[HIVE-11234] CBO: Calcite Operator To Hive Operator (Calcite Return Path): reconsider the cross product in HiveJoinToMultiJoinRule

• Type: Sub-task
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Component/s: CBO
• Assignee: Pengcheng Xiong
• Reporter: Pengcheng Xiong
• Created: 13/Jul/15 03:54
• Updated: 27/Feb/24 22:23
• Resolved: 30/Jul/15 18:47

Description

Cross product in HiveJoinToMultiJoinRule needs to be reconsidered. Now the following explain query will fail at L1116 StatsRulesProcFactory : FAILED: IndexOutOfBoundsException Index: 0, Size: 0 because the key in one of the branches of the join is empty.

explain select *  from src 
where src.key not in (select key  from src1);

If we run

select *  from src 
where src.key not in (select key  from src1);

we will have "FAILED: Execution Error, return code 2 from org.apache.hadoop.hive.ql.exec.mr.MapRedTask". The root cause is the same. Other queries that will also fail include cbo_subq_not_in.q etc.

Issue Links

• Is contained by: HIVE-11257 CBO: Calcite Operator To Hive Operator (Calcite Return Path): Method isCombinablePredicate in HiveJoinToMultiJoinRule should be extended to support MultiJoin operators merge