Improve planner estimates for size of tuple hash tables.

src/backend/optimizer/plan/planner.c

@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/optimizer/plan/planner.c,v 1.199 2006/03/05 15:58:29 momjian Exp $
+ *	  $PostgreSQL: pgsql/src/backend/optimizer/plan/planner.c,v 1.200 2006/06/28 20:04:38 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -1298,7 +1298,7 @@ choose_hashed_grouping(PlannerInfo *root, double tuple_fraction,
 	}
 
 	/* Estimate per-hash-entry space at tuple width... */
-	hashentrysize = cheapest_path_width;
+	hashentrysize = MAXALIGN(cheapest_path_width) + MAXALIGN(sizeof(MinimalTupleData));
 	/* plus space for pass-by-ref transition values... */
 	hashentrysize += agg_counts->transitionSpace;
 	/* plus the per-hash-entry overhead */

src/backend/optimizer/plan/subselect.c

@@ -7,7 +7,7 @@
  * Portions Copyright (c) 1994, Regents of the University of California
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/optimizer/plan/subselect.c,v 1.107 2006/05/03 00:24:56 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/optimizer/plan/subselect.c,v 1.108 2006/06/28 20:04:38 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -585,11 +585,13 @@ subplan_is_hashable(SubLink *slink, SubPlan *node)
 		return false;
 
 	/*
-	 * The estimated size of the subquery result must fit in work_mem. (XXX
-	 * what about hashtable overhead?)
+	 * The estimated size of the subquery result must fit in work_mem.
+	 * (Note: we use sizeof(HeapTupleHeaderData) here even though the tuples
+	 * will actually be stored as MinimalTuples; this provides some fudge
+	 * factor for hashtable overhead.)
 	 */
 	subquery_size = node->plan->plan_rows *
-		(MAXALIGN(node->plan->plan_width) + MAXALIGN(sizeof(HeapTupleData)));
+		(MAXALIGN(node->plan->plan_width) + MAXALIGN(sizeof(HeapTupleHeaderData)));
 	if (subquery_size > work_mem * 1024L)
 		return false;