Skip to content

Support and extend residual conditions in native shuffled hash joins #2540

Description

@Deegue

Is your feature request related to a problem? Please describe.

Native shuffled hash joins still restrict residual conditions to inner joins, although the shared native hash join implementation now supports condition-aware matching for additional join types.

The same support is needed when sort-merge joins are converted to shuffled hash joins.

Describe the solution you'd like

  • Remove the inner-only condition restriction from native shuffled hash join planning.
  • Reuse the shared native hash join implementation for residual-condition evaluation.
  • Apply common condition validation to direct shuffled hash joins and forced sort-merge-to-shuffled-hash conversions.
  • Return no output ordering on Spark 3.1 for build-preserving join orientations that Spark's inherited ordering implementation cannot handle.
  • Add Spark regression tests for join semantics, projections, configuration, and output ordering.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions