Parsimonious stubs

[jpy-git]

Parsimonious is a PEG parser package.

PyPI link: https://pypi.org/project/parsimonious/

I'm making use of this package in a couple projects and there's also an open issue about adding type hints so figured I'd pick this up.

[jpy-git]

@JelleZijlstra Thanks for the thorough review 😄

I believe I've addressed the comments. I'm having some trouble with getting Check third party stubs with stubtest to pass since it's complaining about some of the subclass types not agreeing with the superclass types, which is what's happening in the code, and stubtest_allowlist doesn't seem to fix things. Any ideas what I'm missing?

[hauntsaninja]

stubtest runs a mypy build first to understand what's in the stubs. The mypy build is failing, so we haven't even gotten to the actual stubtest logic, which is what stubtest allowlists affect.

Assuming your subclass types are actually correct, adding type ignore will allow the build to proceed. For the "Type application has too many types", it's probably some mypy bug, try using an old style typing.Union + typing.Tuple

[JelleZijlstra]

The errors are from mypy itself, not stubtest; they'll need # type: ignore. I'll try to figure out why the regular mypy test didn't catch this.

[JelleZijlstra]

@hauntsaninja it's not expected that the mypy test in CI is passing though, right? Looks like it runs only on the stdlib.

[hauntsaninja]

Yeah, looks like mypy_test is only running against the stdlib. It's not the first time I've seen stubtest third party catch mypy errors that mypy_test hasn't caught, so the test has probably been buggy for a while.

[JelleZijlstra]

It's likely related to the mypy feature where it does os._exit to exit faster.

[JelleZijlstra]

Fix in #7478