Fix up hash functions for datetime datatypes so that they don't take

unwarranted liberties with int8 vs float8 values for these types.
Specifically, be sure to apply either hashint8 or hashfloat8 depending
on HAVE_INT64_TIMESTAMP.  Per my gripe of even date.

9 files changed, 88 insertions, 34 deletions

diff --git a/src/backend/utils/adt/date.c b/src/backend/utils/adt/date.c
index 19b7884e03..4257ede403 100644
--- a/src/backend/utils/adt/date.c
+++ b/src/backend/utils/adt/date.c
@@ -1198,6 +1198,17 @@ time_cmp(PG_FUNCTION_ARGS)
 }
 
 Datum
+time_hash(PG_FUNCTION_ARGS)
+{
+	/* We can use either hashint8 or hashfloat8 directly */
+#ifdef HAVE_INT64_TIMESTAMP
+	return hashint8(fcinfo);
+#else
+	return hashfloat8(fcinfo);
+#endif
+}
+
+Datum
 time_larger(PG_FUNCTION_ARGS)
 {
 	TimeADT		time1 = PG_GETARG_TIMEADT(0);
@@ -1960,20 +1971,27 @@ timetz_cmp(PG_FUNCTION_ARGS)
 	PG_RETURN_INT32(timetz_cmp_internal(time1, time2));
 }
 
-/*
- * timetz, being an unusual size, needs a specialized hash function.
- */
 Datum
 timetz_hash(PG_FUNCTION_ARGS)
 {
 	TimeTzADT  *key = PG_GETARG_TIMETZADT_P(0);
+	uint32		thash;
 
 	/*
-	 * Specify hash length as sizeof(double) + sizeof(int4), not as
-	 * sizeof(TimeTzADT), so that any garbage pad bytes in the structure won't
-	 * be included in the hash!
+	 * To avoid any problems with padding bytes in the struct,
+	 * we figure the field hashes separately and XOR them.  This also
+	 * provides a convenient framework for dealing with the fact that
+	 * the time field might be either double or int64.
 	 */
-	return hash_any((unsigned char *) key, sizeof(key->time) + sizeof(key->zone));
+#ifdef HAVE_INT64_TIMESTAMP
+	thash = DatumGetUInt32(DirectFunctionCall1(hashint8,
+											   Int64GetDatumFast(key->time)));
+#else
+	thash = DatumGetUInt32(DirectFunctionCall1(hashfloat8,
+											   Float8GetDatumFast(key->time)));
+#endif
+	thash ^= DatumGetUInt32(hash_uint32(key->zone));
+	PG_RETURN_UINT32(thash);
 }
 
 Datum
diff --git a/src/backend/utils/adt/timestamp.c b/src/backend/utils/adt/timestamp.c
index 974e4a9d10..8cc69ef625 100644
--- a/src/backend/utils/adt/timestamp.c
+++ b/src/backend/utils/adt/timestamp.c
@@ -1839,6 +1839,17 @@ timestamp_cmp(PG_FUNCTION_ARGS)
 	PG_RETURN_INT32(timestamp_cmp_internal(dt1, dt2));
 }
 
+Datum
+timestamp_hash(PG_FUNCTION_ARGS)
+{
+	/* We can use either hashint8 or hashfloat8 directly */
+#ifdef HAVE_INT64_TIMESTAMP
+	return hashint8(fcinfo);
+#else
+	return hashfloat8(fcinfo);
+#endif
+}
+
 
 /*
  * Crosstype comparison functions for timestamp vs timestamptz
@@ -2110,21 +2121,32 @@ interval_cmp(PG_FUNCTION_ARGS)
 	PG_RETURN_INT32(interval_cmp_internal(interval1, interval2));
 }
 
-/*
- * interval, being an unusual size, needs a specialized hash function.
- */
 Datum
 interval_hash(PG_FUNCTION_ARGS)
 {
 	Interval   *key = PG_GETARG_INTERVAL_P(0);
+	uint32		thash;
+	uint32		mhash;
 
 	/*
-	 * Specify hash length as sizeof(double) + sizeof(int4), not as
-	 * sizeof(Interval), so that any garbage pad bytes in the structure won't
-	 * be included in the hash!
+	 * To avoid any problems with padding bytes in the struct,
+	 * we figure the field hashes separately and XOR them.  This also
+	 * provides a convenient framework for dealing with the fact that
+	 * the time field might be either double or int64.
 	 */
-	return hash_any((unsigned char *) key,
-				  sizeof(key->time) + sizeof(key->day) + sizeof(key->month));
+#ifdef HAVE_INT64_TIMESTAMP
+	thash = DatumGetUInt32(DirectFunctionCall1(hashint8,
+											   Int64GetDatumFast(key->time)));
+#else
+	thash = DatumGetUInt32(DirectFunctionCall1(hashfloat8,
+											   Float8GetDatumFast(key->time)));
+#endif
+	thash ^= DatumGetUInt32(hash_uint32(key->day));
+	/* Shift so "k days" and "k months" don't hash to the same thing */
+	mhash = DatumGetUInt32(hash_uint32(key->month));
+	thash ^= mhash << 24;
+	thash ^= mhash >> 8;
+	PG_RETURN_UINT32(thash);
 }
 
 /* overlaps_timestamp() --- implements the SQL92 OVERLAPS operator.
diff --git a/src/include/catalog/catversion.h b/src/include/catalog/catversion.h
index 60986c57fa..e32c35c5a8 100644
--- a/src/include/catalog/catversion.h
+++ b/src/include/catalog/catversion.h
@@ -53,6 +53,6 @@
  */
 
 /*							yyyymmddN */
-#define CATALOG_VERSION_NO	200706271
+#define CATALOG_VERSION_NO	200707051
 
 #endif
diff --git a/src/include/catalog/pg_amproc.h b/src/include/catalog/pg_amproc.h
index 2320d80d7d..6eb5c664f1 100644
--- a/src/include/catalog/pg_amproc.h
+++ b/src/include/catalog/pg_amproc.h
@@ -147,11 +147,11 @@ DATA(insert (	1987   19 19 1 455 ));
 DATA(insert (	1990   26 26 1 453 ));
 DATA(insert (	1992   30 30 1 457 ));
 DATA(insert (	1995   25 25 1 400 ));
-DATA(insert (	1997   1083 1083 1 452 ));
+DATA(insert (	1997   1083 1083 1 1688 ));
 DATA(insert (	1998   1700 1700 1 432 ));
-DATA(insert (	1999   1184 1184 1 452 ));
+DATA(insert (	1999   1184 1184 1 2039 ));
 DATA(insert (	2001   1266 1266 1 1696 ));
-DATA(insert (	2040   1114 1114 1 452 ));
+DATA(insert (	2040   1114 1114 1 2039 ));
 DATA(insert (	2222   16 16 1 454 ));
 DATA(insert (	2223   17 17 1 456 ));
 DATA(insert (	2224   22 22 1 398 ));
diff --git a/src/include/catalog/pg_proc.h b/src/include/catalog/pg_proc.h
index b911a1723f..3ed0b26205 100644
--- a/src/include/catalog/pg_proc.h
+++ b/src/include/catalog/pg_proc.h
@@ -2474,6 +2474,8 @@ DESCR("greater-than-or-equal");
 DATA(insert OID = 1693 (  btboolcmp			PGNSP PGUID 12 1 0 f f t f i 2 23 "16 16" _null_ _null_ _null_	btboolcmp - _null_ ));
 DESCR("btree less-equal-greater");
 
+DATA(insert OID = 1688 (  time_hash			PGNSP PGUID 12 1 0 f f t f i 1 23 "1083" _null_ _null_ _null_ time_hash - _null_ ));
+DESCR("hash");
 DATA(insert OID = 1696 (  timetz_hash		PGNSP PGUID 12 1 0 f f t f i 1 23 "1266" _null_ _null_ _null_ timetz_hash - _null_ ));
 DESCR("hash");
 DATA(insert OID = 1697 (  interval_hash		PGNSP PGUID 12 1 0 f f t f i 1 23 "1186" _null_ _null_ _null_ interval_hash - _null_ ));
@@ -3043,6 +3045,8 @@ DATA(insert OID = 2037 (  timezone			PGNSP PGUID 12 1 0 f f t f v 2 1266 "25 126
 DESCR("adjust time with time zone to new zone");
 DATA(insert OID = 2038 (  timezone			PGNSP PGUID 12 1 0 f f t f i 2 1266 "1186 1266" _null_ _null_ _null_	timetz_izone - _null_ ));
 DESCR("adjust time with time zone to new zone");
+DATA(insert OID = 2039 (  timestamp_hash	PGNSP PGUID 12 1 0 f f t f i 1	23 "1114" _null_ _null_ _null_ timestamp_hash - _null_ ));
+DESCR("hash");
 DATA(insert OID = 2041 ( overlaps			PGNSP PGUID 12 1 0 f f f f i 4 16 "1114 1114 1114 1114" _null_ _null_ _null_	overlaps_timestamp - _null_ ));
 DESCR("SQL92 interval comparison");
 DATA(insert OID = 2042 ( overlaps			PGNSP PGUID 14 1 0 f f f f i 4 16 "1114 1186 1114 1186" _null_ _null_ _null_	"select ($1, ($1 + $2)) overlaps ($3, ($3 + $4))" - _null_ ));
diff --git a/src/include/utils/date.h b/src/include/utils/date.h
index 508f490baa..ec108fdfbb 100644
--- a/src/include/utils/date.h
+++ b/src/include/utils/date.h
@@ -148,6 +148,7 @@ extern Datum time_le(PG_FUNCTION_ARGS);
 extern Datum time_gt(PG_FUNCTION_ARGS);
 extern Datum time_ge(PG_FUNCTION_ARGS);
 extern Datum time_cmp(PG_FUNCTION_ARGS);
+extern Datum time_hash(PG_FUNCTION_ARGS);
 extern Datum overlaps_time(PG_FUNCTION_ARGS);
 extern Datum time_larger(PG_FUNCTION_ARGS);
 extern Datum time_smaller(PG_FUNCTION_ARGS);
diff --git a/src/include/utils/timestamp.h b/src/include/utils/timestamp.h
index 99f6a3aac1..ab37a58692 100644
--- a/src/include/utils/timestamp.h
+++ b/src/include/utils/timestamp.h
@@ -211,6 +211,7 @@ extern Datum timestamp_ge(PG_FUNCTION_ARGS);
 extern Datum timestamp_gt(PG_FUNCTION_ARGS);
 extern Datum timestamp_finite(PG_FUNCTION_ARGS);
 extern Datum timestamp_cmp(PG_FUNCTION_ARGS);
+extern Datum timestamp_hash(PG_FUNCTION_ARGS);
 extern Datum timestamp_smaller(PG_FUNCTION_ARGS);
 extern Datum timestamp_larger(PG_FUNCTION_ARGS);
 
diff --git a/src/test/regress/expected/opr_sanity.out b/src/test/regress/expected/opr_sanity.out
index 8b5f64473d..247c8c9583 100644
--- a/src/test/regress/expected/opr_sanity.out
+++ b/src/test/regress/expected/opr_sanity.out
@@ -980,10 +980,9 @@ WHERE p3.opfmethod = (SELECT oid FROM pg_am WHERE amname = 'btree')
 -- For hash we can also do a little better: the support routines must be
 -- of the form hash(lefttype) returns int4.  There are several cases where
 -- we cheat and use a hash function that is physically compatible with the
--- datatype even though there's no cast, so for now we can't check that.
-SELECT p1.amprocfamily, p1.amprocnum,
-	p2.oid, p2.proname,
-	p3.opfname
+-- datatype even though there's no cast, so this check does find a small
+-- number of entries.
+SELECT p1.amprocfamily, p1.amprocnum, p2.proname, p3.opfname
 FROM pg_amproc AS p1, pg_proc AS p2, pg_opfamily AS p3
 WHERE p3.opfmethod = (SELECT oid FROM pg_am WHERE amname = 'hash')
     AND p1.amprocfamily = p3.oid AND p1.amproc = p2.oid AND
@@ -991,11 +990,20 @@ WHERE p3.opfmethod = (SELECT oid FROM pg_am WHERE amname = 'hash')
      OR proretset
      OR prorettype != 'int4'::regtype
      OR pronargs != 1
---   OR NOT physically_coercible(amproclefttype, proargtypes[0])
-     OR amproclefttype != amprocrighttype);
- amprocfamily | amprocnum | oid | proname | opfname 
---------------+-----------+-----+---------+---------
-(0 rows)
+     OR NOT physically_coercible(amproclefttype, proargtypes[0])
+     OR amproclefttype != amprocrighttype)
+ORDER BY 1;
+ amprocfamily | amprocnum |    proname     |      opfname       
+--------------+-----------+----------------+--------------------
+          435 |         1 | hashint4       | date_ops
+         1999 |         1 | timestamp_hash | timestamptz_ops
+         2222 |         1 | hashchar       | bool_ops
+         2223 |         1 | hashvarlena    | bytea_ops
+         2225 |         1 | hashint4       | xid_ops
+         2226 |         1 | hashint4       | cid_ops
+         2229 |         1 | hashvarlena    | text_pattern_ops
+         2231 |         1 | hashvarlena    | bpchar_pattern_ops
+(8 rows)
 
 -- Support routines that are primary members of opfamilies must be immutable
 -- (else it suggests that the index ordering isn't fixed).  But cross-type
diff --git a/src/test/regress/sql/opr_sanity.sql b/src/test/regress/sql/opr_sanity.sql
index f93a71ddf9..54c60c1a6d 100644
--- a/src/test/regress/sql/opr_sanity.sql
+++ b/src/test/regress/sql/opr_sanity.sql
@@ -790,11 +790,10 @@ WHERE p3.opfmethod = (SELECT oid FROM pg_am WHERE amname = 'btree')
 -- For hash we can also do a little better: the support routines must be
 -- of the form hash(lefttype) returns int4.  There are several cases where
 -- we cheat and use a hash function that is physically compatible with the
--- datatype even though there's no cast, so for now we can't check that.
+-- datatype even though there's no cast, so this check does find a small
+-- number of entries.
 
-SELECT p1.amprocfamily, p1.amprocnum,
-	p2.oid, p2.proname,
-	p3.opfname
+SELECT p1.amprocfamily, p1.amprocnum, p2.proname, p3.opfname
 FROM pg_amproc AS p1, pg_proc AS p2, pg_opfamily AS p3
 WHERE p3.opfmethod = (SELECT oid FROM pg_am WHERE amname = 'hash')
     AND p1.amprocfamily = p3.oid AND p1.amproc = p2.oid AND
@@ -802,8 +801,9 @@ WHERE p3.opfmethod = (SELECT oid FROM pg_am WHERE amname = 'hash')
      OR proretset
      OR prorettype != 'int4'::regtype
      OR pronargs != 1
---   OR NOT physically_coercible(amproclefttype, proargtypes[0])
-     OR amproclefttype != amprocrighttype);
+     OR NOT physically_coercible(amproclefttype, proargtypes[0])
+     OR amproclefttype != amprocrighttype)
+ORDER BY 1;
 
 -- Support routines that are primary members of opfamilies must be immutable
 -- (else it suggests that the index ordering isn't fixed).  But cross-type