Prevent Parallel Hash Join for JOIN_UNIQUE_INNER.

WHERE EXISTS (...) queries cannot be executed by Parallel Hash Join
with jointype JOIN_UNIQUE_INNER, because there is no way to make a
partial plan totally unique.  The consequence of allowing such plans
was duplicate results from some EXISTS queries.

Back-patch to 11.  Bug #15857.

Author: Thomas Munro
Reviewed-by: Tom Lane
Reported-by: Vladimir Kriukov
Discussion: https://postgr.es/m/15857-d1ba2a64bce0795e%40postgresql.org

diff --git a/src/backend/optimizer/path/joinpath.c b/src/backend/optimizer/path/joinpath.c
index 501ad775cbe9ce05fcbc182e21cd30e22883d403..dc28b56e7427e13aa5db455d840f1bfcea6565a3 100644 (file)
--- a/src/backend/optimizer/path/joinpath.c
+++ b/src/backend/optimizer/path/joinpath.c
@@ -1867,9 +1867,12 @@ hash_inner_and_outer(PlannerInfo *root,
 
                        /*
                         * Can we use a partial inner plan too, so that we can build a
-                        * shared hash table in parallel?
+                        * shared hash table in parallel?  We can't handle
+                        * JOIN_UNIQUE_INNER because we can't guarantee uniqueness.
                         */
-                       if (innerrel->partial_pathlist != NIL && enable_parallel_hash)
+                       if (innerrel->partial_pathlist != NIL &&
+                               save_jointype != JOIN_UNIQUE_INNER &&
+                               enable_parallel_hash)
                        {
                                cheapest_partial_inner =
                                        (Path *) linitial(innerrel->partial_pathlist);