How to sort a numpy array based on the values in a specific row?

Question

I was wondering how I would be able to sort a whole array by the values in one of its columns.

I have :

array([5,2,8,2,4])

and:

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]])

I want to append the first array to the second one like this:

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],
       [5,  2,  8,  2,  4]])

And then sort the array by the appended row to get either this:

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

or this:

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

And then remove the appended column to get:

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

or:

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

Is there a code to carry out this procedure. I am very new to python. Thanks a lot!

pezcode · Accepted Answer · 2014-06-20 16:28:17Z

19

You can use numpy.argsort to get a list with the sorted indices of your array. Using that you can then rearrange the columns of the matrix.

import numpy as np

c = np.array([5,2,8,2,4])    
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]])

i = np.argsort(c)
a = a[:,i]

answered Jun 20, 2014 at 16:28

pezcode

5,8292 gold badges26 silver badges41 bronze badges

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

3 Comments

tooty44 Over a year ago

Thanks, it works like magic! Would you happen to know how it works as I can't seem to figure it out from the code and documentation? Thanks!

pezcode Over a year ago

argsort returns an array of indices that would sort the array. The first element in i is the position in c with the lowest value, etc. You can use lists to index numpy arrays, so c[i] would give you the sorted version of c. Extending that to 2D arrays, you can use the list to select a bunch of rows or (in this case) columns. You want the columns in the order specified by i but you want all the rows untouched, hence : (that's a slice that covers all the rows). You should find a tutorial about slicing and indexing in numpy because it's really essential stuff.

fuenfundachtzig Over a year ago

I tried to achieve the same with a = np.take_along_axis(a, i, axis = 0) but this gives ValueError: indices and arr must have the same number of dimensions. What is the correct syntax?

will · Accepted Answer · 2014-06-20 16:27:51Z

You don't need numpy to do this; (although if you are using numpy, you can just use the .transpose() method of the array class.

What this essentially does, is transpose your array so that it's array[column][row], and then takes each columns, and pairs them with the sortKeys you provided in a list of tuples (the zip(sortKeys, a) bit). Then it sorts this list of tuples. By default, sort with sort tuples based on their first value, then their second value, then third, etc. Now you have the columns in order.

then aNew = [...] just extracts the columns and creates your new array, still in the array[column][row] format, and then transposes it again.

a = [[ 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]]

#transpose a
a = zip(*a)

sortKeys = [5,2,8,2,4]

b = zip(sortKeys, a)

aNew = [row[1] for row in sorted(b)]

#transpose a back
aNew = zip(*aNew)

print aNew

Collectives™ on Stack Overflow

How to sort a numpy array based on the values in a specific row?

2 Answers 2

3 Comments

Comments

Linked

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

3 Comments

Comments

Linked

Related