I have two arrays:

arr1 = np.array((
    np.array([ 32,  32,  32,  32,  32,  39], dtype=np.int64),
    np.array([449, 451, 452, 453, 454, 463], dtype=np.int64)))
arr2 = np.array((
    np.array([ 39,  34,  32,  32,  37,  32], dtype=np.int64),
    np.array([463, 393, 453, 452, 261, 449], dtype=np.int64)))

In these 2D arrays, the:

• First array (arr1[0], arr2[0]) are x-axis values
• Second array (arr1[1], arr2[1]) are y-axis values

I would like to find the xy pairs that match between the two arrays.

Some clarifications:

• arr1 and arr2 will not necessarily be of equal length. They could be different lengths
• X value and Y value pairs could be in any ordering. Sorting or alignment between arrays is not expected
• Duplicates of the same X value and Y value pairs will not occur in the same array

In the above examples, the pairs that are the same between the two arrays are:

• X = 32, Y = 449
• X = 32, Y = 452
• X = 32, Y = 453
• X = 39, Y = 463

I've tried to use np.intersect1d and some other functions I found.

CodePudding user response：

Use a structured array.

import numpy as np

# Define a dtype with x and y integers    
arr1 = np.empty(6, dtype=[('x', int), ('y', int)])
arr2 = np.empty(6, dtype=[('x', int), ('y', int)])

# Add the data to the structured array
arr1['x'] = np.array([ 32,  32,  32,  32,  32,  39])
arr1['y'] = np.array([449, 451, 452, 453, 454, 463])

arr2['x'] = np.array([ 39,  34,  32,  32,  37,  32])
arr2['y'] = np.array([463, 393, 453, 452, 261, 449])

Use intersect1d:

>>> np.intersect1d(arr1, arr2)
array([(32, 449), (32, 452), (32, 453), (39, 463)],
  dtype=[('x', '<i8'), ('y', '<i8')])

CodePudding user response：

Since you need to find the intersection of couples, in my opinion it's better to use the set data structure instead of numpy.array. Here's how I would do:

np.array(list(
    set(zip(arr1[0,:], arr1[1,:])).intersection(
        set(zip(arr2[0,:], arr2[1,:])))
))

or wjandrea's more concise version:

np.array(list(
    set(zip(*arr1)).intersection(zip(*arr2))
))