gh-112509: Fix keys being present in both required_keys and optional_keys in TypedDict

[JelleZijlstra]

• Issue: TypedDict: __required_keys__ and __optional_keys__ can be wrong in the presence of inheritance #112509

[alicederyn]

I don't think this change is required; see #112509 (comment)

[AlexWaygood]

Thanks, I agree that this is a problem that's worth fixing.

I don't think this change is required; see #112509 (comment)

I actually think whether or not static type checkers should consider this behaviour illegal according to PEP-589 has little bearing on what we should do at runtime here. I don't want us to start raising a TypeError if a TypedDict incompatibly overrides a specific key to make it Required when the subclass specified the same key as being NotRequired. Doing that now would break backwards compatibility at runtime, and I don't think there's a strong motivation for doing so: it's the job of type checkers to catch this kind of error and emit warnings about it. So, if we've ruled out raising a TypeError in this kind of situation, we just have to do the thing at runtime that makes the most sense -- and the status quo doesn't really make any sense, I don't think :)

[AlexWaygood]

Is it correct that the last base should win? Doesn't that go against how multiple inheritance in Python usually works, where earlier bases in the __bases__ tuple generally have priority?

Python 3.13.0a2+ (heads/main:e9d1360c9a, Nov 24 2023, 11:23:45) [MSC v.1932 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license" for more information.
>>> class Foo:
...     x = 1
...
>>> class Bar:
...     x = 2
...
>>> class Baz(Foo, Bar): pass
...
>>> Baz.x
1

[JelleZijlstra]

It's arguably not correct as you say, but it's the current behavior for __annotations__ and changing that behavior seems difficult.

>>> class A(TypedDict):
...     a: int
... 
>>> class B(TypedDict):
...     a: str
... 
>>> class C(A, B): pass
... 
>>> C.__annotations__
{'a': <class 'str'>}

[AlexWaygood]

Oof, that seems unfortunate. As you say, though, better to be internally consistent for now, I guess!

[miss-islington-app]

Thanks @JelleZijlstra for the PR 🌮🎉.. I'm working now to backport this PR to: 3.11, 3.12.
🐍🍒⛏🤖

[bedevere-app]

GH-112530 is a backport of this pull request to the 3.12 branch.

[bedevere-app]

GH-112531 is a backport of this pull request to the 3.11 branch.