How to annotate kwargs with Unpack and TypedDict?

[93578237]

Consider this example:
foo should require either x or y is required, not both at the same time.

class Kwargs(TypedDict, total=False):
    x: Required[int]
    y: Required[int]
    z: NotRequired[str]


def foo(**kwargs: Unpack[Kwargs]):
   ....
   
foo(x=1) # should be ok
foo(y=1) # should be ok
foo(z=1) # raise error

[JelleZijlstra]

This isn't something you can do with a TypedDict, and I don't think the upcoming PEP for TypedDict + kwargs will support it. You should use overloads instead.

[93578237]

Hmm, but is it possible to combine overload with Unpack[TypedDict] for other NotRequired kwargs? Using only @overload makes lots of combinations.

[JelleZijlstra]

I feel like that should be OK. Perhaps try it with pyright, which already supports this feature.

[93578237]

like this?

class Kwargs(TypedDict, total=False):
    z: NotRequired[str]
    c: NotRequired[str]
    d: NotRequired[str]

class _X(Kwargs):
    x: Required[int]

class _Y(Kwargs):
    y: Required[int]


@overload
def foo(**kwargs: Unpack[_X]) -> None:
    ...


@overload
def foo(**kwargs: Unpack[_Y]) -> None:
    ...


def foo(**kwargs):
    print(kwargs)

[erictraut]

Yes, that's correct. This type checks as expected in pyright. In particular, it allows foo(x=1) and foo(y=1) but rejects foo(z=1).

[93578237]

Yes, pyright works, thx!