Details
-
Bug
-
Resolution: Done
-
Critical
-
8.0
-
None
Description
As discussed in https://developer.jboss.org/message/939309#939309, an inconsistent behavior of the NOT IN clause occurs when NULL values are returned by the subquery.
For example, let us suppose to have a table defined as:
CREATE TABLE test1 (col1 character varying(10)); insert into test1 values ('a'); insert into test1 values ('b'); insert into test1 values ('c'); insert into test1 values ('d');
If I enumerate values in the NOT IN clause and one of them is a NULL value:
select * from pg.test1 where col1 NOT IN ('a', NULL, 'b');
the result is not empty (as it would be expected) but 'c' and 'd' are returned.
Looking at the query plan, the NULL is wrongly removed from the list:
---------------------------------------------------------------------------- OPTIMIZATION COMPLETE: PROCESSOR PLAN: AccessNode(0) output=[pg.test1.col1] SELECT g_0.col1 FROM pg.test1 AS g_0 WHERE g_0.col1 NOT IN ('a', 'b') ============================================================================
