Indexing and boolean masks

Selecting with a condition

Arrays slice like lists — a[1:4], a[-1], a[::2] all behave as you would expect. What they add is boolean indexing: put a boolean array inside the brackets and you get back only the elements where it was True.

a = np.array([3, 9, 1, 7, 5])
mask = a > 4          # array([False, True, False, True, True])
a[mask]               # array([9, 7, 5])
a[a > 4]              # the same thing, written the way people actually write it

Conditions combine with & (and), | (or) and ~ (not) — not the and, or, not keywords, which cannot work element by element. Each condition needs its own brackets:

a[(a > 2) & (a < 8)]    # array([3, 7, 5])

Forgetting those brackets is the single most common NumPy error, because & binds tighter than >.

Your turn: from scores, select the passing scores (50 or more) and the middling ones (40 up to but not including 70).

Replacing values with np.where

np.where(condition, if_true, if_false) builds a new array by choosing between two values element by element:

a = np.array([-3, 5, -1, 8])
np.where(a < 0, 0, a)     # array([0, 5, 0, 8]) — floor negatives at zero

It is the vectorised if. You can also assign through a mask, which changes the array in place:

a[a < 0] = 0

Your turn: write clip_negatives(values) that returns a new array with every negative replaced by zero, leaving the input untouched.