prohibit_parallel_limit_subselect_v2_rel96.patch

application/octet-stream

Filename: prohibit_parallel_limit_subselect_v2_rel96.patch
Type: application/octet-stream
Part: 1
Message: Re: BUG #15324: Non-deterministic behaviour from parallelised sub-query

Patch

Same data as JSON: GET /api/v1/attachments/:id/patch the parsed metadata as JSON — format, series position, per-file stats; never the diff bytes. API reference →
Format: unified
Series: patch v2
File+
src/backend/optimizer/path/allpaths.c 10 0
src/backend/optimizer/plan/planner.c 1 2
src/include/optimizer/planner.h 2 0
src/test/regress/expected/select_parallel.out 19 0
src/test/regress/sql/select_parallel.sql 5 0
diff --git a/src/backend/optimizer/path/allpaths.c b/src/backend/optimizer/path/allpaths.c
index b4a1e8b..df0ee2f 100644
--- a/src/backend/optimizer/path/allpaths.c
+++ b/src/backend/optimizer/path/allpaths.c
@@ -585,7 +585,17 @@ set_rel_consider_parallel(PlannerInfo *root, RelOptInfo *rel,
 			 * the SubqueryScanPath as not parallel-safe.  (Note that
 			 * set_subquery_pathlist() might push some of these quals down
 			 * into the subquery itself, but that doesn't change anything.)
+			 *
+			 * We can't push sub-select containing LIMIT/OFFSET to workers as
+			 * that can change the results depending on what is beneath LIMIT
+			 * node.  So, consider that case as parallel-unsafe.
 			 */
+			{
+				Query	*subquery = castNode(Query, rte->subquery);
+
+				if (limit_needed(subquery))
+					return;
+			}
 			break;
 
 		case RTE_JOIN:
diff --git a/src/backend/optimizer/plan/planner.c b/src/backend/optimizer/plan/planner.c
index 73eb307..f348493 100644
--- a/src/backend/optimizer/plan/planner.c
+++ b/src/backend/optimizer/plan/planner.c
@@ -98,7 +98,6 @@ static void preprocess_rowmarks(PlannerInfo *root);
 static double preprocess_limit(PlannerInfo *root,
 				 double tuple_fraction,
 				 int64 *offset_est, int64 *count_est);
-static bool limit_needed(Query *parse);
 static void remove_useless_groupby_columns(PlannerInfo *root);
 static List *preprocess_groupclause(PlannerInfo *root, List *force);
 static List *extract_rollup_sets(List *groupingSets);
@@ -2492,7 +2491,7 @@ preprocess_limit(PlannerInfo *root, double tuple_fraction,
  * a key distinction: here we need hard constants in OFFSET/LIMIT, whereas
  * in preprocess_limit it's good enough to consider estimated values.
  */
-static bool
+bool
 limit_needed(Query *parse)
 {
 	Node	   *node;
diff --git a/src/include/optimizer/planner.h b/src/include/optimizer/planner.h
index d9790d7..b4546b1 100644
--- a/src/include/optimizer/planner.h
+++ b/src/include/optimizer/planner.h
@@ -46,6 +46,8 @@ extern bool is_dummy_plan(Plan *plan);
 extern RowMarkType select_rowmark_type(RangeTblEntry *rte,
 					LockClauseStrength strength);
 
+extern bool limit_needed(Query *parse);
+
 extern void mark_partial_aggref(Aggref *agg, AggSplit aggsplit);
 
 extern Path *get_cheapest_fractional_path(RelOptInfo *rel,
diff --git a/src/test/regress/expected/select_parallel.out b/src/test/regress/expected/select_parallel.out
index 43801e0..ba1949f 100644
--- a/src/test/regress/expected/select_parallel.out
+++ b/src/test/regress/expected/select_parallel.out
@@ -182,6 +182,25 @@ select count(*) from tenk1;
 (1 row)
 
 reset role;
+-- LIMIT/OFFSET within sub-selects can't be pushed to workers.
+explain (costs off)
+select * from tenk1 where two in
+       (select two from tenk1 where stringu1 like '%AAAA' limit 3);
+                          QUERY PLAN                           
+---------------------------------------------------------------
+ Hash Semi Join
+   Hash Cond: (tenk1.two = tenk1_1.two)
+   ->  Gather
+         Workers Planned: 4
+         ->  Parallel Seq Scan on tenk1
+   ->  Hash
+         ->  Limit
+               ->  Gather
+                     Workers Planned: 4
+                     ->  Parallel Seq Scan on tenk1 tenk1_1
+                           Filter: (stringu1 ~~ '%AAAA'::text)
+(11 rows)
+
 explain (costs off)
   select stringu1::int2 from tenk1 where unique1 = 1;
                   QUERY PLAN                   
diff --git a/src/test/regress/sql/select_parallel.sql b/src/test/regress/sql/select_parallel.sql
index 7defc34..dafc4ad 100644
--- a/src/test/regress/sql/select_parallel.sql
+++ b/src/test/regress/sql/select_parallel.sql
@@ -83,6 +83,11 @@ drop role regress_parallel_worker;
 select count(*) from tenk1;
 reset role;
 
+-- LIMIT/OFFSET within sub-selects can't be pushed to workers.
+explain (costs off)
+select * from tenk1 where two in
+       (select two from tenk1 where stringu1 like '%AAAA' limit 3);
+
 explain (costs off)
   select stringu1::int2 from tenk1 where unique1 = 1;