NPY001 should not replace np.bool by bool (at least in type hints)

[bersbersbers]

bug.py

import numpy as np
import numpy.typing as npt

bad1: npt.NDArray[np.bool_] = np.array([True])
good: npt.NDArray[np.bool] = np.array([True])
bad2: npt.NDArray[bool] = np.array([True])

For mypy v1.9.0, only np.bool is acceptable:

bug.py:4: error: Name "np.bool_" is not defined  [name-defined]
bug.py:6: error: Type argument "bool" of "NDArray" must be a subtype of "generic"  [type-var]
bug.py:6: note: See https://mypy.rtfd.io/en/stable/_refs.html#code-type-var for more info
Found 2 errors in 1 file (checked 1 source file)

However, NPY001 does not like that solution:

ruff check --isolated --select NPY001 bug.py

bug.py:5:19: NPY001 [*] Type alias `np.bool` is deprecated, replace with builtin type
Found 1 error.
[*] 1 fixable with the `--fix` option.

--fix gives

import numpy as np
import numpy.typing as npt

bad1: npt.NDArray[np.bool_] = np.array([True])
good: npt.NDArray[bool] = np.array([True])
bad2: npt.NDArray[bool] = np.array([True])

Keywords: numpy 2, deprecation, bool, bool_

ruff 0.4.1
mypy 1.9.0
numpy 2.0.0rc1

[bersbersbers]

See also numpy/numpy#25080

[charliermarsh]

Is this considered "recommended" by NumPy, or is it a bug or limitation in the type definitions?

[charliermarsh]

\cc @mtsokol if you know!

[bersbersbers]

Is this considered "recommended" by NumPy

This. See https://numpy.org/devdocs/release/2.0.0-notes.html#changes, 5th bullet in "Changes".

[tmke8]

So, it seems np.bool used to be an alias for built-in bool while np.bool_ was a separate type. The np.bool alias was then deprecated in numpy 1.20, which is what NPY001 is about. But then numpy 2.0 renamed np.bool_ to np.bool (while keeping np.bool_ as an alias for the new np.bool for backwards compatibility with numpy 1.x), so now the depreciation from numpy 1.20 conflicts with the new recommended name in numpy 2.0 for np.bool_.