Release 0.950 planning

[JukkaL]

I'm planning to make a 0.950 release by Apr 22 (tentative timeline).

Please post here any issues you'd like to see fixed in the release, or PRs you'd like to have merged. I've already cut the release branch, but we can still cherry-pick low-risk features and bug fixes until Apr 18. After Apr 18 only critical bug fixes and documentation updates should be cherry-picked.

[AlexWaygood]

For typeshed, there's a trio of PEP-604 type-alias bugs which it would be great to see fixed, if at all possible:

• False-positive errors when using type[T] (or Type[T]) in a PEP 604 type alias #12392
• False-positive error when using Callable in a PEP 604 type alias #12393
• (🐞) False positive "error: Type application has too many types (1 expected)" on tuple when annotated as a type or in a TypeAlias with new union syntax (PEP 604) #11098

(There are a few others as well, but these are the ones which are a blocker for typeshed to use PEP-604 syntax everywhere.)

[cdce8p]

For ParamSpec, these two small PRs would be useful:

• Fix propagated Any constraint #12548
• SubtypeVisitor: treat ParamSpec as if it was a covariant TypeVar #12096

[97littleleaf11]

• Add support for assert_type #12584
• FindModuleCache: leverage BuildSourceSet #9478

[JukkaL]

This regression seems pretty serious, and I'm trying to prepare a fix in time for the release:

• Invalid type inference for zip in a generic class #12588

[JukkaL]

This is blocking some typeshed improvements:

• Overloading staticmethods #7781

[huguesb]

#12032

[hauntsaninja]

I missed the branch cut date. Thoughts on cherry picking an additional typeshed sync #12596 ? Should be pretty low risk (relatively few changes, no mypy-primer utput, includes assert_type, better ExceptionGroup support)

[JelleZijlstra]

• Add support for assert_type #12584

This was merged, would appreciate cherry-picking it into the release branch.

Edit: please also cherry-pick #12612

[JelleZijlstra]

#12610 is claimed to be a regression since 0.942, so we should fix it before the release.

[AlexWaygood]

• more enum-related speedups #12032

@huguesb mentioned this one above — I haven't looked at it in depth, but it looks like it might help with the enum-related performance regressions you've recently been working on @JukkaL

[huguesb]

* [FindModuleCache: leverage BuildSourceSet #9478](https://github.com/python/mypy/pull/9478)

I apologize for the inconvenience. I accidentally deleted the underlying branch which closed #9478
Re-opened as #12616 after rebasing on master.

[JukkaL]

Cherry-picked these PRs:

• Use class name as namespace for type variables #12590
• Sync typeshed #12596

This has also been fixed:

• add test case: regression w/ mypy mis-interpreting generator comprehension #12610

If the latter doesn't cause issues I can, cherry-pick the assert_type PR afterwards.

I've been investigating a few other potential regressions (no GitHub issues yes) so I'm a little behind. I can still merge new features until tomorrow. Bug fixes can be merged afterwards.

[cdce8p]

I can still merge new features until tomorrow.

If still possible, it would be awesome if this one could be included.

• Fix nested overload merging #12607

[huguesb]

@JukkaL I believe your recent changes to make_simplified_union actually caused performance regressions. I have a proposed solution, with clarifying comments in #12630

[JukkaL]

Cherry-picked assert_type PRs:

• Add support for assert_type #12584
• Fix type context for assert_type() #12612

[JukkaL]

Cherry-picked enum/union optimizations:

• more enum-related speedups #12032
• make_simplified_union: simpler and faster #12630

[huguesb]

Speeding up error-counting: #12631

[hauntsaninja]

Regression from the change to namespace typevars: #12633

[huguesb]

Per-file timing stats #12639

[cdce8p]

#12668 is an awesome PR which addresses some outstanding issues with P.args and P.kwargs.

[JukkaL]

I'm a little behind schedule, since there were a few tricky regressions that I'm still working on. Only fixes to regressions and low-risk fixes to (significant) bugs can be merged until the release (+ doc updates).

It may be okay to include #12668, but no promises.

[JukkaL]

Mypy 0.950 is out: http://mypy-lang.blogspot.com/2022/04/mypy-0950-released.html

Some PRs were left out, unfortunately, but we'll probably have a bugfix release in 1-2 weeks (or earlier if there is a major regression) which can include fixes to regressions and low-risk bug fixes. Feel free to suggest PRs to include in a bugfix release here.

The next feature release will likely go out in the second half of May or early June.

[hauntsaninja]

#12685 is a crash that bisects to #12656 (Fix types of inherited attributes in generic dataclasses).
Edit: now has a minimal repro

[cdce8p]

Noticed #12548 wasn't included. It fixed a crash with ParamSpec

• "Parameters cannot be constrained" with generic ParamSpec #12542

[JukkaL]

I'm going to skip the bugfix release, since 0.960 should be out soon.

[intgr]

I noticed that the "News" section on the official website http://mypy-lang.org/news.html is missing the 0.950 release (related: #12811)

[JukkaL]

@intgr Oops, I will update the the website when 0.960 goes out (should happen any time now!).