			}
 
+		case T_MergeSupportFunc:
+			{
+				/* must be in a MERGE, else something messed up */
+				if (!state->parent ||
+					!IsA(state->parent, ModifyTableState) ||
+					((ModifyTableState *) state->parent)->operation != CMD_MERGE)
+					elog(ERROR, "MergeSupportFunc found in non-merge plan node");
+
+				scratch.opcode = EEOP_MERGE_SUPPORT_FUNC;
+				ExprEvalPushStep(state, &scratch);
+				break;
+			}
+
 		case T_SubscriptingRef:
 			{
 				SubscriptingRef *sbsref = (SubscriptingRef *) node;
diff --git a/src/backend/executor/execExprInterp.c b/src/backend/executor/execExprInterp.c
new file mode 100644
index 7c1f51e..a25ab75
--- a/src/backend/executor/execExprInterp.c
+++ b/src/backend/executor/execExprInterp.c
@@ -484,6 +484,7 @@ ExecInterpExpr(ExprState *state, ExprCon
 		&&CASE_EEOP_AGGREF,
 		&&CASE_EEOP_GROUPING_FUNC,
 		&&CASE_EEOP_WINDOW_FUNC,
+		&&CASE_EEOP_MERGE_SUPPORT_FUNC,
 		&&CASE_EEOP_SUBPLAN,
 		&&CASE_EEOP_AGG_STRICT_DESERIALIZE,
 		&&CASE_EEOP_AGG_DESERIALIZE,
@@ -1592,6 +1593,14 @@ ExecInterpExpr(ExprState *state, ExprCon
 			EEO_NEXT();
 		}
 
+		EEO_CASE(EEOP_MERGE_SUPPORT_FUNC)
+		{
+			/* too complex/uncommon for an inline implementation */
+			ExecEvalMergeSupportFunc(state, op, econtext);
+
+			EEO_NEXT();
+		}
+
 		EEO_CASE(EEOP_SUBPLAN)
 		{
 			/* too complex for an inline implementation */
@@ -4246,6 +4255,45 @@ ExecEvalGroupingFunc(ExprState *state, E
 }
 
 /*
+ * ExecEvalMergeSupportFunc
+ *
+ * Returns information about the current MERGE action for its RETURNING list.
+ */
+void
+ExecEvalMergeSupportFunc(ExprState *state, ExprEvalStep *op,
+						 ExprContext *econtext)
+{
+	ModifyTableState *mtstate = castNode(ModifyTableState, state->parent);
+	MergeActionState *relaction = mtstate->mt_merge_action;
+
+	if (!relaction)
+		elog(ERROR, "no merge action in progress");
+
+	/* Return the MERGE action ("INSERT", "UPDATE", or "DELETE") */
+	switch (relaction->mas_action->commandType)
+	{
+		case CMD_INSERT:
+			*op->resvalue = PointerGetDatum(cstring_to_text_with_len("INSERT", 6));
+			*op->resnull = false;
+			break;
+		case CMD_UPDATE:
+			*op->resvalue = PointerGetDatum(cstring_to_text_with_len("UPDATE", 6));
+			*op->resnull = false;
+			break;
+		case CMD_DELETE:
+			*op->resvalue = PointerGetDatum(cstring_to_text_with_len("DELETE", 6));
+			*op->resnull = false;
+			break;
+		case CMD_NOTHING:
+			elog(ERROR, "unexpected merge action: DO NOTHING");
+			break;
+		default:
+			elog(ERROR, "unrecognized commandType: %d",
+				 (int) relaction->mas_action->commandType);
+	}
+}
+
+/*
  * Hand off evaluation of a subplan to nodeSubplan.c
  */
 void
diff --git a/src/backend/executor/execPartition.c b/src/backend/executor/execPartition.c
new file mode 100644
index 8ca512d..64fcb01
--- a/src/backend/executor/execPartition.c
+++ b/src/backend/executor/execPartition.c
@@ -609,8 +609,8 @@ ExecInitPartitionInfo(ModifyTableState *
 	 * Build the RETURNING projection for the partition.  Note that we didn't
 	 * build the returningList for partitions within the planner, but simple
 	 * translation of varattnos will suffice.  This only occurs for the INSERT
-	 * case or in the case of UPDATE tuple routing where we didn't find a
-	 * result rel to reuse.
+	 * case or in the case of UPDATE/MERGE tuple routing where we didn't find
+	 * a result rel to reuse.
 	 */
 	if (node && node->returningLists != NIL)
 	{
@@ -619,12 +619,14 @@ ExecInitPartitionInfo(ModifyTableState *
 		List	   *returningList;
 
 		/* See the comment above for WCO lists. */
-		/* (except no RETURNING support for MERGE yet) */
 		Assert((node->operation == CMD_INSERT &&
 				list_length(node->returningLists) == 1 &&
 				list_length(node->resultRelations) == 1) ||
 			   (node->operation == CMD_UPDATE &&
 				list_length(node->returningLists) ==
+				list_length(node->resultRelations)) ||
+			   (node->operation == CMD_MERGE &&
+				list_length(node->returningLists) ==
 				list_length(node->resultRelations)));
 
 		/*
diff --git a/src/backend/executor/functions.c b/src/backend/executor/functions.c
new file mode 100644
index 6e926ef..539cd0a
--- a/src/backend/executor/functions.c
+++ b/src/backend/executor/functions.c
@@ -1662,8 +1662,8 @@ check_sql_fn_retval(List *queryTreeLists
 
 	/*
 	 * If it's a plain SELECT, it returns whatever the targetlist says.
-	 * Otherwise, if it's INSERT/UPDATE/DELETE with RETURNING, it returns
-	 * that. Otherwise, the function return type must be VOID.
+	 * Otherwise, if it's INSERT/UPDATE/DELETE/MERGE with RETURNING, it
+	 * returns that. Otherwise, the function return type must be VOID.
 	 *
 	 * Note: eventually replace this test with QueryReturnsTuples?	We'd need
 	 * a more general method of determining the output type, though.  Also, it
@@ -1681,7 +1681,8 @@ check_sql_fn_retval(List *queryTreeLists
 	else if (parse &&
 			 (parse->commandType == CMD_INSERT ||
 			  parse->commandType == CMD_UPDATE ||
-			  parse->commandType == CMD_DELETE) &&
+			  parse->commandType == CMD_DELETE ||
+			  parse->commandType == CMD_MERGE) &&
 			 parse->returningList)
 	{
 		tlist = parse->returningList;
@@ -1695,7 +1696,7 @@ check_sql_fn_retval(List *queryTreeLists
 				(errcode(ERRCODE_INVALID_FUNCTION_DEFINITION),
 				 errmsg("return type mismatch in function declared to return %s",
 						format_type_be(rettype)),
-				 errdetail("Function's final statement must be SELECT or INSERT/UPDATE/DELETE RETURNING.")));
+				 errdetail("Function's final statement must be SELECT or INSERT/UPDATE/DELETE/MERGE RETURNING.")));
 		return false;			/* keep compiler quiet */
 	}
 
diff --git a/src/backend/executor/nodeModifyTable.c b/src/backend/executor/nodeModifyTable.c
new file mode 100644
index 9351fbc..4abfe82
--- a/src/backend/executor/nodeModifyTable.c
+++ b/src/backend/executor/nodeModifyTable.c
@@ -36,8 +36,7 @@
  *		RETURNING tuple after completing each row insert, update, or delete.
  *		It must be called again to continue the operation.  Without RETURNING,
  *		we just loop within the node until all the work is done, then
- *		return NULL.  This avoids useless call/return overhead.  (MERGE does
- *		not support RETURNING.)
+ *		return NULL.  This avoids useless call/return overhead.
  */
 
 #include "postgres.h"
@@ -85,9 +84,6 @@ typedef struct ModifyTableContext
 	 */
 	TupleTableSlot *planSlot;
 
-	/* MERGE specific */
-	MergeActionState *relaction;	/* MERGE action in progress */
-
 	/*
 	 * Information about the changes that were made concurrently to a tuple
 	 * being updated or deleted
@@ -150,14 +146,15 @@ static TupleTableSlot *ExecMerge(ModifyT
 								 HeapTuple oldtuple,
 								 bool canSetTag);
 static void ExecInitMerge(ModifyTableState *mtstate, EState *estate);
-static bool ExecMergeMatched(ModifyTableContext *context,
-							 ResultRelInfo *resultRelInfo,
-							 ItemPointer tupleid,
-							 HeapTuple oldtuple,
-							 bool canSetTag);
-static void ExecMergeNotMatched(ModifyTableContext *context,
-								ResultRelInfo *resultRelInfo,
-								bool canSetTag);
+static TupleTableSlot *ExecMergeMatched(ModifyTableContext *context,
+										ResultRelInfo *resultRelInfo,
+										ItemPointer tupleid,
+										HeapTuple oldtuple,
+										bool canSetTag,
+										bool *matched);
+static TupleTableSlot *ExecMergeNotMatched(ModifyTableContext *context,
+										   ResultRelInfo *resultRelInfo,
+										   bool canSetTag);
 
 
 /*
@@ -977,7 +974,7 @@ ExecInsert(ModifyTableContext *context,
 		if (mtstate->operation == CMD_UPDATE)
 			wco_kind = WCO_RLS_UPDATE_CHECK;
 		else if (mtstate->operation == CMD_MERGE)
-			wco_kind = (context->relaction->mas_action->commandType == CMD_UPDATE) ?
+			wco_kind = (mtstate->mt_merge_action->mas_action->commandType == CMD_UPDATE) ?
 				WCO_RLS_UPDATE_CHECK : WCO_RLS_INSERT_CHECK;
 		else
 			wco_kind = WCO_RLS_INSERT_CHECK;
@@ -1831,7 +1828,7 @@ ExecCrossPartitionUpdate(ModifyTableCont
 		 * additional rechecking, and might end up executing a different
 		 * action entirely).
 		 */
-		if (context->relaction != NULL)
+		if (mtstate->operation == CMD_MERGE)
 			return *tmresult == TM_Ok;
 		else if (TupIsNull(epqslot))
 			return true;
@@ -2072,7 +2069,7 @@ lreplace:
 		 * No luck, a retry is needed.  If running MERGE, we do not do so
 		 * here; instead let it handle that on its own rules.
 		 */
-		if (context->relaction != NULL)
+		if (context->mtstate->operation == CMD_MERGE)
 			return result;
 
 		/*
@@ -2713,6 +2710,7 @@ static TupleTableSlot *
 ExecMerge(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 		  ItemPointer tupleid, HeapTuple oldtuple, bool canSetTag)
 {
+	TupleTableSlot *rslot = NULL;
 	bool		matched;
 
 	/*-----
@@ -2761,19 +2759,18 @@ ExecMerge(ModifyTableContext *context, R
 	 */
 	matched = tupleid != NULL || oldtuple != NULL;
 	if (matched)
-		matched = ExecMergeMatched(context, resultRelInfo, tupleid, oldtuple,
-								   canSetTag);
+		rslot = ExecMergeMatched(context, resultRelInfo, tupleid, oldtuple,
+								 canSetTag, &matched);
 
 	/*
-	 * Either we were dealing with a NOT MATCHED tuple or ExecMergeMatched()
-	 * returned "false", indicating the previously MATCHED tuple no longer
-	 * matches.
+	 * Deal with the NOT MATCHED case (either a NOT MATCHED tuple from the
+	 * join, or a previously MATCHED tuple for which ExecMergeMatched() set
+	 * "matched" to false, indicating that it no longer matches).
 	 */
 	if (!matched)
-		ExecMergeNotMatched(context, resultRelInfo, canSetTag);
+		rslot = ExecMergeNotMatched(context, resultRelInfo, canSetTag);
 
-	/* No RETURNING support yet */
-	return NULL;
+	return rslot;
 }
 
 /*
@@ -2785,8 +2782,8 @@ ExecMerge(ModifyTableContext *context, R
  * We start from the first WHEN MATCHED action and check if the WHEN quals
  * pass, if any. If the WHEN quals for the first action do not pass, we
  * check the second, then the third and so on. If we reach to the end, no
- * action is taken and we return true, indicating that no further action is
- * required for this tuple.
+ * action is taken and "matched" is set to true, indicating that no further
+ * action is required for this tuple.
  *
  * If we do find a qualifying action, then we attempt to execute the action.
  *
@@ -2795,16 +2792,18 @@ ExecMerge(ModifyTableContext *context, R
  * with individual actions are evaluated by this routine via ExecQual, while
  * EvalPlanQual checks for the join quals. If EvalPlanQual tells us that the
  * updated tuple still passes the join quals, then we restart from the first
- * action to look for a qualifying action. Otherwise, we return false --
- * meaning that a NOT MATCHED action must now be executed for the current
- * source tuple.
+ * action to look for a qualifying action. Otherwise, "matched" is set to
+ * false -- meaning that a NOT MATCHED action must now be executed for the
+ * current source tuple.
  */
-static bool
+static TupleTableSlot *
 ExecMergeMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
-				 ItemPointer tupleid, HeapTuple oldtuple, bool canSetTag)
+				 ItemPointer tupleid, HeapTuple oldtuple, bool canSetTag,
+				 bool *matched)
 {
 	ModifyTableState *mtstate = context->mtstate;
-	TupleTableSlot *newslot;
+	TupleTableSlot *newslot = NULL;
+	TupleTableSlot *rslot = NULL;
 	EState	   *estate = context->estate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
 	bool		isNull;
@@ -2815,7 +2814,10 @@ ExecMergeMatched(ModifyTableContext *con
 	 * If there are no WHEN MATCHED actions, we are done.
 	 */
 	if (resultRelInfo->ri_matchedMergeAction == NIL)
-		return true;
+	{
+		*matched = true;
+		return NULL;
+	}
 
 	/*
 	 * Make tuple and any needed join variables available to ExecQual and
@@ -2905,12 +2907,15 @@ lmerge_matched:
 				 */
 				newslot = ExecProject(relaction->mas_proj);
 
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				if (!ExecUpdatePrologue(context, resultRelInfo,
 										tupleid, NULL, newslot, &result))
 				{
 					if (result == TM_Ok)
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 					break;		/* concurrent update/delete */
 				}
 
@@ -2920,7 +2925,10 @@ lmerge_matched:
 				{
 					if (!ExecIRUpdateTriggers(estate, resultRelInfo,
 											  oldtuple, newslot))
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 				}
 				else
 				{
@@ -2933,12 +2941,15 @@ lmerge_matched:
 					 * cross-partition update was done, then there's nothing
 					 * else for us to do --- the UPDATE has been turned into a
 					 * DELETE and an INSERT, and we must not perform any of
-					 * the usual post-update tasks.
+					 * the usual post-update tasks.  Also, the RETURNING tuple
+					 * (if any) has been projected, so we can just return
+					 * that.
 					 */
 					if (updateCxt.crossPartUpdate)
 					{
 						mtstate->mt_merge_updated += 1;
-						return true;
+						*matched = true;
+						return context->cpUpdateReturningSlot;
 					}
 				}
 
@@ -2951,12 +2962,15 @@ lmerge_matched:
 				break;
 
 			case CMD_DELETE:
-				context->relaction = relaction;
+				mtstate->mt_merge_action = relaction;
 				if (!ExecDeletePrologue(context, resultRelInfo, tupleid,
 										NULL, NULL, &result))
 				{
 					if (result == TM_Ok)
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 					break;		/* concurrent update/delete */
 				}
 
@@ -2966,7 +2980,10 @@ lmerge_matched:
 				{
 					if (!ExecIRDeleteTriggers(estate, resultRelInfo,
 											  oldtuple))
-						return true;	/* "do nothing" */
+					{
+						*matched = true;
+						return NULL;	/* "do nothing" */
+					}
 				}
 				else
 					result = ExecDeleteAct(context, resultRelInfo, tupleid,
@@ -3046,7 +3063,8 @@ lmerge_matched:
 				 * If the tuple was already deleted, return to let caller
 				 * handle it under NOT MATCHED clauses.
 				 */
-				return false;
+				*matched = false;
+				return NULL;
 
 			case TM_Updated:
 				{
@@ -3092,13 +3110,19 @@ lmerge_matched:
 							 * NOT MATCHED actions.
 							 */
 							if (TupIsNull(epqslot))
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							(void) ExecGetJunkAttribute(epqslot,
 														resultRelInfo->ri_RowIdAttNo,
 														&isNull);
 							if (isNull)
-								return false;
+							{
+								*matched = false;
+								return NULL;
+							}
 
 							/*
 							 * When a tuple was updated and migrated to
@@ -3133,7 +3157,8 @@ lmerge_matched:
 							 * tuple already deleted; tell caller to run NOT
 							 * MATCHED actions
 							 */
-							return false;
+							*matched = false;
+							return NULL;
 
 						case TM_SelfModified:
 
@@ -3161,13 +3186,13 @@ lmerge_matched:
 
 							/* This shouldn't happen */
 							elog(ERROR, "attempted to update or delete invisible tuple");
-							return false;
+							return NULL;
 
 						default:
 							/* see table_tuple_lock call in ExecDelete() */
 							elog(ERROR, "unexpected table_tuple_lock status: %u",
 								 result);
-							return false;
+							return NULL;
 					}
 				}
 
@@ -3179,6 +3204,31 @@ lmerge_matched:
 				break;
 		}
 
+		/* Process RETURNING if present */
+		if (resultRelInfo->ri_projectReturning)
+		{
+			switch (commandType)
+			{
+				case CMD_UPDATE:
+					rslot = ExecProcessReturning(resultRelInfo, newslot,
+												 context->planSlot);
+					break;
+
+				case CMD_DELETE:
+					rslot = ExecProcessReturning(resultRelInfo,
+												 resultRelInfo->ri_oldTupleSlot,
+												 context->planSlot);
+					break;
+
+				case CMD_NOTHING:
+					break;
+
+				default:
+					elog(ERROR, "unrecognized commandType: %d",
+						 (int) commandType);
+			}
+		}
+
 		/*
 		 * We've activated one of the WHEN clauses, so we don't search
 		 * further. This is required behaviour, not an optimization.
@@ -3189,19 +3239,22 @@ lmerge_matched:
 	/*
 	 * Successfully executed an action or no qualifying action was found.
 	 */
-	return true;
+	*matched = true;
+
+	return rslot;
 }
 
 /*
  * Execute the first qualifying NOT MATCHED action.
  */
-static void
+static TupleTableSlot *
 ExecMergeNotMatched(ModifyTableContext *context, ResultRelInfo *resultRelInfo,
 					bool canSetTag)
 {
 	ModifyTableState *mtstate = context->mtstate;
 	ExprContext *econtext = mtstate->ps.ps_ExprContext;
 	List	   *actionStates = NIL;
+	TupleTableSlot *rslot = NULL;
 	ListCell   *l;
 
 	/*
@@ -3251,10 +3304,10 @@ ExecMergeNotMatched(ModifyTableContext *
 				 * so we don't need to map the tuple here.
 				 */
 				newslot = ExecProject(action->mas_proj);
-				context->relaction = action;
+				mtstate->mt_merge_action = action;
 
-				(void) ExecInsert(context, mtstate->rootResultRelInfo, newslot,
-								  canSetTag, NULL, NULL);
+				rslot = ExecInsert(context, mtstate->rootResultRelInfo,
+								   newslot, canSetTag, NULL, NULL);
 				mtstate->mt_merge_inserted += 1;
 				break;
 			case CMD_NOTHING:
@@ -3270,6 +3323,8 @@ ExecMergeNotMatched(ModifyTableContext *
 		 */
 		break;
 	}
+
+	return rslot;
 }
 
 /*
@@ -3732,9 +3787,17 @@ ExecModifyTable(PlanState *pstate)
 				{
 					EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-					ExecMerge(&context, node->resultRelInfo, NULL, NULL,
-							  node->canSetTag);
-					continue;	/* no RETURNING support yet */
+					slot = ExecMerge(&context, node->resultRelInfo,
+									 NULL, NULL, node->canSetTag);
+
+					/*
+					 * If we got a RETURNING result, return it to the caller.
+					 * We'll continue the work on next call.
+					 */
+					if (slot)
+						return slot;
+
+					continue;	/* continue with the next tuple */
 				}
 
 				elog(ERROR, "tableoid is NULL");
@@ -3811,9 +3874,17 @@ ExecModifyTable(PlanState *pstate)
 					{
 						EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-						ExecMerge(&context, node->resultRelInfo, NULL, NULL,
-								  node->canSetTag);
-						continue;	/* no RETURNING support yet */
+						slot = ExecMerge(&context, node->resultRelInfo,
+										 NULL, NULL, node->canSetTag);
+
+						/*
+						 * If we got a RETURNING result, return it to the
+						 * caller.  We'll continue the work on next call.
+						 */
+						if (slot)
+							return slot;
+
+						continue;	/* continue with the next tuple */
 					}
 
 					elog(ERROR, "ctid is NULL");
@@ -3860,9 +3931,17 @@ ExecModifyTable(PlanState *pstate)
 					{
 						EvalPlanQualSetSlot(&node->mt_epqstate, context.planSlot);
 
-						ExecMerge(&context, node->resultRelInfo, NULL, NULL,
-								  node->canSetTag);
-						continue;	/* no RETURNING support yet */
+						slot = ExecMerge(&context, node->resultRelInfo,
+										 NULL, NULL, node->canSetTag);
+
+						/*
+						 * If we got a RETURNING result, return it to the
+						 * caller.  We'll continue the work on next call.
+						 */
+						if (slot)
+							return slot;
+
+						continue;	/* continue with the next tuple */
 					}
 
 					elog(ERROR, "wholerow is NULL");
@@ -3924,7 +4003,6 @@ ExecModifyTable(PlanState *pstate)
 				}
 				slot = ExecGetUpdateNewTuple(resultRelInfo, context.planSlot,
 											 oldSlot);
-				context.relaction = NULL;
 
 				/* Now apply the update. */
 				slot = ExecUpdate(&context, resultRelInfo, tupleid, oldtuple,
diff --git a/src/backend/executor/spi.c b/src/backend/executor/spi.c
new file mode 100644
index 8585771..a97a7e3
--- a/src/backend/executor/spi.c
+++ b/src/backend/executor/spi.c
@@ -2032,6 +2032,8 @@ SPI_result_code_string(int code)
 			return "SPI_OK_TD_REGISTER";
 		case SPI_OK_MERGE:
 			return "SPI_OK_MERGE";
+		case SPI_OK_MERGE_RETURNING:
+			return "SPI_OK_MERGE_RETURNING";
 	}
 	/* Unrecognized code ... return something useful ... */
 	sprintf(buf, "Unrecognized SPI code %d", code);
@@ -2885,7 +2887,10 @@ _SPI_pquery(QueryDesc *queryDesc, bool f
 				res = SPI_OK_UPDATE;
 			break;
 		case CMD_MERGE:
-			res = SPI_OK_MERGE;
+			if (queryDesc->plannedstmt->hasReturning)
+				res = SPI_OK_MERGE_RETURNING;
+			else
+				res = SPI_OK_MERGE;
 			break;
 		default:
 			return SPI_ERROR_OPUNKNOWN;
diff --git a/src/backend/jit/llvm/llvmjit_expr.c b/src/backend/jit/llvm/llvmjit_expr.c
new file mode 100644
index 0c44842..2a7d84f
--- a/src/backend/jit/llvm/llvmjit_expr.c
+++ b/src/backend/jit/llvm/llvmjit_expr.c
@@ -1986,6 +1986,12 @@ llvm_compile_expr(ExprState *state)
 					break;
 				}
 
+			case EEOP_MERGE_SUPPORT_FUNC:
+				build_EvalXFunc(b, mod, "ExecEvalMergeSupportFunc",
+								v_state, op, v_econtext);
+				LLVMBuildBr(b, opblocks[opno + 1]);
+				break;
+
 			case EEOP_SUBPLAN:
 				build_EvalXFunc(b, mod, "ExecEvalSubPlan",
 								v_state, op, v_econtext);
diff --git a/src/backend/jit/llvm/llvmjit_types.c b/src/backend/jit/llvm/llvmjit_types.c
new file mode 100644
index 47c9daf..7d7aeee
--- a/src/backend/jit/llvm/llvmjit_types.c
+++ b/src/backend/jit/llvm/llvmjit_types.c
@@ -155,6 +155,7 @@ void	   *referenced_functions[] =
 	ExecEvalFuncExprFusage,
 	ExecEvalFuncExprStrictFusage,
 	ExecEvalGroupingFunc,
+	ExecEvalMergeSupportFunc,
 	ExecEvalMinMax,
 	ExecEvalNextValueExpr,
 	ExecEvalParamExec,
diff --git a/src/backend/nodes/nodeFuncs.c b/src/backend/nodes/nodeFuncs.c
new file mode 100644
index 6ba8e73..5b70280
--- a/src/backend/nodes/nodeFuncs.c
+++ b/src/backend/nodes/nodeFuncs.c
@@ -66,6 +66,9 @@ exprType(const Node *expr)
 		case T_WindowFunc:
 			type = ((const WindowFunc *) expr)->wintype;
 			break;
+		case T_MergeSupportFunc:
+			type = ((const MergeSupportFunc *) expr)->msftype;
+			break;
 		case T_SubscriptingRef:
 			type = ((const SubscriptingRef *) expr)->refrestype;
 			break;
@@ -809,6 +812,9 @@ exprCollation(const Node *expr)
 		case T_WindowFunc:
 			coll = ((const WindowFunc *) expr)->wincollid;
 			break;
+		case T_MergeSupportFunc:
+			coll = ((const MergeSupportFunc *) expr)->msfcollid;
+			break;
 		case T_SubscriptingRef:
 			coll = ((const SubscriptingRef *) expr)->refcollid;
 			break;
@@ -1084,6 +1090,9 @@ exprSetCollation(Node *expr, Oid collati
 		case T_WindowFunc:
 			((WindowFunc *) expr)->wincollid = collation;
 			break;
+		case T_MergeSupportFunc:
+			((MergeSupportFunc *) expr)->msfcollid = collation;
+			break;
 		case T_SubscriptingRef:
 			((SubscriptingRef *) expr)->refcollid = collation;
 			break;
@@ -1342,6 +1351,9 @@ exprLocation(const Node *expr)
 			/* function name should always be the first thing */
 			loc = ((const WindowFunc *) expr)->location;
 			break;
+		case T_MergeSupportFunc:
+			loc = ((const MergeSupportFunc *) expr)->location;
+			break;
 		case T_SubscriptingRef:
 			/* just use container argument's location */
 			loc = exprLocation((Node *) ((const SubscriptingRef *) expr)->refexpr);
@@ -2034,6 +2046,7 @@ expression_tree_walker_impl(Node *node,
 		case T_RangeTblRef:
 		case T_SortGroupClause:
 		case T_CTESearchClause:
+		case T_MergeSupportFunc:
 			/* primitive node types with no expression subnodes */
 			break;
 		case T_WithCheckOption:
@@ -2868,6 +2881,7 @@ expression_tree_mutator_impl(Node *node,
 		case T_RangeTblRef:
 		case T_SortGroupClause:
 		case T_CTESearchClause:
+		case T_MergeSupportFunc:
 			return (Node *) copyObject(node);
 		case T_WithCheckOption:
 			{
@@ -3832,6 +3846,7 @@ raw_expression_tree_walker_impl(Node *no
 		case T_ParamRef:
 		case T_A_Const:
 		case T_A_Star:
+		case T_MergeSupportFunc:
 			/* primitive node types with no subnodes */
 			break;
 		case T_Alias:
@@ -4052,6 +4067,8 @@ raw_expression_tree_walker_impl(Node *no
 					return true;
 				if (WALK(stmt->mergeWhenClauses))
 					return true;
+				if (WALK(stmt->returningList))
+					return true;
 				if (WALK(stmt->withClause))
 					return true;
 			}
diff --git a/src/backend/optimizer/plan/subselect.c b/src/backend/optimizer/plan/subselect.c
new file mode 100644
index 47e1472..b9e0c96
--- a/src/backend/optimizer/plan/subselect.c
+++ b/src/backend/optimizer/plan/subselect.c
@@ -1855,7 +1855,8 @@ convert_EXISTS_to_ANY(PlannerInfo *root,
 /*
  * Replace correlation vars (uplevel vars) with Params.
  *
- * Uplevel PlaceHolderVars and aggregates are replaced, too.
+ * Uplevel PlaceHolderVars, aggregates, GROUPING() expressions, and
+ * MergeSupportFuncs are replaced, too.
  *
  * Note: it is critical that this runs immediately after SS_process_sublinks.
  * Since we do not recurse into the arguments of uplevel PHVs and aggregates,
@@ -1909,6 +1910,12 @@ replace_correlation_vars_mutator(Node *n
 		if (((GroupingFunc *) node)->agglevelsup > 0)
 			return (Node *) replace_outer_grouping(root, (GroupingFunc *) node);
 	}
+	if (IsA(node, MergeSupportFunc))
+	{
+		if (root->parse->commandType != CMD_MERGE)
+			return (Node *) replace_outer_merge_support(root,
+														(MergeSupportFunc *) node);
+	}
 	return expression_tree_mutator(node,
 								   replace_correlation_vars_mutator,
 								   (void *) root);
diff --git a/src/backend/optimizer/util/paramassign.c b/src/backend/optimizer/util/paramassign.c
new file mode 100644
index a58da7c..f461fed
--- a/src/backend/optimizer/util/paramassign.c
+++ b/src/backend/optimizer/util/paramassign.c
@@ -308,6 +308,57 @@ replace_outer_grouping(PlannerInfo *root
 }
 
 /*
+ * Generate a Param node to replace the given MergeSupportFunc expression
+ * which is expected to be in the RETURNING list of an upper-level MERGE
+ * query.  Record the need for the MergeSupportFunc in the proper upper-level
+ * root->plan_params.
+ */
+Param *
+replace_outer_merge_support(PlannerInfo *root, MergeSupportFunc *msf)
+{
+	Param	   *retval;
+	PlannerParamItem *pitem;
+	Oid			ptype = exprType((Node *) msf);
+
+	Assert(root->parse->commandType != CMD_MERGE);
+
+	/*
+	 * The parser should have ensured that the MergeSupportFunc is in the
+	 * RETURNING list of an upper-level MERGE query, so find that query.
+	 */
+	do
+	{
+		root = root->parent_root;
+		if (root == NULL)
+			elog(ERROR, "MergeSupportFunc found outside MERGE");
+	} while (root->parse->commandType != CMD_MERGE);
+
+	/*
+	 * It does not seem worthwhile to try to de-duplicate references to outer
+	 * MergeSupportFunc expressions.  Just make a new slot every time.
+	 */
+	msf = copyObject(msf);
+
+	pitem = makeNode(PlannerParamItem);
+	pitem->item = (Node *) msf;
+	pitem->paramId = list_length(root->glob->paramExecTypes);
+	root->glob->paramExecTypes = lappend_oid(root->glob->paramExecTypes,
+											 ptype);
+
+	root->plan_params = lappend(root->plan_params, pitem);
+
+	retval = makeNode(Param);
+	retval->paramkind = PARAM_EXEC;
+	retval->paramid = pitem->paramId;
+	retval->paramtype = ptype;
+	retval->paramtypmod = -1;
+	retval->paramcollid = InvalidOid;
+	retval->location = msf->location;
+
+	return retval;
+}
+
+/*
  * Generate a Param node to replace the given Var,
  * which is expected to come from some upper NestLoop plan node.
  * Record the need for the Var in root->curOuterParams.
diff --git a/src/backend/parser/analyze.c b/src/backend/parser/analyze.c
new file mode 100644
index 7f23d18..40ea19e
--- a/src/backend/parser/analyze.c
+++ b/src/backend/parser/analyze.c
@@ -72,7 +72,6 @@ static void determineRecursiveColTypes(P
 									   Node *larg, List *nrtargetlist);
 static Query *transformReturnStmt(ParseState *pstate, ReturnStmt *stmt);
 static Query *transformUpdateStmt(ParseState *pstate, UpdateStmt *stmt);
-static List *transformReturningList(ParseState *pstate, List *returningList);
 static Query *transformPLAssignStmt(ParseState *pstate,
 									PLAssignStmt *stmt);
 static Query *transformDeclareCursorStmt(ParseState *pstate,
@@ -551,7 +550,8 @@ transformDeleteStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -978,7 +978,8 @@ transformInsertStmt(ParseState *pstate,
 	/* Process RETURNING, if any. */
 	if (stmt->returningList)
 		qry->returningList = transformReturningList(pstate,
-													stmt->returningList);
+													stmt->returningList,
+													EXPR_KIND_RETURNING);
 
 	/* done building the range table and jointree */
 	qry->rtable = pstate->p_rtable;
@@ -2454,7 +2455,8 @@ transformUpdateStmt(ParseState *pstate,
 	qual = transformWhereClause(pstate, stmt->whereClause,
 								EXPR_KIND_WHERE, "WHERE");
 
-	qry->returningList = transformReturningList(pstate, stmt->returningList);
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_RETURNING);
 
 	/*
 	 * Now we are done with SELECT-like processing, and can get on with
@@ -2551,10 +2553,11 @@ transformUpdateTargetList(ParseState *ps
 
 /*
  * transformReturningList -
- *	handle a RETURNING clause in INSERT/UPDATE/DELETE
+ *	handle a RETURNING clause in INSERT/UPDATE/DELETE/MERGE
  */
-static List *
-transformReturningList(ParseState *pstate, List *returningList)
+List *
+transformReturningList(ParseState *pstate, List *returningList,
+					   ParseExprKind exprKind)
 {
 	List	   *rlist;
 	int			save_next_resno;
@@ -2571,7 +2574,7 @@ transformReturningList(ParseState *pstat
 	pstate->p_next_resno = 1;
 
 	/* transform RETURNING identically to a SELECT targetlist */
-	rlist = transformTargetList(pstate, returningList, EXPR_KIND_RETURNING);
+	rlist = transformTargetList(pstate, returningList, exprKind);
 
 	/*
 	 * Complain if the nonempty tlist expanded to nothing (which is possible
diff --git a/src/backend/parser/gram.y b/src/backend/parser/gram.y
new file mode 100644
index c6e2f67..6419b32
--- a/src/backend/parser/gram.y
+++ b/src/backend/parser/gram.y
@@ -733,7 +733,7 @@ static Node *makeRecursiveViewSelect(cha
 	LEADING LEAKPROOF LEAST LEFT LEVEL LIKE LIMIT LISTEN LOAD LOCAL
 	LOCALTIME LOCALTIMESTAMP LOCATION LOCK_P LOCKED LOGGED
 
-	MAPPING MATCH MATCHED MATERIALIZED MAXVALUE MERGE METHOD
+	MAPPING MATCH MATCHED MATERIALIZED MAXVALUE MERGE MERGE_ACTION METHOD
 	MINUTE_P MINVALUE MODE MONTH_P MOVE
 
 	NAME_P NAMES NATIONAL NATURAL NCHAR NEW NEXT NFC NFD NFKC NFKD NO NONE
@@ -12374,6 +12374,7 @@ MergeStmt:
 			USING table_ref
 			ON a_expr
 			merge_when_list
+			returning_clause
 				{
 					MergeStmt  *m = makeNode(MergeStmt);
 
@@ -12382,6 +12383,7 @@ MergeStmt:
 					m->sourceRelation = $6;
 					m->joinCondition = $8;
 					m->mergeWhenClauses = $9;
+					m->returningList = $10;
 
 					$$ = (Node *) m;
 				}
@@ -15795,6 +15797,14 @@ func_expr_common_subexpr:
 					n->location = @1;
 					$$ = (Node *) n;
 				}
+			| MERGE_ACTION '(' ')'
+				{
+					MergeSupportFunc *m = makeNode(MergeSupportFunc);
+
+					m->msftype = TEXTOID;
+					m->location = @1;
+					$$ = (Node *) m;
+				}
 			;
 
 
@@ -17492,6 +17502,7 @@ col_name_keyword:
 			| JSON_SCALAR
 			| JSON_SERIALIZE
 			| LEAST
+			| MERGE_ACTION
 			| NATIONAL
 			| NCHAR
 			| NONE
@@ -17881,6 +17892,7 @@ bare_label_keyword:
 			| MATERIALIZED
 			| MAXVALUE
 			| MERGE
+			| MERGE_ACTION
 			| METHOD
 			| MINVALUE
 			| MODE
diff --git a/src/backend/parser/parse_agg.c b/src/backend/parser/parse_agg.c
new file mode 100644
index 9d151a8..bee7d83
--- a/src/backend/parser/parse_agg.c
+++ b/src/backend/parser/parse_agg.c
@@ -468,6 +468,7 @@ check_agglevels_and_constraints(ParseSta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
@@ -915,6 +916,7 @@ transformWindowFuncCall(ParseState *psta
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_cte.c b/src/backend/parser/parse_cte.c
new file mode 100644
index 3c88c9a..6826d4f
--- a/src/backend/parser/parse_cte.c
+++ b/src/backend/parser/parse_cte.c
@@ -126,13 +126,6 @@ transformWithClause(ParseState *pstate,
 		CommonTableExpr *cte = (CommonTableExpr *) lfirst(lc);
 		ListCell   *rest;
 
-		/* MERGE is allowed by parser, but unimplemented. Reject for now */
-		if (IsA(cte->ctequery, MergeStmt))
-			ereport(ERROR,
-					errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
-					errmsg("MERGE not supported in WITH query"),
-					parser_errposition(pstate, cte->location));
-
 		for_each_cell(rest, withClause->ctes, lnext(withClause->ctes, lc))
 		{
 			CommonTableExpr *cte2 = (CommonTableExpr *) lfirst(rest);
@@ -153,7 +146,8 @@ transformWithClause(ParseState *pstate,
 			/* must be a data-modifying statement */
 			Assert(IsA(cte->ctequery, InsertStmt) ||
 				   IsA(cte->ctequery, UpdateStmt) ||
-				   IsA(cte->ctequery, DeleteStmt));
+				   IsA(cte->ctequery, DeleteStmt) ||
+				   IsA(cte->ctequery, MergeStmt));
 
 			pstate->p_hasModifyingCTE = true;
 		}
diff --git a/src/backend/parser/parse_expr.c b/src/backend/parser/parse_expr.c
new file mode 100644
index 9300c7b..d44b1f2
--- a/src/backend/parser/parse_expr.c
+++ b/src/backend/parser/parse_expr.c
@@ -54,6 +54,7 @@ static Node *transformAExprDistinct(Pars
 static Node *transformAExprNullIf(ParseState *pstate, A_Expr *a);
 static Node *transformAExprIn(ParseState *pstate, A_Expr *a);
 static Node *transformAExprBetween(ParseState *pstate, A_Expr *a);
+static Node *transformMergeSupportFunc(ParseState *pstate, MergeSupportFunc *f);
 static Node *transformBoolExpr(ParseState *pstate, BoolExpr *a);
 static Node *transformFuncCall(ParseState *pstate, FuncCall *fn);
 static Node *transformMultiAssignRef(ParseState *pstate, MultiAssignRef *maref);
@@ -227,6 +228,11 @@ transformExprRecurse(ParseState *pstate,
 			result = transformGroupingFunc(pstate, (GroupingFunc *) expr);
 			break;
 
+		case T_MergeSupportFunc:
+			result = transformMergeSupportFunc(pstate,
+											   (MergeSupportFunc *) expr);
+			break;
+
 		case T_NamedArgExpr:
 			{
 				NamedArgExpr *na = (NamedArgExpr *) expr;
@@ -541,6 +547,7 @@ transformColumnRef(ParseState *pstate, C
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CHECK_CONSTRAINT:
@@ -1354,6 +1361,31 @@ transformAExprBetween(ParseState *pstate
 }
 
 static Node *
+transformMergeSupportFunc(ParseState *pstate, MergeSupportFunc *f)
+{
+	/*
+	 * All we need to do is check that we're in the RETURNING list of a MERGE
+	 * command.  If so, we just return the node as-is.
+	 */
+	if (pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+	{
+		ParseState *parent_pstate = pstate->parentParseState;
+
+		while (parent_pstate &&
+			   parent_pstate->p_expr_kind != EXPR_KIND_MERGE_RETURNING)
+			parent_pstate = parent_pstate->parentParseState;
+
+		if (!parent_pstate)
+			ereport(ERROR,
+					errcode(ERRCODE_SYNTAX_ERROR),
+					errmsg("MERGE_ACTION() can only be used in the RETURNING list of a MERGE command"),
+					parser_errposition(pstate, f->location));
+	}
+
+	return (Node *) f;
+}
+
+static Node *
 transformBoolExpr(ParseState *pstate, BoolExpr *a)
 {
 	List	   *args = NIL;
@@ -1767,6 +1799,7 @@ transformSubLink(ParseState *pstate, Sub
 		case EXPR_KIND_LIMIT:
 		case EXPR_KIND_OFFSET:
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
 		case EXPR_KIND_CYCLE_MARK:
@@ -3115,6 +3148,7 @@ ParseExprKindName(ParseExprKind exprKind
 		case EXPR_KIND_OFFSET:
 			return "OFFSET";
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			return "RETURNING";
 		case EXPR_KIND_VALUES:
 		case EXPR_KIND_VALUES_SINGLE:
diff --git a/src/backend/parser/parse_func.c b/src/backend/parser/parse_func.c
new file mode 100644
index fdb3e6d..0cbc950
--- a/src/backend/parser/parse_func.c
+++ b/src/backend/parser/parse_func.c
@@ -2599,6 +2599,7 @@ check_srf_call_placement(ParseState *pst
 			errkind = true;
 			break;
 		case EXPR_KIND_RETURNING:
+		case EXPR_KIND_MERGE_RETURNING:
 			errkind = true;
 			break;
 		case EXPR_KIND_VALUES:
diff --git a/src/backend/parser/parse_merge.c b/src/backend/parser/parse_merge.c
new file mode 100644
index a7d8ba7..04ed5e6
--- a/src/backend/parser/parse_merge.c
+++ b/src/backend/parser/parse_merge.c
@@ -234,6 +234,10 @@ transformMergeStmt(ParseState *pstate, M
 	 */
 	qry->jointree = makeFromExpr(pstate->p_joinlist, joinExpr);
 
+	/* Transform the RETURNING list, if any */
+	qry->returningList = transformReturningList(pstate, stmt->returningList,
+												EXPR_KIND_MERGE_RETURNING);
+
 	/*
 	 * We now have a good query shape, so now look at the WHEN conditions and
 	 * action targetlists.
@@ -391,9 +395,6 @@ transformMergeStmt(ParseState *pstate, M
 
 	qry->mergeActionList = mergeActionList;
 
-	/* RETURNING could potentially be added in the future, but not in SQL std */
-	qry->returningList = NULL;
-
 	qry->hasTargetSRFs = false;
 	qry->hasSubLinks = pstate->p_hasSubLinks;
 
diff --git a/src/backend/parser/parse_relation.c b/src/backend/parser/parse_relation.c
new file mode 100644
index 6f5d9e2..427b732
--- a/src/backend/parser/parse_relation.c
+++ b/src/backend/parser/parse_relation.c
@@ -2341,9 +2341,10 @@ addRangeTableEntryForCTE(ParseState *pst
 		cte->cterefcount++;
 
 	/*
-	 * We throw error if the CTE is INSERT/UPDATE/DELETE without RETURNING.
-	 * This won't get checked in case of a self-reference, but that's OK
-	 * because data-modifying CTEs aren't allowed to be recursive anyhow.
+	 * We throw error if the CTE is INSERT/UPDATE/DELETE/MERGE without
+	 * RETURNING.  This won't get checked in case of a self-reference, but
+	 * that's OK because data-modifying CTEs aren't allowed to be recursive
+	 * anyhow.
 	 */
 	if (IsA(cte->ctequery, Query))
 	{
diff --git a/src/backend/parser/parse_target.c b/src/backend/parser/parse_target.c
new file mode 100644
index 5b92502..ea522b9
--- a/src/backend/parser/parse_target.c
+++ b/src/backend/parser/parse_target.c
@@ -1820,6 +1820,10 @@ FigureColnameInternal(Node *node, char *
 			/* make GROUPING() act like a regular function */
 			*name = "grouping";
 			return 2;
+		case T_MergeSupportFunc:
+			/* make MERGE_ACTION() act like a regular function */
+			*name = "merge_action";
+			return 2;
 		case T_SubLink:
 			switch (((SubLink *) node)->subLinkType)
 			{
diff --git a/src/backend/rewrite/rewriteHandler.c b/src/backend/rewrite/rewriteHandler.c
new file mode 100644
index 7a46e8b..9fd05b1
--- a/src/backend/rewrite/rewriteHandler.c
+++ b/src/backend/rewrite/rewriteHandler.c
@@ -3833,9 +3833,9 @@ RewriteQuery(Query *parsetree, List *rew
 	ListCell   *lc1;
 
 	/*
-	 * First, recursively process any insert/update/delete statements in WITH
-	 * clauses.  (We have to do this first because the WITH clauses may get
-	 * copied into rule actions below.)
+	 * First, recursively process any insert/update/delete/merge statements in
+	 * WITH clauses.  (We have to do this first because the WITH clauses may
+	 * get copied into rule actions below.)
 	 */
 	foreach(lc1, parsetree->cteList)
 	{
@@ -3860,7 +3860,8 @@ RewriteQuery(Query *parsetree, List *rew
 			if (!(ctequery->commandType == CMD_SELECT ||
 				  ctequery->commandType == CMD_UPDATE ||
 				  ctequery->commandType == CMD_INSERT ||
-				  ctequery->commandType == CMD_DELETE))
+				  ctequery->commandType == CMD_DELETE ||
+				  ctequery->commandType == CMD_MERGE))
 			{
 				/*
 				 * Currently it could only be NOTIFY; this error message will
diff --git a/src/backend/rewrite/rowsecurity.c b/src/backend/rewrite/rowsecurity.c
new file mode 100644
index 450af27..59fd305
--- a/src/backend/rewrite/rowsecurity.c
+++ b/src/backend/rewrite/rowsecurity.c
@@ -384,10 +384,10 @@ get_row_security_policies(Query *root, R
 	 * on the final action we take.
 	 *
 	 * We already fetched the SELECT policies above, to check existing rows,
-	 * but we must also check that new rows created by UPDATE actions are
-	 * visible, if SELECT rights are required for this relation. We don't do
-	 * this for INSERT actions, since an INSERT command would only do this
-	 * check if it had a RETURNING list, and MERGE does not support RETURNING.
+	 * but we must also check that new rows created by INSERT/UPDATE actions
+	 * are visible, if SELECT rights are required. For INSERT actions, we only
+	 * do this if RETURNING is specified, to be consistent with a plain INSERT
+	 * command, which can only require SELECT rights when RETURNING is used.
 	 *
 	 * We don't push the UPDATE/DELETE USING quals to the RTE because we don't
 	 * really want to apply them while scanning the relation since we don't
@@ -409,6 +409,8 @@ get_row_security_policies(Query *root, R
 		List	   *merge_delete_restrictive_policies;
 		List	   *merge_insert_permissive_policies;
 		List	   *merge_insert_restrictive_policies;
+		List	   *merge_select_permissive_policies = NIL;
+		List	   *merge_select_restrictive_policies = NIL;
 
 		/*
 		 * Fetch the UPDATE policies and set them up to execute on the
@@ -446,9 +448,6 @@ get_row_security_policies(Query *root, R
 		 */
 		if (perminfo->requiredPerms & ACL_SELECT)
 		{
-			List	   *merge_select_permissive_policies;
-			List	   *merge_select_restrictive_policies;
-
 			get_policies_for_relation(rel, CMD_SELECT, user_id,
 									  &merge_select_permissive_policies,
 									  &merge_select_restrictive_policies);
@@ -497,6 +496,21 @@ get_row_security_policies(Query *root, R
 							   withCheckOptions,
 							   hasSubLinks,
 							   false);
+
+		/*
+		 * Add ALL/SELECT policies as WCO_RLS_INSERT_CHECK WCOs, to ensure
+		 * that the inserted row is visible when executing an INSERT action,
+		 * if RETURNING is specified and SELECT rights are required for this
+		 * relation.
+		 */
+		if (perminfo->requiredPerms & ACL_SELECT && root->returningList)
+			add_with_check_options(rel, rt_index,
+								   WCO_RLS_INSERT_CHECK,
+								   merge_select_permissive_policies,
+								   merge_select_restrictive_policies,
+								   withCheckOptions,
+								   hasSubLinks,
+								   true);
 	}
 
 	table_close(rel, NoLock);
diff --git a/src/backend/tcop/utility.c b/src/backend/tcop/utility.c
new file mode 100644
index 83f86a4..fa66b80
--- a/src/backend/tcop/utility.c
+++ b/src/backend/tcop/utility.c
@@ -2138,11 +2138,10 @@ QueryReturnsTuples(Query *parsetree)
 		case CMD_SELECT:
 			/* returns tuples */
 			return true;
-		case CMD_MERGE:
-			return false;
 		case CMD_INSERT:
 		case CMD_UPDATE:
 		case CMD_DELETE:
+		case CMD_MERGE:
 			/* the forms with RETURNING return tuples */
 			if (parsetree->returningList)
 				return true;
diff --git a/src/backend/utils/adt/ruleutils.c b/src/backend/utils/adt/ruleutils.c
new file mode 100644
index 2a1ee69..e2b46b6
--- a/src/backend/utils/adt/ruleutils.c
+++ b/src/backend/utils/adt/ruleutils.c
@@ -7196,8 +7196,13 @@ get_merge_query_def(Query *query, depars
 			appendStringInfoString(buf, "DO NOTHING");
 	}
 
-	/* No RETURNING support in MERGE yet */
-	Assert(query->returningList == NIL);
+	/* Add RETURNING if present */
+	if (query->returningList)
+	{
+		appendContextKeyword(context, " RETURNING",
+							 -PRETTYINDENT_STD, PRETTYINDENT_STD, 1);
+		get_target_list(query->returningList, context, NULL, colNamesVisible);
+	}
 }
 
 
@@ -8297,6 +8302,7 @@ isSimpleNode(Node *node, Node *parentNod
 		case T_Aggref:
 		case T_GroupingFunc:
 		case T_WindowFunc:
+		case T_MergeSupportFunc:
 		case T_FuncExpr:
 		case T_JsonConstructorExpr:
 			/* function-like: name(..) or name[..] */
@@ -8651,6 +8657,10 @@ get_rule_expr(Node *node, deparse_contex
 			get_windowfunc_expr((WindowFunc *) node, context);
 			break;
 
+		case T_MergeSupportFunc:
+			appendStringInfoString(buf, "MERGE_ACTION()");
+			break;
+
 		case T_SubscriptingRef:
 			{
 				SubscriptingRef *sbsref = (SubscriptingRef *) node;
diff --git a/src/bin/psql/common.c b/src/bin/psql/common.c
new file mode 100644
index 76e01b0..2830bde
--- a/src/bin/psql/common.c
+++ b/src/bin/psql/common.c
@@ -983,13 +983,17 @@ PrintQueryResult(PGresult *result, bool
 			else
 				success = true;
 
-			/* if it's INSERT/UPDATE/DELETE RETURNING, also print status */
+			/*
+			 * If it's INSERT/UPDATE/DELETE/MERGE RETURNING, also print
+			 * status.
+			 */
 			if (last || pset.show_all_results)
 			{
 				cmdstatus = PQcmdStatus(result);
 				if (strncmp(cmdstatus, "INSERT", 6) == 0 ||
 					strncmp(cmdstatus, "UPDATE", 6) == 0 ||
-					strncmp(cmdstatus, "DELETE", 6) == 0)
+					strncmp(cmdstatus, "DELETE", 6) == 0 ||
+					strncmp(cmdstatus, "MERGE", 5) == 0)
 					PrintQueryStatus(result, printStatusFout);
 			}
 
diff --git a/src/include/executor/execExpr.h b/src/include/executor/execExpr.h
new file mode 100644
index a28ddcd..8953d76
--- a/src/include/executor/execExpr.h
+++ b/src/include/executor/execExpr.h
@@ -243,6 +243,7 @@ typedef enum ExprEvalOp
 	EEOP_AGGREF,
 	EEOP_GROUPING_FUNC,
 	EEOP_WINDOW_FUNC,
+	EEOP_MERGE_SUPPORT_FUNC,
 	EEOP_SUBPLAN,
 
 	/* aggregation related nodes */
@@ -810,6 +811,8 @@ extern void ExecEvalJsonConstructor(Expr
 									ExprContext *econtext);
 extern void ExecEvalJsonIsPredicate(ExprState *state, ExprEvalStep *op);
 extern void ExecEvalGroupingFunc(ExprState *state, ExprEvalStep *op);
+extern void ExecEvalMergeSupportFunc(ExprState *state, ExprEvalStep *op,
+									 ExprContext *econtext);
 extern void ExecEvalSubPlan(ExprState *state, ExprEvalStep *op,
 							ExprContext *econtext);
 extern void ExecEvalWholeRowVar(ExprState *state, ExprEvalStep *op,
diff --git a/src/include/executor/spi.h b/src/include/executor/spi.h
new file mode 100644
index c13e1f6..48b8773
--- a/src/include/executor/spi.h
+++ b/src/include/executor/spi.h
@@ -97,6 +97,7 @@ typedef struct _SPI_plan *SPIPlanPtr;
 #define SPI_OK_REL_UNREGISTER	16
 #define SPI_OK_TD_REGISTER		17
 #define SPI_OK_MERGE			18
+#define SPI_OK_MERGE_RETURNING	19
 
 #define SPI_OPT_NONATOMIC		(1 << 0)
 
diff --git a/src/include/nodes/execnodes.h b/src/include/nodes/execnodes.h
new file mode 100644
index 27614ab..9259352
--- a/src/include/nodes/execnodes.h
+++ b/src/include/nodes/execnodes.h
@@ -1325,6 +1325,9 @@ typedef struct ModifyTableState
 	/* Flags showing which subcommands are present INS/UPD/DEL/DO NOTHING */
 	int			mt_merge_subcommands;
 
+	/* For MERGE, the action currently being executed */
+	MergeActionState *mt_merge_action;
+
 	/* tuple counters for MERGE */
 	double		mt_merge_inserted;
 	double		mt_merge_updated;
diff --git a/src/include/nodes/parsenodes.h b/src/include/nodes/parsenodes.h
new file mode 100644
index aadaf67..2c648a6
--- a/src/include/nodes/parsenodes.h
+++ b/src/include/nodes/parsenodes.h
@@ -1939,6 +1939,7 @@ typedef struct MergeStmt
 	Node	   *sourceRelation; /* source relation */
 	Node	   *joinCondition;	/* join condition between source and target */
 	List	   *mergeWhenClauses;	/* list of MergeWhenClause(es) */
+	List	   *returningList;	/* list of expressions to return */
 	WithClause *withClause;		/* WITH clause */
 } MergeStmt;
 
diff --git a/src/include/nodes/primnodes.h b/src/include/nodes/primnodes.h
new file mode 100644
index 4a15460..8df8884
--- a/src/include/nodes/primnodes.h
+++ b/src/include/nodes/primnodes.h
@@ -572,6 +572,27 @@ typedef struct WindowFunc
 } WindowFunc;
 
 /*
+ * MergeSupportFunc
+ *
+ * A MergeSupportFunc is a merge support function expression that can only
+ * appear in the RETURNING list of a MERGE command.  It returns information
+ * about the currently executing merge action.
+ *
+ * Currently, the only supported function is MERGE_ACTION(), which returns the
+ * command executed ("INSERT", "UPDATE", or "DELETE").
+ */
+typedef struct MergeSupportFunc
+{
+	Expr		xpr;
+	/* type Oid of result */
+	Oid			msftype;
+	/* OID of collation, or InvalidOid if none */
+	Oid			msfcollid;
+	/* token location, or -1 if unknown */
+	int			location;
+} MergeSupportFunc;
+
+/*
  * SubscriptingRef: describes a subscripting operation over a container
  * (array, etc).
  *
diff --git a/src/include/optimizer/paramassign.h b/src/include/optimizer/paramassign.h
new file mode 100644
index 501f728..4026b74
--- a/src/include/optimizer/paramassign.h
+++ b/src/include/optimizer/paramassign.h
@@ -20,6 +20,8 @@ extern Param *replace_outer_placeholderv
 										   PlaceHolderVar *phv);
 extern Param *replace_outer_agg(PlannerInfo *root, Aggref *agg);
 extern Param *replace_outer_grouping(PlannerInfo *root, GroupingFunc *grp);
+extern Param *replace_outer_merge_support(PlannerInfo *root,
+										  MergeSupportFunc *msf);
 extern Param *replace_nestloop_param_var(PlannerInfo *root, Var *var);
 extern Param *replace_nestloop_param_placeholdervar(PlannerInfo *root,
 													PlaceHolderVar *phv);
diff --git a/src/include/parser/analyze.h b/src/include/parser/analyze.h
new file mode 100644
index 5ecdd10..28b66fc
--- a/src/include/parser/analyze.h
+++ b/src/include/parser/analyze.h
@@ -44,6 +44,8 @@ extern List *transformInsertRow(ParseSta
 								bool strip_indirection);
 extern List *transformUpdateTargetList(ParseState *pstate,
 									   List *origTlist);
+extern List *transformReturningList(ParseState *pstate, List *returningList,
+									ParseExprKind exprKind);
 extern Query *transformTopLevelStmt(ParseState *pstate, RawStmt *parseTree);
 extern Query *transformStmt(ParseState *pstate, Node *parseTree);
 
diff --git a/src/include/parser/kwlist.h b/src/include/parser/kwlist.h
new file mode 100644
index 2331aca..0993534
--- a/src/include/parser/kwlist.h
+++ b/src/include/parser/kwlist.h
@@ -266,6 +266,7 @@ PG_KEYWORD("matched", MATCHED, UNRESERVE
 PG_KEYWORD("materialized", MATERIALIZED, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("maxvalue", MAXVALUE, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("merge", MERGE, UNRESERVED_KEYWORD, BARE_LABEL)
+PG_KEYWORD("merge_action", MERGE_ACTION, COL_NAME_KEYWORD, BARE_LABEL)
 PG_KEYWORD("method", METHOD, UNRESERVED_KEYWORD, BARE_LABEL)
 PG_KEYWORD("minute", MINUTE_P, UNRESERVED_KEYWORD, AS_LABEL)
 PG_KEYWORD("minvalue", MINVALUE, UNRESERVED_KEYWORD, BARE_LABEL)
diff --git a/src/include/parser/parse_node.h b/src/include/parser/parse_node.h
new file mode 100644
index 99d6515..5b781d8
--- a/src/include/parser/parse_node.h
+++ b/src/include/parser/parse_node.h
@@ -61,7 +61,8 @@ typedef enum ParseExprKind
 	EXPR_KIND_DISTINCT_ON,		/* DISTINCT ON */
 	EXPR_KIND_LIMIT,			/* LIMIT */
 	EXPR_KIND_OFFSET,			/* OFFSET */
-	EXPR_KIND_RETURNING,		/* RETURNING */
+	EXPR_KIND_RETURNING,		/* RETURNING in INSERT/UPDATE/DELETE */
+	EXPR_KIND_MERGE_RETURNING,	/* RETURNING in MERGE */
 	EXPR_KIND_VALUES,			/* VALUES */
 	EXPR_KIND_VALUES_SINGLE,	/* single-row VALUES (in INSERT only) */
 	EXPR_KIND_CHECK_CONSTRAINT, /* CHECK constraint for a table */
diff --git a/src/pl/plpgsql/src/pl_exec.c b/src/pl/plpgsql/src/pl_exec.c
new file mode 100644
index ed51694..6947575
--- a/src/pl/plpgsql/src/pl_exec.c
+++ b/src/pl/plpgsql/src/pl_exec.c
@@ -4267,9 +4267,9 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 	/*
 	 * If we have INTO, then we only need one row back ... but if we have INTO
 	 * STRICT or extra check too_many_rows, ask for two rows, so that we can
-	 * verify the statement returns only one.  INSERT/UPDATE/DELETE are always
-	 * treated strictly. Without INTO, just run the statement to completion
-	 * (tcount = 0).
+	 * verify the statement returns only one.  INSERT/UPDATE/DELETE/MERGE are
+	 * always treated strictly. Without INTO, just run the statement to
+	 * completion (tcount = 0).
 	 *
 	 * We could just ask for two rows always when using INTO, but there are
 	 * some cases where demanding the extra row costs significant time, eg by
@@ -4307,10 +4307,11 @@ exec_stmt_execsql(PLpgSQL_execstate *est
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 			Assert(stmt->mod_stmt);
 			exec_set_found(estate, (SPI_processed != 0));
 			break;
@@ -4489,10 +4490,11 @@ exec_stmt_dynexecute(PLpgSQL_execstate *
 		case SPI_OK_INSERT:
 		case SPI_OK_UPDATE:
 		case SPI_OK_DELETE:
+		case SPI_OK_MERGE:
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
-		case SPI_OK_MERGE:
+		case SPI_OK_MERGE_RETURNING:
 		case SPI_OK_UTILITY:
 		case SPI_OK_REWRITTEN:
 			break;
diff --git a/src/pl/tcl/pltcl.c b/src/pl/tcl/pltcl.c
new file mode 100644
index 6187e15..18d14a2
--- a/src/pl/tcl/pltcl.c
+++ b/src/pl/tcl/pltcl.c
@@ -2456,6 +2456,7 @@ pltcl_process_SPI_result(Tcl_Interp *int
 		case SPI_OK_INSERT_RETURNING:
 		case SPI_OK_DELETE_RETURNING:
 		case SPI_OK_UPDATE_RETURNING:
+		case SPI_OK_MERGE_RETURNING:
 
 			/*
 			 * Process the tuples we got
diff --git a/src/test/regress/expected/merge.out b/src/test/regress/expected/merge.out
new file mode 100644
index e3ebf46..f9f6c2a
--- a/src/test/regress/expected/merge.out
+++ b/src/test/regress/expected/merge.out
@@ -124,20 +124,20 @@ ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
 ERROR:  name "target" specified more than once
 DETAIL:  The name is used both as MERGE target table and data source.
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
-ERROR:  MERGE not supported in WITH query
-LINE 1: WITH foo AS (
-             ^
--- used in COPY
+ERROR:  WITH query "foo" does not have a RETURNING clause
+LINE 4: ) SELECT * FROM foo;
+                        ^
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) TO stdout;
-ERROR:  MERGE not supported in COPY
+ERROR:  COPY query must have a RETURNING clause
 -- unsupported relation types
 -- materialized view
 CREATE MATERIALIZED VIEW mv AS SELECT * FROM target;
@@ -1255,7 +1255,7 @@ BEGIN;
 MERGE INTO sq_target
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1270,7 +1270,7 @@ INSERT INTO sq_source (sid, balance, del
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1294,7 +1294,7 @@ WITH targq AS (
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -1302,21 +1302,207 @@ WHEN MATCHED AND tid < 2 THEN
 	DELETE;
 ROLLBACK;
 -- RETURNING
+SELECT * FROM sq_source ORDER BY sid;
+ delta | sid | balance 
+-------+-----+---------
+    10 |   1 |       0
+    20 |   2 |       0
+    40 |   4 |       0
+(3 rows)
+
+SELECT * FROM sq_target ORDER BY tid;
+ tid | balance 
+-----+---------
+   1 |     100
+   2 |     200
+   3 |     300
+(3 rows)
+
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
+CREATE TABLE merge_actions(action text, abbrev text);
+INSERT INTO merge_actions VALUES ('INSERT', 'ins'), ('UPDATE', 'upd'), ('DELETE', 'del');
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
-ERROR:  syntax error at or near "RETURNING"
-LINE 10: RETURNING *;
-         ^
+    DELETE
+RETURNING (SELECT abbrev FROM merge_actions
+            WHERE action = merge_action()) AS action,
+          t.*,
+          CASE merge_action()
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ action | tid | balance |     description     
+--------+-----+---------+---------------------
+ del    |   1 |     100 | Removed (1,100)
+ upd    |   2 |     220 | Added 20 to balance
+ ins    |   4 |      40 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- error when using merge_action() outside MERGE
+SELECT merge_action() FROM sq_target;
+ERROR:  MERGE_ACTION() can only be used in the RETURNING list of a MERGE command
+LINE 1: SELECT merge_action() FROM sq_target;
+               ^
+UPDATE sq_target SET balance = balance + 1 RETURNING merge_action();
+ERROR:  MERGE_ACTION() can only be used in the RETURNING list of a MERGE command
+LINE 1: ...ATE sq_target SET balance = balance + 1 RETURNING merge_acti...
+                                                             ^
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING merge_action() AS action, t.*,
+              CASE merge_action()
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING action, merge_action() AS log_action, l.*
+)
+SELECT * FROM m2;
+ action | log_action | tid |     last_change     
+--------+------------+-----+---------------------
+ DELETE | UPDATE     |   1 | Removed (1,100)
+ UPDATE | INSERT     |   2 | Added 20 to balance
+ INSERT | INSERT     |   4 | Inserted (4,40)
+(3 rows)
+
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ tid |     last_change     
+-----+---------------------
+   1 | Removed (1,100)
+   2 | Added 20 to balance
+   4 | Inserted (4,40)
+(3 rows)
+
+ROLLBACK;
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING merge_action(), t.*
+) TO stdout;
+DELETE	1	100
+UPDATE	2	220
+INSERT	4	40
+ROLLBACK;
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING merge_action(), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ action | tid | new_balance 
+--------+-----+-------------
+ DELETE |   1 |         100
+ UPDATE |   3 |         320
+ INSERT |   4 |         110
+(3 rows)
+
+ROLLBACK;
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING merge_action(), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ action | tid | balance 
+--------+-----+---------
+ DELETE |   1 |     100
+ UPDATE |   2 |     220
+ INSERT |   4 |      40
+(3 rows)
+
+ROLLBACK;
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING merge_action(), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ r_action | r_tid | r_balance 
+----------+-------+-----------
+ DELETE   |     1 |       100
+ UPDATE   |     3 |       320
+ INSERT   |     4 |       110
+(3 rows)
+
 ROLLBACK;
 -- EXPLAIN
 CREATE TABLE ex_mtarget (a int, b int)
@@ -1513,7 +1699,7 @@ SELECT * FROM sq_target WHERE tid = 1;
 (1 row)
 
 ROLLBACK;
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 NOTICE:  drop cascades to view v
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1639,6 +1825,32 @@ SELECT * FROM pa_target ORDER BY tid;
 (14 rows)
 
 ROLLBACK;
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING merge_action(), t.*;
+ merge_action | tid | balance |           val            
+--------------+-----+---------+--------------------------
+ UPDATE       |   2 |     110 | initial updated by merge
+(1 row)
+
+SELECT * FROM pa_target ORDER BY tid;
+ tid | balance |           val            
+-----+---------+--------------------------
+   2 |     110 | initial updated by merge
+   3 |     300 | initial
+   5 |     500 | initial
+   7 |     700 | initial
+   9 |     900 | initial
+  11 |    1100 | initial
+  13 |    1300 | initial
+(7 rows)
+
+ROLLBACK;
 DROP TABLE pa_target CASCADE;
 -- The target table is partitioned in the same way, but this time by attaching
 -- partitions which have columns in different order, dropped columns etc.
@@ -1888,7 +2100,21 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING merge_action(), t.*;
+ merge_action |          logts           | tid | balance |           val            
+--------------+--------------------------+-----+---------+--------------------------
+ UPDATE       | Tue Jan 31 00:00:00 2017 |   1 |     110 | initial updated by merge
+ UPDATE       | Tue Feb 28 00:00:00 2017 |   2 |     220 | initial updated by merge
+ INSERT       | Sun Jan 15 00:00:00 2017 |   3 |      30 | inserted by merge
+ UPDATE       | Tue Jan 31 00:00:00 2017 |   4 |     440 | initial updated by merge
+ UPDATE       | Tue Feb 28 00:00:00 2017 |   5 |     550 | initial updated by merge
+ INSERT       | Sun Jan 15 00:00:00 2017 |   6 |      60 | inserted by merge
+ UPDATE       | Tue Jan 31 00:00:00 2017 |   7 |     770 | initial updated by merge
+ UPDATE       | Tue Feb 28 00:00:00 2017 |   8 |     880 | initial updated by merge
+ INSERT       | Sun Jan 15 00:00:00 2017 |   9 |      90 | inserted by merge
+(9 rows)
+
 SELECT * FROM pa_target ORDER BY tid;
           logts           | tid | balance |           val            
 --------------------------+-----+---------+--------------------------
diff --git a/src/test/regress/expected/rowsecurity.out b/src/test/regress/expected/rowsecurity.out
new file mode 100644
index 4538f0c..d507a2c
--- a/src/test/regress/expected/rowsecurity.out
+++ b/src/test/regress/expected/rowsecurity.out
@@ -2302,6 +2302,35 @@ WHEN MATCHED THEN
 	UPDATE SET dnotes = dnotes || ' notes added by merge8 '
 WHEN NOT MATCHED THEN
 	INSERT VALUES (13, 44, 1, 'regress_rls_bob', 'new manga');
+SELECT * FROM document WHERE did = 13;
+ did | cid | dlevel | dauthor | dtitle | dnotes 
+-----+-----+--------+---------+--------+--------
+(0 rows)
+
+-- but not OK if RETURNING is used
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge9 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 44, 1, 'regress_rls_bob', 'new manga')
+RETURNING *;
+ERROR:  new row violates row-level security policy for table "document"
+-- but OK if new row is visible
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge10 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 11, 1, 'regress_rls_bob', 'new novel')
+RETURNING *;
+ sdid | did | cid | dlevel |     dauthor     |  dtitle   | dnotes 
+------+-----+-----+--------+-----------------+-----------+--------
+   14 |  14 |  11 |      1 | regress_rls_bob | new novel | 
+(1 row)
+
 RESET SESSION AUTHORIZATION;
 -- drop the restrictive SELECT policy so that we can look at the
 -- final state of the table
@@ -2325,7 +2354,8 @@ SELECT * FROM document;
   12 |  11 |      1 | regress_rls_bob   | another novel                    | 
    1 |  11 |      1 | regress_rls_bob   | my first novel                   |  notes added by merge2  notes added by merge3  notes added by merge4  notes added by merge7 
   13 |  44 |      1 | regress_rls_bob   | new manga                        | 
-(15 rows)
+  14 |  11 |      1 | regress_rls_bob   | new novel                        | 
+(16 rows)
 
 --
 -- ROLE/GROUP
diff --git a/src/test/regress/expected/rules.out b/src/test/regress/expected/rules.out
new file mode 100644
index 0cd2c64..84e359f
--- a/src/test/regress/expected/rules.out
+++ b/src/test/regress/expected/rules.out
@@ -3638,7 +3638,7 @@ MERGE INTO rule_merge1 t USING (SELECT 1
 -- test deparsing
 CREATE TABLE sf_target(id int, data text, filling int[]);
 CREATE FUNCTION merge_sf_test()
- RETURNS void
+ RETURNS TABLE(action text, a int, b text, id int, data text, filling int[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -3675,11 +3675,13 @@ WHEN NOT MATCHED
    VALUES (s.a, s.b, DEFAULT)
 WHEN NOT MATCHED
    THEN INSERT (filling[1], id)
-   VALUES (s.a, s.a);
+   VALUES (s.a, s.a)
+RETURNING
+   merge_action() AS action, *;
 END;
 \sf merge_sf_test
 CREATE OR REPLACE FUNCTION public.merge_sf_test()
- RETURNS void
+ RETURNS TABLE(action text, a integer, b text, id integer, data text, filling integer[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -3716,7 +3718,13 @@ BEGIN ATOMIC
       VALUES (s.a, s.b, DEFAULT)
     WHEN NOT MATCHED
      THEN INSERT (filling[1], id)
-      VALUES (s.a, s.a);
+      VALUES (s.a, s.a)
+   RETURNING MERGE_ACTION() AS action,
+     s.a,
+     s.b,
+     t.id,
+     t.data,
+     t.filling;
 END
 DROP FUNCTION merge_sf_test;
 DROP TABLE sf_target;
diff --git a/src/test/regress/expected/updatable_views.out b/src/test/regress/expected/updatable_views.out
new file mode 100644
index 794cf9c..1062c34
--- a/src/test/regress/expected/updatable_views.out
+++ b/src/test/regress/expected/updatable_views.out
@@ -479,7 +479,15 @@ MERGE INTO rw_view1 t
                 (2, 'ROW 2'), (3, 'ROW 3')) AS v(a,b) ON t.a = v.a
   WHEN MATCHED AND t.a <= 1 THEN UPDATE SET b = v.b
   WHEN MATCHED THEN DELETE
-  WHEN NOT MATCHED AND a > 0 THEN INSERT (a) VALUES (v.a);
+  WHEN NOT MATCHED AND a > 0 THEN INSERT (a) VALUES (v.a)
+  RETURNING merge_action(), v.*, t.*;
+ merge_action | a |   b   | a |      b      
+--------------+---+-------+---+-------------
+ UPDATE       | 1 | ROW 1 | 1 | ROW 1
+ DELETE       | 3 | ROW 3 | 3 | Row 3
+ INSERT       | 2 | ROW 2 | 2 | Unspecified
+(3 rows)
+
 SELECT * FROM base_tbl ORDER BY a;
  a  |      b      
 ----+-------------
@@ -631,7 +639,15 @@ MERGE INTO rw_view2 t
   USING (VALUES (3, 'R3'), (4, 'R4'), (5, 'R5')) AS v(a,b) ON aaa = v.a
   WHEN MATCHED AND aaa = 3 THEN DELETE
   WHEN MATCHED THEN UPDATE SET bbb = v.b
-  WHEN NOT MATCHED THEN INSERT (aaa) VALUES (v.a);
+  WHEN NOT MATCHED THEN INSERT (aaa) VALUES (v.a)
+  RETURNING merge_action(), v.*, t.*;
+ merge_action | a | b  | aaa |     bbb     
+--------------+---+----+-----+-------------
+ DELETE       | 3 | R3 |   3 | Row 3
+ UPDATE       | 4 | R4 |   4 | R4
+ INSERT       | 5 | R5 |   5 | Unspecified
+(3 rows)
+
 SELECT * FROM rw_view2 ORDER BY aaa;
  aaa |     bbb     
 -----+-------------
@@ -1071,7 +1087,15 @@ MERGE INTO rw_view2 t
   USING (SELECT x, 'R'||x FROM generate_series(0,3) x) AS s(a,b) ON t.a = s.a
   WHEN MATCHED AND t.a <= 1 THEN DELETE
   WHEN MATCHED THEN UPDATE SET b = s.b
-  WHEN NOT MATCHED AND s.a > 0 THEN INSERT VALUES (s.a, s.b);
+  WHEN NOT MATCHED AND s.a > 0 THEN INSERT VALUES (s.a, s.b)
+  RETURNING merge_action(), s.*, t.*;
+ merge_action | a | b  | a |   b   
+--------------+---+----+---+-------
+ DELETE       | 1 | R1 | 1 | Row 1
+ UPDATE       | 2 | R2 | 2 | R2
+ INSERT       | 3 | R3 | 3 | R3
+(3 rows)
+
 SELECT * FROM base_tbl ORDER BY a;
  a  |   b    
 ----+--------
diff --git a/src/test/regress/expected/with.out b/src/test/regress/expected/with.out
new file mode 100644
index 7d796ea..6de347b
--- a/src/test/regress/expected/with.out
+++ b/src/test/regress/expected/with.out
@@ -3493,6 +3493,16 @@ SELECT * FROM t;
 ERROR:  WITH query "t" does not have a RETURNING clause
 LINE 4: SELECT * FROM t;
                       ^
+-- RETURNING tries to return its own output
+WITH RECURSIVE t(action, a) AS (
+	MERGE INTO y USING (VALUES (11)) v(a) ON y.a = v.a
+		WHEN NOT MATCHED THEN INSERT VALUES (v.a)
+		RETURNING merge_action(), (SELECT a FROM t)
+)
+SELECT * FROM t;
+ERROR:  recursive query "t" must not contain data-modifying statements
+LINE 1: WITH RECURSIVE t(action, a) AS (
+                       ^
 -- data-modifying WITH allowed only at the top level
 SELECT * FROM (
 	WITH t AS (UPDATE y SET a=a+1 RETURNING *)
diff --git a/src/test/regress/sql/merge.sql b/src/test/regress/sql/merge.sql
new file mode 100644
index 79a1868..dfcc243
--- a/src/test/regress/sql/merge.sql
+++ b/src/test/regress/sql/merge.sql
@@ -88,12 +88,12 @@ MERGE INTO target
 USING target
 ON tid = tid
 WHEN MATCHED THEN DO NOTHING;
--- used in a CTE
+-- used in a CTE without RETURNING
 WITH foo AS (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
 ) SELECT * FROM foo;
--- used in COPY
+-- used in COPY without RETURNING
 COPY (
   MERGE INTO target USING source ON (true)
   WHEN MATCHED THEN DELETE
@@ -817,7 +817,7 @@ BEGIN;
 MERGE INTO sq_target
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -830,7 +830,7 @@ INSERT INTO sq_source (sid, balance, del
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -848,7 +848,7 @@ WITH targq AS (
 MERGE INTO sq_target t
 USING v
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
 	INSERT (balance, tid) VALUES (balance + delta, sid)
@@ -857,18 +857,149 @@ WHEN MATCHED AND tid < 2 THEN
 ROLLBACK;
 
 -- RETURNING
+SELECT * FROM sq_source ORDER BY sid;
+SELECT * FROM sq_target ORDER BY tid;
+
 BEGIN;
-INSERT INTO sq_source (sid, balance, delta) VALUES (-1, -1, -10);
+CREATE TABLE merge_actions(action text, abbrev text);
+INSERT INTO merge_actions VALUES ('INSERT', 'ins'), ('UPDATE', 'upd'), ('DELETE', 'del');
 MERGE INTO sq_target t
-USING v
+USING sq_source s
 ON tid = sid
-WHEN MATCHED AND tid > 2 THEN
+WHEN MATCHED AND tid >= 2 THEN
     UPDATE SET balance = t.balance + delta
 WHEN NOT MATCHED THEN
-	INSERT (balance, tid) VALUES (balance + delta, sid)
+    INSERT (balance, tid) VALUES (balance + delta, sid)
 WHEN MATCHED AND tid < 2 THEN
-	DELETE
-RETURNING *;
+    DELETE
+RETURNING (SELECT abbrev FROM merge_actions
+            WHERE action = merge_action()) AS action,
+          t.*,
+          CASE merge_action()
+              WHEN 'INSERT' THEN 'Inserted '||t
+              WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+              WHEN 'DELETE' THEN 'Removed '||t
+          END AS description;
+ROLLBACK;
+
+-- error when using merge_action() outside MERGE
+SELECT merge_action() FROM sq_target;
+UPDATE sq_target SET balance = balance + 1 RETURNING merge_action();
+
+-- RETURNING in CTEs
+CREATE TABLE sq_target_merge_log (tid integer NOT NULL, last_change text);
+INSERT INTO sq_target_merge_log VALUES (1, 'Original value');
+BEGIN;
+WITH m AS (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING merge_action() AS action, t.*,
+              CASE merge_action()
+                  WHEN 'INSERT' THEN 'Inserted '||t
+                  WHEN 'UPDATE' THEN 'Added '||delta||' to balance'
+                  WHEN 'DELETE' THEN 'Removed '||t
+              END AS description
+), m2 AS (
+    MERGE INTO sq_target_merge_log l
+    USING m
+    ON l.tid = m.tid
+    WHEN MATCHED THEN
+        UPDATE SET last_change = description
+    WHEN NOT MATCHED THEN
+        INSERT VALUES (m.tid, description)
+    RETURNING action, merge_action() AS log_action, l.*
+)
+SELECT * FROM m2;
+SELECT * FROM sq_target_merge_log ORDER BY tid;
+ROLLBACK;
+
+-- COPY (MERGE ... RETURNING) TO ...
+BEGIN;
+COPY (
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING merge_action(), t.*
+) TO stdout;
+ROLLBACK;
+
+-- SQL function with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT action text, OUT tid int, OUT new_balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING merge_action(), t.*;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT action, tid, new_balance FROM merge_into_sq_target(sid, balance, delta)) m;
+ROLLBACK;
+
+-- SQL SRF with MERGE ... RETURNING
+BEGIN;
+CREATE FUNCTION merge_sq_source_into_sq_target()
+RETURNS TABLE (action text, tid int, balance int)
+LANGUAGE sql AS
+$$
+    MERGE INTO sq_target t
+    USING sq_source s
+    ON tid = sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (balance + delta, sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING merge_action(), t.*;
+$$;
+SELECT * FROM merge_sq_source_into_sq_target();
+ROLLBACK;
+
+-- PL/pgSQL function with MERGE ... RETURNING ... INTO
+BEGIN;
+CREATE FUNCTION merge_into_sq_target(sid int, balance int, delta int,
+                                     OUT r_action text, OUT r_tid int, OUT r_balance int)
+LANGUAGE plpgsql AS
+$$
+BEGIN
+    MERGE INTO sq_target t
+    USING (VALUES ($1, $2, $3)) AS v(sid, balance, delta)
+    ON tid = v.sid
+    WHEN MATCHED AND tid >= 2 THEN
+        UPDATE SET balance = t.balance + v.delta
+    WHEN NOT MATCHED THEN
+        INSERT (balance, tid) VALUES (v.balance + v.delta, v.sid)
+    WHEN MATCHED AND tid < 2 THEN
+        DELETE
+    RETURNING merge_action(), t.* INTO r_action, r_tid, r_balance;
+END;
+$$;
+SELECT m.*
+FROM (VALUES (1, 0, 0), (3, 0, 20), (4, 100, 10)) AS v(sid, balance, delta),
+LATERAL (SELECT r_action, r_tid, r_balance FROM merge_into_sq_target(sid, balance, delta)) m;
 ROLLBACK;
 
 -- EXPLAIN
@@ -967,7 +1098,7 @@ WHEN MATCHED THEN
 SELECT * FROM sq_target WHERE tid = 1;
 ROLLBACK;
 
-DROP TABLE sq_target, sq_source CASCADE;
+DROP TABLE sq_target, sq_target_merge_log, sq_source CASCADE;
 
 CREATE TABLE pa_target (tid integer, balance float, val text)
 	PARTITION BY LIST (tid);
@@ -1034,6 +1165,17 @@ SELECT merge_func();
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
+-- update partition key to partition not initially scanned
+BEGIN;
+MERGE INTO pa_target t
+  USING pa_source s
+  ON t.tid = s.sid AND t.tid = 1
+  WHEN MATCHED THEN
+    UPDATE SET tid = tid + 1, balance = balance + delta, val = val || ' updated by merge'
+  RETURNING merge_action(), t.*;
+SELECT * FROM pa_target ORDER BY tid;
+ROLLBACK;
+
 DROP TABLE pa_target CASCADE;
 
 -- The target table is partitioned in the same way, but this time by attaching
@@ -1210,7 +1352,8 @@ MERGE INTO pa_target t
   WHEN MATCHED THEN
     UPDATE SET balance = balance + delta, val = val || ' updated by merge'
   WHEN NOT MATCHED THEN
-    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge');
+    INSERT VALUES (slogts::timestamp, sid, delta, 'inserted by merge')
+  RETURNING merge_action(), t.*;
 SELECT * FROM pa_target ORDER BY tid;
 ROLLBACK;
 
diff --git a/src/test/regress/sql/rowsecurity.sql b/src/test/regress/sql/rowsecurity.sql
new file mode 100644
index dec7340..1d5ed0a
--- a/src/test/regress/sql/rowsecurity.sql
+++ b/src/test/regress/sql/rowsecurity.sql
@@ -992,6 +992,27 @@ WHEN MATCHED THEN
 	UPDATE SET dnotes = dnotes || ' notes added by merge8 '
 WHEN NOT MATCHED THEN
 	INSERT VALUES (13, 44, 1, 'regress_rls_bob', 'new manga');
+SELECT * FROM document WHERE did = 13;
+
+-- but not OK if RETURNING is used
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge9 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 44, 1, 'regress_rls_bob', 'new manga')
+RETURNING *;
+
+-- but OK if new row is visible
+MERGE INTO document d
+USING (SELECT 14 as sdid) s
+ON did = s.sdid
+WHEN MATCHED THEN
+	UPDATE SET dnotes = dnotes || ' notes added by merge10 '
+WHEN NOT MATCHED THEN
+	INSERT VALUES (14, 11, 1, 'regress_rls_bob', 'new novel')
+RETURNING *;
 
 RESET SESSION AUTHORIZATION;
 -- drop the restrictive SELECT policy so that we can look at the
diff --git a/src/test/regress/sql/rules.sql b/src/test/regress/sql/rules.sql
new file mode 100644
index 6924012..27340ba
--- a/src/test/regress/sql/rules.sql
+++ b/src/test/regress/sql/rules.sql
@@ -1294,7 +1294,7 @@ MERGE INTO rule_merge1 t USING (SELECT 1
 CREATE TABLE sf_target(id int, data text, filling int[]);
 
 CREATE FUNCTION merge_sf_test()
- RETURNS void
+ RETURNS TABLE(action text, a int, b text, id int, data text, filling int[])
  LANGUAGE sql
 BEGIN ATOMIC
  MERGE INTO sf_target t
@@ -1331,7 +1331,9 @@ WHEN NOT MATCHED
    VALUES (s.a, s.b, DEFAULT)
 WHEN NOT MATCHED
    THEN INSERT (filling[1], id)
-   VALUES (s.a, s.a);
+   VALUES (s.a, s.a)
+RETURNING
+   merge_action() AS action, *;
 END;
 
 \sf merge_sf_test
diff --git a/src/test/regress/sql/updatable_views.sql b/src/test/regress/sql/updatable_views.sql
new file mode 100644
index ae11e46..afdf331
--- a/src/test/regress/sql/updatable_views.sql
+++ b/src/test/regress/sql/updatable_views.sql
@@ -175,7 +175,8 @@ MERGE INTO rw_view1 t
                 (2, 'ROW 2'), (3, 'ROW 3')) AS v(a,b) ON t.a = v.a
   WHEN MATCHED AND t.a <= 1 THEN UPDATE SET b = v.b
   WHEN MATCHED THEN DELETE
-  WHEN NOT MATCHED AND a > 0 THEN INSERT (a) VALUES (v.a);
+  WHEN NOT MATCHED AND a > 0 THEN INSERT (a) VALUES (v.a)
+  RETURNING merge_action(), v.*, t.*;
 SELECT * FROM base_tbl ORDER BY a;
 
 EXPLAIN (costs off) UPDATE rw_view1 SET a=6 WHERE a=5;
@@ -246,7 +247,8 @@ MERGE INTO rw_view2 t
   USING (VALUES (3, 'R3'), (4, 'R4'), (5, 'R5')) AS v(a,b) ON aaa = v.a
   WHEN MATCHED AND aaa = 3 THEN DELETE
   WHEN MATCHED THEN UPDATE SET bbb = v.b
-  WHEN NOT MATCHED THEN INSERT (aaa) VALUES (v.a);
+  WHEN NOT MATCHED THEN INSERT (aaa) VALUES (v.a)
+  RETURNING merge_action(), v.*, t.*;
 SELECT * FROM rw_view2 ORDER BY aaa;
 
 EXPLAIN (costs off) UPDATE rw_view2 SET aaa=5 WHERE aaa=4;
@@ -458,7 +460,8 @@ MERGE INTO rw_view2 t
   USING (SELECT x, 'R'||x FROM generate_series(0,3) x) AS s(a,b) ON t.a = s.a
   WHEN MATCHED AND t.a <= 1 THEN DELETE
   WHEN MATCHED THEN UPDATE SET b = s.b
-  WHEN NOT MATCHED AND s.a > 0 THEN INSERT VALUES (s.a, s.b);
+  WHEN NOT MATCHED AND s.a > 0 THEN INSERT VALUES (s.a, s.b)
+  RETURNING merge_action(), s.*, t.*;
 SELECT * FROM base_tbl ORDER BY a;
 
 EXPLAIN (costs off) UPDATE rw_view2 SET a=3 WHERE a=2;
diff --git a/src/test/regress/sql/with.sql b/src/test/regress/sql/with.sql
new file mode 100644
index f8a213e..037bc0a
--- a/src/test/regress/sql/with.sql
+++ b/src/test/regress/sql/with.sql
@@ -1591,6 +1591,14 @@ WITH t AS (
 )
 SELECT * FROM t;
 
+-- RETURNING tries to return its own output
+WITH RECURSIVE t(action, a) AS (
+	MERGE INTO y USING (VALUES (11)) v(a) ON y.a = v.a
+		WHEN NOT MATCHED THEN INSERT VALUES (v.a)
+		RETURNING merge_action(), (SELECT a FROM t)
+)
+SELECT * FROM t;
+
 -- data-modifying WITH allowed only at the top level
 SELECT * FROM (
 	WITH t AS (UPDATE y SET a=a+1 RETURNING *)
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
new file mode 100644
index aa7a25b..7a102e8
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -1577,6 +1577,7 @@ MergeJoinState
 MergePath
 MergeScanSelCache
 MergeStmt
+MergeSupportFunc
 MergeWhenClause
 MetaCommand
 MinMaxAggInfo