Allows to use type[T] in stubs

[sobolevn]

Refs #10303

CC @JelleZijlstra @AlexWaygood

[AlexWaygood]

I don't know my way around the mypy code base very well, but it looks like this solves using type[T] in annotations, but not in other uses such as type aliases and subclasses — is that correct? 🙂 (Totally understand if you want to leave those use cases to another PR!)

[sobolevn]

Yes, other cases are not covered yet 🙂
Only the one case I've tested.

Example:

(I prefer to make a lot of small PRs instead of one big one, because they are a lot harder to get merged)

[AlexWaygood]

(I prefer to make a lot of small PRs instead of one big one, because they are a lot harder to get merged)

Of course! Just wanted to clarify 🙂

[hauntsaninja]

Thanks, this is great.

I think the names here could use some cleaning up, if you're willing to do a little more refactoring.
self.allow_new_syntax is misleading, since it seems to be exactly the same as is_stub_file.
It would be great if we could set self.always_allow_new_syntax = self.is_stub_file or self.api.is_future_flag_set('annotations').
We could then get rid of the new method and have checks like if self.always_allow_new_syntax or self.options.python_version >= (3, 9), which seems the most readable to me.

[sobolevn]

Sounds like a good idea! I was confused with allow_new_syntax as well. Will do tomorrow.

[sobolevn]

@hauntsaninja done!

I've also added a TODO for myself for later. We need to double check that it handles __future__ import correctly. Ok, it only check rhs, which is not affected by __future__.

[hauntsaninja]

Hooray!