Closed
Description
It would be nice to have an alternative to numpy.random.shuffle that accepts an axis argument, and that independently shuffles the one-dimensional slices. Here's an implementation that I'll call disarrange. It works, but it would be nice to have a more efficient C implementation.
def disarrange(a, axis=-1):
"""
Shuffle `a` in-place along the given axis.
Apply numpy.random.shuffle to the given axis of `a`.
Each one-dimensional slice is shuffled independently.
"""
b = a.swapaxes(axis, -1)
# Shuffle `b` in-place along the last axis. `b` is a view of `a`,
# so `a` is shuffled in place, too.
shp = b.shape[:-1]
for ndx in np.ndindex(shp):
np.random.shuffle(b[ndx])
return
Example:
In [156]: a = np.arange(20).reshape(4,5)
In [157]: a
Out[157]:
array([[ 0, 1, 2, 3, 4],
[ 5, 6, 7, 8, 9],
[10, 11, 12, 13, 14],
[15, 16, 17, 18, 19]])
In [158]: disarrange(a, axis=-1)
In [159]: a
Out[159]:
array([[ 2, 0, 4, 3, 1],
[ 8, 6, 7, 9, 5],
[11, 14, 13, 10, 12],
[19, 18, 16, 17, 15]])
In [160]: a = np.arange(20).reshape(4,5)
In [161]: disarrange(a, axis=0)
In [162]: a
Out[162]:
array([[ 5, 11, 7, 13, 14],
[ 0, 6, 2, 3, 4],
[10, 1, 17, 18, 19],
[15, 16, 12, 8, 9]])
This request was motivated by this question on stackoverflow: http://stackoverflow.com/questions/26310346/quickly-calculate-randomized-3d-numpy-array-from-2d-numpy-array/