[LON-CAPA-dev] Internationalization

[Gerd Kortemeyer]

Hi,

As you might have seen from a number of CVS commits in the last two 
days, we have started to internationalize LON-CAPA, so that it can run 
in different languages. We have also started localizing for German and 
Portuguese. Internalizing is the bulk of the work right now, localizing 
can be done anytime, and involves no programming.

The internationalization process involves putting a wrapper around 
on-screen user messages and menus and turning them into keys, which the 
Maketext library translates into the desired language output using a 
look-up table ("lexicon").

As keys we are currently using the plain English messages, and Maketext 
is configured to replace the message by its own key if no translation 
is found. This makes it easy to phase in the internationalization 
without disturbing the screen output.

Internationalization is tedious, and the more we can distribute the 
work, the better.

*** New software should be written in internationalized format from now 
on.

*** If you feel like doing something brainless but productive, grab a 
module and internationalize it.

*** As you come across some code during normal programming, just insert 
the wrappers.

*** HERE'S WHAT TO DO:

Into the "use" section of a module, we need to insert

  use Apache::lonlocal;

- note that there are NO BRACKETS, we *want* to pollute our namespace.

Inside might be something like this

sub message {
    my $status=shift;
    my $message='Status unknown';
    if ($status eq 'WON') {
       $message='You have won.';
    } elsif ($status eq 'LOST') {
       $message='You are a total looser.';
    }
    return $message;
}
...
$r->print('<h3>Gamble your Homework Points</h3>');
...
$r->print(<<ENDMSG);
<font size="1">Rules:</font>
<font size="0">No purchase necessary. Illegal where not allowed.</font>
ENDMSG

We have to now wrap the subroutine &mt() ("maketext") around our 
messages, but not around markup, etc. We also want minimal disturbance. 
The first two examples are easy:

sub message {
    my $status=shift;
    my $message='Status unknown';
    if ($status eq 'WON') {
       $message='You have won.';
    } elsif ($status eq 'LOST') {
       $message='You are a total looser.';
    }
    return &mt($message);
}
...
$r->print('<h3>'.&mt('Gamble your Homework Points').'</h3>');

The last one is a bummer, since you cannot call subroutines inside of 
(<<MARKER). I have written a little subroutine to generate a translated 
hash for that purpose:

my %lt=&Apache::lonlocal::texthash('header' => 'Rules', 'disclaimer' => 
'No purchase necessary. Illegal where not allowed.');
$r->print(<<ENDMSG);
<font size="1">$lt{'header'}:</font>
<font size="0">$lt{'disclaimer'}</font>
ENDMSG

As a programmer, your job is done here. If everything worked, you 
should see no changes on the screen.

As a translator, you need to provide the lexicon for the keys, which in 
this case is the plain text message. The lexicons sit in 
loncom/localize/localize, with the language code as filename, for 
example de.pm for the German translation. The file then simply looks 
like this:

    'You have won.'
=> 'Sie haben gewonnen.',

    'You are a total looser.'
=> 'Sie sind der totale Verlierer.',

    'Rules'
=> 'Regeln',

    'No purchase necessary. Illegal where not allowed.'
=> 'Es ist erlaubt, einfach zu verlieren, und das ist Ihre Schuld.'

The German translation lexicon is in pretty okay shape, but not 
complete yet. Portuguese currently only covers the login screen. 
Russian is purely experimental. Looks like UTF-8 is the way to encode 
this, at least for latin/greek-based languages, but we still have to 
learn a lot.

This is a relatively easy programming task, and any help is appreciated.

Maketext can do a whole lot more, see

  http://search.cpan.org/dist/Locale-Maketext/lib/Locale/Maketext.pod

but for most purposes, we do not have to mess with that.

- Gerd.