This question is an extension of shift numpy array by row
If I shift (from scipy.ndimage.interpolation) using a test 3 x 5 x 5 array like so everything works as expected:
arr = np.ones([3,5,5])
array([[[ 1., 1., 1., 1., 1.],
[ 1., 1., 1., 1., 1.],
[ 1., 1., 1., 1., 1.],
[ 1., 1., 1., 1., 1.],
[ 1., 1., 1., 1., 1.]],
[[ 1., 1., 1., 1., 1.],
[ 1., 1., 1., 1., 1.],
[ 1., 1., 1., 1., 1.],
[ 1., 1., 1., 1., 1.],
[ 1., 1., 1., 1., 1.]],
[[ 1., 1., 1., 1., 1.],
[ 1., 1., 1., 1., 1.],
[ 1., 1., 1., 1., 1.],
[ 1., 1., 1., 1., 1.],
[ 1., 1., 1., 1., 1.]]])
shift(arr,(1,0,0), cval=np.nan)
array([[[ nan, nan, nan, nan, nan],
[ nan, nan, nan, nan, nan],
[ nan, nan, nan, nan, nan],
[ nan, nan, nan, nan, nan],
[ nan, nan, nan, nan, nan]],
[[ 1., 1., 1., 1., 1.],
[ 1., 1., 1., 1., 1.],
[ 1., 1., 1., 1., 1.],
[ 1., 1., 1., 1., 1.],
[ 1., 1., 1., 1., 1.]],
[[ 1., 1., 1., 1., 1.],
[ 1., 1., 1., 1., 1.],
[ 1., 1., 1., 1., 1.],
[ 1., 1., 1., 1., 1.],
[ 1., 1., 1., 1., 1.]]])
HOWEVER, if I perform the same shift on my 3 x 5 x 5 data array I get all np.nan values:
array([[[ 0. , nan, nan, nan, nan],
[ nan, 0. , nan, nan, -1.07346633],
[ nan, nan, 0. , nan, nan],
[ nan, nan, nan, 0. , nan],
[ nan, 1.07346633, nan, nan, 0. ]],
[[ 0. , nan, nan, nan, nan],
[ nan, 0. , nan, nan, nan],
[ nan, nan, 0. , -1.44470265, nan],
[ nan, nan, 1.44470265, 0. , nan],
[ nan, nan, nan, nan, 0. ]],
[[ 0. , nan, 1.80965682, nan, nan],
[ nan, 0. , nan, nan, nan],
[-1.80965682, nan, 0. , nan, nan],
[ nan, nan, nan, 0. , nan],
[ nan, nan, nan, nan, 0. ]]])
shift(stats1_arr,(1,0,0), cval=np.nan)
array([[[ nan, nan, nan, nan, nan],
[ nan, nan, nan, nan, nan],
[ nan, nan, nan, nan, nan],
[ nan, nan, nan, nan, nan],
[ nan, nan, nan, nan, nan]],
[[ nan, nan, nan, nan, nan],
[ nan, nan, nan, nan, nan],
[ nan, nan, nan, nan, nan],
[ nan, nan, nan, nan, nan],
[ nan, nan, nan, nan, nan]],
[[ nan, nan, nan, nan, nan],
[ nan, nan, nan, nan, nan],
[ nan, nan, nan, nan, nan],
[ nan, nan, nan, nan, nan],
[ nan, nan, nan, nan, nan]]])
Am I doing something wrong (mis-using shift?) or is this a bug? Seems like a bug in scipy.ndimage.interpolation.shift
It's not a bug. According to the docs, it's using spline interpolation of order 3 (by default), and your sparse matrix just ends up full of np.nan values because you can't really interpolate it.
You can essentially turn the interpolation 'feature' off by using order=0:
shift(stats1_arr, (1, 0, 0), cval=np.nan, order=0)
Which results in:
array([[[ nan, nan, nan, nan, nan],
[ nan, nan, nan, nan, nan],
[ nan, nan, nan, nan, nan],
[ nan, nan, nan, nan, nan],
[ nan, nan, nan, nan, nan]],
[[ 0. , nan, nan, nan, nan],
[ nan, 0. , nan, nan, -1.07347],
[ nan, nan, 0. , nan, nan],
[ nan, nan, nan, 0. , nan],
[ nan, 1.07347, nan, nan, 0. ]],
[[ 0. , nan, nan, nan, nan],
[ nan, 0. , nan, nan, nan],
[ nan, nan, 0. , -1.4447 , nan],
[ nan, nan, 1.4447 , 0. , nan],
[ nan, nan, nan, nan, 0. ]]])