Do not allow type[] to contain Literal types
#18276
Conversation
for more information, see https://pre-commit.ci
| @@ -81,14 +82,14 @@ def is_bad_type_type_item(item: Type) -> bool: | |||
| Such types are explicitly prohibited by PEP 484. Also, they cause problems | |||
| with recursive types like T = Type[T], because internal representation of | |||
| TypeType item is normalized (i.e. always a proper type). | |||
|
|
|||
| Also forbids `Type[Literal[...]]`, because typing spec does not allow it. | |||
There was a problem hiding this comment.
I think it might be better to frame this in terms of what is allowed. Here are some other things mypy allows that I don't think make sense: https://mypy-play.net/?mypy=latest&python=3.12&gist=aa900cb6c86fd55901b21ee182aa35ec
There was a problem hiding this comment.
I suspected that this will be brought up :)
I would prefer to keep this PR focused on just one thing. Because, for example, I only agree with the first example: type[TypeGuard[int]], but it is a completely different thing to fix.
There was a problem hiding this comment.
I think that we can clarify this in https://github.com/python/typing/tree/main/docs/spec
There was a problem hiding this comment.
This is already specified in the grammar at https://typing.readthedocs.io/en/latest/spec/annotations.html#type-and-annotation-expressions . (Though I noticed we're missing unions there, which should be supported.)
There was a problem hiding this comment.
I'm fine with deferring other similar changes to other PRs, though.
This comment has been minimized.
This comment has been minimized.
for more information, see https://pre-commit.ci
|
Diff from mypy_primer, showing the effect of this PR on open source code: bokeh (https://github.com/bokeh/bokeh)
+ src/bokeh/core/property/enum.py: note: In function "__init__":
+ src/bokeh/core/property/enum.py:64:30: error: type[...] can't contain "Literal[...]" [valid-type]
+ src/bokeh/core/property/enum.py:70:36: error: type[...] can't contain "Literal[...]" [valid-type]
|
Closes #18196