Feature or enhancement

Currently, locale.getlocale() does not support locales with @-modifiers. If the current locale has an @-modifier, it will be silently dropped in the locale.getlocale() result. For example, ca_ES.utf8@valencia will result in ('ca_ES', 'UTF-8'), sr_RS.utf8@latin -- ('sr_RS', 'UTF-8'), uz_UZ.utf8@cyrillic -- ('uz_UZ', 'UTF-8'). They are different locales! The @euro modifier, the only one that handled specially, is also removed, even if locales with and without @euro can be different locales, and @euro affects collation in some locales. And for the el_gr@euro locale it returns wrong encoding ISO8859-15 (the correct encoding for this locale on Linux is ISO8859-7).

Accordingly, locale.setlocale() does not support @-modifiers if the locale is specified as a language code/encoding pair. It is supported if specify locale as string, but then the purpose of the wrapper is lost.

See also #87281.

The only way to support @-modifiers in the format compatible with this functions is to include it in the language code. So locale.getlocale() will return ('ca_ES@valencia', 'UTF-8') for locale ca_ES.utf8@valencia, and passing ('ca_ES@valencia', 'UTF-8') to locale.setlocale() will set locale to ca_ES.utf8@valencia. This is mildly breaking change, but it is better to get error in program that do not expect @ in the locale.getlocale() result than get a wrong result for that locale.

Linked PRs

• gh-137729: Fix support for locales with @-modifiers #137253