Use gettext-po parser for imports

Here is the file that is causing us performance problems

Which features are missing that you require?

Hmm, checking what we have right now seems like it has most of the required functionality, at least I didn't find anything that we cannot workaround except for a limitation itself in gettextpo, the fact that we don't have a way to parse a .po file from memory, it requires that the file exist in the filesystem.

Taking another look to the code, I see that it could read from standard input if you set filename to '-' so maybe we could add something to the PoFile.__init__ method to accept a buffer and to provide it to libgettext using standard input.

What do you think? Would that be possible?
If that's done, I think we could try to switch to this other parser.

Parsing from stdin does not sound like a good idea. Are you planning on doing dup2() tricks to open a new file descriptor over the top of it?

If you need a named file, just use tempfile.NamedTemporaryFile. Something like:

     fp = tempfile.NamedTemporaryFile()
     fp.write('data')
     fp.flush()

Now you have fp.name as the filename, and the file will be removed when the NamedTemporaryFile instance goes away.

Reduced to high, we will try to do this as part of our urgent tasks before Gutsy is open to translations

We are not doing this in the near future. We implemented an easy fix that improve the speed a lot. From more than 2 hours to 12 minutes!