Consider disk-based hash aggregation to implement DISTINCT.

jeff-davis · jeff-davis · commit dd8e19132acf · 2020-03-24T18:30:04.000-07:00
Correct oversight in 1f39bce. If enable_hashagg_disk=true, we should consider hash aggregation for DISTINCT when applicable.
diff --git a/src/backend/optimizer/plan/planner.c b/src/backend/optimizer/plan/planner.c
@@ -4868,8 +4868,8 @@ create_distinct_paths(PlannerInfo *root,
 		Size		hashentrysize = hash_agg_entry_size(
 			0, cheapest_input_path->pathtarget->width, 0);
 
-		/* Allow hashing only if hashtable is predicted to fit in work_mem */
-		allow_hash = (hashentrysize * numDistinctRows <= work_mem * 1024L);
+		allow_hash = enable_hashagg_disk ||
+			(hashentrysize * numDistinctRows <= work_mem * 1024L);
 	}
 
 	if (allow_hash && grouping_is_hashable(parse->distinctClause))

Original file line number	Diff line number	Diff line change
`@@ -4868,8 +4868,8 @@ create_distinct_paths(PlannerInfo *root,`
`4868`	`4868`	`Size hashentrysize = hash_agg_entry_size(`
`4869`	`4869`	`0, cheapest_input_path->pathtarget->width, 0);`
`4870`	`4870`
`4871`		`- /* Allow hashing only if hashtable is predicted to fit in work_mem */`
`4872`		`- allow_hash = (hashentrysize * numDistinctRows <= work_mem * 1024L);`
	`4871`	`+ allow_hash = enable_hashagg_disk \|\|`
	`4872`	`+ (hashentrysize * numDistinctRows <= work_mem * 1024L);`
`4873`	`4873`	`}`
`4874`	`4874`
`4875`	`4875`	`if (allow_hash && grouping_is_hashable(parse->distinctClause))`