I've been looking at options for internationalizing an open source project:
Gettext, which everyone seems to recommend, apparently needs locales to be "installed on your system" to be used. See this note on the PHP manual, which reflects exactly my situation. This SO question also asks about the same problem. This isn't suitable for an open source project because I can't trust that the end users have the appropriate locales installed on their system. Besides, it's very very weird that you have to install locales just to use strings that you've translated (IMO).
Zend_Translate is also recommended in favour of gettext sometimes, but I'm not using the Zend framework so I don't think it's an option for me. Some people have said you can split it from the Zend framework, but I'd have no clue how. If anyone can tell me what files I need (I've downloaded a tarball of the Zend framework) to pick out, I would be open to using Zend_Translate.
Arrays. That's what I'm doing now, but it's not ideal because:
- It will be using up lots of memory to define every translation, when most won't be used by the current page.
- I am having problems with finding myself duplicating the keys in the array, which has already become 1000 lines of code long and I've barely added anything yet...
- It means non-programmers can't really translate, whereas POedit is the standard that everyone expects to use.
Can I somehow read .mo
files without Gettext or Zend_Translate, or must I use Gettext? If so, how can I make all locales work, like in the question I linked to above?
EDIT: I'm now willing to use Zend_Translate. I just need to work out what files I need (it would be great if they could be combined into one file) - I don't want the entire Zend Framework in my project.
Update: I was interested to see how big open source projects handle i18n:
- Moodle uses arrays.
- Drupal seems to use arrays, with a database fallback. They also run
setlocale('C')
to get a generic locale and circumvent the entiresetlocale
mess. - Wordpress uses heaps of their own functions for i18n, although underneath does use gettext.
So none of those three random projects use Zend_Translate, nor gettext directly, as far as I can see.
Maybe it's a good idea to use the C
locale, store the name of the language in the text domain name, and go from there.
So here's as far as I've got with that:
$lang = 'de'; //debug
setlocale( LC_ALL, 'C' );
bindtextdomain( 'default', PATH . "/locale/$lang" );
bind_textdomain_codeset( 'default', 'UTF-8' );
textdomain( 'default' );
var_dump( file_exists( PATH . "/locale/$lang/C/LC_MESSAGES/default.mo" ) ); //bool(true)
But I still just get the English string, even though I've used poedit, msgfmt etc. to make the appropriate files. I've also tried restarting Apache.
.mo
files and set the appropriate locale. Its really not as scary as you make it out to be. – Finniganlocale -a
to see what locales are available to use (and runsudo apt-get install language-pack-langcode-base
to install extra language packs). My own tests also reflect this. This isn't suitable for an open source project as I can't expect users to do this. – DissatisfiedZend_Translate
class, done. – AntefixZend_Translate
working, go removing classes/folders and see if it still works... until you end with a small enough (for you) set of classes. – DomesdayZend/I18n/Translator/Translator.php
, or Zend Loader or something...? – Dissatisfied