Many internationalization engines like Rails i18n support complex value types.

We need to support something like t('en_US', 'date.abbr_day_names') with:

date:
  abbr_day_names:
  - Sun
  - Mon
  - Tue
  - Wed
  - Thu
  - Fri
  - Sat

But when we try to compile it using Babelfish errors thrown.

Take a look at this dictionary entries:

datetime:
  distance_in_words:
    less_than_x_minutes:
      one: "less than a minute"
      other: "less than #{count} minutes"

Rails pluralization backend supports selection of key based on count value.

In addition to typical one, few, manu and other suffixes it supports optional zero suffix.

This cannot be repeated by Babelfish syntax because here a case when no value should be printed.

Fixes #7: Zero-count in plurals.

Since achor was already checked for being undefined, I see no reason to use String(anchor || '') instead of just String(anchor). Otherwise, it will produce wrong result with values like false, null and, mostly important, 0.

I also want to discuss changing check from typeof to .hasOwnProperty() since undefined could also be wanted in the translated string (imo, it's also more semantically correct since we are not intrested in variable type or value, but presence):

// lib/babelfish.js:100
if ('variable' === node.type) {
  translator.push(
    "str += ( !params.hasOwnProperty('" + node.anchor + "') )" +  // here
    " ? '[missed variable: " + node.anchor + "]'" +
    " : String(" + anchor + ");"
  );
  return;
}

Hello.

I would like to ask question about using country variants:

var BabelFish = require('babelfish');
var i18n = new BabelFish('en-GB');
i18n.addPhrase('pt', 'test.a', 'Apelido')
i18n.addPhrase('pt-BR', 'test.a', 'Sobrenome')
i18n.addPhrase('pt', 'test.b', 'Título')
> i18n.t('pt', 'test.a')
'Apelido'
> i18n.t('pt-BR', 'test.a')
'Sobrenome'
> i18n.t('pt-BR', 'test.b')
'pt-BR: No translation for [test.b]'

Shouldn't last line fallback to 'pt' ? Or i should manually add every possible fallback with .setFallback?

Thanks.

README.md contains examples which call getTranslation() on BabelFish instance. This is probably an artefact left from some previous version(s) and should be changed to translation() or t() call.

locale in .getCompiledData() should default to fallback locale, but following adds store['undefined']:

var BabelFish = require('babelfish');
var i18n = new BabelFish('ru');

i18n.addPhrase('ru', 'i', 'я');

console.dir(i18n.getCompiledData());
console.dir(i18n._storage);

Outputs:

{ ru: 
   { i: 
      { type: 'string',
        locale: 'ru',
        translation: 'я' } },
  undefined: 
   { i: 
      { type: 'string',
        locale: 'ru',
        translation: 'я' } } }

Following

var translator = require('babelfish').create('ru');
translator.addPhrase('ru', 'nailsCount', 'У меня #{n} ((гвоздь|гвоздя|гвоздей)):n');
[0, 1, 2, 5].forEach(function (count) {
  console.log(translator.t('ru', 'nailsCount', {n: count}));
});

outputs

У меня  гвоздей
У меня 1 гвоздь
У меня 2 гвоздя
У меня 5 гвоздей

I also think that ability to define different form for zero-count should be added. ("No items" instead of "0 items".) Same for singular ("One item"). That could be optional last two forms inside (()).

This library worth to be wrap into angular module with directive, service, filter stuff. Is there any wip in that way? Maybe someone from community started to develop such module for its own.

NB There is already some angular i18n modules but I wasn't able find one which fit me. At least it lacks of normal namespacing and such powerfull stringify method.

Pluralizer rules can be autogenerated automatically from CLDR data.

References:

• CLDR downloads: http://cldr.unicode.org/index/downloads (json.zip -> supplemental -> plurals)
• Latest chart: http://www.unicode.org/cldr/charts/latest/supplemental/language_plural_rules.html
• Syntax description http://www.unicode.org/reports/tr35/tr35-numbers.html#Language_Plural_Rules

More to investigate:

• search by "CLDR" & "LDML"
• https://www.npmjs.com/packages/make-plural
• https://www.npmjs.com/package/cldr-data

Just note that we plan to release CPAN (open source) Perl module as counterpart for your's one.

Release date planned to ~11 Jun 2014.

Also this pair will be disputed on YAPC::Russia 2014 conference (Spb, 13-14 Jun 2014).

https://github.com/nodeca/babelfish/blob/master/lib/babelfish.js#L236

1. Check, if we really need that property in public?
2. Can we replace it with getter?
3. Do we need to add fallback, or use es5-shams?

If one need piped processing for variables to do additional l10n, it's possible to use this notation:

#{variable:processor|param1|param2|...}

Examples:

• #{count:localize_number}
• #{date:format|short}
• #{another.translation:phrase} - insert another translation by this key

The same can be done for selector notation, to change selection rule:

• ((male|female))[:gender]
• ((male|female))[:ordinal]

That's not planned right now, because no real requests exist. Also syntax is subject to discuss.

Probably, that should not be done at all, because data can be prepared before passing to translate()