Skip to content

[Feature Request] Iceberg: rewrite_equality_delete_files procedure to compact equality deletes into position deletes #73947

@eshishki

Description

@eshishki

Problem

StarRocks reads Iceberg v2 tables with equality-delete files by applying them as a LEFT ANTI JOIN at query time. Equality deletes accumulate over time, so every read keeps paying the anti-join cost, and there is no way from StarRocks to compact them away.

Proposal

Add an Iceberg table procedure rewrite_equality_delete_files(), invoked via:

ALTER TABLE <iceberg_table> EXECUTE rewrite_equality_delete_files();

It converts the live equality-delete files of the current snapshot into position-delete files. The set of visible rows stays identical (deleted keys, NULL included, must not resurrect), but subsequent reads no longer pay the equality-delete anti-join.

Tracked by #73911.

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