What is the best way to perform an anti-transpose in python?

Question

Lets say I have an array

a = np.arange(16).reshape((4,4))

0   1  2   3
4   5  6   7
8   9  10  11
12  13 14  15

But I want

15  11  7  3
14  10  6  2
13  9   5  1
12  8   4  0

which is a flip across the secondary diagonal, or a kind of anti-transpose.

How can I do this in numpy?

Have you attempted anything or are you simply getting your homework done by us... — Vaibhav Bajaj, Jun 27 '17 at 04:58

Jaden Travnik · Accepted Answer · 2017-06-27T05:43:09.013

18

One could do one of the following:

rot90(a,2).T

rot90(flipud(a),1)

rot90(fliplr(a), -1)

or as hpaulj suggested in the comments (thanks hpaulj)

a[::-1,::-1].T

Here are the speed rankings as ratios of the slowest method after anti-transposing 1000 random 10000x10000 arrays.

edited Jun 27 '17 at 05:43

answered Jun 27 '17 at 04:50

Jaden Travnik

I'm hoping someone knows a secret numpy way of doing this. – Jaden Travnik Jun 27 '17 at 04:51
7

`a[::-1,::-1].T`; this is a `view`, sharing the same data buffer, so it will be as fast as possible. To test for view, try changing one element and see if it changes the original. The double reverse slice is the same as `np.rot90(a,2)`. Look at it's code. – hpaulj Jun 27 '17 at 04:56
@hpaulj so you should add that as an answer – Antti Haapala -- Слава Україні Jun 27 '17 at 05:10

score 2 · Answer 2 · answered Sep 29 '22 at 05:39

2

np.flip(a).T

From the np.flip documentation & @hpaulj 's comment:

flip(m) corresponds to m[::-1,::-1,...,::-1] with ::-1 at all positions.

answered Sep 29 '22 at 05:39

johnDanger

score 1 · Answer 3 · answered Jun 27 '17 at 06:39

1

Here's another to throw into the mix.

a.ravel('F')[::-1].reshape(a.shape)

answered Jun 27 '17 at 06:39

piRSquared

Although I like this because of ravel, its also about 200x slower than `rot90(fliplr(a),-1)`. :/ – Jaden Travnik Jun 27 '17 at 09:02
No worries, I was just exploring other ways of doing it (-: – piRSquared Jun 27 '17 at 16:05

score 0 · Answer 4 · answered Jun 27 '17 at 07:09

0

Try it in this manner,

np=np[::-1] #reverse the array
a = np.arange(16).reshape((4,4))

answered Jun 27 '17 at 07:09

Omi Harjani

4 Answers4