Creating new array from array indexing

Question

Suppose I have the following array:

print(my_array)
(array([[[[5,  7,  3,  1],
          [-6, 0, -8, -2],
          [ 9 -7, -5, -9],
          [-1,  6,  0,  1],
          [-7, -8 , -3, 4]]],

        [[[-1,  5, -2, 2],
          [4,  -3, -1, 2],
          [-9,  0,  7, 1],
          [-4,  6, -5, -8],
          [-7, -3,  0 , 1]]]]),

 array([[[[ 7, 9 , 4, -3 ],
          [-4, 7, -1, -9],
          [6,  0, -3,  -7],
          [ 1,  6,  9, -3],
          [-4, -1, -9 , -6]]],

        [[[ 0, 8,  2, 6],
          [4,  5,  1,  2],
          [3,  7,  5, 2],
          [6, -1,  9,  5],
          [ 0,  5, 7,  7]]]]))

Then I want to form four new arrays, where the first new array is created from first columns of all nested arrays in my_array, the second from second columns etc.. such that

A = array([5,-6,9,-1,-7,-1,4,-9,-4,-7,7,-4,6,1,-4,0,4,3,6,0])

And the second array formed by second columns of each nested, like so:

B = array([7,0,-7,6,-8,5,-3,0,6,-3,9,7,0,6,-1,8,5,7,-1,5])

How can I do this?

Do you mind using numpy arrays? If so, the solution is pretty straightforward. — liorr
– liorr, Commented Jun 3, 2020 at 0:08

liorr · Accepted Answer · 2020-06-03 00:49:22Z

1

import numpy as np

# Create example arrays
arr1 = np.random.randint(10,size=(5, 4))
arr2 = np.random.randint(10,size=(5, 4))
arr3 = np.random.randint(10,size=(5, 4))
arr4 = np.random.randint(10,size=(5, 4))

# Combine all the arrays to match the dimensions mentioned in the question
arr_comb = np.array([[arr1, arr2],[arr3, arr4]])
print("Old array:")
print(arr_comb)

new_column_length = np.shape(arr_comb)[0] * np.shape(arr_comb)[1] * np.shape(arr_comb)[2]
# Reshape the array into a new array. The columns of the new array are what you requested
new_comb_arr = arr_comb.reshape(new_column_length,-1)

# e.g
print("First column of new array")
new_comb_arr[:,0]

Which results in:

Old array:
[[[[2 7 0 0]
   [0 1 7 3]
   [5 7 4 2]
   [4 5 3 6]
   [4 8 2 0]]

  [[4 6 3 4]
   [6 5 3 7]
   [6 2 7 4]
   [9 1 8 2]
   [3 9 2 2]]]


 [[[6 8 3 4]
   [0 6 6 6]
   [1 8 1 2]
   [4 2 6 4]
   [7 8 0 2]]

  [[2 6 4 1]
   [4 7 1 4]
   [8 0 5 6]
   [4 7 4 7]
   [8 9 4 5]]]]
First column of new array
array([2, 0, 5, 4, 4, 4, 6, 6, 9, 3, 6, 0, 1, 4, 7, 2, 4, 8, 4, 8])

edited Jun 3, 2020 at 0:49

answered Jun 3, 2020 at 0:30

liorr

8005 silver badges22 bronze badges

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

5 Comments

arilwan Over a year ago

Is there a way to reshape the new array without explicitly stating total number of rows as in arr_comb.reshape(16,-1)?

liorr Over a year ago

Will your arrays always be square?

arilwan Over a year ago

It is not square, a (5,4) array.

liorr Over a year ago

Oh, you're right. My bad :). Will the size of your arrays always be the same?

liorr Over a year ago

Anyway, I have generalized my answer. Now it supports what you've asked for. But the arrays have to have the same size (not necessarily square).

Collectives™ on Stack Overflow

Creating new array from array indexing

1 Answer 1

5 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

5 Comments

Your Answer

Sign up or log in

Post as a guest

Related