Locale::Maketext::Gettext(3) Joins the gettext and Maketext frameworks

SYNOPSIS

In your localization class:


package MyPackage::L10N;
use base qw(Locale::Maketext::Gettext);
return 1;

In your application:

  use MyPackage::L10N;
  $LH = MyPackage::L10N->get_handle or die "What language?";
  $LH->bindtextdomain("mypackage", "/home/user/locale");
  $LH->textdomain("mypackage");
  $LH->maketext("Hello, world!!");

If you want to have more control to the detail:

  # Change the output encoding
  $LH->encoding("UTF-8");
  # Stick with the Maketext behavior on lookup failures
  $LH->die_for_lookup_failures(1);
  # Flush the MO file cache and re-read your updated MO files
  $LH->reload_text;
  # Set the encoding of your maketext keys, if not in English
  $LH->key_encoding("Big5");
  # Set the action when encode fails
  $LH->encode_failure(Encode::FB_HTMLCREF);

Use Locale::Maketext::Gettext to read and parse the MO file:

  use Locale::Maketext::Gettext;
  %Lexicon = read_mo($MOfile);

DESCRIPTION

Locale::Maketext::Gettext joins the GNU gettext and Maketext frameworks. It is a subclass of Locale::Maketext(3) that follows the way GNU gettext works. It works seamlessly, both in the sense of GNU gettext and Maketext. As a result, you enjoy both their advantages, and get rid of both their problems, too.

You start as an usual GNU gettext localization project: Work on PO files with the help of translators, reviewers and Emacs. Turn them into MO files with msgfmt. Copy them into the appropriate locale directory, such as /usr/share/locale/de/LC_MESSAGES/myapp.mo.

Then, build your Maketext localization class, with your base class changed from Locale::Maketext(3) to Locale::Maketext::Gettext. That is all.

METHODS

$LH->bindtextdomain(DOMAIN, LOCALEDIR)
Register a text domain with a locale directory. Returns "LOCALEDIR" itself. If "LOCALEDIR" is omitted, the registered locale directory of "DOMAIN" is returned. This method always success.
$LH->textdomain(DOMAIN)
Set the current text domain. Returns the "DOMAIN" itself. If "DOMAIN" is omitted, the current text domain is returned. This method always success.
$text = $LH->maketext($key, @param...)
Lookup the $key in the current lexicon and return a translated message in the language of the user. This is the same method in Locale::Maketext(3), with a wrapper that returns the text message "encode"d according to the current "encoding". Refer to Locale::Maketext(3) for the maketext plural notation.
$text = $LH->pmaketext($ctxt, $key, @param...)
Lookup the $key in a particular context in the current lexicon and return a translated message in the language of the user. Use ``--keyword=pmaketext:1c,2'' for the xgettext utility.
$LH->language_tag
Retrieve the language tag. This is the same method in Locale::Maketext(3). It is readonly.
$LH->encoding(ENCODING)
Set or retrieve the output encoding. The default is the same encoding as the gettext MO file. You can specify "undef", to return the result in unencoded UTF-8.
$LH->key_encoding(ENCODING)
Specify the encoding used in your original text. The "maketext" method itself is not multibyte-safe to the _AUTO lexicon. If you are using your native non-English language as your original text and you are having troubles like:

Unterminated bracket group, in:

Then, specify the "key_encoding" to the encoding of your original text. Returns the current setting.

WARNING: You should always use US-ASCII text keys. Using non-US-ASCII keys is always discouraged and is not guaranteed to be working.

$LH->encode_failure(CHECK)
Set the action when encode fails. This happens when the output text is out of the scope of your output encoding. For exmaple, output Chinese into US-ASCII. Refer to Encode(3) for the possible values of this "CHECK". The default is "FB_DEFAULT", which is a safe choice that never fails. But part of your text may be lost, since that is what "FB_DEFAULT" does. Returns the current setting.
$LH->die_for_lookup_failures(SHOULD_I_DIE)
Maketext dies for lookup failures, but GNU gettext never fails. By default Lexicon::Maketext::Gettext follows the GNU gettext behavior. But if you are Maketext-styled, or if you need a better control over the failures (like me :p), set this to 1. Returns the current setting.

Note that lookup failure handler you registered with fail_with() only work when die_for_lookup_failures() is enabled. if you disable die_for_lookup_failures(), maketext() never fails and lookup failure handler will be ignored.

$LH->reload_text
Purge the MO text cache. It purges the MO text cache from the base class Locale::Maketext::Gettext. The next time "maketext" is called, the MO file will be read and parse from the disk again. This is used when your MO file is updated, but you cannot shutdown and restart the application. For example, when you are a co-hoster on a mod_perl-enabled Apache, or when your mod_perl-enabled Apache is too vital to be restarted for every update of your MO file, or if you are running a vital daemon, such as an X display server.

FUNCTIONS

%Lexicon = read_mo($MOfile);
Read and parse the MO file. Returns the read %Lexicon. The returned lexicon is in its original encoding.

If you need the meta infomation of your MO file, parse the entry $Lexicon{""}. For example:

  /^Content-Type: text\/plain; charset=(.*)$/im;
  $encoding = $1;

"read_mo()" is exported by default, but you need to "use Locale::Maketext::Gettext" in order to use it. It is not exported from your localization class, but from the Locale::Maketext::Gettext package.

NOTES

WARNING: do not try to put any lexicon in your language subclass. When the "textdomain" method is called, the current lexicon will be replaced, but not appended. This is to accommodate the way "textdomain" works. Messages from the previous text domain should not stay in the current text domain.

An essential benefit of this Locale::Maketext::Gettext over the original Locale::Maketext(3) is that: GNU gettext is multibyte safe, but Perl source is not. GNU gettext is safe to Big5 characters like \xa5\x5c (Gong1). But if you follow the current Locale::Maketext(3) document and put your lexicon as a hash in the source of a localization subclass, you have to escape bytes like \x5c, \x40, \x5b, etc., in the middle of some natural multibyte characters. This breaks these characters in halves. Your non-technical translators and reviewers will be presented with unreadable mess, ``Luan4Ma3''. Sorry to say this, but it is weird for a localization framework to be not multibyte-safe. But, well, here comes Locale::Maketext::Gettext to rescue. With Locale::Maketext::Gettext, you can sit back and relax now, leaving all this mess to the excellent GNU gettext framework.

The idea of Locale::Maketext::Getttext came from Locale::Maketext::Lexicon(3), a great work by Autrijus. But it has several problems at that time (version 0.16). I was first trying to write a wrapper to fix it, but finally I dropped it and decided to make a solution towards Locale::Maketext(3) itself. Locale::Maketext::Lexicon(3) should be fine now if you obtain a version newer than 0.16.

Locale::Maketext::Gettext also solved the problem of lack of the ability to handle the encoding in Locale::Maketext(3). I implement this since this is what GNU gettext does. When %Lexicon is read from MO files by "read_mo()", the encoding tagged in gettext MO files is used to "decode" the text into the internal encoding of Perl. Then, when extracted by "maketext", it is "encode"d by the current "encoding" value. The "encoding" can be set at run time, so that you can run a daemon and output to different encoding according to the language settings of individual users, without having to restart the application. This is an improvement to the Locale::Maketext(3), and is essential to daemons and "mod_perl" applications.

You should trust the encoding of your gettext MO file. GNU gettext "msgfmt" checks the illegal characters for you when you compile your MO file from your PO file. The encoding form your MO files are always good. If you try to output to a wrong encoding, part of your text may be lost, as "FB_DEFAULT" does. If you do not like this "FB_DEFAULT", change the failure behavior with the method "encode_failure".

If you need the behavior of auto Traditional Chinese/Simplfied Chinese conversion, as GNU gettext smartly does, do it yourself with Encode::HanExtra(3), too. There may be a solution for this in the future, but not now.

If you set "textdomain" to a domain that is not "bindtextdomain" to specific a locale directory yet, it will try search system locale directories. The current system locale directory search order is: /usr/share/locale, /usr/lib/locale, /usr/local/share/locale, /usr/local/lib/locale. Suggestions for this search order are welcome.

NOTICE: MyPackage::L10N::en->maketext(...) is not available anymore, as the "maketext" method is no more static. That is a sure result, as %Lexicon is imported from foreign sources dynamically, but not statically hardcoded in Perl sources. But the documentation of Locale::Maketext(3) does not say that you can use it as a static method anyway. Maybe you were practicing this before. You had better check your existing code for this. If you try to invoke it statically, it returns "undef".

"dgettext" and "dcgettext" in GNU gettext are not implemented. It is not possible to temporarily change the current text domain in the current design of Locale::Maketext::Gettext. Besides, it is meaningless. Locale::Maketext is object-oriented. You can always raise a new language handle for another text domain. This is different from the situation of GNU gettext. Also, the category is always "LC_MESSAGES". Of course it is. We are gettext and Maketext.

Avoid creating different language handles with different textdomain on the same localization subclass. This currently works, but it violates the basic design of Locale::Maketext(3). In Locale::Maketext(3), %Lexicon is saved as a class variable, in order for the lexicon inheritance system to work. So, multiple language handles to a same localization subclass shares a same lexicon space. Their lexicon space clash. I tried to avoid this problem by saving a copy of the current lexicon as an instance variable, and replacing the class lexicon with the current instance lexicon whenever it is changed by another language handle instance. But this involves large scaled memory copy, which affects the proformance seriously. This is discouraged. You are advised to use a single textdomain for a single localization class.

The "key_encoding" is a workaround, not a solution. There is no solution to this problem yet. You should avoid using non-English language as your original text. You will get yourself into trouble if you mix several original text encodings, for example, joining several pieces of code from programmers all around the world, with their messages written in their own language and encodings. Solution suggestions are welcome.

"pgettext" in GNU gettext is implemented as "pmaketext", in order to look up the text message translation in a particular context. Thanks to the suggestion from Chris Travers.

BUGS

GNU gettext never fails. I tries to achieve it as long as possible. The only reason that maketext may die unexpectedly now is ``Unterminated bracket group''. I cannot get a better solution to it currently. Suggestions are welcome.

You are welcome to fix my English. I have done my best to this documentation, but I am not a native English speaker after all. ^^;

AUTHOR

imacat <[email protected]>

COPYRIGHT

Copyright (c) 2003-2008 imacat. All rights reserved. This program is free software; you can redistribute it and/or modify it under the same terms as Perl itself.