requests: type RequestException members (not Any)

[ErikBjare]

Thought it was weird that mypy didn't catch the following:

import requests

try:
    r = requests.get("http://localhost:666")
except requests.RequestException as e:
    # e.response can be None for some subtypes of RequestException (like ConnectionError, Timeout, etc.)
    # basically for all except HTTPError (I think?)
    # Therefore, it crashes with:
    #   AttributeError: 'NoneType' object has no attribute 'status_code'
    if e.response.status_code == 500:
        pass

Hopefully this is a reasonable change.

[ErikBjare]

Looks like this triggered a bunch of issues in repos, should be possible to make more precise by specifying that HTTPError always has them set (not sure about this assumption, though).

[ErikBjare]

If psf/requests#6270 is considered mergeable/correct, that means we can further specify ConnectionError.request to not be None, which would get rid of the final complaints from mypy_primer.

[JelleZijlstra]

Thanks! Let's wait for resolution on the requests PR so that we can merge this without the mypy-primer regression.

[intgr]

Let's wait for resolution on the requests PR

Oct 26, 2022

[github-actions]

Diff from mypy_primer, showing the effect of this PR on open source code:

schemathesis (https://github.com/schemathesis/schemathesis)
+ src/schemathesis/models.py: note: In member "call" of class "Case":
+ src/schemathesis/models.py:340: error: Argument 2 to "_get_code_message" of "Case" has incompatible type "Optional[Request]"; expected "PreparedRequest"  [arg-type]
+ src/schemathesis/models.py: note: At top level:
+ src/schemathesis/runner/events.py: note: In member "from_exc" of class "InternalError":
+ src/schemathesis/runner/events.py:195: error: Item "None" of "Optional[Request]" has no attribute "url"  [union-attr]

[intgr]

Maybe we can live with the regression and merge this as is?

As long as psf/requests#6270 isn't merged, ConnectionError.request is truly None'able, the additional errors from mypy are correct and may help fix bugs in user code.

[JelleZijlstra]

Given that it's been almost a year, I agree. I'll wait a day or so in case other typeshed maintainers disagree, then merge this PR.