[VL] improve Sort Merge Join on Velox backend
- Dominant language
- Scala
- Stars
- 1.6k
- Forks
- 657
- Avg merge
- 2d 14h
- Merged PRs (30d)
- 80
Description
### Description
By default gluten with Velox backend will use Shuffled Hash Join to get better performance. Sort Merge Join is still useful in some cases 1) very big table on build side 2) bucket scan 3) inner join with redundant keys
the plan is to
- Adding support for SMJ join types other than inner and left join, and join with conditions
- improve the performance of SMJ/Sort
thanks,
-yuan
Contributor guide
Research direction
The issue mentions the Velox backend, Sort Merge Join, join types, join conditions, and sorting, but names no files, tests, or entry points. Start by locating the existing Sort Merge Join implementation and its inner and left join handling; done means supporting the listed additional cases and improving SMJ/sort performance with validation for the affected joins.
Written by the indexing model from the issue text.
Assessment
- Domain
- backend
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100