Rewrite interval_hash() so that the hashcodes are equal for values that

interval_eq() considers equal.  I'm not sure how that fundamental requirement
escaped us through multiple revisions of this hash function, but there it is;
it's been wrong since interval_hash was first written for PG 7.1.
Per bug #4748 from Roman Kononov.

Backpatch to all supported releases.

This patch changes the contents of hash indexes for interval columns.  That's
no particular problem for PG 8.4, since we've broken on-disk compatibility
of hash indexes already; but it will require a migration warning note in
the next minor releases of all existing branches: "if you have any hash
indexes on columns of type interval, REINDEX them after updating".

src/backend/utils/adt/timestamp.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/utils/adt/timestamp.c,v 1.169.2.2 2008/07/07 18:10:03 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/utils/adt/timestamp.c,v 1.169.2.3 2009/04/04 04:53:43 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -1728,31 +1728,35 @@ timestamptz_cmp_timestamp(PG_FUNCTION_ARGS)
  *
  *		collate invalid interval at the end
  */
+#ifdef HAVE_INT64_TIMESTAMP
+typedef int64 TimeOffset;
+#else
+typedef double TimeOffset;
+#endif
+
+static inline TimeOffset
+interval_cmp_value(const Interval *interval)
+{
+	TimeOffset	span;
+
+	span = interval->time;
+
+#ifdef HAVE_INT64_TIMESTAMP
+	span += interval->month * INT64CONST(30) * USECS_PER_DAY;
+	span += interval->day * INT64CONST(24) * USECS_PER_HOUR;
+#else
+	span += interval->month * ((double) DAYS_PER_MONTH * SECS_PER_DAY);
+	span += interval->day * ((double) HOURS_PER_DAY * SECS_PER_HOUR);
+#endif
+
+	return span;
+}
+
 static int
 interval_cmp_internal(Interval *interval1, Interval *interval2)
 {
-#ifdef HAVE_INT64_TIMESTAMP
-	int64		span1,
-				span2;
-#else
-	double		span1,
-				span2;
-#endif
-
-	span1 = interval1->time;
-	span2 = interval2->time;
-
-#ifdef HAVE_INT64_TIMESTAMP
-	span1 += interval1->month * INT64CONST(30) * USECS_PER_DAY;
-	span1 += interval1->day * INT64CONST(24) * USECS_PER_HOUR;
-	span2 += interval2->month * INT64CONST(30) * USECS_PER_DAY;
-	span2 += interval2->day * INT64CONST(24) * USECS_PER_HOUR;
-#else
-	span1 += interval1->month * ((double) DAYS_PER_MONTH * SECS_PER_DAY);
-	span1 += interval1->day * ((double) HOURS_PER_DAY * SECS_PER_HOUR);
-	span2 += interval2->month * ((double) DAYS_PER_MONTH * SECS_PER_DAY);
-	span2 += interval2->day * ((double) HOURS_PER_DAY * SECS_PER_HOUR);
-#endif
+	TimeOffset	span1 = interval_cmp_value(interval1);
+	TimeOffset	span2 = interval_cmp_value(interval2);
 
 	return ((span1 < span2) ? -1 : (span1 > span2) ? 1 : 0);
 }
@@ -1821,20 +1825,28 @@ interval_cmp(PG_FUNCTION_ARGS)
 }
 
 /*
- * interval, being an unusual size, needs a specialized hash function.
+ * Hashing for intervals
+ *
+ * We must produce equal hashvals for values that interval_cmp_internal()
+ * considers equal.  So, compute the net span the same way it does,
+ * and then hash that, using either int64 or float8 hashing.
  */
 Datum
 interval_hash(PG_FUNCTION_ARGS)
 {
-	Interval   *key = PG_GETARG_INTERVAL_P(0);
+	Interval   *interval = PG_GETARG_INTERVAL_P(0);
+	TimeOffset	span = interval_cmp_value(interval);
+	uint32		thash;
 
-	/*
-	 * 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!
-	 */
-	return hash_any((unsigned char *) key,
-				  sizeof(key->time) + sizeof(key->day) + sizeof(key->month));
+#ifdef HAVE_INT64_TIMESTAMP
+	thash = DatumGetUInt32(DirectFunctionCall1(hashint8,
+											   Int64GetDatumFast(span)));
+#else
+	thash = DatumGetUInt32(DirectFunctionCall1(hashfloat8,
+											   Float8GetDatumFast(span)));
+#endif
+
+	PG_RETURN_UINT32(thash);
 }
 
 /* overlaps_timestamp() --- implements the SQL92 OVERLAPS operator.

src/test/regress/expected/interval.out

@@ -349,3 +349,16 @@ SELECT '5.5 seconds 3 milliseconds'::interval;      -- error
 ERROR:  invalid input syntax for type interval: "5.5 seconds 3 milliseconds"
 SELECT '1:20:05 5 microseconds'::interval;          -- error
 ERROR:  invalid input syntax for type interval: "1:20:05 5 microseconds"
+-- check that '30 days' equals '1 month' according to the hash function
+select '30 days'::interval = '1 month'::interval as t;
+ t 
+---
+ t
+(1 row)
+
+select interval_hash('30 days'::interval) = interval_hash('1 month'::interval) as t;
+ t 
+---
+ t
+(1 row)
+

src/test/regress/sql/interval.sql

@@ -127,3 +127,7 @@ SELECT '1 second 2 seconds'::interval;              -- error
 SELECT '10 milliseconds 20 milliseconds'::interval; -- error
 SELECT '5.5 seconds 3 milliseconds'::interval;      -- error
 SELECT '1:20:05 5 microseconds'::interval;          -- error
+
+-- check that '30 days' equals '1 month' according to the hash function
+select '30 days'::interval = '1 month'::interval as t;
+select interval_hash('30 days'::interval) = interval_hash('1 month'::interval) as t;