[PATCH 0 of 7] Gettext infrastructure and Danish translation

[Martin Geisler]

Hello everybody!

This patch series fixes a bunch of translation problems and finally
adds code to lookup translations from the generated XX.mo files.

* Patch 1 removes some code that tried to make plural forms by
  appending 's'.

* Patch 2 does almost the same, but for templatefilters. You might
  want to *skip* this patch since it looks strange that part of the
  web interface is translated when the rest is not.

* Patch 3 marks some extra strings for translation.

* Patch 4 is sort of unrelated to the translation business -- but the
  context for next patch overlaps with it, so I left it in.

* Patch 5 makes it easier to find the XX.mo files. The gettext module
  would lookup XX.mo in

    /usr/share/locale/XX/LC_MESSAGES/hg.mo

  or something like that. But that makes it annoying to test new
  translations and difficult for people who does not install to that
  location. So I changed i18n.py to do the lookup itself, based on
  code from gettext.py from Python, which should be okay since Python
  code is under a GPL-compatible license. So for language XX
  locale/XX.mo and ../locale/XX.mo are searched for relative to the
  location of i18n.py. This is the same way templater.py uses to find
  the templates directory.

* Patch 6 adds new targets to the Makefile for updating the hg.pot (PO
  template) file and to generate XX.mo files. I don't particularly
  like the update-mo target name I chose -- maybe someone can come up
  with something better.

* Patch 7 has the first part of a Danish translation containing 200
  translations (out of almost 1300). I am definitely not the right
  person to do the entire Danish translation: I prefer English output
  and only did this to test the infrastructure...


Problems:

* Many short strings make little sense in isolation, but this is
  difficult to fix.

* Some strings use abbreviations ('repo', 'rev') which can be
  difficult to carry over.

* Some strings build larger strings from smaller pieces, such as this
  one from mercurial/commands.py:2208:

    ui.warn(_('not removing %s: file %s (use -f to force removal)\n')
            % (m.rel(f), reason))

* Much of the help is in docstrings -- and pygettext.py can extract
  those. But we don't want to translate all docstrings...

  Also, the record extension runs *part* of a docstring through _()
  which is clever, but not such a good idea from a translation point
  of view.

* If this goes in, all tests should be run in the C locale to ensure
  deterministic output.

* The dist target in the Makefile should probably depend on the
  update-pot target.

* It needs testing :-) I don't know if the py2exe stuff works...


But please give it a try and let me know what you think! You can
create a new language by going to the po directory and running

  msginit -l xx

where xx is your language code (not country code: da vs dk). Then edit
away and run

  make update-mo

from the root to compile the .po file into a .mo file. Then set
LANG=xx and run hg.