[HIVE-20826] Enhance HiveSemiJoin rule to convert join + group by on left side to Left Semi Join

• Type: Improvement
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 4.0.0-alpha-1
• Target Version/s: 4.0.0
• Assignee: Vineet Garg
• Reporter: Vineet Garg
• Created: 27/Oct/18 22:05
• Updated: 17/Nov/22 08:55
• Resolved: 11/Nov/18 02:46

Description

Currently HiveSemiJoin rule looks for pattern where group by is on right side.

We can convert joins which have group by on left side (assuming group by keys are same as join keys and none of the columns are being projected from left side) to LEFT SEMI JOIN by swapping the inputs. e.g. queries such as:

explain select pp.p_partkey from (select distinct p_name from part) p join part pp on pp.p_name = p.p_name;

Issue Links

• links to: ReivewBoard