Is is_typeddict with Generics supposed to eval to false?

[semohr]

Is is_typeddict supposed to return False for generic TypedDict specializations, or is this a bug? The documentation does not mention this.

Example:

from typing import is_typeddict, Generic, TypeVar, TypedDict

A = TypeVar("A", bound=TypedDict)

class Attributes(TypedDict, total=False):
    title: str

class Resource(TypedDict, Generic[A]):
    a: A 
    
print(is_typeddict(Resource[Attributes]))
# False
# Would have expected True...

Live demo: https://www.online-python.com/Pa6C0sWjIh

Hope this is the right place to report this.

[semohr]

@srittau Since you have turned into a discussion, do you have an opinion on this? My impression is that this is a bug, although I understand that properly validating specializations to ensure they are still TypedDicts could be quite a headache.

[davidhalter]

I personally somewhat agree, but I'm not sure the typing discussions are the right place for this. is_typeddict is mostly relevant for runtime checks and I would therefore recommend you to open an issue on the cpython issue tracker, where this could be fixed or closed.

[semohr]

FYI: I opened an issue in CPython’s issue tracker. The current behavior appears to be intentional and is unlikely to change.

[TayfurYldz]

This is expected for the specialized form, rather than a validation failure of the TypedDict itself.

Resource is the TypedDict class, while Resource[Attributes] is a generic alias whose origin is Resource. Current CPython documentation makes this distinction explicit: is_typeddict() returns true for TypedDict classes, but not for generic aliases of them (the docs use the analogous GenericFilm[int] example).

So these differ intentionally:

from typing import get_origin, is_typeddict

assert is_typeddict(Resource)
assert not is_typeddict(Resource[Attributes])

specialized = Resource[Attributes]
assert get_origin(specialized) is Resource
assert is_typeddict(get_origin(specialized))

If the goal is to recognize both an unspecialized TypedDict and a specialization, checking the origin is therefore a reasonable normalization step (while handling get_origin(tp) is None for ordinary classes).

The current docs are here: https://docs.python.org/3/library/typing.html#typing.is_typeddict

[semohr]

I opened the issue and pushed for documentation additions. The time I opened this issue/discussion it was not documented yet but thanks ;)