Add warning in makemessages command if the localecode with `l` flag is not correct

[Sanyam Khurana]

Hey Calude,

What about normalizing the directory name to something that would just work.

For example,

No matter, if the developer is doing all these:

python manage.py makemessages -l zh_cn
python manage.py makemessages -l zh_CN
python manage.py makemessages -l ZH_CN
python manage.py makemessages -l ZH-CN

etc.

we, just normalize the directory name to zh_CN and it would work.

I'm still about to read the code of makemessages command and probably if there are any more checks than just this, then we'll have to figure out another way all together.

[Claude Paroz]

We have to be careful in deciding what is a "good" and a "bad" code. However I agree that we can avoid some mistakes, notably the confusion between IETF language tags [1] and ISO/IEC 15897 (Posix) [2] codes generally expected by Django.

[1] ​https://en.wikipedia.org/wiki/IETF_language_tag
[2] ​https://en.wikipedia.org/wiki/Locale_(computer_software)

[Claude Paroz]

​PR

[Vishvajit Pathak]

​PR

[Claude Paroz]

I think I would still output a warning when a language code is normalized, just to inform the user that its input has been corrected.

[Vishvajit Pathak]

Warning added when a language code is normalized.

[Nick Pope]

Trying to coerce any input to something like a ​language tag only to hack it back to a ​locale using to_locale() feels like a kludge. It would be better to improve documentation.

[Claude Paroz]

Improving documentation is welcome, but silently accepting a wrong language code also look a bit suspicious. I think I would be happy with a warning without coercing anything.

[Vishvajit Pathak]

We definitely need to improve the documentations. Coercing the language code is something we have to take call on.

[Nick Pope]

Replying to Claude Paroz:

Improving documentation is welcome, but silently accepting a wrong language code also look a bit suspicious. I think I would be happy with a warning without coercing anything.

I agree. I think a warning would make sense, without coercion.

It is still possible to provide a locale to makemessages where there are no actual message catalogs in any of the paths in settings.LOCALE_PATHS.
We should probably scrap all the normalization stuff and just output a warning message if a locale specified by the user is not in all_locales.
At the moment we output a "processing locale xx_XX" message if verbosity > 0 which should be fixed to only happen for valid, existing locales.

As an aside, this is checking --locale for makemessages, but what about compilemessages? (And are their any others?)

[Vishvajit Pathak]

Now I am thinking more to avoid the coercing and to put just a warning message.

[Vishvajit Pathak]

Tim

Current implementation involved just adding the warning message. We are not normalizing locale now.

[Vishvajit Pathak]

Tim,

I would like to understand better what needs to be done next ?

[Claude Paroz]

As commented on the pull request, you cannot use the all_locales variable as a reference as it does not contain new (valid) language codes. The modified test in your patch shows a 'Invalid locale en' message which is obviously wrong, isn't it?
You may try using the django.utils.translation.trans_real.language_code_re to check the locale code syntax.

[Sanyam Khurana]

Hi Vishvajit!

Are you still working on the bug? Do you need any help?

[Vishvajit Pathak]

Hi Sanyam,

Yes, I am working on this ticket but your help would be very appreciated.

[Herbert Fortes]

Hi,

Here my thoughts:

language_code_re gets 'pt_BR'. And it does not get 'ZH-CN'. I did this:
r'[a-z]{2}(_[A-Z]{2})?'

It seems to work. But one test prints:

System check identified no issues (0 silenced).
Invalid locale d
Invalid locale e
.................................................
----------------------------------------------------------------------
Ran 49 tests in 1.425s

OK

I did a check and there is only one line - 740 - that's use locale=LOCALE. All
others use locale=[LOCALE].

What do you think Mark?

[Herbert Fortes]

I forgot to say that the regrex I sent fails to:

ja_JP_JP
no_NO_NY
th_TH_TH

[Carlton Gibson]

I'm going to de-assign this because there's been progress, so that someone looking can pick it up.

[Manav Agarwal]

Can I claim this? I think I can move this forward

[Carlton Gibson]

Hi Manav, please do, no need to ask (but can I advise that you take just one at a time to begin :)

[Manav Agarwal]

Can we use LANG_INFO from django.conf.locale in order to solve the conflict? If yes, then please confirm so that I may submit a patch, and also please suggest some way to use the same

[Claude Paroz]

The problem with using LANG_INFO is that you exclude some extra languages added specifically for a project.

You can try to use language_code_re, but you'll have to take into account the language code vs locale code difference.

[Manav Agarwal]

​PR

[Manav Agarwal]

If the patch in ​PR seems fine then please update me so that I may start working on tests.

[Manav Agarwal]

As per the ​comment by felixxm on ​PR, I have created a check for the valid locales to not have hyphens.

[Mariusz Felisiak <felisiak.mariusz@…>]

In f63f3cdf:

Fixed #29712 -- Made makemessages warn if locales have hyphens and skip them.