Locale code validation too strict

[Michal Čihař]

The locale code validation is too strict as it refuses locale variants which are used by some. See ​https://www.gnu.org/software/gettext/manual/html_node/Locale-Names.html for more information. Most frequently it can be seen with languages which can be written in both Latin and Cyrillic.

I will provide pull request for this.

[Michal Čihař]

Pull requests for master and 1.7.x:

​https://github.com/django/django/pull/3815

​https://github.com/django/django/pull/3816

[Claude Paroz]

I think this issue is a little more complicated. There are several families of language codes. But basically, we are handling two different ones in Django:

• IETF Language tags coming from the browser, defined in ​http://tools.ietf.org/html/rfc5646
• POSIX locale codes (​http://en.wikipedia.org/wiki/Locale) which is the format used by gettext translations file names.

We should not mix those. See also django.utils.translation.trans_real.to_locale/to_language which is the code trying to map strings from one code to the other. To correctly handle script variants, there is surely some more clever mapping work to do (e.g. Cyrl in RFC5646 should map to @cyrillic in the POSIX locale).

[Michal Čihař]

Indeed handling Language tags coming from the browser is not covered by my patch. What I've tried to address is to allow Django to set such locale manually. Automatic mapping of browser locales would be indeed nice.

[Tim Graham]

Claude, could you check the pull request now?

[Tim Graham <timograham@…>]

In 76d26d89:

Fixed #24063 -- Allowed locale variants supported by gettext.

The locale code can contain a variant after @, so allowed that.