Better solution when using guvectorize

[HannoSpreeuw]

The solution posted here :

@guvectorize([(complex128[:, :], int64, int64[:, :])],
             "(n,m),()->(n,m)",
             nopython=True)

uses 2D arrays - complex128[:,:] - while guvectorize will turn a function for 1D arrays into a function suitable for 2D arrays when a 2D argument is used in the call.

So the function and its decorator should be written as operating on 1D arrays.
Or written for 2D and use a 3D argument in the call.

See the example from the Numba documentation:

The nice thing is that NumPy will automatically dispatch over more complicated inputs, depending on their shapes: