4

So I have two 2D numpy arrays, data containing the actual sample data:

[[12 15  5  0]
 [ 3 11  3  7]
 [ 9  3  5  2]
 [ 4  7  6  8]]

and the other, location, containing a map, another 2D array of unique, non-overlapping int values corresponding to the spaces in a new 1D array of the same size as the two:

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

The only way I've been able to run the transfer so far is with a simple for loop:

arr = np.zeros(4*4, dtype = int)

for i in range(4):
    for j in range(4):
        mapval = location[i, j]
        arr[mapval] = data[i, j]

Which does correctly output [ 4 7 3 9 3 12 15 11 3 5 0 7 2 5 6 8]

This is fine with the simple 4*4 array, but the actual dataset clocks in at 512*512, and this method takes rather very long to complete. So my question is, are there any functions, or methods exploiting ufuncs/numpy's fast processing capabilities to get this done more efficiently?

Improve this question
3
  • Does your map describe a Hilbert curve ? Commented Jul 9, 2017 at 16:20
  • Yes, it does actually! Commented Jul 9, 2017 at 16:21
  • How about simply the assignment arr[location] = data after the zeros initialization? Commented Jul 9, 2017 at 16:43

1 Answer 1

Reset to default
2

You need the index that sorts the location array to reorder the data array, which can be calculated using argsort:

data.ravel()[location.ravel().argsort()]
# array([ 4,  7,  3,  9,  3, 12, 15, 11,  3,  5,  0,  7,  2,  5,  6,  8])

import numpy as np
data = np.array([[12, 15,  5,  0],
 [ 3, 11,  3,  7],
 [ 9,  3,  5,  2],
 [ 4,  7,  6,  8]])

location = np.array([[ 5,  6,  9, 10],
 [ 4,  7,  8, 11],
 [ 3,  2, 13, 12],
 [ 0,  1, 14, 15]])
Improve this answer
Sign up to request clarification or add additional context in comments.

1 Comment

Haven't encountered the ravel method before, but this works great, and only more reason now to dive into the documentation. Thanks!

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.