Add support for pep585 with postponed evaluation

[cdce8p]

Steps

• Add yourself to CONTRIBUTORS if you are a new contributor.
• Add a ChangeLog entry describing what your PR does.
• If it's a new feature or an important bug fix, add a What's New entry in doc/whatsnew/<current release.rst>.
• Write a good description on what the PR does.

Description

With PEP 585 and postponed evaluation of annotations it's possible to use list[int] and others even in versions 3.7 and 3.8 although PEP 585 was implemented in 3.9.

Currently, running pylint in a 3.7 or 3.8 environment always returns unsubscriptable-object errors. However the following example is valid

from __future__ import annotations

var: list[int]

def func(arg: dict[str, int]) -> list[int]:
  ...

The error is correct only if the type is evaluated instantly as for example with type aliases

from __future__ import annoations

InvalidAlias = list[int]

The latest mypy version 0.800 just added support for PEP 585 so pylint is the only thing holding it back for many.

--
Since I don't have much experience developing for pylint, some guidance would be much appreciated.
There is probably a better way to do this 😄

Type of Changes

	Type
✓	🐛 Bug fix
✓	✨ New feature
	🔨 Refactoring
	📜 Docs

Related Issue

Closes: #3320

[Pierre-Sassoulas]

Thank you for the MR, very clean. This change look reasonable, would it be possible maybe to check that from __future__ import annotations is present if we're under python 3.9 ? Ie. The same functional file without from __future__ import annotations would raise Error for python under 3.9 ? Also sorry for the master branch failing that also makes your CI fail, no idea how that happened, maybe a flaky test.

[cdce8p]

@Pierre-Sassoulas Thanks for the quick review. Regarding your comments

would it be possible maybe to check that from future import annotations is present if we're under python 3.9 ? Ie. The same functional file without from future import annotations would raise Error for python under 3.9 ?

I believe it already does. I've added if is_postponed_evaluation_enabled(node) to supports_getitem which should already check for it. I can a a new test case though, just to make sure.

https://github.com/PyCQA/pylint/blob/1f34e3194bcfa281b9eb4b0595ec654b782405ee/pylint/checkers/utils.py#L1325-L1331

Also sorry for the master branch failing that also makes your CI fail, no idea how that happened, maybe a flaky test.

I noticed that the master branch is now passing, I will rebase the MR. Maybe that solves it.

[cdce8p]

Also sorry for the master branch failing that also makes your CI fail, no idea how that happened, maybe a flaky test.

I noticed that the master branch is now passing, I will rebase the MR. Maybe that solves it.

Didn't work. I was already using the latest master. Maybe it will work for the next commit.

[cdce8p]

I believe it already does. I've added if is_postponed_evaluation_enabled(node) to supports_getitem which should already check for it. I can a a new test case though, just to make sure.

It worked, as expected. I've included the checks in my last commit.

[cdce8p]

Sorry for the many updates, the MR should be ready now.

[Pierre-Sassoulas]

Thanks a lot ! I'll merge this as soon as the build is fixed. Apparently the problem come from the current astroid master branch. Can you please allow edit from maintainer so I can rebase your branch myself once it's fixed ? :)

[cdce8p]

Can you please allow edit from maintainer so I can rebase your branch myself once it's fixed ? :)

@Pierre-Sassoulas Already checked. If it doesn't work for whatever reason, just let me know. I usually can get back to it within a day.

[cdce8p]

@Pierre-Sassoulas I saw that there are plans to do a new release soon pylint-dev/astroid#886 (comment).
Can you add the 2.6.1 milestone, just so it will be included?

[cdce8p]

@Pierre-Sassoulas I noticed today that I was missing checks for TypedDict, NamedTuple and dataclass. Can you take a look at it again, since adding those required a bigger code change?

[cdce8p]

Sorry for the back and forth, I find it quite difficult to get it right with checking so many different things which aren't really documented anywhere. However I think I got it now.

Turns out my last code change was wrong, with from __future__ import annotations TypeDict, NamedTuple and Dataclasses work mostly as expected. In particular the following does indeed work and is not a TypeError:

# Valid in Python 3.7/3.8 (TypedDict)
from __future__ import annotations

class CustomNamedTuple(NamedTuple):
    my_var: list[int]

class CustomTypedDict(TypedDict):
    my_var: list[int]

@dataclass
class CustomDataClass:
    my_var: list[int]

What doesn't work and I guess caused my confusion is:

CustomNamedTuple = typing.NamedTuple(
    "CustomNamedTuple", [("my_var", list[int])])  # [unsubscriptable-object]

CustomTypedDict = TypedDict("CustomTypedDict", my_var=list[int])  # [unsubscriptable-object]

CustomTypedDict2 = TypedDict("CustomTypedDict2", {"my_var": list[int]})  # [unsubscriptable-object]

Just to be sure I also checked it with Pylance / pyright, but, as it turns out, even it doesn't recognize all cases correctly so I also did some manual checks again.

[Pierre-Sassoulas]

No worry, mypy 0.8 has only been released 2 weeks ago, this is cutting edge statistical analysis on multiple python interpreters after all ;)

[PCManticore]

Great PR, thank you @cdce8p !

[coveralls]

Coverage decreased (-0.02%) to 90.563% when pulling d877dd0 on cdce8p:pep585-postponed-evaluation into 005cf2f on PyCQA:master.

[coveralls]

Coverage decreased (-0.02%) to 90.563% when pulling d877dd0 on cdce8p:pep585-postponed-evaluation into 005cf2f on PyCQA:master.

[cdce8p]

@Pierre-Sassoulas This one is ready to be merged.