Applying functions to generic functions doesn't seem to work properly

[ecprice]

from typing import Function, typevar

A = typevar('A')
B = typevar('B')

def square(n: int) -> int:
    return n*n

def id(f: Function[[A], B]) -> Function[[A], B]:
    return f

g = id(square)

print(g(1))

the internal representation of the type of g seems to be "def ('None') -> builtins.int*" which isn't valid.

[jswalden]

ExpressionChecker.infer_function_type_arguments decides for id(square), via ExpressionChecker.get_arg_infer_passes, that the function argument type should be resolved in the second pass. Or something along those lines. It then proceeds to use None as argument type, which "type" propagates via return value until the embedded argument type gets compared against int for the 1.

It's not immediately clear to me, looking at this code, why it shouldn't iteratively infer types until a fixed point is reached.

[jswalden]

Reduced further:

from typing import Function, typevar

A = typevar('A')

def voidify(n: int) -> None:
    pass

def identity(f: Function[[A], None]) -> Function[[A], None]:
    return f

identity(voidify)(3)

The return type doesn't matter, can be a typevar or a builtin type or None. What specifically matters is that an argument be a typevar.