Justifying NumPy Arrays with Vectorized Functions

NumPy provides efficient ways to justify arrays using vectorized functions, offering improved performance and code simplicity compared to traditional Python loops.

Problem Statement

Given a NumPy array, the task is to shift its non-zero elements to the left, right, up, or down while maintaining its shape.

Numpy Solution

The following NumPy implementation performs efficient justification:

import numpy as np

def justify(a, invalid_val=0, axis=1, side='left'):    
    if invalid_val is np.nan:
        mask = ~np.isnan(a)
    else:
        mask = a!=invalid_val
    justified_mask = np.sort(mask,axis=axis)
    if (side=='up') | (side=='left'):
        justified_mask = np.flip(justified_mask,axis=axis)
    out = np.full(a.shape, invalid_val) 
    if axis==1:
        out[justified_mask] = a[mask]
    else:
        out.T[justified_mask.T] = a.T[mask.T]
    return out

This function justifies a 2D array along the specified axis and side (left, right, up, down). It works by identifying non-zero elements using mask, sorting them using sort, flipping the mask if justifying upwards or leftwards, and finally overwriting the original array with the justified values.

Example Usage

Here's a usage example that covers non-zero elements to the left:

a = np.array([[1,0,2,0], 
              [3,0,4,0], 
              [5,0,6,0], 
              [0,7,0,8]])

# Cover left
covered_left = justify(a, axis=1, side='left')

print("Original Array:")
print(a)
print("\nCovered Left:")
print(covered_left)

Output:

Original Array:
[[1 0 2 0]
 [3 0 4 0]
 [5 0 6 0]
 [0 7 0 8]]

Covered Left:
[[1 2 0 0]
 [3 4 0 0]
 [5 6 0 0]
 [7 8 0 0]]

Justifying for a Generic N-Dimensional Array

For justifying an N-dimensional array, the following function can be used:

def justify_nd(a, invalid_val, axis, side):    
    pushax = lambda a: np.moveaxis(a, axis, -1)
    if invalid_val is np.nan:
        mask = ~np.isnan(a)
    else:
        mask = a!=invalid_val
    justified_mask = np.sort(mask,axis=axis)
    
    if side=='front':
        justified_mask = np.flip(justified_mask,axis=axis)
            
    out = np.full(a.shape, invalid_val)
    if (axis==-1) or (axis==a.ndim-1):
        out[justified_mask] = a[mask]
    else:
        pushax(out)[pushax(justified_mask)] = pushax(a)[pushax(mask)]
    return out

This function supports more complex scenarios by justifying an N-dimensional array along an arbitrary axis and to either the 'front' or 'end' of the array.