Code marked as unreachable after numpy.cross()

[liam-clink]

Environment data

• Language Server version: v2022.8.50
• OS and version: Windows 10
• Python version (& distribution if applicable, e.g. Anaconda): 3.10

Code Snippet

def foo(vector1, vector2):
    perp = numpy.cross(vector1, vector2)
    normed_perp = perp/numpy.linalg.norm(perp) # This line and everything after is marked as unreachable

Logs

XXX

Unclear how to do this, I changed the settings.json as instructed and no additional output was generated.

[erictraut]

This appears to be a bug in the type stubs that ship with numpy. It includes the following (first) overload for the cross method:

@overload
def cross(
    a: _ArrayLikeBool_co,
    b: _ArrayLikeBool_co,
    axisa: int = ...,
    axisb: int = ...,
    axisc: int = ...,
    axis: None | int = ...,
) -> NoReturn: ...

If the provided arguments have an unknown type, this first overload will be selected, and a NoReturn return type will be assumed. That means the call will generate an exception or will never return control back to the caller, hence the grayed-out text.

You can work around the issue by providing appropriate type annotations for vector1 and vector2.

The real fix will need to be made by the maintainers of numpy. They should delete the first overload, the one that returns a NoReturn.

[liam-clink]

Ok I'll close the issue in that case

[Korijn]

Has a corresponding issue been opened with the numpy maintainers?

[jabarragann]

Code snippet to work around the problem

import numpy as np

#option 1
cross1 = lambda x,y:np.cross(x,y)
#option 2
def cross2(a:np.ndarray,b:np.ndarray)->np.ndarray:
    return np.cross(a,b)

def cross_test():

    y_ax1 = cross1(np.array([1, 0, 0]), np.array([0, 1, 0]))
    y_ax2 = cross2(np.array([1, 0, 0]), np.array([0, 1, 0]))
    print(y_ax1)
    return y_ax1,y_ax2

if __name__ == "__main__":
    cross_test()

[janosh]

@Korijn Yes, see numpy/numpy#22146.