Skip to content

numpy/element-loop

Rule element-loop
Pack NumPy
Severity warning
Fix none, edit by hand
Library versions every version

A loop that computes one element of an array per pass runs Python once for every value, which was 70 to 530 times slower. Write it as one expression on whole arrays.

out = np.empty_like(a)
for i in range(len(a)):
out[i] = a[i] * 2 + np.sqrt(b[i])
out = a * 2 + np.sqrt(b)

The rule reports a for loop over range, np.arange or np.ndindex, and a while i < n loop with i += 1, when the loop's only statement writes one element of a NumPy array from arithmetic on the loop position. It also reports a loop over np.nditer or np.ndenumerate without external_loop. The rule leaves a loop alone whose body calls another function, reads the array it writes, or writes a slice.

xpd has no automatic fix, so make the change by hand.

xpd prints this message with the issue. A name that starts with $ stands for code from your file.

This loop computes $OUT one element per pass in Python. Write it as one expression on whole arrays, with slices or index arrays in place of $I, which ran 70 to 530 times faster on 1,000,000 values.

Browse the other rules of the NumPy pack.