Skip to content

Performance: Reduce refcount churn in csv.reader #156748

Description

@brittanyrey

Minor change to reduce churn in parse_save_field(). It currently calls PyList_Append(self->fields, field); Py_DECREF(field) for every field of every row. PyList_Append internally does Py_NewRef(newitem), so each field pays an incref immediately followed by a decref, plus a critical-section enter/exit.
Since the reference is already about to be dropped, just pass the reference directly to the caller.

(PR is written & ready to attach)

Linked PRs

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

    extension-modulesC modules in the Modules dirperformancePerformance or resource usagetype-featureA feature request or enhancement

    Projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions