prefer-mypy-num-workers

Detects repositories that install mypy 2.0 or newer but run it without parallel workers.

What it detects

• mypy 2.0+ declared in dependency files (pyproject.toml, requirements*.txt, poetry.lock, setup.cfg, setup.py, Pipfile), and
• a CI command that invokes mypy without --num-workers / -n, and
• no parallel configuration in mypy.ini, .mypy.ini, pyproject.toml ([tool.mypy]), setup.cfg ([mypy]), or the MYPY_NUM_WORKERS environment variable.

Why it matters

mypy 2.0 supports experimental parallel type checking. With --num-workers mypy type-checks independent module groups in separate processes and has shown up to 5x speedups with 8 workers on large projects. Parallel checking is opt-in (num_workers defaults to 0, i.e. disabled), so upgrading to mypy 2.0 alone does not capture the speedup.

Suggested action

For mypy 2.0–2.3, use a fixed integer, such as --num-workers 8 or num_workers = 8. mypy 2.4+ also supports --num-workers auto. In TOML, quote the automatic value: [tool.mypy] with num_workers = "auto"; INI files use num_workers = auto. Tune fixed counts from 3-4 upward; using more workers than physical CPU cores is not beneficial.

Commands are checked individually, including a mypy invocation following an install in the same step. Worker environment values apply only to the workflow/job/step that inherits them.

Measurement

Compare mypy wall-clock time with and without --num-workers on the same CI runner.