scipy.stats.qmc.scale does not accept numpy array with typed shape as input when type checked with pyright

[jenshnielsen]

The following example type checks with sample being a numpy array without an explicit shape type
but fails to type check with a given shape

numpy 2.1.3
pyright 1.1.388
python 3.12.7

from typing import Literal

import numpy as np
from scipy.stats import qmc

l_bounds, u_bounds = 0, 1


# sample = np.random.rand(10, 10)  # works
# sample: np.ndarray[tuple[int, int], np.dtype[np.float64]] = np.random.rand(10, 10) # fails
sample: np.ndarray[tuple[Literal[10], Literal[10]], np.dtype[np.float64]] = (  # fails
    np.random.rand(10, 10)
)


setpoint_values = qmc.scale(
    sample,
    l_bounds=l_bounds,
    u_bounds=u_bounds,
)

The error returned is

d:\source\repos\debug_typing.py
  d:\source\repos\debug_typing.py:16:5 - error: Argument of type "NDArray[float64]" cannot be assigned to parameter "sample" of type "_Any2D_f" in function "scale"
    Type "NDArray[float64]" is not assignable to type "_Any2D_f"
      "ndarray[tuple[Literal[10], Literal[10]], dtype[float64]]" is incompatible with protocol "_CanLenArray[floating[Any]]"
        "__array__" is an incompatible type
          One or more overloads of "__array__" is not assignable
            No overloaded function matches type "(dtype: _DT@__array__, /) -> ndarray[_ShapeT_co@__array__, _DT@__array__]"
      "ndarray[tuple[Literal[10], Literal[10]], dtype[float64]]" is not assignable to "Sequence[Sequence[float | floating[Any]]]"
      "ndarray[tuple[Literal[10], Literal[10]], dtype[float64]]" is not assignable to "Sequence[_Any1D_f]" (reportArgumentType)
1 error, 0 warnings, 0 informations

[jorenham]

I can reproduce this 👌🏻

from typing import Any

import numpy as np
from scipy.stats import qmc

_f8_xd: np.ndarray[Any, np.dtype[np.float64]]
qmc.scale(_f8_xd, 0, 1)  # ok

_f8_nd: np.ndarray[tuple[int, ...], np.dtype[np.float64]]
qmc.scale(_f8_nd, 0, 1)  # E: reportArgumentType

[jorenham]

It turns out that this is caused by a bug in optype.numpy.CanArray when numpy>=2.1.

Luckily, working around this is easy.