Bug report

Bug description:

textwrap.wrap() takes quadratic time on a single long word. TextWrapper._handle_long_word() breaks off one line's worth of the word at a time and stores the rest with reversed_chunks[-1] = chunk[end:], which copies the whole rest of the word for every output line.

import textwrap, time
for n in (400_000, 800_000, 1_600_000, 3_200_000):
    t = time.perf_counter()
    textwrap.wrap('x' * n, 70)
    print(f'{n:>9,} chars: {time.perf_counter() - t:.2f}s')

Doubling the length makes it about four times slower:

This is a different cause from bpo-22687 (gh-66877), which fixed the word-splitting regex.

CPython versions tested on:

CPython main branch

Operating systems tested on:

macOS

Linked PRs

• gh-161977: Fix quadratic time in textwrap when breaking a long word #161978