assertIsNotNone(foo) in tests is not treated the same as assert foo is not None

[timabbott]

As part of Zulip's strict-optional effort, I had to convert a test that was using assertIsNotNone(foo) to use assert foo is not None in order to get mypy to pass. I think mypy should treat those two the same.

[JelleZijlstra]

Is assertIsNotNone here a custom function? One way to implement this could be to type assertisNotNone as:

@overload
def assertisNotNone(obj: None) -> NoReturn: ...
@overload
def assertIsNotNone(obj: Any) -> None: ...

Haven't tested it though.

[elazarg]

@JelleZijlstra that's #4063. It is not implemented.

[timabbott]

It's part of the standard unittest library: https://docs.python.org/2/library/unittest.html#unittest.TestCase.debug

[ethanhs]

I think this can be closed in favor of #4063 then.

[1kastner]

Is assertIsNotNone here a custom function? One way to implement this could be to type assertisNotNone as:

@overload
def assertisNotNone(obj: None) -> NoReturn: ...
@overload
def assertIsNotNone(obj: Any) -> None: ...

Haven't tested it though.

Is that something that should be added to the typeshed description of unittest then? I checked https://github.com/python/typeshed/blob/master/stdlib/unittest/case.pyi#L120 which reads this:

def assertIsNotNone(self, obj: object, msg: Any = ...) -> None: ...

I have seen the forward to #4063. However, that issue is open for 5 years with no result. And currently I prefer to have this issue solved. I just do not want to double-check everything with both an assert statement and a unittest check. I also do not want to refactor all unittests of a project just because of a missing functionality of mypy. I hope that makes it understandable why it might be more reasonable to pick up this track. Thank you!