Allow @final on TypedDict
#13557
Allow @final on TypedDict
#13557
Conversation
This comment has been minimized.
This comment has been minimized.
| if isinstance(decorator, RefExpr): | ||
| if decorator.fullname in FINAL_DECORATOR_NAMES: | ||
| self.fail("@final cannot be used with TypedDict", decorator) | ||
| if decorator.fullname in FINAL_DECORATOR_NAMES and info is not None: |
There was a problem hiding this comment.
What if the info is None? Does that happen if there's a circular reference? It would be good to get a test covering that code path.
There was a problem hiding this comment.
It could be FakeInfo for some reason.
There was a problem hiding this comment.
What if the info is None? Does that happen if there's a circular reference?
When there's a forward declaration in the TypedDict body, analyze_typeddict_classdef gets called multiple times. info is initially None in that case, but eventually it's a TypeInfo.
It would be good to get a test covering that code path.
Done.
It could be
FakeInfofor some reason.
When would that happen?
|
Also the CI failures look like there are some duplicate error messages. |
Allow a `TypedDict` to be decorated with `@final`. Like a regular class, mypy will emit an error if a final `TypedDict` is subclassed. Relates-to: python#7981
godlygeek
force-pushed
the
final_typeddicts
branch
from
3c0a1ed to
623a332
Compare
That's what I get for only running the |
This comment has been minimized.
This comment has been minimized.
|
Huh. This seems to be causing a mypyc segfault, judging by The mypyc tests ( |
|
The segfaults may be unrelated, we're seeing them on a bunch of PRs. Not sure what's causing them :( |
hauntsaninja
left a comment
hauntsaninja
left a comment
There was a problem hiding this comment.
This looks good to me!
Although it's a little surprising to me that we don't have the correct MRO (in which case we could just fallback to the existing final logic in checker.py). Not familiar enough with mypy's TypedDict impl to know if that's intentional.
|
According to mypy_primer, this change has no effect on the checked open source code. 🤖🎉 |
4 participants
Allow a
TypedDictto be decorated with@final. Like a regular class, mypy will emit an error if a finalTypedDictis subclassed.Relates-to: #7981
Description
Allow
@finalto be applied to aTypedDict, and have mypy emit an error if class is derived from a finalTypedDict. This goes some way towards closing #7981 and closing a feature gap with pyright, though not the whole way, as #7981 also asks for additional type narrowing for a finalTypedDict.Test Plan
testCannotUseFinalDecoratorWithTypedDictwas removed, andtestCannotSubclassFinalTypedDictadded.