Mypy reports error for different argument, than the one which is actually wrong

[jan-spurny]

Bug Report

When using generic Callables as parameters, the error is correctly reported, but it's reported for different argument than the one which is in fact wrong.

To Reproduce

from typing import Callable, TypeVar, List

T1 = TypeVar('T1')
T2 = TypeVar('T2')

def bigfunc(
    a: int,
    b: T1,
    c: Callable[[T1], T2],
    d: Callable[[T2], str],
) -> None:
    res = d(c(b))
    print(f'{a=}, {b=}, {res=}')


# should be:
#   c_func(x: float) -> List[float]:
def c_func(x: float) -> None:
    pass

def d_func(y: List[float]) -> str:
    return str(y)

bigfunc(1, 0.1, c_func, d_func)

Expected Behavior

I would expect mypy to complain about 3rd argument (c - c_func).

Actual Behavior

Mypy complains about 1st argument, which is 100% correct (a: int)

a.py:24: error: Cannot infer type argument 1 of "bigfunc"  [misc]
Found 1 error in 1 file (checked 1 source file)

Your Environment

• Mypy version used: 1.16.0
• Mypy command-line flags: None
• Mypy configuration options from mypy.ini (and other config files): None
• Python version used: 3.11.2

[brianschubert]

To clarify, it's complaining about the first type argument, not the type of the first argument. bigfunc takes two type arguments, one for each of T1 and T2. The error is stating that it failed to infer a value for the first one, T1.

However, that's still wrong. It actually infers T1 fine and wittles the signature of bigfunc down to

def [T2] (a: builtins.int, b: builtins.float, c: def (builtins.float) -> T2, d: def (T2) -> builtins.str)

mypy then fails to infer a value for T2, which it sees as being the first type parameter of the refined signature.

[sterliakov]

Can we just use tvar name instead of index? Names don't change when some generics are substituted, and are no less helpful for identification.

[jan-spurny]

@sterliakov that would be great!