CODEWITHSUNDEEP
pythonarraysnumpy
ktitimbo
ktitimbo

Reputation: 83

Numpy reorganize array

Let's say I have got the following numpy array

A = np.array([[0,1,2,3,4],[5,6,7,8,9],[10,11,12,13,14],[15,16,17,18,19],[20,21,22,23,24],[25,26,27,28,29]])

out[]: array([[ 0,  1,  2,  3,  4],
       [ 5,  6,  7,  8,  9],
       [10, 11, 12, 13, 14],
       [15, 16, 17, 18, 19],
       [20, 21, 22, 23, 24],
       [25, 26, 27, 28, 29]])

I would like to reorganize it in such a way that the output is

out[] : array([[0,5],
         [1,6],
         [2,7],
         [3,8],
         [4,9],
         [10,15],
         [11,16],
         [12,17],
         [13,18],
         [14,19],
         [20,25],
         [21,26],
         ....,
         [24,29]])

I have been trying different combinations of np.reshape, tranpose, flatten, np.swapaxes, but with no success.

The real array has tens or sometimes hundreds of rows.

Originally, the data is given as DataFrame, but I realized that converting to numpy array could be a better alternative... Can it be done directly using pandas?

Upvotes: 2

Views: 118

Answers (1)

MSS
MSS

Reputation: 3623

Yes this can be done in this way using as_strided.

from numpy.lib.stride_tricks import as_strided

A = np.array([[0,1,2,3,4],[5,6,7,8,9],[10,11,12,13,14],[15,16,17,18,19],[20,21,22,23,24],[25,26,27,28,29]])
A = A.reshape(3, 2, 5)
#print(A)
S = A.itemsize
out = as_strided(A, shape=(3,5,2), strides=(2*A.shape[1]*S ,S, A.shape[1]*S)).copy()
out = out.reshape(15,2)
print(out)
>> [[ 0  5]
 [ 1  6]
 [ 2  7]
 [ 3  8]
 [ 4  9]
 [10 15]
 [11 16]
 [12 17]
 [13 18]
 [14 19]
 [20 25]
 [21 26]
 [22 27]
 [23 28]
 [24 29]]

Upvotes: 1

Related Questions