Deprecate re.LOCALE

Feature or enhancement

Proposal:

Deprecate the re.LOCALE flag and the inline (?L) modifier: emit a DeprecationWarning when compiling a pattern that uses them, for removal no earlier than Python 3.18.

re.LOCALE only works with bytes patterns and only makes sense in 8-bit locales, which are rare today (it does not work in UTF-8 or other multibyte locales). It makes matching an order of magnitude slower: it excludes compile-time optimization of character sets, and at matching time it calls the locale-dependent C functions for every character instead of a table lookup. It also complicates the engine: every character class and case-insensitive operation needs a third variant beside the plain and Unicode ones, and every new optimization must special-case it.

A scan of the top 5,000 PyPI projects found 14 matches in 9 projects, and in the observed cases the flag was unnecessary. The replacement is to decode input and use str patterns, or to spell the needed byte semantics explicitly in the pattern.

This was discussed in 2023 in https://discuss.python.org/t/what-to-do-with-re-locale/32600 with a consensus for deprecation; matching str patterns with re.LOCALE has already been an error since Python 3.6 (bpo-22407).

Linked PRs

• gh-153178

Bắt đầu bằng cách lần theo cách CPython biên dịch cờ re.LOCALE và bộ sửa đổi inline (?L), sau đó kiểm tra gh-153178 để xem những công việc nào đang được tiến hành. Được xem là hoàn tất khi việc biên dịch một trong hai dạng phát ra DeprecationWarning, đồng thời vẫn giữ nguyên mốc thời gian loại bỏ đã nêu và hành vi hiện có cho đến khi loại bỏ.