What are these ordering code doing

[deanhystad]

order = mothFit.argsort() creates an array or matrix the same size as mothFit which contains indices that will result in a sorted list. Since there is no "axis" argument specified, mothFit is sorted along the "last" axis. Axis 0 sorts the values in colomns (along the rows axis), Axis 1 sorts values in rows (along the columns axis). The "last" axis is Axis 1.

This is a complete, though short example.

import numpy as np
x = np.array([[1,6,3],[2,5,4]])
i = x.argsort()
y = np.take_along_axis(x, i, axis=1)
print('Sort each row\nIndices\n', i)
print('Sorted Values\n', y)

Output:
Sort each row
Indices
 [[0 2 1] # In first row, the middle number is moved to the end
 [0 2 1]] # Same for the second row
Sorted Values
 [[1 3 6] # Values in a row are sorted in increasing value
 [2 4 5]]

You can specify the sorting axis = 0 and sort the values in a column.

i = x.argsort(axis=0)
y = np.take_along_axis(x, i, axis=0)
print('\nSort each column\nIndices\n', i)
print('Sorted Values\n', y)

Output:
Sort each column
Indices
 [[0 1 0] # Need to flip middle column
 [1 0 1]]
Sorted Values
 [[1 5 3] # All columns sorted in increasing order
 [2 6 4]]

If you set axis = None, the matrix is sorted as a flat list.

i = x.argsort(axis=None)
y = np.take_along_axis(x, i, axis=None)
print('\nFlat sort\nIndices\n', i)
print('Sorted Values\n', y)

Output:
Flat sort
Indices
 [0 3 2 5 4 1]
Sorted Values
 [1 2 3 4 5 6]

Sorting the values to to still be in matrix form requires the additional step of splitting the array into two parts.

i = x.argsort(axis=None)
y = np.take_along_axis(x, i, axis=None)
z = np.array(np.array_split(y, 2))
print('\nFlat sort\nIndices\n', i)
print('Sorted Values\n', z)

Output:
Flat sort
Indices
 [0 3 2 5 4 1]
Sorted Values
 [[1 2 3]
 [4 5 6]]