incubator-gluten
incubator-gluten copied to clipboard
apache
[GLUTEN-4745][CH] support Sort Merge Join
Support inner and outer joins
What changes were proposed in this pull request?
(Fixes: #GLUTEN-4745)
How was this patch tested?
unit tests and integration tests
![github-actions[bot] avatar](https://avatars.githubusercontent.com/in/15368?v=4 "Published by a pub.dev verified publisher"){kind=small}
seems unrelated failed in static-build-centos7-test :
error: Failed to download from mirror set
error: https://web.mit.edu/kerberos/dist/krb5/1.20/krb5-1.20.tar.gz: curl failed to download with exit code 18
@zzcclp This PR is ready for review.Thanks. In order to support join keys with null correctly, we need change CH's MergeJoinTransform, will add that to CH later.
From https://opencicd.kyligence.com/job/gluten/job/gluten-ci/7371/console i can see all test passed. Not sure why report faliure.
ERROR: script returned exit code 1
PR for support nulls first. https://github.com/Kyligence/ClickHouse/pull/483 May be can make nulls order configurable in ClickHouse later.
From https://opencicd.kyligence.com/job/gluten/job/gluten-ci/7371/console i can see all test passed. Not sure why report faliure.
ERROR: script returned exit code 1
@liuneng1994 could you see if ci is running ok?
PR for support nulls first. Kyligence/ClickHouse#483 May be can make nulls order configurable in ClickHouse later.
Is there any side effect about this change ? Pls try to complete it in upstream ClickHouse/ClickHouse firstly instead of directly modify src/* files in Kyligence/ClickHouse. cc @baibaichen
@zzcclp This PR is ready for review.Thanks. In order to support join keys with null correctly, we need change CH's MergeJoinTransform, will add that to CH later.
Is there already a PR in CH?
@zzcclp This PR is ready for review.Thanks. In order to support join keys with null correctly, we need change CH's MergeJoinTransform, will add that to CH later.
Is there already a PR in CH?
After this PR merged, we can set nulls smallest in Gluten. https://github.com/ClickHouse/ClickHouse/pull/60896
For now we can disable outer joins whose join keys may be null. And only support INNER join first.