Release 1.7 planning

[JukkaL]

I'm planning to make a 1.7 release around Nov 2 to Nov 6 (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'll cut the release branch around Oct 30.

This release will be the first one where we create the release blog post in the repo (either in CHANGELOG.md or in a temporary file which we'll add to CHANGELOG.md at the time of the release).

Here are some other major changes I'm hoping to include:

• Enable --new-type-inference by default (subject to no major known regressions).
• Enable TypeVarTuple by default.

[JukkaL]

I've encountered these (minor) regressions or potential regressions so far:

• Regression: inferred type too wide when using type variable bound #16331
• Regression: false positive about overload item never being matched #16338
• Changed strict-equality behavior with dict/Mapping #16340

[KRRT7]

i'm not sure if this the right place to mention this but, you might need to add setuptools as a dependency for python 3.12+

[AlexWaygood]

i'm not sure if this the right place to mention this but, you might need to add setuptools as a dependency for python 3.12+

Already done in ff7ac75

[ilevkivskyi]

It would be great to have:

• Fix dmypy inspect on Windows #16355
• Fix dmypy inspect for namespace packages #16357
• Fix file reloading in dmypy with --export-types #16359

These are all fixes for dmypy inspect, they should be all low-risk, as they don't touch "regular" daemon logic.

[JukkaL]

#15913 please

This was merged a while ago, so it will be included.

[JukkaL]

I've now cut the release branch: https://github.com/python/mypy/tree/release-1.7

We can still cherry-pick important/low-risk fixes and documentation updates.

[ilevkivskyi]

May be good to add #16381 (Unpack crash fix)

[JukkaL]

Cherry-picked these PRs:

• Fix crash on unpack call special-casing #16381
• Fix file reloading in dmypy with --export-types #16359

[hauntsaninja]

I wouldn't mind a cherry pick of #16407

[JukkaL]

I wrote 1.7 release notes in the changelog, which will be copied to the release blog post: https://github.com/python/mypy/blob/master/CHANGELOG.md

Please review. I'll leave a comment here when I've copied it to the blog, after which we shouldn't make any major changes to the changelog.

The release will probably go out tomorrow, 10 Nov, if I don't get distracted.

[ilevkivskyi]

@JukkaL Two comments on the blog post:

• I think it is worth mentioning new --enable-incomplete-feature=PreciseTupleTypes. And maybe also in general mention that from now on we will try to use --enable-incomplete-feature as a more systematic way to try new features. Note there are some docs https://mypy.readthedocs.io/en/latest/command_line.html#enabling-incomplete-experimental-features, feel free to edit them if you want.
• You refer to PEP 692 in description of TypedDict unpack. I think it may be worth mentioning that there are some intentional deviations from the PEP. The author of the PEP actually agrees it should be amended, but PEP is already accepted, so I don't know how they will sort it out. The summary is that unlike in the current edition of the PEP mypy always treats signatures with Unpack[SomeTypedDict] as equivalent to their expanded forms.

[ilevkivskyi]

Another thing is that there are two crashes with TypeVarTuple (one came with a PR, I will merge it soon). Not sure what to do with this:

• Trying to define a tuple with at least 1 element causes a crash if you try to define a tuple with *exactly* 1 element #16425
• Overloaded def including TypeVarTuple and variable length Tuple crash #16427

[ilevkivskyi]

FWIW the fix for the second crash is very low risk: it is just an obvious typo that I forgot to test: #16428

[ilevkivskyi]

Fix for the other crash looks also low-risk #16444

[JukkaL]

Cherry-picked these PRs:

• Fix type narrowing in lambda expressions #16407
• Handle TypeVarTupleType when checking overload constraints #16428

I will cherry-pick #16444 once it's merged.

[ilevkivskyi]

Merged #16444

[JukkaL]

Cherry-picked #16444.

[JukkaL]

Changelog updates: #16446 (I will leave it open for a while but will merge it fairly soon)

[JukkaL]

Mypy 1.7 is out: https://mypy-lang.blogspot.com/2023/11/mypy-17-released.html

[DanielNoord]

Congrats with the release guys!

Sadly, I think I found a regression: #16451. Let me know if I can provide more details.

[jessemyers-lettuce]

FYI, this release "regressed" pydantic usage. This snippet:

from pydantic import BaseModel

class Foo(BaseModel):
    pass

Now raises:

test.py:3: error: Explicit "Any" is not allowed  [misc]
    class Foo(BaseModel):
    ^~~~~~~~~~~~~~~~~~~~
test.py:3: error: Type of decorated function contains type "Any"
("Callable[[DefaultNamedArg(dict[str, Any] | None, '__pydantic_extra__'), DefaultNamedArg(set[str], '__pydantic_fields_set__'), DefaultNamedArg(dict[str, Any] | None, '__pydantic_private__')], None]")  [misc]
    class Foo(BaseModel):
    ^~~~~~~~~~~~~~~~~~~~
Found 2 errors in 1 file (checked 1 source file)

I'd guess that this behavior is correct... there is, in fact, an explicit Any in the BaseModel type definition, but it's going to mean a lot of code has to be updated to ignore this behavior.

[hauntsaninja]

@jessemyers-lettuce could you open a new issue with more details? I can't reproduce.

[jessemyers-lettuce]

@hauntsaninja Done. See #16454

[ilevkivskyi]

Possible candidates for 1.7.1:

• Fix crash on Callable self in __call__ #16453
• Special-case unions in polymorphic inference #16461
• Exclude private attributes from override checks #16464

[ilevkivskyi]

Two more potential candidates for 1.7.1:

• Fix missing meet case exposed by len narrowing #16470 (technically can be seen a regression)
• Ignore position if imprecise arguments are matched by name #16471 (less important, but one of the issues fixed is a regression in 1.6.0)

[ilevkivskyi]

Another candidate from me for 1.7.1:

• Fix crash on strict-equality with recursive types #16483

[JukkaL]

There may be a recent 1-2% perf regression, possibly related to some of the fixes that we are planning to include in 1.7.1. Perf data: https://github.com/mypyc/mypyc-benchmark-results/blob/master/reports/benchmarks/mypy_self_check.md

I'm planning to run some performance comparisons locally to get a better idea of whether there actually is a regression, since the automatically generated performance metrics are a bit noisy. The regression is small enough that even if it's real, we can probably proceed, but it seems worth it first check if there is some trivial fix.

[mr-c]

I nominate #16484 for 1.7.1

[JukkaL]

Cherry-picked these PRs for the point release:

• Fix crash on Callable self in __call__ #16453
• Special-case unions in polymorphic inference #16461
• Exclude private attributes from override checks #16464
• Fix missing meet case exposed by len narrowing #16470
• Ignore position if imprecise arguments are matched by name #16471
• Fix crash on strict-equality with recursive types #16483
• Fix polymorphic application for callback protocols #16514
• [mypyc] Fix regression with nested functions #16484

Thank you @ilevkivskyi for all the fixes!

I'm planning to publish 1.7.1 soon (either later today or tomorrow).

[JukkaL]

1.7.1 is out now.