Numpy recarray.argpartition() function | Python

In numpy, arrays may have a data-types containing fields, analogous to columns in a spreadsheet. An example is [(a, int), (b, float)], where each entry in the array is a pair of (int, float). Normally, these attributes are accessed using dictionary lookups such as arr['a'] and arr['b'].

Record arrays allow the fields to be accessed as members of the array, using arr.a and arr.b. numpy.recarray.argpartition() function returns the indices that would partition this array.

Syntax : numpy.recarray.argpartition(kth, axis=-1, kind='introselect', order=None)

Parameters:
kth : [int or sequence of ints ] Element index to partition by.
axis : [int or None] Axis along which to sort. If None, the array is flattened before sorting. The default is -1, which sorts along the last axis.
kind : Selection algorithm. Default is ‘introselect’.
order : [str or list of str] When arr is an array with fields defined, this argument specifies which fields to compare first, second, etc.

Return : [index_array, ndarray] Array of indices that partition arr along the specified axis.

Code #1 :

# Python program explaining
# numpy.recarray.argpartition() method

# importing numpy as geek
import numpy as geek

# creating input array with 2 different field
in_arr = geek.array([[(5.0, 2), (3.0, -4), (6.0, 9)],
                     [(9.0, 1), (5.0, 4), (-12.0, -7)]],
                     dtype =[('a', float), ('b', int)])
print ("Input array : ", in_arr)

# convert it to a record array,
# using arr.view(np.recarray)
rec_arr = in_arr.view(geek.recarray)
print("Record array of float: ", rec_arr.a)
print("Record array of int: ", rec_arr.b)

# applying recarray.argpartition methods
# to float record array along axis 1
out_arr = geek.recarray.argpartition(rec_arr.a, kth = 1, axis = 1)
print ("Output partitioned array indices along axis 1: ", out_arr)

# applying recarray.argpartition methods
# to int record array along axis 0
out_arr = geek.recarray.argpartition(rec_arr.b, kth = 1, axis = 0)
print ("Output partitioned array indices array along axis 0: ", out_arr)

Output:

Input array :  [[(5.0, 2) (3.0, -4) (6.0, 9)]
 [(9.0, 1) (5.0, 4) (-12.0, -7)]]
Record array of float:  [[  5.   3.   6.]
 [  9.   5. -12.]]
Record array of int:  [[ 2 -4  9]
 [ 1  4 -7]]
Output partitioned array indices along axis 1:  [[1 0 2]
 [2 1 0]]
Output partitioned array indices array along axis 0:  [[1 0 1]
 [0 1 0]]

Code #2 :

We are applying numpy.recarray.argpartition() to whole record array.

# Python program explaining
# numpy.recarray.argpartition() method

# importing numpy as geek
import numpy as geek

# creating input array with 2 different field
in_arr = geek.array([[(5.0, 2), (3.0, 4), (6.0, -7)],
                     [(9.0, 1), (6.0, 4), (-2.0, -7)]],
                     dtype =[('a', float), ('b', int)])
print ("Input array : ", in_arr)

# convert it to a record array,
# using arr.view(np.recarray)
rec_arr = in_arr.view(geek.recarray)

# applying recarray.argpartition methods to  record array
out_arr = geek.recarray.argpartition(rec_arr, kth = 2)

print ("Output array : ", out_arr)

Output:

Input array :  [[(5.0, 2) (3.0, 4) (6.0, -7)]
 [(9.0, 1) (6.0, 4) (-2.0, -7)]]
Output array :  [[1 0 2]
 [2 1 0]]