As part of localization, we have to pull all of the "strings" out of the code and templates. Strings are just copy, labels, and other words shown to users.
Our current script scripts/extract_po.sh is okay, but lacks the following features:
If you read extract_po.sh you'll be horrified to find that we parse JS files via xgettext in Perl mode. Similarly, we EJS files are parsed via xgettext in PHP mode.
Ideally we'd either use Translate Toolkit, patch Translate Toolkit with Node.js support, patch GNU Gettext (good luck with that) or ... maybe you have some ideas.
Gettext comments mean that I can do this
/* l10n: This %s will be the user's name */
var msg = "Hello %s, welcome back!";
After string extraction, the po file has not only our string, but a special comment, so the localizer knows what %s is supposed to mean.
@ozten is happy to mentor first time contributors on this issue.
When I extract new strings using extract_po.sh, I get different POTs generated if I leave existing POTs in the templates directory than if I empty the old POTs.
The difference is that strings that we remove for the new releae don't get removed from the POT if I keep the old POTs, but they do get removed if I empty the POTs before running extract_po.sh. So what I do now is every time before I run extract_po.sh, I empty messages.pot and client.pot in the templates folder.
(Should I file another bug?)
@mathjazz - please file a new bug. You basically want to make client.pot and messages.pot an empty file at the beginning of extract_po.sh, right? I can do that ASAP.
I started a proof-of-concept here.
@zaach - you finished this work up, didn't you?
@shane-tomlinson Yes, it's currently being used in the standalone i18n-abide. #1322 tracks when we'll pull that into browserid (post beta1).
Closing this in favor of #1322, thanks @zaach
OK. Verifying as such.