Allow correct translation of plurals

[Richard Mansfield]

With the current language pack format, it's impossible to translate many plurals correctly unless the language uses the same rules as English. We need to let translations specify plurals via a function over the number, as described in http://www.gnu.org/s/hello/manual/gettext/Plural-forms.html

The following emails have more detail about how we propose to fix this in Mahara:

---------------------------- Original Message ----------------------------
Subject: Re: Using PluralForms in Mahara strings + DWOO problem/question
From: "Richard Mansfield"
Date: Tue, September 13, 2011 12:16 pm
To: Gregor Anzelj, Francois Marier
--------------------------------------------------------------------------

Thanks very much for this Gregor, I've now had time to have a good look
at those links you sent, and done a bit of playing around on Mahara.

I'd like to propose that we change Mahara along the lines of this commit:

https://github.com/richardmansfield/richardms-mahara/commit/157c384b2860553ab78fad88f7a32b30d4af0ec2

And the following commit shows how the strings would look in the default
langpack:

https://github.com/richardmansfield/richardms-mahara/commit/b0aba7671669ab996008a0c65cc11538dc529bda

It's all pretty much the same as your scheme, but just closer to the
gettext/PO format, using a php array to store the different plural forms
rather than special codes inside the string. So to use your example,
this string:

$string['notsosimpleplural'] = "%d
{{PLURAL:%d|sporocilo|sporocili|sporocila|sporocil}}";

would be specified like this in the php langpack:

$string['notsosimpleplural'] = array(
    '%d sporocilo',
    '%d sporocili',
    '%d sporocila',
    '%d sporocil',
);

and like this in our PO files:

#: lang/en.utf8/mahara.php notsosimpleplural
msgctxt "lang/en.utf8/mahara.php notsosimpleplural"
msgid "%d message"
msgid_plural "%d messages"
msgstr[0] "%d sporocilo"
msgstr[1] "%d sporocili"
msgstr[2] "%d sporocila"
msgstr[3] "%d sporocil"

Keeping our scheme similar to PO & gettext has the advantage that it'll
make the conversion of langpacks to/from PO a lot easier, and the other
advantage is that we won't need to rely on special characters like { and
| inside the translated strings themselves (these characters could
legitimately appear in strings too, and one day that might require us to
dream up some escaping rules).

The disadvantage I see, compared to your scheme, is that we're removing
the option to put multiple plural forms anywhere inside a string. For
example in the second patch above, I had to split a string up, because
"Created %s folders and %s files" contains two plurals. However, I
don't think PO can deal with that nicely (correct me if I'm wrong on
that). If we're expecting people to translate Mahara using standard PO
editing tools and Launchpad, I don't see that we've got much choice but
to limit ourselves to one plural per string.

What do you guys think?

On 29/08/11 19:49, Gregor Anzelj wrote:
> > Regarding plural forms and gettext/ .po files...
> >
> > I think these pages contain all the info needed:
> >
> > * http://www.gnu.org/s/hello/manual/gettext/Plural-forms.html
> > *
http://www.gnu.org/s/hello/manual/gettext/Translating-plural-forms.html#Translating-plural-forms
> >
> > But I think we need to be careful to cover all possible cases:
> >
> > * simple one, e.g.: $string['simpleplural'] =
> > "{{PLURAL:%d|view|views}}"; as well as
> > * not so simple one, e.g.: $string['notsosimpleplural'] = "%d
> > {{PLURAL:%d|message|messages}}";
> >
> > or in Slovenian:
> >
> > * simple one, e.g.: $string['simpleplural'] =
> > "{{PLURAL:%d|pogled|pogleda|pogledi|pogledov}}"; as well as
> > * not so simple one, e.g.: $string['notsosimpleplural'] = "%d
> > {{PLURAL:%d|sporocilo|sporocili|sporocila|sporocil}}";
> >
> > HTH,
> > Gregor
> > --
> > ----------
> > Gregor Anzelj, prof.
> > Gimnazija Ledina, Ljubljana

[François Marier (fmarier)]

https://reviews.mahara.org/922

[Mahara Bot (dev-mahara)]

Reviewed: https://reviews.mahara.org/923
Committed: http://gitorious.org/mahara/mahara/commit/526e469835876d958e1f1f1dcf5dad680168ffec
Submitter: Richard Mansfield (<email address hidden>)
Branch: master

commit 526e469835876d958e1f1f1dcf5dad680168ffec
Author: Richard Mansfield <email address hidden>
Date: Tue Sep 13 10:32:37 2011 +1200

    Use the plural form translations for selected strings (bug #901051)

    Minor changes to several strings in the default language pack to use
    plural forms. This commit is intended as an example to demonstrate
    the use of plural forms.

    Change-Id: I425fbdaa10496c747b415e07e18aeb23736967b8
    Signed-off-by: Richard Mansfield <email address hidden>

[Mahara Bot (dev-mahara)]

Reviewed: https://reviews.mahara.org/922
Committed: http://gitorious.org/mahara/mahara/commit/52a6e7e19632665fe037b0a44134045773e5bb78
Submitter: Richard Mansfield (<email address hidden>)
Branch: master

commit 52a6e7e19632665fe037b0a44134045773e5bb78
Author: Richard Mansfield <email address hidden>
Date: Tue Sep 13 10:25:57 2011 +1200

    Modify get_string to take plural forms (bug #901051)

    Allows language packs to specify an array of strings (one for each
    plural form in the language) instead of a single string.

    The first of the variable (sprintf) arguments is assumed to be the
    number we need to check to choose a plural form. This argument is
    passed into a language-specific function defined in langconfig.php,
    which returns the appropriate plural form based on a formula.

    A similar scheme is used for strings formatted by the javascript
    get_string function in mahara.js.

    See http://www.gnu.org/s/hello/manual/gettext/Plural-forms.html

    Change-Id: Ifb58ac4f1e13a54edbc57e5c0a9faaf8454e53a4
    Signed-off-by: Richard Mansfield <email address hidden>