Examples fail mypy-checks in unexpected ways

[julian-goettingen]

What you are doing looks very cool and could be helpful to me and a lot of people.

However, the examples from the readme do not work for me (I have tried mypy with python3.10 and python3.11)

For example, with this one:

from typing import Any

import numpy as np

from phantom_tensors import parse
from phantom_tensors.numpy import NDArray
from phantom_tensors.alphabet import A, B  # these are just NewType(..., int) types

def func_on_2d(x: NDArray[Any, Any]): ...
def func_on_3d(x: NDArray[Any, Any, Any]): ...
def func_on_any_arr(x: np.ndarray): ...

# runtime: ensures shape of arr_3d matches (A, B, A) patterns
arr_3d = parse(np.ones((3, 5, 3)), NDArray[A, B, A])

func_on_2d(arr_3d)  # static type checker: Error!  # expects 2D arr, got 3D

func_on_3d(arr_3d)  # static type checker: OK
func_on_any_arr(arr_3d)  # static type checker: OK

I get the following errors:

phan_tensor_test.py:11: error: "NDArray" expects no type arguments, but 2 given  [type-arg]
phan_tensor_test.py:12: error: "NDArray" expects no type arguments, but 3 given  [type-arg]
phan_tensor_test.py:15: error: The type "Type[NDArray]" is not generic and not indexable  [misc]

I am not deep enough into python-typing to understand how this issue could be related to a misconfiguration. I know variadic generics are a very recent feature, which is why I tried to run it with python 3.11 but it still showed this error.

I am using the latest pypi-version of your library (I did not clone your repo)

[julian-goettingen]

(edited the title because the example is supposed to fail mypy, but on another line)

[rsokl]

mypy does not support PEP 646 yet, but pyright does. You can run pyright on the following examples to see that they do, indeed type-check as expected!

[julian-goettingen]

Thanks!

mypy is working on support: python/mypy#12280
I wont switch typechecker to use phantom_tensors, but I expect to start using it once mypy catches up.