Let's say I have a multidimensional array with a shape that I don't know until runtime.
How can I reverse it along a given axis k, also not known until runtime?
The notation somearray[:,:,::-1,:,:] relies on static dimension references, as in this other SO question, so I can't use it here.
You can either construct a tuple of slice objects such as @ali_m suggests, or do something like this:
reversed_arr = np.swapaxes(np.swapaxes(arr, 0, k)[::-1], 0, k)
This places the desired axis at the front of the shape tuple, then reverses that first axis, and then returns it to its original position.
Some people think this approach lacks readability, but I disagree.
axis.I would use a tuple of slice objects for this:
def reversedim(M,k=0):
idx = tuple((slice(None,None,-1) if ii == k else slice(None)
for ii in xrange(M.ndim)))
return M[idx]
