Skip to content

cache dictionary arc ptr in groupColumn #23921

Description

@Rich-T-kid

Same single-element create_hashes pattern here, but this one is on the streaming hot path (called once per appended group). Since consecutive append_val calls usually see the same dict_values array, caching the batch's hashes keyed on Arc::as_ptr(dict.values()) would reduce this to one vectorized hash pass per batch. Fine as a follow-up.

Originally posted by @zhuqi-lucas in #23187 (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