Diff from mypy_primer, showing the effect of this PR on open source code:

pandera (https://github.com/pandera-dev/pandera)
+ pandera/checks.py:74: error: Argument 1 to "ChainMap" has incompatible type "MappingProxyType[str, Any]"; expected "MutableMapping[str, Callable[..., Any]]"  [arg-type]
+ pandera/model.py:362: error: Need type annotation for "attrs" (hint: "attrs: Dict[<type>, <type>] = ...")  [var-annotated]

pydantic (https://github.com/samuelcolvin/pydantic)
+ pydantic/class_validators.py:332: error: Need type annotation for "all_attributes"  [var-annotated]
+ pydantic/class_validators.py:332: error: Argument 1 to "ChainMap" has incompatible type "*List[MappingProxyType[str, Any]]"; expected "MutableMapping[<nothing>, <nothing>]"  [arg-type]

Diff from mypy_primer, showing the effect of this PR on open source code:

pandera (https://github.com/pandera-dev/pandera)
+ pandera/checks.py:74: error: Argument 1 to "ChainMap" has incompatible type "MappingProxyType[str, Any]"; expected "MutableMapping[str, Callable[..., Any]]"  [arg-type]
+ pandera/model.py:362: error: Need type annotation for "attrs" (hint: "attrs: Dict[<type>, <type>] = ...")  [var-annotated]

pydantic (https://github.com/samuelcolvin/pydantic)
+ pydantic/class_validators.py:332: error: Need type annotation for "all_attributes"  [var-annotated]
+ pydantic/class_validators.py:332: error: Argument 1 to "ChainMap" has incompatible type "*List[MappingProxyType[str, Any]]"; expected "MutableMapping[<nothing>, <nothing>]"  [arg-type]

Unfortunate. I think this has more to do with the stub for ChainMap (specifically, it's hard for type checkers to reason about the class given that the maps could be mutable or immutable) than the proposed changes to type.__dict__ in this PR. Nevertheless, maybe it would be better to go back to using

@property
def __dict__(self) -> dict[str, Any]

for type.__dict__.

What do you think, @BvB93?

In my opinion this is more of a problem with ChainMap potentially being too strict (well, this depends a bit on whether or not one is interested in mutating the ChainMap I suppose...), rather than an issue with type.__dict__ itself.

While the discussion on whether ChainMap should or should not accept immutable mappings is a bit more philosophical in nature, setting the return type of type.__dict__ to a dict[str, Any] on the other end is just straight up incorrect based on runtime behavior, especially since types.MappingProxy is neither a sub- nor a supertype of dict.

I suppose this was discussed in depth in:

• collections.ChainMap has wrong type hint for self.maps, rendering it "immutable" #6042
• change chained maps to MutableMapping; fix #6042 #6044

and the decision then was that this kind of mypy_primer hit was acceptable.