Hello gettext

[James T. Dennis]

 You'd think that using things like gettext would be easy.  Superficially
 it seems well documented in the Library Reference(*).  However, it can
 be surprisingly difficult to get the external details right.

	* http://docs.python.org/lib/node738.html 

 Here's what I finally came up with as the simplest instructions, suitable
 for an "overview of Python programming" class:

 Start with the venerable "Hello, World!" program ... slightly modified
 to make it ever-so-slightly more "functional:"



	#!/usr/bin/env python
	import sys

	def hello(s="World"):
	    print "Hello,", s

	if __name__ == "__main__":
	    args = sys.argv[1:]
	    if len(args):
		for each in args:
		    hello(each)
	    else:
		hello()

 ... and add gettext support (and a little os.path handling on the
 assumption that our message object files will not be readily
 installable into the system /usr/share/locale tree):

	#!/usr/bin/env python
	import sys, os, gettext

	_ = gettext.lgettext
	mydir = os.path.realpath(os.path.dirname(sys.argv[0]))
	localedir = os.path.join(mydir, "locale")
	gettext.bindtextdomain('HelloPython', localedir)
	gettext.textdomain('HelloPython')

	def hello(s=_("World")):
	    print _("Hello,"), s

	if __name__ == "__main__":
	    args = sys.argv[1:]
	    if len(args):
		for each in args:
		    hello(each)
	    else:
		hello()

 Note that I've only added five lines, the two modules to my import
 line, and wrapped two strings with the conventional _() function.

 This part is easy, and well-documented.

 Running pygettext or GNU xgettext (-L or --language=Python) is
 also easy and gives us a file like:


	# SOME DESCRIPTIVE TITLE.
	# Copyright (C) YEAR ORGANIZATION
	# FIRST AUTHOR <EMAIL at ADDRESS>, YEAR.
	#
	msgid ""
	msgstr ""
	"Project-Id-Version: PACKAGE VERSION\n"
	"POT-Creation-Date: 2007-05-14 12:19+PDT\n"
	"PO-Revision-Date: YEAR-MO-DA HO:MI+ZONE\n"
	"Last-Translator: FULL NAME <EMAIL at ADDRESS>\n"
	"Language-Team: LANGUAGE <LL at li.org>\n"
	"MIME-Version: 1.0\n"
	"Content-Type: text/plain; charset=CHARSET\n"
	"Content-Transfer-Encoding: ENCODING\n"
	"Generated-By: pygettext.py 1.5\n"


	#: HelloWorld.py:10
	msgid "World"
	msgstr ""

	#: HelloWorld.py:11
	msgid "Hello,"
	msgstr ""

 ... I suppose I should add the appropriate magic package name,
 version, author and other values to my source.  Anyone remember
 where those are documented?  Does pygettext extract them from the
 sources and insert them into the .pot?

 Anyway, I minimally have to change one line thus:

	"Content-Type: text/plain; charset=utf-8\n"

 ... and I suppose there are other ways to do this more properly.
 (Documented where?)

 I did find that I could either change that in the .pot file or
 in the individual .po files.  However, if I failed to change it
 then my translations would NOT work and would throw an exception.

 (Where is the setting to force the _() function to fail gracefully
 --- falling back to no-translation and NEVER raise exceptions?
 I seem to recall there is one somewhere --- but I just spent all
 evening reading the docs and various Google hits to get this far; so
 please excuse me if it's a blur right now).

 Now we just copy these templates to individual .po files and
 make our LC_MESSAGES directories:

        mkdir locale && mv HelloPython.pot locale
	cd locale

	for i in es_ES fr_FR # ...
	    do
	        cp HelloPython.pot HelloPython_$i.po
                mkdir -p $i/LC_MESSAGES
	    done

 ... and finally we can work on the translations.

 We edit each of the _*.po files inserting "Hola" and "Bonjour" and
 "Mundo" and "Monde" in the appropriate places.  And then process
 these into .mo files and move them into place as follows:

  	for i in *_*.po; do 
		i=${i#*_} 
		msgfmt -o ./${i%.po}/LC_MESSAGES/HelloPython.mo 
		done

 ... in other words HelloPython_es_ES.po is written to 
 ./es_ES/LC_MESSAGES/HelloPython.mo, etc.

 This last part was the hardest to get right. 

 To test this we simply run:

	$HELLO_PATH/HelloPython.py
	Hello, World

	export LANG=es_ES
	$HELLO_PATH/HelloPython.py
	Hola, Mundo

	export LANG=fr_FR
	$HELLO_PATH/HelloPython.py
	Bonjour, Monde

	export LANG=zh_ZH
	$HELLO_PATH/HelloPython.py
	Hello, World

 ... and we find that our Spanish and French translations work. (With
 apologies if my translations are technically wrong).

 Of course I realize this only barely scratches the surface of I18n and
 L10n issues.  Also I don't know, offhand, how much effort would be
 required to make even this trivial example work on an MS Windows box.
 It would be nice to find a document that would cover the topic in more
 detail while still giving a sufficiently clear and concise set of examples
 that one could follow them without getting hung up on something stupid
 like: "Gee! You have to create $LANG/LC_MESSAGES/ directories and put
 the .mo files thereunder; the Python won't find them under directly
 under $LANG nor under LC_MESSAGES/$LANG" ... and "Gee!  For reasons
 I don't yet understand you need call both the .bindtextdomain() AND 
 the .textdomain() functions."  ... and even "Hmmm ... seems that we
 don't need to import locale and call local.setlocale() despite what
 some examples in Google seem to suggest"(*)

	* http://www.pixelbeat.org/programming/i18n.html

 (So, when to you need that and when is gettext.install() really
 useful?)

 (I gather that the setlocale() stuff is not for simple string
 translations but for things like numeric string formatting 
 with "%d" % ... for example).
 
 

-- 
Jim Dennis,
Starshine: Signed, Sealed, Delivered