How to replace NumPy array values conditionally

Data-cleaning rules often need to replace threshold breaches while leaving every other array element and the original ordering intact. NumPy applies that rule across the array with a boolean mask instead of a Python loop.

The np.where(condition, x, y) function builds a new array by choosing values from x at true positions and y elsewhere. The condition and branch values must broadcast to a common shape, so a scalar replacement works without constructing a full replacement array.

Boolean mask assignment changes the selected elements of its target array, while a separate working copy preserves the original values for later code. That distinction makes the mutation boundary explicit before either replacement method is chosen.

1. Create array-replace-conditional.py with the source array and its low-score mask.

   array-replace-conditional.py

   import numpy as np
    
   scores = np.array([35, 72, 88, 41, 93])
   low_score = scores < 50

2. Append the np.where() replacement section below low_score.

   array-replace-conditional.py

   adjusted = np.where(low_score, 50, scores)
   labels = np.where(scores >= 70, "pass", "review")

   np.where() returns new arrays. Both scalar branches broadcast across the five positions in scores.

3. Append the mask-assignment verification section below labels.

   array-replace-conditional.py

   in_place = scores.copy()
   in_place[low_score] = 50
    
   np.testing.assert_array_equal(adjusted, [50, 72, 88, 50, 93])
   np.testing.assert_array_equal(in_place, adjusted)
   np.testing.assert_array_equal(scores, [35, 72, 88, 41, 93])
    
   print("low score mask:", low_score)
   print("np.where adjusted:", adjusted)
   print("mask assignment adjusted:", in_place)
   print("labels:", labels)
   print("source array:", scores)

   The assertions stop the script if either replacement changes the wrong position, the methods disagree, or mask assignment alters the source array.

4. Run the completed conditional replacement script.

   $ python3 array-replace-conditional.py
   low score mask: [ True False False  True False]
   np.where adjusted: [50 72 88 50 93]
   mask assignment adjusted: [50 72 88 50 93]
   labels: ['review' 'pass' 'pass' 'review' 'pass']
   source array: [35 72 88 41 93]