reducing NUMERIC size for 9.1, take two

diff --git a/src/backend/utils/adt/numeric.c b/src/backend/utils/adt/numeric.c
index 51a46fd..2568fdc 100644
--- a/src/backend/utils/adt/numeric.c
+++ b/src/backend/utils/adt/numeric.c
@@ -35,38 +35,6 @@
 #include "utils/int8.h"
 #include "utils/numeric.h"
 
-/*
- * Sign values and macros to deal with packing/unpacking n_sign_dscale
- */
-#define NUMERIC_SIGN_MASK	0xC000
-#define NUMERIC_POS			0x0000
-#define NUMERIC_NEG			0x4000
-#define NUMERIC_NAN			0xC000
-#define NUMERIC_DSCALE_MASK 0x3FFF
-#define NUMERIC_SIGN(n)		((n)->n_sign_dscale & NUMERIC_SIGN_MASK)
-#define NUMERIC_DSCALE(n)	((n)->n_sign_dscale & NUMERIC_DSCALE_MASK)
-#define NUMERIC_IS_NAN(n)	(NUMERIC_SIGN(n) != NUMERIC_POS &&	\
-							 NUMERIC_SIGN(n) != NUMERIC_NEG)
-#define NUMERIC_HDRSZ	(VARHDRSZ + sizeof(uint16) + sizeof(int16))
-
-
-/*
- * The Numeric data type stored in the database
- *
- * NOTE: by convention, values in the packed form have been stripped of
- * all leading and trailing zero digits (where a "digit" is of base NBASE).
- * In particular, if the value is zero, there will be no digits at all!
- * The weight is arbitrary in that case, but we normally set it to zero.
- */
-struct NumericData
-{
-	int32		vl_len_;		/* varlena header (do not touch directly!) */
-	uint16		n_sign_dscale;	/* Sign + display scale */
-	int16		n_weight;		/* Weight of 1st digit	*/
-	char		n_data[1];		/* Digits (really array of NumericDigit) */
-};
-
-
 /* ----------
  * Uncomment the following to enable compilation of dump_numeric()
  * and dump_var() and to get a dump of any result produced by make_result().
@@ -120,6 +88,122 @@ typedef signed char NumericDigit;
 typedef int16 NumericDigit;
 #endif
 
+/*
+ * The Numeric type as stored on disk.
+ *
+ * If the high bits of the first word of a NumericChoice (n_header, or
+ * n_short.n_header, or n_long.n_sign_dscale) are NUMERIC_SHORT, then the
+ * numeric follows the NumericShort format; if they are NUMERIC_POS or
+ * NUMERIC_NEG, it follows the NumericLong format.  If they are NUMERIC_NAN,
+ * it is a NaN.  We currently always store a NaN using just two bytes (i.e.
+ * only n_header), but previous releases used only the NumericLong format,
+ * so we might find 4-byte NaNs on disk if a database has been migrated using
+ * pg_upgrade.  In either case, when the high bits indicate a NaN, the
+ * remaining bits are never examined.  Currently, we always initialize these
+ * to zero, but it might be possible to use them for some other purpose in
+ * the future.
+ * 
+ * In the NumericShort format, the remaining 14 bits of the header word
+ * (n_short.n_header) are allocated as follows: 1 for sign (positive or
+ * negative), 6 for dynamic scale, and 7 for weight.  In practice, most
+ * commonly-encountered values can be represented this way.
+ *
+ * In the NumericLong format, the remaining 14 bits of the header word
+ * (n_long.n_sign_dscale) represent the display scale; and the weight is
+ * stored separately in n_weight.
+ *
+ * NOTE: by convention, values in the packed form have been stripped of
+ * all leading and trailing zero digits (where a "digit" is of base NBASE).
+ * In particular, if the value is zero, there will be no digits at all!
+ * The weight is arbitrary in that case, but we normally set it to zero.
+ */
+
+struct NumericShort
+{
+	uint16		n_header;		/* Sign + display scale + weight */
+	NumericDigit n_data[1];		/* Digits */
+};
+
+struct NumericLong
+{
+	uint16		n_sign_dscale;	/* Sign + display scale */
+	int16		n_weight;		/* Weight of 1st digit	*/
+	NumericDigit n_data[1];		/* Digits */
+};
+
+union NumericChoice
+{
+	uint16		n_header;		/* Header word */
+	struct NumericLong	n_long;	/* Long form (4-byte header) */
+	struct NumericShort	n_short;	/* Short form (2-byte header) */
+};
+
+struct NumericData
+{
+	int32		vl_len_;		/* varlena header (do not touch directly!) */
+	union NumericChoice	choice;	/* choice of format */
+};
+
+
+/*
+ * Interpretation of high bits.
+ */
+
+#define NUMERIC_SIGN_MASK	0xC000
+#define NUMERIC_POS			0x0000
+#define NUMERIC_NEG			0x4000
+#define NUMERIC_SHORT		0x8000
+#define NUMERIC_NAN			0xC000
+
+#define NUMERIC_FLAGBITS(n) ((n)->choice.n_header & NUMERIC_SIGN_MASK)
+#define NUMERIC_IS_NAN(n)		(NUMERIC_FLAGBITS(n) == NUMERIC_NAN)
+#define NUMERIC_IS_SHORT(n)		(NUMERIC_FLAGBITS(n) == NUMERIC_SHORT)
+
+#define NUMERIC_HDRSZ	(VARHDRSZ + sizeof(uint16) + sizeof(int16))
+#define NUMERIC_HDRSZ_SHORT	(VARHDRSZ + sizeof(uint16))
+
+/*
+ * If the flag bits are NUMERIC_SHORT or NUMERIC_NAN, we want the short header;
+ * otherwise, we want the long one.  Instead of testing against each value, we
+ * can just look at the high bit, for a slight efficiency gain.
+ */
+#define NUMERIC_HEADER_SIZE(n) \
+	(VARHDRSZ + sizeof(uint16) + \
+		(((NUMERIC_FLAGBITS(n) & 0x8000) == 0) ? sizeof(int16) : 0))
+
+/*
+ * Short format definitions.
+ */
+
+#define NUMERIC_SHORT_SIGN_MASK			0x2000
+#define NUMERIC_SHORT_DSCALE_MASK		0x1F80
+#define NUMERIC_SHORT_DSCALE_SHIFT		7
+#define NUMERIC_SHORT_DSCALE_MAX		\
+	(NUMERIC_SHORT_DSCALE_MASK >> NUMERIC_SHORT_DSCALE_SHIFT)
+#define NUMERIC_SHORT_WEIGHT_SIGN_MASK	0x0040
+#define NUMERIC_SHORT_WEIGHT_MASK		0x003F
+#define NUMERIC_SHORT_WEIGHT_MAX		NUMERIC_SHORT_WEIGHT_MASK
+#define NUMERIC_SHORT_WEIGHT_MIN		(-(NUMERIC_SHORT_WEIGHT_MASK+1))
+
+/*
+ * Extract sign, display scale, weight.
+ */
+
+#define NUMERIC_DSCALE_MASK 		0x3FFF
+
+#define NUMERIC_SIGN(n) \
+	(NUMERIC_IS_SHORT(n) ? \
+		(((n)->choice.n_short.n_header & NUMERIC_SHORT_SIGN_MASK) ? \
+		NUMERIC_NEG : NUMERIC_POS) : NUMERIC_FLAGBITS(n))
+#define NUMERIC_DSCALE(n)	(NUMERIC_IS_SHORT((n)) ? \
+	((n)->choice.n_short.n_header & NUMERIC_SHORT_DSCALE_MASK) \
+		>> NUMERIC_SHORT_DSCALE_SHIFT \
+	: ((n)->choice.n_long.n_sign_dscale & NUMERIC_DSCALE_MASK))
+#define NUMERIC_WEIGHT(n)	(NUMERIC_IS_SHORT((n)) ? \
+	(((n)->choice.n_short.n_header & NUMERIC_SHORT_WEIGHT_SIGN_MASK ? \
+		~NUMERIC_SHORT_WEIGHT_MASK : 0) \
+	 | ((n)->choice.n_short.n_header & NUMERIC_SHORT_WEIGHT_MASK)) \
+	: ((n)->choice.n_long.n_weight))
 
 /* ----------
  * NumericVar is the format we use for arithmetic.	The digit-array part
@@ -266,9 +350,14 @@ static void dump_var(const char *str, NumericVar *var);
 
 #define init_var(v)		MemSetAligned(v, 0, sizeof(NumericVar))
 
-#define NUMERIC_DIGITS(num) ((NumericDigit *)(num)->n_data)
+#define NUMERIC_DIGITS(num) (NUMERIC_IS_SHORT(num) ? \
+	(num)->choice.n_short.n_data : (num)->choice.n_long.n_data)
 #define NUMERIC_NDIGITS(num) \
-	((VARSIZE(num) - NUMERIC_HDRSZ) / sizeof(NumericDigit))
+	((VARSIZE(num) - NUMERIC_HEADER_SIZE(num)) / sizeof(NumericDigit))
+#define NUMERIC_CAN_BE_SHORT(scale,weight) \
+	((scale) <= NUMERIC_SHORT_DSCALE_MAX && \
+	(weight) <= NUMERIC_SHORT_WEIGHT_MAX && \
+	(weight) >= NUMERIC_SHORT_WEIGHT_MIN)
 
 static void alloc_var(NumericVar *var, int ndigits);
 static void free_var(NumericVar *var);
@@ -652,15 +741,23 @@ numeric		(PG_FUNCTION_ARGS)
 	/*
 	 * If the number is certainly in bounds and due to the target scale no
 	 * rounding could be necessary, just make a copy of the input and modify
-	 * its scale fields.  (Note we assume the existing dscale is honest...)
+	 * its scale fields, unless the larger scale forces us to abandon the
+	 * short representation.  (Note we assume the existing dscale is honest...)
 	 */
-	ddigits = (num->n_weight + 1) * DEC_DIGITS;
-	if (ddigits <= maxdigits && scale >= NUMERIC_DSCALE(num))
+	ddigits = (NUMERIC_WEIGHT(num) + 1) * DEC_DIGITS;
+	if (ddigits <= maxdigits && scale >= NUMERIC_DSCALE(num)
+		&& (NUMERIC_CAN_BE_SHORT(scale, NUMERIC_WEIGHT(num))
+		|| !NUMERIC_IS_SHORT(num)))
 	{
 		new = (Numeric) palloc(VARSIZE(num));
 		memcpy(new, num, VARSIZE(num));
-		new->n_sign_dscale = NUMERIC_SIGN(new) |
-			((uint16) scale & NUMERIC_DSCALE_MASK);
+		if (NUMERIC_IS_SHORT(num))
+			new->choice.n_short.n_header =
+				(num->choice.n_short.n_header & ~NUMERIC_SHORT_DSCALE_MASK)
+				| (scale << NUMERIC_SHORT_DSCALE_SHIFT);
+		else
+			new->choice.n_long.n_sign_dscale = NUMERIC_SIGN(new) |
+				((uint16) scale & NUMERIC_DSCALE_MASK);
 		PG_RETURN_NUMERIC(new);
 	}
 
@@ -766,7 +863,11 @@ numeric_abs(PG_FUNCTION_ARGS)
 	res = (Numeric) palloc(VARSIZE(num));
 	memcpy(res, num, VARSIZE(num));
 
-	res->n_sign_dscale = NUMERIC_POS | NUMERIC_DSCALE(num);
+	if (NUMERIC_IS_SHORT(num))
+		res->choice.n_short.n_header =
+			num->choice.n_short.n_header & ~NUMERIC_SHORT_SIGN_MASK;
+	else
+		res->choice.n_long.n_sign_dscale = NUMERIC_POS | NUMERIC_DSCALE(num);
 
 	PG_RETURN_NUMERIC(res);
 }
@@ -795,13 +896,18 @@ numeric_uminus(PG_FUNCTION_ARGS)
 	 * we can identify a ZERO by the fact that there are no digits at all.	Do
 	 * nothing to a zero.
 	 */
-	if (VARSIZE(num) != NUMERIC_HDRSZ)
+	if (NUMERIC_NDIGITS(num) != 0)
 	{
 		/* Else, flip the sign */
-		if (NUMERIC_SIGN(num) == NUMERIC_POS)
-			res->n_sign_dscale = NUMERIC_NEG | NUMERIC_DSCALE(num);
+		if (NUMERIC_IS_SHORT(num))
+			res->choice.n_short.n_header =
+				num->choice.n_short.n_header ^ NUMERIC_SHORT_SIGN_MASK;
+		else if (NUMERIC_SIGN(num) == NUMERIC_POS)
+			res->choice.n_long.n_sign_dscale =
+				NUMERIC_NEG | NUMERIC_DSCALE(num);
 		else
-			res->n_sign_dscale = NUMERIC_POS | NUMERIC_DSCALE(num);
+			res->choice.n_long.n_sign_dscale =
+				NUMERIC_POS | NUMERIC_DSCALE(num);
 	}
 
 	PG_RETURN_NUMERIC(res);
@@ -845,7 +951,7 @@ numeric_sign(PG_FUNCTION_ARGS)
 	 * The packed format is known to be totally zero digit trimmed always. So
 	 * we can identify a ZERO by the fact that there are no digits at all.
 	 */
-	if (VARSIZE(num) == NUMERIC_HDRSZ)
+	if (NUMERIC_NDIGITS(num) == 0)
 		set_var_from_var(&const_zero, &result);
 	else
 	{
@@ -1283,9 +1389,9 @@ cmp_numerics(Numeric num1, Numeric num2)
 	else
 	{
 		result = cmp_var_common(NUMERIC_DIGITS(num1), NUMERIC_NDIGITS(num1),
-								num1->n_weight, NUMERIC_SIGN(num1),
+								NUMERIC_WEIGHT(num1), NUMERIC_SIGN(num1),
 								NUMERIC_DIGITS(num2), NUMERIC_NDIGITS(num2),
-								num2->n_weight, NUMERIC_SIGN(num2));
+								NUMERIC_WEIGHT(num2), NUMERIC_SIGN(num2));
 	}
 
 	return result;
@@ -1302,12 +1408,13 @@ hash_numeric(PG_FUNCTION_ARGS)
 	int			end_offset;
 	int			i;
 	int			hash_len;
+	NumericDigit   *digits;
 
 	/* If it's NaN, don't try to hash the rest of the fields */
 	if (NUMERIC_IS_NAN(key))
 		PG_RETURN_UINT32(0);
 
-	weight = key->n_weight;
+	weight = NUMERIC_WEIGHT(key);
 	start_offset = 0;
 	end_offset = 0;
 
@@ -1317,9 +1424,10 @@ hash_numeric(PG_FUNCTION_ARGS)
 	 * zeros are suppressed, but we're paranoid. Note that we measure the
 	 * starting and ending offsets in units of NumericDigits, not bytes.
 	 */
+	digits = NUMERIC_DIGITS(key);
 	for (i = 0; i < NUMERIC_NDIGITS(key); i++)
 	{
-		if (NUMERIC_DIGITS(key)[i] != (NumericDigit) 0)
+		if (digits[i] != (NumericDigit) 0)
 			break;
 
 		start_offset++;
@@ -1340,7 +1448,7 @@ hash_numeric(PG_FUNCTION_ARGS)
 
 	for (i = NUMERIC_NDIGITS(key) - 1; i >= 0; i--)
 	{
-		if (NUMERIC_DIGITS(key)[i] != (NumericDigit) 0)
+		if (digits[i] != (NumericDigit) 0)
 			break;
 
 		end_offset++;
@@ -2536,7 +2644,7 @@ numeric_avg(PG_FUNCTION_ARGS)
 
 	/* SQL92 defines AVG of no values to be NULL */
 	/* N is zero iff no digits (cf. numeric_uminus) */
-	if (VARSIZE(N) == NUMERIC_HDRSZ)
+	if (NUMERIC_NDIGITS(N) == 0)
 		PG_RETURN_NULL();
 
 	PG_RETURN_DATUM(DirectFunctionCall2(numeric_div,
@@ -2974,7 +3082,8 @@ dump_numeric(const char *str, Numeric num)
 
 	ndigits = NUMERIC_NDIGITS(num);
 
-	printf("%s: NUMERIC w=%d d=%d ", str, num->n_weight, NUMERIC_DSCALE(num));
+	printf("%s: NUMERIC w=%d d=%d ", str,
+		   NUMERIC_WEIGHT(num), NUMERIC_DSCALE(num));
 	switch (NUMERIC_SIGN(num))
 	{
 		case NUMERIC_POS:
@@ -3265,11 +3374,11 @@ set_var_from_num(Numeric num, NumericVar *dest)
 
 	alloc_var(dest, ndigits);
 
-	dest->weight = num->n_weight;
+	dest->weight = NUMERIC_WEIGHT(num);
 	dest->sign = NUMERIC_SIGN(num);
 	dest->dscale = NUMERIC_DSCALE(num);
 
-	memcpy(dest->digits, num->n_data, ndigits * sizeof(NumericDigit));
+	memcpy(dest->digits, NUMERIC_DIGITS(num), ndigits * sizeof(NumericDigit));
 }
 
 
@@ -3561,11 +3670,11 @@ make_result(NumericVar *var)
 
 	if (sign == NUMERIC_NAN)
 	{
-		result = (Numeric) palloc(NUMERIC_HDRSZ);
+		result = (Numeric) palloc(NUMERIC_HDRSZ_SHORT);
 
-		SET_VARSIZE(result, NUMERIC_HDRSZ);
-		result->n_weight = 0;
-		result->n_sign_dscale = NUMERIC_NAN;
+		SET_VARSIZE(result, NUMERIC_HDRSZ_SHORT);
+		result->choice.n_header = NUMERIC_NAN;
+		/* the header word is all we need */
 
 		dump_numeric("make_result()", result);
 		return result;
@@ -3592,16 +3701,33 @@ make_result(NumericVar *var)
 	}
 
 	/* Build the result */
-	len = NUMERIC_HDRSZ + n * sizeof(NumericDigit);
-	result = (Numeric) palloc(len);
-	SET_VARSIZE(result, len);
-	result->n_weight = weight;
-	result->n_sign_dscale = sign | (var->dscale & NUMERIC_DSCALE_MASK);
+	if (NUMERIC_CAN_BE_SHORT(var->dscale, weight))
+	{
+		len = NUMERIC_HDRSZ_SHORT + n * sizeof(NumericDigit);
+		result = (Numeric) palloc(len);
+		SET_VARSIZE(result, len);
+		result->choice.n_short.n_header = 
+			(sign == NUMERIC_NEG ? (NUMERIC_SHORT | NUMERIC_SHORT_SIGN_MASK)
+				: NUMERIC_SHORT)
+			| (var->dscale << NUMERIC_SHORT_DSCALE_SHIFT)
+			| (weight < 0 ? NUMERIC_SHORT_WEIGHT_SIGN_MASK : 0)
+			| (weight & NUMERIC_SHORT_WEIGHT_MASK);
+	}
+	else
+	{
+		len = NUMERIC_HDRSZ + n * sizeof(NumericDigit);
+		result = (Numeric) palloc(len);
+		SET_VARSIZE(result, len);
+		result->choice.n_long.n_sign_dscale =
+			sign | (var->dscale & NUMERIC_DSCALE_MASK);
+		result->choice.n_long.n_weight = weight;
+	}
 
-	memcpy(result->n_data, digits, n * sizeof(NumericDigit));
+	memcpy(NUMERIC_DIGITS(result), digits, n * sizeof(NumericDigit));
+	Assert(NUMERIC_NDIGITS(result) == n);
 
 	/* Check for overflow of int16 fields */
-	if (result->n_weight != weight ||
+	if (NUMERIC_WEIGHT(result) != weight ||
 		NUMERIC_DSCALE(result) != var->dscale)
 		ereport(ERROR,
 				(errcode(ERRCODE_NUMERIC_VALUE_OUT_OF_RANGE),