Skip to content

Supports skipping partial aggregates #327

@richox

Description

@richox

Is your feature request related to a problem? Please describe.
as described in SPARK-31973, skipping partial aggregates where data cardinality is high (like group by user_id) sufficiently improves performance.

Describe the solution you'd like
implements partial agg skipping strategy in agg_tables.rs:

  1. mark an AggExec as skippable. (in spark side, where requiredDistribution is empty)
  2. process the first N records.
  3. check the number of input records and the number of aggregated records, if reached threshold, directly outputs all in-memory, spilled and rest records.

Describe alternatives you've considered
A clear and concise description of any alternative solutions or features you've considered.

Additional context
Add any other context or screenshots about the feature request here.

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