[HIVE-4927] When we merge two MapJoin MapRedTasks, the TableScanOperator of the second one should be removed

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 0.12.0
• Fix Version/s: 0.12.0
• Component/s: Query Processor
• Assignee: Yin Huai
• Reporter: Yin Huai
• Created: 24/Jul/13 21:08
• Updated: 15/Oct/13 23:30
• Resolved: 26/Jul/13 15:54

Description

set hive.auto.convert.join=true;
set hive.auto.convert.join.noconditionaltask=true;
EXPLAIN
SELECT x1.key AS key FROM src x1 JOIN src1 y1 ON (x1.key = y1.key) JOIN src1 y2 ON (x1.value = y2.value) GROUP BY x1.key;

We will get a NPE from MetadataOnlyOptimizer. The reason is that the operator tree of the MapRedTask evaluating two MapJoins is

TS1->MapJoin1->TS2->MapJoin2->...

We should remove the TS2...

Issue Links

• relates to: HIVE-4827 Merge a Map-only task to its child task
• links to: Phabricator