Re: Bug in Swedish ordinal messages

[Claude Paroz]

Le samedi 18 mai 2013 à 07:38 -0700, Horst Gutmann a écrit :
> Hi :-)
>
> There is currently a bug in the Swedish ordinal messages for
> django.contrib.humanize which is documented here:
> https://code.djangoproject.com/ticket/20441
>
> The patch includes the changes messages and should probably be
> duplicated in Transifex. apollo13 suggested that - if possible - one
> of you guys should merge this change and apply it to the Transifex
> repository in order to avoid confusion.

Hi Horst,

You'll see that I closed the ticket. Translation bugs should be fixed
directly on Transifex and will be automatically included in the next
Django release.

Claude
--
www.2xlibre.net

[Horst Gutmann]

Hi :-)

Yes, I saw that but I'm not completely agreeing here due to one detail: I totally agree that localisations should not be tested, but with the ordinal tag we are actually talking about functionality that depends on the i18n system and there exists a test for the English version of it. Given that the functionality should differ depending on the language, a unittest is in my opinion warranted.

Best wishes,
Horst

--
You received this message because you are subscribed to the Google Groups "Django internationalization and localization" group.
To unsubscribe from this group and stop receiving emails from it, send an email to django-i18n...@googlegroups.com.
To post to this group, send email to djang...@googlegroups.com.
Visit this group at http://groups.google.com/group/django-i18n?hl=en.
For more options, visit https://groups.google.com/groups/opt_out.

[Claude Paroz]

Le samedi 18 mai 2013 à 22:54 +0200, Horst Gutmann a écrit :
> Hi :-)
>

> Yes, I saw that but I'm not completely agreeing here due to one
> detail: I totally agree that localisations should not be tested, but
> with the ordinal tag we are actually talking about functionality that
> depends on the i18n system and there exists a test for the English
> version of it. Given that the functionality should differ depending on
> the language, a unittest is in my opinion warranted.

OK, then I interpret your test as "test that ordinal is properly using
translations when language is not en". This is not directly related to
the ticket title and the functionality wasn't broken, but it doesn't
hurt to add this test. But then I'd rather use a language that is
currently not broken, or show me that Swedish is fixed on Transifex.

Claude

[Horst Gutmann]

On Saturday, May 18, 2013 11:23:05 PM UTC+2, claudep wrote:

Le samedi 18 mai 2013 à 22:54 +0200, Horst Gutmann a écrit :
> Hi :-)
>
> Yes, I saw that but I'm not completely agreeing here due to one
> detail: I totally agree that localisations should not be tested, but
> with the ordinal tag we are actually talking about functionality that
> depends on the i18n system and there exists a test for the English
> version of it. Given that the functionality should differ depending on
> the language, a unittest is in my opinion warranted.

OK, then I interpret your test as "test that ordinal is properly using
translations when language is not en". This is not directly related to
the ticket title and the functionality wasn't broken, but it doesn't
hurt to add this test. But then I'd rather use a language that is
currently not broken, or show me that Swedish is fixed on Transifex.

Claude

Well, the test is there to prevent a regression and to document that
there is an issue with the ordinal tag when the language is set
explicitly to Swedish. In the end I think there should be a test for
every language here that has special cases (I guess every language
except for those where you can put a full-stop (or any other sign) after
the number and be done with it).  That the actual implementation uses
gettext messages is unfortunate here simply because it prevents the
method from working reliably independent of the language :-) IMO the
situation is quite similar to localflavors where usually at least the
postalcode validator has a unittest.

That being said, in the current implementation the test tests messages,
but hopefully in the future it will test the functionality that replaces
the current implementation of the tag (related to #15156) so it would
even more become a regression test.

That was also my intention when writing the test but I definitely won't
put up a fight if it is not merged ;-) The broken localization of the
ordinal numbering tag should IMHO be fixed, no matter if it comes with a
unittest or not :-)

Another aspect here is, that I don't know Swedish but noticed this issue
when trying to come up with a solution for #15156 and therefore wanted
to delegate the actual Transifex fix to someone on the relevant i18n
team :-)

Cheers,
Horst