Stubs replace : Any | None with : Incomplete | None
#9565
Conversation
This comment has been minimized.
This comment has been minimized.
| @@ -191,7 +191,7 @@ class Values(Generative, FromClause): | |||
| name: Any | |||
| literal_binds: Any | |||
| def __init__(self, *columns, **kw) -> None: ... | |||
| def alias(self: Self, name: Any | None, **kw) -> Self: ... # type: ignore[override] | |||
| def alias(self: Self, name: Incomplete | None, **kw) -> Self: ... # type: ignore[override] | |||
There was a problem hiding this comment.
None of the ones in this file are stubgen artefacts; they were all added in https://github.com/python/typeshed/pull/7603/files. But, there's no source-code comment indicating why they used Any here (and the comments on the PR thread give no insight), so I think it's fine to change the Any uses in this file to Incomplete.
(Same comment applies to a few other files touched by this PR.)
|
|
||
| class _SupportsGraphident(Protocol): | ||
| graphident: str | ||
|
|
||
| # code, filename and packagepath are always initialized to None. But they can be given a value later. | ||
| class Node: | ||
| # Compiled code. See stdlib.builtins.compile | ||
| code: Any | None | ||
| code: Incomplete | None |
There was a problem hiding this comment.
You only added this one recently and, from the comment, it looks like it's meant to be Any ;)
There was a problem hiding this comment.
Definitely, I don't see anything else obvious after manually reviewing as well.
There was a problem hiding this comment.
I haven't checked the code but it looks like this should just be types.CodeType.
There was a problem hiding this comment.
This PR no longer touches this line, so I think that can be deferred to another PR
There was a problem hiding this comment.
It's the return of stdlib/builtins.pyi > compile, which is currently typed as Any, but also has this comment:
TODO:
compilehas a more precise return type in reality; work on a way of expressing that?
https://github.com/python/typeshed/blob/main/stdlib/builtins.pyi#L1234
There was a problem hiding this comment.
There was a problem hiding this comment.
Right, I'd be surprised if PyInstaller is using PyCF_ONLY_AST here.
There was a problem hiding this comment.
(technically it's modulegraph, pyinstaller vendors it) It is using PyCF_ONLY_AST as well, but it's the non-ast call that is stored.
| # TODO: change this to packaging.markers.Marker | None once we can import | ||
| # packaging.markers | ||
| marker: Any | None | ||
| marker: Incomplete | None |
There was a problem hiding this comment.
Oh, we should, you know, take action on this comment now (in a separate PR)
|
(The only actionable thing in my review is the |
|
According to mypy_primer, this change has no effect on the checked open source code. 🤖🎉 |
First commit does
: Any | None\nto: Incomplete | None\nSecond commit does
: Any | Noneto: Incomplete | NoneSplit them up for ease of review, just in case. With #9558 I think we can effectively say that all
: Any | Nonewere to be replaced.The only instances of
Any | Noneleft are ClassVar type parameter, return types and 2 Aliases because of name clash:tAny | Nonein stubs\protobuf\google\protobuf\internal\well_known_types.pyi_Any | Nonein stubs\SQLAlchemy\sqlalchemy\dialects\postgresql\array.pyiRef #9550