Skip to content

perf: enable matching optimizations by default - #1165

Open
lpukatch wants to merge 2 commits into
devfrom
perf/1163-default-matching-optimizations
Open

lpukatch wants to merge 2 commits into
devfrom
perf/1163-default-matching-optimizations

Conversation

@lpukatch

Copy link
Copy Markdown
Collaborator

Enable candidate token indexing and negative match caching when TEAMARR_TOKEN_INDEX and TEAMARR_NEGATIVE_CACHE are unset. Existing explicit overrides keep their behavior, including =0 and empty values disabling either optimization.

Update default-on and boolean parsing tests, make disabled parity paths set =0 explicitly, and exercise negative-cache regressions with the variable unset. New regressions verify that a default-cached failure is immediately retried after alias creation or team-data invalidation. The existing scoped invalidation fix (#1153) is already in the base branch.

README, advanced-settings documentation, and both agent instruction files describe the defaults and opt-outs. This PR is independent of catalog reuse (#1162).

Validation:

  • Ruff and Pyright passed.
  • 96 targeted matching/invalidation tests passed.
  • Full suite: 4,867 tests passed (pytest tests/ -v -n 4).
  • npm run build passed.

The initial parallel suite exposed an unrelated SSE initial-frame race, tracked separately in #1164; its isolated rerun and the complete full-suite rerun passed.

Refs #1163. Bead: teamarr-27yp. Leave open for review.

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant