`type[Literal['']]` does not accept `str` as the type

[sobolevn]

I am not even sure that type[Literal] is correct :)

from typing import Literal

def a(x: type[Literal['a']]):
    reveal_type(x)
    
a(str)  # Argument 1 to "a" has incompatible type "type[str]"; expected "type[Literal['a']]"  [arg-type]

https://mypy-play.net/?mypy=latest&python=3.12&gist=6b8a1d7e6f98c64e7f751c58d356bf16

[JelleZijlstra]

type[Literal[...]] is invalid and should be rejected.

[jorenham]

type[Literal[...]] is invalid and should be rejected.

>>> from typing import Literal
... class A(str):
...     def __new__(cls, /) -> Literal["a"]:
...         return super().__new__(cls, "a")
...      
>>> A()
'a'

[JelleZijlstra]

@jorenham I'm not sure what point you are making. type[Literal[...]] isn't meaningful under the typing specification. The upcoming TypeForm PEP will provide a way to express this type.

[jorenham]

Wow that was a quick response 😋.

A is a type that only produces Literal["a"], so it appears to be compatible with a type[Literal["a"]]. And as far as I can tell, this isn't (potentially) type-unsafe. So I guess that I'm trying to say that I don't see why it's necessarily "invalid".

[KotlinIsland]

of course type[Literal[...]] is valid? why wouldn't it be?

[sobolevn]

Isn't Literal[1] is just 1 and type(1) is int?

[JelleZijlstra]

type[] represents an instance of type. It doesn't work with arbitrary type annotation objects.

[jorenham]

Isn't Literal[1] is just 1 and type(1) is int?

I think it's more accurate to describe it as a refinement type, i.e. a T: int s.t. T in {1}

[jorenham]

type[] represents an instance of type

type[S | T] doesn't represent an instance of type, but it's a valid statement.

[JelleZijlstra]

It is syntactic sugar for type[S] | type[T].

[jorenham]

It is syntactic sugar for type[S] | type[T].

yes, which doesn't describe an instance of type