Wanting to figure out how to support multiple languages in my applications I read the gettext documentation and got to the part saying one should use pygettext.py to create .po files.

After copying that program from the python SVN repository I later found out that it cannot cope with plurals, i.e. ngettext(). xgettext can do that just fine. Since pygettext.py was not modified in 5 years I propose changing the documentation of gettext to encourage xgettext rather than pygettext.py.

My proposed changes are attached (edited as text file; I do not know how to convert it to a webpage and hence did not test that).
I used the 2.7b1 documentation, but the patch appears to apply just as well to the 3.1.2 documentation.

Does the fixing of #5464 solve your problem?

It looks like msgfmt.py now parses 'msgid_plural' but pygettext.py does not produce these. It is still oblivious to plurals as produced by ngettext(). My originally proposed change to the documentation was to point people to the GNU xgettext/msgfmt tools rather than the pygettext.py/msgfmt.py pair because of this limitation.

This has not changed.

What about the more ambitious approach, fixing pygettext?

Since the GNU tools work fine I had no reason to look into pygettext.py.
Now I did it anyway...

I added ngettext to the default keywords.
Any function that is keyworded and supplied more than 2 keywords is treated like ngettext.
Also simple constructs like _("foo" + 10*"bar") are now possible.

Patch attached.

A simple test case for the earlier changes.

run pygettext.py and msgfmt.py and move the .mo file to 
./locale/de/LC_MESSAGES/helloworld.mo 
to see that it prints the localized strings.

Thank you for making a patch.  I’ve been busy and could not find time to look at it, but after I add a testing machinery for scripts like pygettext I will look into this.