Skip to content

Allow co-partitioned range inputs to satisfy inner partitioned hash joins #23183

Description

@gene-bordegaray

Is your feature request related to a problem or challenge?

Part of #22395.

DataFusion can represent source-declared Partitioning::Range, but inner partitioned hash joins still require hash-partitioned inputs. This can add unnecessary hash repartitions even when both join inputs are already partitioned by compatible ranges on the join keys.

Describe the solution you'd like

Allow inner partitioned hash joins to accept inputs that are co-partitioned on their join keys, including compatible range-partitioned inputs.

  • adds a key-based distribution requirement that can be satisfied by hash or range partitioning
  • checks that both join sides can be paired by partition index
  • keeps incompatible range bounds, mismatched hash counts, and non-inner joins on the existing hash repartition path
  • keeps partitioned dynamic filters restricted to hash-compatible routing

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions