error in guess_language in pot.py

[dahool]

I assume this function will extract the language code in a file named like 'thisisapofile.en.po', in that case, line 74 in POTManager:

return os.path.basename(filename[:-3:])

Should be:

return os.path.basename(filename[:-3:][-2:])

File: translations/lib/types/pot.py

Version: 0.8.0

[glezos]

Maybe it would be even better if we used a regular expression?

[dahool]

Replying to glezos:

Maybe it would be even better if we used a regular expression?

Yes. As the proposed solution won't work with languages like pt_br :)

[glezos]

Gabriel, can you maybe attach a patch or point to a bitbucket fork of yours with a commit? I can push it right away.

[dahool]

according to my tests, the attached patch should do the trick.

[glezos]

Seeing the quite restrictive (but OK-looking) first part [-_@\w]+, I was wondering whether this restricts the query too much. Wanting to see it being run on a large Tx instance and check whether any file path of any large project will have some files blocked from this, I run something similar on Txn.

REGEX = "[-_@\w]+\.(['po']+)"
filelist = [(p.object.project, p.filename) for p in POFile.objects.all() if not re.match(REGEX, os.path.basename(p.filename)) and not os.path.basename(p.filename).endswith('.pot')]
print filelist

[(<Project: django-l10n-portuguese>, u'pt_BR/ref/index.txt.po'),
 (<Project: translation of man-pages>, u'ru/LC_MESSAGES/getc_unlocked.3.po'),
 (<Project: greenmachine>, u'base/locale/kr/LC_MESSAGES/Copy of django.po')]

The last two are handled by the first part of the if clause, and the first one is quite a rare occasion.

[dahool]

Right. It could be replaced with something like (.+)\.(?P<lang>[-_@\w]+)\.(po?+)

Of course, the first example, 'pt_BR/ref/index.txt.po' is a totally different scenario. I wonder, how it was managed?

[dahool]

[.+]\.(?P<lang>[-_@\w]+)\.['po']+

Sorry about that. wiki formatting issues :P

[diegobz]

Fixed in r1481.