May 2025

1 commit
Revert "Don't lock partitions pruned by initial pruning"
22 May 2025Amit Langotemaster18beta2REL_19_STABLE18beta2REL_18_STABLE18beta2revert
LLM analysis, 99% suremodified before commitlight changesfrom patchset #2180 of 1 reviewer saw the final version

The attached revert in #218 supplies the reversal; the commit additionally removes release-note text and relocates a plancache header include. #220 confirms the revert landed while the pruning prerequisites remained.

  • Tom Lanesaw an earlier versionLLM analysis, 91% sureHe demonstrated statement-list invalidation failures and demanded reversal in #214 and #217, but did not review the subsequently attached revert #218.

Jan - Feb 2025

7 commits
Fix bug in cbc127917 to handle nested Append correctly
25 Feb 2025Amit Langotemaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% suremodified before commitlight changesfrom patchset #2111 of 1 reviewer saw the final version

The backend changes in src/backend/partitioning/partprune.c and src/backend/executor/execPartition.c match #211. The commit revises comments and expands the regression case to exercise nested MergeAppend pruning.

  • Tender Wangsaw the final versiontrailerHe diagnosed the non-leaf partition classification error in #210 and judged the attached fix good in #212, before comment and test refinements.
Remove unstable test suite added by 525392d57
22 Feb 2025Amit Langotemaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% surewritten from the discussion, no matching patch0 of 1 reviewer saw the final version

No attachment proposes this deletion: Tom Lane reports CLOBBER_CACHE_ALWAYS failures in #203, recommends removing the fragile cached-plan-inval suite in #206, and Amit accepts in #207.

  • Tom Lanesaw an earlier versionLLM analysis, 87% sureHe diagnosed the unstable test and proposed its removal in #203 and #206; this discussion-only change predates the later redesigns and final thread patchset.
Don't lock partitions pruned by initial pruning
20 Feb 2025Amit Langotemaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% suremodified before commitmoderate changesfrom patchset #1970 of 5 reviewers saw the final version

Matches the deferred-locking patch in #197, including ExecutorStartCachedPlan(), UpdateCachedPlan(), and delay_execution tests. Besides commentary edits, the commit moves is_reused assignment from CheckCachedPlan() into GetCachedPlan().

  • Robert Haassaw an earlier versiontrailerHe reviewed successive designs and retry semantics in #7, #142, and #154, but explicitly had not seriously reviewed the final locking/invalidation components in #175.
  • David Rowleysaw an earlier versiontrailerHe proposed compact pruning metadata, challenged mutable CachedPlan state, and supplied fixups in #37, #52, and #54, before substantial redesigns.
  • Tom Lanesaw an earlier versiontrailerHe reviewed earlier locking and executor designs in #98 and #114, and explicitly denied reviewing the landed implementation before its commit in #217.
  • Tomas Vondrasaw an earlier versiontrailerHe benchmarked and reviewed the series in #180 and discussed hook safety afterward, before the boolean startup API and later invalidation changes.
  • Junwang Zhaosaw an earlier versiontrailerHe caught the custom-versus-generic BuildCachedPlan flag error in #168, well before the subsequent pruning and retry redesigns.
Fix an oversight in cbc127917 to handle MERGE correctly
17 Feb 2025Amit Langotemaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% suremodified before commitmoderate changesfrom patchset #1991 of 1 reviewer saw the final version

Extends #199 in src/backend/executor/nodeModifyTable.c with the reviewed pruned-mergeActionLists guard and more consistent resultRelations use. The commit also adds all-pruned UPDATE and MERGE regression cases.

  • Junwang Zhaosaw the final versiontrailerHe tested the patch and supplied the missing pruned-list guard in #200, which Amit accepted in #201; remaining refinements were equivalent cleanup and tests.
Track unpruned relids to avoid processing pruned relations
7 Feb 2025Amit Langotemaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 99% surepart of a posted patchsetlight changesfrom patchset #1960 of 2 reviewers saw the final version

Takes the unpruned-relation tracking component of #196, leaving deferred locking for #197. Executable changes match that component; the commit expands comments explaining leafpart_rti_map and pruning metadata.

  • Robert Haassaw an earlier versiontrailerHe reviewed precursor pruning-metadata and executor designs in #65, #142, and #154, before unpruned-relation tracking was revised and split into its own component.
  • Tomas Vondrasaw an earlier versiontrailerHe reviewed pruning and ModifyTable handling in #180, before later changes to unpruned-state initialization and propagation into parallel execution.
Perform runtime initial pruning outside ExecInitNode()
31 Jan 2025Amit Langotemaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% surepart of a posted patchsetmoderate changesfrom patchset #1940 of 2 reviewers saw the final version

Takes the early-pruning component of #194. In execPartition.c, the committed version removes PartitionPruneContext.initialized bookkeeping and guards expression-context reset with !initial_prune instead.

  • Robert Haassaw an earlier versiontrailerHe supported moving pruning before node initialization but questioned context initialization in #175, before lazy initialization and its replacement were reworked.
  • Tomas Vondrasaw an earlier versiontrailerHe reviewed initialization flags, ExprContexts, and early-pruning placement in #180, before the lazy-context approach was abandoned in #194.
Move PartitionPruneInfo out of plan nodes into PlannedStmt
30 Jan 2025Amit Langotemaster18beta1REL_19_STABLE18beta1REL_18_STABLE18beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #1942 of 3 reviewers saw the final version

The diff matches the first component of #194, moving pruning descriptors into PlannedStmt and registering their indexes in setrefs.c. The other two components are absent, making this a series subset.

  • Alvaro Herrerasaw an earlier versiontrailerHe reviewed the original descriptor relocation and requested identity checks in #77 and #79, before its revert and later reimplementation.
  • Robert Haassaw the final versiontrailerHe reviewed the relocation component and requested clearer relids naming in #175, with naming, diagnostic, and rebase adjustments afterward.
  • Tomas Vondrasaw the final versiontrailerHe reviewed the relocation component's diagnostics and comments in #180, which were addressed without changing its architecture.

Sep 2023

1 commit
Remove obsolete executor cleanup code
28 Sep 2023Amit Langotemaster17beta1REL_19_STABLE17beta1REL_18_STABLE17beta1REL_17_STABLE17beta1
LLM analysis, 100% surepart of a posted patchsetlight changesfrom patchset #1391 of 1 reviewer saw the final version

Takes only the first component of #139, removing redundant executor cleanup calls and obsolete end-node routines. Its code matches modulo comment indentation and blank lines.

  • Robert Haassaw the final versiontrailerHe separated cleanup from invalidation scaffolding, caught unsafe tuple-slot cleanup removal in #135, and approved the corrected component in #137.

May 2023

1 commit
Revert "Move PartitionPruneInfo out of plan nodes into PlannedStmt"
4 May 2023Alvaro Herreramaster16beta1REL_19_STABLE16beta1REL_18_STABLE16beta1REL_17_STABLE16beta1REL_16_STABLE16beta1reverts 1 commit hereAlso in: "variable not found in subplan target list" (written from the discussion, no matching patch)

Dec 2022

1 commit
Move PartitioPruneInfo out of plan nodes into PlannedStmt
1 Dec 2022Alvaro Herreramaster16beta1REL_19_STABLE16beta1REL_18_STABLE16beta1REL_17_STABLE16beta1REL_16_STABLE16beta1
LLM analysis, 96% suremodified before commitlight changesfrom patchset #781 of 3 reviewers saw the final version

Combines the descriptor-relocation component of #76 with the latest attached delta #78, explicitly merged in #79. Relative to that combined patch, edits are diagnostic formatting and a catalog-version bump; the source identifies the final delta, not a standalone complete patch.

  • David Rowleysaw an earlier versionLLM analysis, 95% sureHis review in #37 proposed collecting pruning descriptors in PlannedStmt and using compact indexes, before the later split and identity-crosscheck additions.
  • Robert Haassaw an earlier versionLLM analysis, 96% sureHe reviewed the relocation prerequisite's executor fields and list flattening in #65, before the identity checks added by #78.
  • Alvaro Herrerasaw the final versioncommittercommitted the patch

Apr 2022

1 commit
Refactor and cleanup runtime partition prune code a little
5 Apr 2022Alvaro Herreramaster15beta1REL_19_STABLE15beta1REL_18_STABLE15beta1REL_17_STABLE15beta1REL_16_STABLE15beta1REL_15_STABLE15beta1
LLM analysis, 100% suremodified before commitlight changesfrom patchset #471 of 3 reviewers saw the final version

Matches the standalone pruning cleanup in #47, with helper renaming, expanded comments, and equivalent condition reordering in execPartition.c. The committer explains those finishing edits in #48.

  • Amul Sulsaw an earlier versionLLM analysis, 93% sureHe reviewed duplicated pruning setup and requested caller-owned temporary state in #4, before that work was substantially refactored and extracted.
  • Robert Haassaw an earlier versionLLM analysis, 97% sureHe requested extracting the pruning cleanup in #18 and reviewed its explanatory comments in #29, before helper consolidation and final cleanup revisions.
  • Alvaro Herrerasaw the final versioncommittercommitted the patch

Credits across 12 commits

Author
Amit Langote8 LLM10
Reviewer
Robert Haas5
Tomas Vondra4
Alvaro Herreracommitter3
Junwang Zhao2
David Rowley1
Tender Wang1
Tom Lane1
Reported by
Alexander Lakhin2
Tom Lane2
Committer
Amit Langote9
Alvaro Herrera3

N LLM means N of the credits come from LLM analysis, not from commit trailers.