gimp/README.i18n

155 lines
6.4 KiB
Plaintext
Raw Normal View History

This document exists to document the important things to care for
because of locale support.
Actually this one is maintained by me, that is Daniel Egger
2000-01-02 00:08:09 +08:00
(Daniel.Egger@rz.fh-muenchen.de).
1. Why localisation?
Many persons from many countries start to get used to Linux.
Unfortunately not everyone is able to understand English. But
2000-01-02 00:08:09 +08:00
even those people sometimes like to use good and free software
without using a dictionary to get the unknown words.
So why not simply localise the software to make it available to
2000-01-02 00:08:09 +08:00
the mass which isn't wholly English native? Of course this also
eases the migration from PhotoX to GIMP. :))
2. How?
GNU provides a very nice package called gettext. This one offers
the possibility to translate chosen messages from the native language
of the program into that one of the users if a necessary catalog is
provided. Gettext therefor provides some easy tools to create and maintain
such catalogs and a few functions which can be called by the program to
enable automatic translation at runtime. The program gets linked to the
gettext library or glibc2 which already provides that functionality
and everything is fine.
By the way: gettext is a fixed part of glibc2 but will be shipped with
GIMP and so can be automatically compiled on every platform GIMP itself
runs on.
3. Deep inside...
GIMP provides header files called gimpintl.h and stdplugins-intl.h in the
libgimp directory which check whether gettext is available on the system
which GIMP is compiled on and will deactivate language support if it's not.
You CAN use such a compiled GIMP even without the catalogs or on a system
2000-01-02 00:08:09 +08:00
which dosen't have language support.
If the gettext system is there it will declare 3 functions which will be
described below.
3.1 _() [more correctly: char * _( char * )]
2000-01-02 00:08:09 +08:00
This one is a macro for the function gettext(). You can wrap any text with
it that is allowed to be a return value of a function. If you use it then
libintl will try to translate it into the native language of the
user according to his/her environmental settings.
2000-01-02 00:08:09 +08:00
The gettext() function will do a lookup in the hashed catalog which contains
all the translated texts.
- If it is found a pointer to the string will be returned to the caller.
2000-01-02 00:08:09 +08:00
- If not, the caller will receive a pointer to the original string.
This way it is ensured that there isn't any harm caused to the program
2000-01-02 00:08:09 +08:00
(i.e. The GIMP) if no useful catalog is installed.
Please note that it is important to use _() directly (and not gettext())
for simple messages because of reasons that will be mentioned below.
NOTE: I know some of the developer like short functions like _() but
for a better source understanding I suggest to use it consistently only
2000-01-02 00:08:09 +08:00
for text (like _("That's text!")) and not for variables (like _(text) ) BUT
gettext(text) instead.
2000-01-06 03:57:02 +08:00
3.2 N_() [more correctly: const char * ( const char * ) ]
This one is a macro for the function gettext_noop(). As you can see and
2000-01-02 00:08:09 +08:00
guess it doesn't really do anything in the programm i.e. it is a dummy
macro but nevertheless important. As it isn't possible to call functions
in a structure as seen here:
struct blurb
{
_("This won't work\n");
}
you have to do it in some other way. In GIMP such structures are often used
2000-01-02 00:08:09 +08:00
to create menus or similar things very simply. Here you have to use the
dummy to allow the generation of the template catalog which will be described
below. This one doesn't do anything but it marks the text as important to
the xgettext extractor.
2000-01-02 00:08:09 +08:00
The text has to be translated manually with the next function.
3.3 gettext()
This function is the same as that mcaro in 3.1. But there is one big
difference: The _()'s and N_()'s are the only expressions which get parsed
by the template generator.
If you have strings that should be translated but are unfortunately in a
structure you have to do that on your own which means that you have to
parse the fields with the messages in a loop and translate the texts with
this gettext() function.
Please note that it may be necessary to free or allocate memory in this
case!
4. Some magic...
As you have seen we only did the programming part until now but this isn't
all by far.
To use catalogs we'll have to create them. Now there are 3 different files
which are importart:
gimp.pot:
This one is the so called template. It contains the messages which are
extracted from the sources and empty fields which have to get filled by the
author. It is used to start a new catalog or to update the an already
available one.
The Makefile will automatically call the program gettext which will extract
all messages that are wrapped by a _() or a N_() (but NOT gettext()) and
concat them to this template.
[language].po:
This file has to be an edited gimp.pot and contains the original messages
plus the translated ones. This file will be delivered together with GIMP
and is the base for the final catalog.
[language].mo:
This file is a compiled version of [language.po] which will be
automatically compiled by the Makefile system and installed in the locale
directory of the system. It contains everything that the .po file
contains except not translated messages, comments and other overhead.
For maximum speed it is also hashed to allow gettext a faster search.
5. Tools and how to use them...
As mentioned the to get translated string are extracted directly from the
source and written to the template.
I guess many of you will now ask if it is necessary to add new strings
directly to the template or if there's a tool to achieve that.
I think I can calm down those of you who fear lots of had work just to
update the language files. There's a program called msgmerge which will
add all strings that are in the template but not in the uncompiled catalog
to it. Msgmerge does this job very nicely and also tries to use some kind
of fuzzy logic method for already translated strings for possible reduction
2000-01-02 00:08:09 +08:00
of translators work: If a original string seems similar to a new one
and it already has a translation, it will be taken over to the new catalog
together with a remark that this one may not necessarily fit.
6. And more?
I hope I mentioned everything that is worth it and hope that this document
will clarify some things. If it doesn't please write me a mail and tell me
2000-01-02 00:08:09 +08:00
what you want to know. This text of course contains errors, so if you find one
tell it to me, too....
Happy Gimping. Yours,
2000-01-06 03:57:02 +08:00
Daniel Egger