How do you rotate the numbers in an numpy array of shape (n,) or (n,1)?

Question

Say I have a numpy array:

>>> a 
array([0,1,2,3,4])

and I want to "rotate" it to get:

>>> b
array([4,0,1,2,3])

What is the best way?

I have been converting to a deque and back (see below) but is there a better way?

b = deque(a)
b.rotate(1)
b = np.array(b)

@askewchan, I think that (n,) and (n,1) arrays [look the same](http://stackoverflow.com/questions/7635237/numpy-syntax-idiom-to-cast-n-array-to-a-n-1-array). Am I wrong? — Lee, Apr 15 '13 at 10:21
You're right that they look the same, and even behave the same in many circumstances, including the `roll` function, but be careful in some cases where `ndim` might matter (for `a.shape` is `(n,)`, `a.ndim` is `1`; but for shape `(n,1)`, `a.ndim` is 2). As you can see from the question you linked to, an axis must be added to get from the 1d to 2d case. — askewchan, Apr 15 '13 at 12:51
Fair point. In hindsight, to be consistent with the question title, I should have shown `a` with an (n,1) shape as `array([[0],[1],[2],[3],[4]])`. However, each of the answers does work for both (n,) and (n,1) shapes. — Lee, Apr 15 '13 at 14:50
Yes, `roll` will be effectively be applied along the non-1 axis if there is only one, which is why my comment was nothing beyond pedantry :). But if your array is `(n,m)` (or higher) it will roll along all the axes (the flattened array) which might give unexpected results. The solution there is to just do `np.roll(a, axis=0)` — askewchan, Apr 15 '13 at 14:54

score 12 · Accepted Answer · edited May 23 '17 at 12:07

12

Just use the numpy.roll function:

a = np.array([0,1,2,3,4])
b = np.roll(a,1)
print(b)
>>> [4 0 1 2 3]

edited May 23 '17 at 12:07

Community

answered Apr 12 '13 at 11:28

score 2 · Answer 2 · answered Apr 12 '13 at 11:19

2

numpy.concatenate([a[-1:], a[:-1]])
>>> array([4, 0, 1, 2, 3])

answered Apr 12 '13 at 11:19

Jakub M.

score 1 · Answer 3 · answered Apr 12 '13 at 11:07

1

Try this one

b = a[-1:]+a[:-1]

answered Apr 12 '13 at 11:07

Jan Vorcak

Thanks. Note that I am using numpy arrays so you have to convert to list first... `b=list(a)[-1:]+list(a)[:-1]` – Lee Apr 12 '13 at 11:12
... and I want a numpy array out so `b=np.array(list(a)[-1:]+list(a)[:-1])` – Lee Apr 12 '13 at 11:14

3 Answers3