BIDI language detection not smart enough

[Ahmad Alhashemi]

one of my projects, I'm using a LANGUAGE_CODE of 'ar-kw' (Arabic-Kuwait). We already have an Arabic translation (ar) and it is being appropriately used. The problem is that my language code is not being detected as a right to left language. It has to be exactly 'ar' or it will not be detected.

Maybe we can bring some of the smartness of gettext into detecting the direction of text too?

There is a function in gettext called _expand_lang. It takes the locale as a single argument and returns a list of possible names for this locale, so that ar_KW will return a list containing ar_KW, ar and other locales.

[Jannis Leidel]

Added proposed support for _expand_lang

[arien]

Changing django.utils.translation.get_language_bidi so that it looks to see if xx is in settings.LANGUAGES_BIDI for a language xx-yy (or xx) seems like a nicer way to do this.

[arien]

Use only first part of language name (i.e., xx for a language xx-yy) for BiDi detection.

[Jannis Leidel]

I don't see why this is nicer to not use a upstream function that does exactly that.

[arien]

gettext._expand_lang expands a locale to a list of locales. That's not doing what is needed here. Besides, you're checking to see if any of those locales is listed in settings.LANGUAGES_BIDI, which contains languages. The reason this works is more or less accident.

All that's needed is to check if xx is in settings.LANGUAGES_BIDI for some language xx-yy. So that's what the patch does.

(You could extract the base_lang function from translation and get_language_bidi to be more DRY, but I'll leave that to a committer.)

[arien]

Didn't mean to close this.

[Jacob]

(In [8596]) Fixed #4389: be more robust in get_language_bidi(). Thanks, arien.

[Jacob]

Milestone 1.0 deleted