[f4hy]

For pure ASCII input, we can skip the regex/generator path in unquote() and decode the result of _unquote_impl() directly. The regex would match the whole string as a single chunk anyway, so the result is identical.

Benchmarks are in the issue but up to 2x faster in the micro-benchmark for the common short ascii input.

Added a test checking that the fast path and the general path decode escapes the same way.

Resolves #160690

• Issue: Speed up urllib.parse.unquote() for ASCII input #160690

[python-cla-bot]

All commit authors signed the Contributor License Agreement.

[eendebakpt]

Suggested change

	if string.isascii():
	return _unquote_impl(string).decode(encoding, errors)
	if encoding in ('utf-8', 'utf8'):
	try:
	data = string.encode('utf-8')
	except UnicodeEncodeError:
	pass # lone surrogates must be passed through verbatim
	else:
	return _unquote_impl(data).decode(encoding, errors)

The fast path introduced by the PR looks good to me. I am not sure whether ascii only covers the most common / important cases. The suggestion here would cover more cases and also gives a speedup.

(and the two fast paths can also be combined)