Handling arrays with array.shape[i] == 0

[JesseLivezey]

When (accidentally) calling fft on an array with one dimension having shape=0, mkl_fft gives a somewhat cryptic error wheras numpy.fft.fft (with no mkl) returns a sensibly shaped array.

with numpy from conda (uses mkl_fft)

import numpy as np
from numpy.fft import fft

fft(np.zeros((10, 0, 5)))
# ValueError: Internal error occurred: b'Intel MKL DFTI ERROR: Inconsistent configuration parameters'

with numpy from pip

import numpy as np
from numpy.fft import fft

fft(np.zeros((10, 0, 5)))
# array([], shape=(10, 0, 5), dtype=complex128)

Conda list

# packages in environment at /home/jesse/anaconda3/envs/fft_test:
#
# Name                    Version                   Build  Channel
_libgcc_mutex             0.1                        main  
blas                      1.0                         mkl  
ca-certificates           2019.11.27                    0  
certifi                   2019.11.28               py38_0  
intel-openmp              2019.4                      243  
libedit                   3.1.20181209         hc058e9b_0  
libffi                    3.2.1                hd88cf55_4  
libgcc-ng                 9.1.0                hdf63c60_0  
libgfortran-ng            7.3.0                hdf63c60_0  
libstdcxx-ng              9.1.0                hdf63c60_0  
mkl                       2019.4                      243  
mkl-service               2.3.0            py38he904b0f_0  
mkl_fft                   1.0.15           py38ha843d7b_0  
mkl_random                1.1.0            py38h962f231_0  
ncurses                   6.1                  he6710b0_1  
numpy                     1.17.4           py38hc1035e2_0  
numpy-base                1.17.4           py38hde5b4d6_0  
openssl                   1.1.1d               h7b6447c_3  
pip                       19.3.1                   py38_0  
python                    3.8.0                h0371630_2  
readline                  7.0                  h7b6447c_5  
setuptools                44.0.0                   py38_0  
six                       1.13.0                   py38_0  
sqlite                    3.30.1               h7b6447c_0  
tk                        8.6.8                hbc83047_0  
wheel                     0.33.6                   py38_0  
xz                        5.2.4                h14c3975_4  
zlib                      1.2.11               h7b6447c_3  

Similar things happen for rfft, fft2, etc.

[oleksandr-pavlyk]

This has been addressed in commits of PR #45:

In [1]: import mkl_fft, numpy as np

In [2]: mkl_fft.fft(np.ones((10,0,5), dtype='d'))
Out[2]: array([], shape=(10, 0, 5), dtype=complex128)