Boolean Masks

Filter with Mask

Keep only the values where a boolean mask is True. A zip loop pairs each value with its mask entry; an if-check inside the loop decides whether to append. The replay shows filtered growing only when m is True, making the selection criterion explicit.

By hand

With NumPy

a[m] selects every element of a where the corresponding element of the boolean array m is True. NumPy calls this boolean indexing. The snapshot shows the input, the mask, and the filtered result with its smaller shape.

naive.py 
values = [3, 7, 2, 9, 1, 5]
mask = [False, True, False, True, False, True]
filtered = []
for v, m in zip(values, mask):
    if m:
        filtered.append(v)
print('RESULT:', filtered)
library.py 
import numpy as np

values = [3, 7, 2, 9, 1, 5]
mask = [False, True, False, True, False, True]
a = np.array(values)
m = np.array(mask)
filtered = a[m]
print('a: shape:', a.shape, 'dtype:', a.dtype, 'values:', a.tolist())
print('mask: shape:', m.shape, 'dtype:', m.dtype, 'values:', m.tolist())
print('filtered: shape:', filtered.shape, 'dtype:', filtered.dtype, 'values:', filtered.tolist())
print('RESULT:', filtered.tolist())

a: shape: (6,) dtype: int64 values: [3, 7, 2, 9, 1, 5]
mask: shape: (6,) dtype: bool values: [False, True, False, True, False, True]
filtered: shape: (3,) dtype: int64 values: [7, 9, 5]
RESULT: [7, 9, 5]

Implementation notes

  • Boolean indexing (a[mask]) returns a copy, not a view. Modifying the result does not affect a.
  • The output shape is (count_of_True,) — determined at runtime, not known from the input shape alone.
  • The mask is typically built by a comparison (a > k) rather than written by hand — see mask-from-threshold for that pattern. Chaining both steps: a[a > k] selects in one expression.
  • For the equivalent pure-Python pattern see filter-by-threshold in the python-data-basics book.
  • Shape, dtype, and values are shown explicitly here because ndarray.__repr__ output varies with NumPy version and print options.