Skip to content

REVIEW.md

Latest commit

 

History

History
27 lines (19 loc) · 3.38 KB

REVIEW.md

File metadata and controls

27 lines (19 loc) · 3.38 KB
Raw

Review instructions

The "PR review" Routine reads this file from the base branch of every pull request to this repository. The contract in claude-tools coding/skills/review/SKILL.md sets what counts as a finding and the shape of the review; this file says what matters here. Keep it short: every line competes with the contract for the reviewer's attention.

Important here

  • A conference entry whose cfp, start, end, place or link disagrees with the conference's own website: fetch link (and cfp_link) and compare. The schema checks formats, not facts, and the site publishes whatever is merged.
  • A year that differs from the year of start, or a timezone that is not an IANA name: the schema accepts both, and static/js/countdown-simple.js then falls back to the visitor's own zone without saying so.
  • An entry that already exists in _data/conferences.yml, _data/archive.yml or _data/legacy.yml under another spelling of the name. CI only warns, and only on an exact name-and-year match.
  • A change under _layouts/, _includes/ or _pages/ that names an include, layout or asset path that does not exist: no pull-request job builds the site for those paths, so the first failure is the deploy of main.
  • A utils/ change that lets sort_yaml.py write a data file after a validation error, or drop an entry on the way. The weekly sort (recurring-sort.yml) merges its own result unattended.

Leave alone

  • Everything pixi run sort enforces or rewrites: the pydantic schema in utils/tidy_conf/schema.py (required fields, date and URL formats, sub against _data/types.yml, a location for in-person events), key order, sort order, years inside conference, redundant link fields and geocoding. After a merge, pr-sort.yml opens the clean-up PR itself.
  • Everything pixi run pre enforces on Python: ruff, black, isort, pyupgrade, docsig and the pre-commit-hooks set. Quote their output under Tests.
  • JavaScript style in static/js/: ESLint runs with || true; the jest coverage thresholds in jest.config.js are the only enforced bar.
  • Translations under _i18n/ other than en: Crowdin owns them (crowdin.yml, TRANSLATION.md), and a direct edit is overwritten by the next sync.

Checks

There is no pixi in the Routine sandbox, so run the same checks through uv, pinned the way pixi.toml pins them (Python 3.11, pandas 2.2):

  • Data: uv run --python 3.11 --with "pandas<2.3",pyyaml,pydantic,pytz,thefuzz,iso3166,icalendar,tqdm,requests python utils/sort_yaml.py --skip_links, then git diff --stat -- _data. A non-zero exit is a finding; a diff is the rewrite the post-merge sort will do, not a finding.
  • Python, when utils/ or tests/ change: uv run --python 3.11 --with "pandas<2.3",pytest,pytest-mock,freezegun,hypothesis,responses,pyyaml,pydantic,pytz,thefuzz,iso3166,icalendar,tqdm,requests pytest tests/ -q -x -p no:cacheprovider.
  • Lint, when Python changes: uvx ruff@0.4.4 check utils tests, the version .pre-commit-config.yaml pins (a newer ruff reports dozens of pre-existing errors). The rest of the pre-commit set is the pre-commit.ci - pr status on the PR: quote it, do not rerun it.
  • JavaScript, when static/js/ or tests/frontend/ change: npm ci && npm test.