Fix for py39 - unsubscriptable-object

[cdce8p]

Description

This PR fixes an issue where typing.Optional and typing.Union where falsely flagged with unsubscriptable-object.

This error seems to be the result of a changed inference value:

# py38
Instance of typing._SpecialForm

# py39
FunctionDef.Union(...)
FunctionDef.Optional(...)

Since I'm new to pylint, this might not be the best place to fix it and I saw that #3894 also addresses this issue in a different spot. Please let me know if I should change something or if you like to continue with the other PR.

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.

Type of Changes

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

Related Issue

Closes: #3882
This PR partially addresses #3895

[coveralls]

Coverage decreased (-0.006%) to 90.832% when pulling be3230a on cdce8p:fix-py39-unsubscriptable-object into fed4438 on PyCQA:master.

[JulienPalard]

I had another take on this one here: #3890 but it's my 1st contrib to pylint, I don't know which one is better.

[amyreese]

I think the solution in #3890 is a more correct fix, and resolves this for more than just typing.Optional and typing.Union.

[cdce8p]

@jreese I'm fine with that, as long as the issue is solved 😄 Please close this PR if you don't need it anymore

[hippo91]

@cdce8p thanks for your submission. IMHO the PR #3890 is indeed a more correct fix for the problem. That's why i close this PR but once again thanks for your work and please do not hesitate to submit another one. We need volunteers!

[ssbarnea]

While this bugs seams closed, the sad reality is that the last release was 6 months ago, so py39 users cannot benefit from the bugs fix.

Any change to tag a new release?

[hippo91]

@ssbarnea we are working on it. By the way the last pylint release was only 4 months ago.

[ssbarnea]

@hippo91 Thanks for the update and please excuse my bad math, there were quite long months for most of us.

The problems I am facing is that these error reproduce only on py39 but that is more and more popular and because I use pre-commit (tool, not hook), I do not have any way to restrict which python version it will use when running on developer machine. The result is an unpredictable behavior with no know temporary workaround.

If the fix is already on master adding a pre-release tag would enable pre-commit users to bump it without even having a full release. In fact it may be possible to do something without a tag but is very ugly, it requires putting a specific revision instead of a tag for version.

[hippo91]

@ssbarnea no worry about the previous release date.
The fix concerning this problem has indeed been merged but there are still other bugs to deal with in order to make pylint compatible with python3.9. Fortunately those bugs have corresponding PR so, theoretically it wont be too long before the new release.