TypeVar yields an error when used with reduce

[egeres]

👉🏻 Describe the bug
When analyzing the code from below, I get:

Expression of type "list[int]" cannot be assigned to return type "T@foo"
  Type "list[int]" cannot be assigned to type "T@foo"Pylance[reportGeneralTypeIssues](...

This seems to happen because, the output of reduce is correctly parsed as a list[int], but then that type would be different from list[T], even if T is bound to list[int]

👉🏻 To Reproduce
Put the code below in VSCode + pylance version v2023.7.20

👉🏻 Expected behavior
If a TypeVar is bound to a type (even if it's a composition using Union etc...), Pylance should consider another type valid if it matches with the bound of such TypeVar

👉🏻 Code

from functools import reduce
from typing import TypeVar

T = TypeVar("T", bound=list[int])

def foo(x: list[T]) -> T:
    return reduce(lambda a, b: a + b, x)

👉🏻 VS Code extension or command-line
Running pylance in VSCode as an extension!

[erictraut]

Pyright is correct to emit an error here. You're using a TypeVar that is bound to list[int]. Here's the applicable overload for reduce:

def reduce(function: Callable[[_T, _T], _T], sequence: Iterable[_T]) -> _T:
    ...

That means the two inputs to the callable must be the same type as the output. In your case, the inputs are type T, but the output is type list[int]. The expression a + b in the lambda maps to the method __add__ in the list class. This is defined to always return another list, not a Self. That means if T happens to be a subclass of list, the a + b operation will not produce the same subtype.

Not surprisingly, mypy generates the same error for this code.

Here's how to adjust your code so it type checks without error:

T = TypeVar("T")

def foo(x: Iterable[list[T]]) -> list[T]:
    return reduce(lambda a, b: a + b, x)

[egeres]

I'm sorry I didn't see this and reported a false positive!! 🙇🏻‍♂️