How to change several nested numpy arrays into one multidimensional array?

Question

I have a numpy a = np.load('test.npy') file with these nested numpy array:

In [21]: a.shape
Out[21]: (6886, 3)

In [22]: a[0].shape
Out[22]: (3,)

In [23]: a[0][0].shape
Out[23]: (787, 6)

Is there a simple way to change a to be a 4 dimensional array with shape: (6886, 3, 787, 6)?

Looks like a.dtype is object. Are you sure all elements have the same shape (787,6)? If not the task is impossible. If they do, you can use np.stack (but have to first reshape the outer array). — hpaulj
– hpaulj, Commented Jan 23, 2019 at 17:44
You are correct. The problem is that not all arrays have the same (786,6) shape. How to add np.nan values for arrays that are short: e.g. (506,6) so that they all become same shape: (787,6)? — A.Razavi
– A.Razavi, Commented Jan 23, 2019 at 17:57

tnknepp · Accepted Answer · 2019-01-23 21:19:20Z

1

I would hate to do it this way, but all that comes to mind is making a second array of the desired shape and slice your data into it. I have to admit that I am having difficulty understanding the shapes of each sub-array...it seems counter intuitive. Anyway, this solution will be slow, but you can do it once and save the array and never do it again.

import numpy as np

a = np.load('test.npy')
b = np.full((6886, 3, 787, 6), np.nan)

for row in range(6886):
    for col in range(3):
        tmp = a[row][col]
        b[row, col, :tmp.shape[0], :tmp.shape[1]] = tmp

Does this make sense/work?

answered Jan 23, 2019 at 21:19

tnknepp

6,3037 gold badges48 silver badges68 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

A.Razavi Over a year ago

Yes, this is what I wanted to do, but I was hoping to find a way to avoid python for loops. Thanks!

Collectives™ on Stack Overflow

How to change several nested numpy arrays into one multidimensional array?

1 Answer 1

1 Comment

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

1 Comment

Your Answer

Sign up or log in

Post as a guest

Related