Narrowing IN with unions of TypedDict.

[JanEricNitschke]

I currently have a situation like below where i have multiple TypedDicts with (overlapping) keys of similar structure and i need to extract values from them. However i would like to avoid writing separate functions for each dict that do the same thing.

So i was trying to do something like this:

from typing import Literal, overload, TypedDict, LiteralString

class Test1(TypedDict)
    test1: str

class Test2(TypedDict):
    test2: str

@overload
def _get_actor_key(
    actor: Literal["1"],
    game_action: Test1,
) -> str:
    ...

@overload
def _get_actor_key(
    actor: Literal[""], game_action: Test2
) -> str:
    ...


def _get_actor_key(
    actor: Literal["1","2"],
    game_action: Test1 | Test2,
) -> str:
    actor_name = "test"+actor
    if actor_name in game_action:
        return game_action[actor_name]
    return ""

If i replace the type hints in the implementation with Any i get the desired behaviour of pyright rejecting invalid function calls like _get_action_key("1", Test2).

But i can not managed to type hint the body with anything except any without it getting rejected. If i do the below it complains that "test1" is not a key for Test2 and vice versa. Is there any reasonable chance that something like this could be added to pyright or is there any way i could type hint this better or is just using Any my best bet here?

[erictraut]

Type narrowing doesn't apply in this case. The type of game_action is Test1 | Test2. Its type cannot be further narrowed based on the condition actor_name in game_action.

I would implement this as follows:

def _get_actor_key(actor: Literal["1", "2"], game_action: Test1 | Test2) -> str:
    return game_action.get("test" + actor, "")