README for libintl-perl
The package libintl-perl is an internationalization library for Perl that aims to be compatible with the Uniforum message translations system as implemented for example in GNU gettext.
See the file COPYING and the source code for licensing.
The library is entirely written in Perl. It should run on every system with a Perl5 interpreter. The minimum required Perl version should be 5.004.
The behavior of the package varies a little depending on the Perl version:
- Perl 5.8 or better
Recommended. Perl 5.8 offers maximum performance and support for various multi-byte encodings (even more if Encode::Han is installed). Additionally the output charset is chosen automatically according to the information provided by I18N::Langinfo. In fact, I18N::Langinfo is already available for Perl 5.7 but this developer version is probably not much in use any more.
- Perl 5.6 or better
Still offers high-performance UTF-8 handling but no support for other multi-byte encodings unless the package Encode is installed.
- Earlier Perl versions
Full UTF-8 support but quiet slow since all conversion routines are written in Perl. More exactly: Encoding from 8 bit charsets into UTF-8 is reasonably fast and usable. Decoding UTF-8 is slow, however.
Note that these are actually the requirements for the users of your software internationalized with libintl-perl.
As a maintainer of a Perl package that uses libintl-perl, you will also need a recent version of GNU gettext (see the file README in the subdirectory "sample/" of the source distribution of libintl-perl). Translators of your software can basically do their job with any text editor, but it usually makes sense for them, too, to have GNU gettext installed. End users of your software, or people that install an internationalized Perl package do not need it, unless they want to add a new language to your software.
If libintl-perl is not installed on your system, you have to build it from the sources, which is a lot easier than you may think. You need the program "make" for that and a command line shell. No C compiler is required.
Unpack the package in a directory of your choice, cd into that directory and then type
perl Makefile.PL make
This will build the package. You can then run the tests with
To install the package, type
You will probably need root permissions to do that.
Of course, you can also use the CPAN module to install the package.
Send negative (and positive!) feedback to me. Bug reports can be send directly to me or you can use the RT bugtracking system.
If you use libintl-perl for your project, private or public, free or commercial, please let me know. I am interested in such information.
The primary design goal of libintl-perl is maximum compatibility with the gettext functions available for other programming languages. It is intended that programmers, translators, and end users can fully benefit from all existing i18n tools like xgettext for message extraction, msgfmt, msgmerge, etc. for catalog manipulation, Emacs PO mode (or KBabel, PO-Edit, ...) for catalog editing and so on.
Another design goal is maximum portability. The library should be functional without any additional software but with a wide range of Perl versions. Wherever possible, hooks have been inserted to benefit from advanced features in the runtime environment, but the basic functionality should be present everywhere.
The core of the library is the module Locale::gettext_pp. It is a pure Perl re-implementation of the module Locale::gettext available on CPAN. However, the XS version Locale::gettext lacks some functions (notably plural handling and output conversion) that are already present in Locale::gettext_pp. Locale::gettext_pp provides the internationalization functions that are available in your system library (libc) or additional C libraries (for example libintl in the case of GNU gettext).
The class Locale::Messages is an additional abstraction layer that is prepared for dynamic switching between different gettext implementations (for example Locale::gettext_pp and Locale::gettext). It provides basically the same interface as Locale::gettext_pp but in an implementation-independent manner.
The module Locale::TextDomain is the only module that you should actually use in your software. It represents the message translation system for a particular text domain (a text domain is a unique identifier for your software package), makes use of Locale::Messages for message translation and catalog location, and it provides additional utility functions, for example common shortcut names for i18n routines, tied hashes for hash-like lookups into the translation database, and finally an interpolation mechanism suitable for internationalized messages.
The package also contains a charset conversion library Locale::Recode. This library is used internally by Locale::gettext_pp to allow on-the-fly charset conversion between the charset in a message catalog and the preferred (end) user charset. Its main advantage about the Encode package available for recent Perl versions is its portability, since it does not require the Unicode capabilities of Perl that were introduced with Perl 5.6. It fully supports UTF-8 with every Perl version and a wealth of common 8 bit encodings. If you have to do charset conversion with older Perl versions, then Locale::Recode may be worth a try although it is really only a helper library, not intended as a competitor to Encode.
For a basic understanding of message translation in Perl with libintl-perl you should read the perldoc of Locale::TextDomain. Don't bother about the documentation of the other modules in the library, you will not need it unless you want to hack the library yourself.
In order to make use of the software, you will also need various tools from GNU gettext savannah. The documentation is located at www.gnu.org. You will find there a language-independent overview of internationalization with GNU gettext, and in the Perl-specific sections you will find details about the parser that extracts translatable messages from your Perl sources.
The subdirectory "sample" of the source distribution of libintl-perl contains a full-fledged example for an internationalized Perl package, including a working Makefile. The README of that subdirectory explains all necessary steps.
However, if you are on a recent GNU/Linux system or similar (cygwin should also do), chances are that you can get the following example to run:
#! /usr/local/bin/perl -w use strict; # This assumes that the textdomain 'libc' is available on your # system. Try "locate libc.mo" or "locate libc.gmo" (or # "find / -type f -name libc.mo" if locate is not available on # your system). # # By the way, the "use Locale::TextDomain (TEXTDOMAIN) is the # equivalent of # # textdomain ("TEXTDOMAIN"); # # in C or similar languages. use Locale::TextDomain ('libc'); # The locale category LC_MESSAGES is not exported by the POSIX # module on older Perl versions. use Locale::Messages qw (LC_MESSAGES); use POSIX ('setlocale'); # Set the locale according to our environment. setlocale (LC_MESSAGES, ''); # This makes the assumption that your system libc defines a # message "No such file or directory". Check the exact # spelling on your system with something like # "ls NON-EXISTANT". # Note the double underscore in front of the string. This is # really a function call to the function __() that is # automagically imported by Locale::TextDomain into your # namespace. This function takes its argument, looks up a # translation for it, and returns that, or the unmodified # string in case of failure. print __"No such file or directory", ".\n"; __END__
Now run the command "locale -a" or "nlsinfo" to get a list of available locales on your system. Try the section "Finding locales" in "perldoc perllocale" if you have problems.
If, for example, the locale "fr_FR" is available on your system, set the environment variable LANG to that value, for a POSIX shell
LANG=fr_FR export LANG
for the C shell
setenv LANG fr_FR
and run your little Perl script. It should tell you what the error message for "No such file or directory" is in French, or whatever language you chose. Not a real example, because we have "stolen" a message from a system catalog. But it should give you the general idea, especially if you are already familiar with gettext in C.
If you still see the English message, this does not necessarily mean a failure, since the string is maybe not translated on your system (try "locate libc.mo" to get a list of available translations). Even for the translations listed there, that particular message might be missing. Try a common locale like "de_DE" or "fr_FR" that are usually fully translated then.
Your next steps should be "perldoc Locale::TextDomain", and then study the example in the subdirectory "sample" of this distribution.
Have fun with libintl-perl!