Re: [Plone-i18n] i18dude improvements

[Hanno S. <pl...@ha...>]

Hi.

Tarek Ziadé wrote:
> i18ndude is cool

Thx :)

> but I'd like to use it with my eggs using several
> distutils extensions,
> instead of a console script. (like pylons does i think)
> 
> so i can call for example:
> 
> python setup.py i18n-merge

I'm not quite sure if this is a good match. Guessing the po/pot files
combinations inside a package seems to involve quite some automagic
work. I think I prefer explicit over implicit here.

My second problem with this, is that it involves making i18ndude
available in some way to distutils (not sure how that works exactly,
though). You cannot do that globally as noted in the other mail, as
i18ndude needs to run in a sandbox to not interfere with other projects.

I would prefer this to be done as a separate extension package to
i18ndude at this point I guess.

> Furthermore, I think there's a missing feature that is going to be helpfull
> for plone 3
> (if it's not already beeing develop in some branch :) ):
> 
> compiling .mo files found in the egg. (This can be done for example with
> po_lib).

See http://dev.plone.org/plone/ticket/7157. I consider it a bug that
only po files inside products are compiled automatically to mo files in
Plone 3. Po files inside packages should also be compiled automatically.
The code that does this currently lives in PTS and might move down to
zope.i18n at one point. The actual compilation uses my python-gettext
package to be cross platform compatible.

For i18ndude I don't quite see this as a missing feature as you can just
easily compile those files yourself with the offical gettext binaries
(in this case 'msgfmt'). Writing a simple script to iterate over all
files inside a folder and calling 'msgfmt foo.po -o foo.mo --no-hash'
seems to be the easier way of doing this. The --no-hash is not needed
but will save a bit of space, as the hash table inside mo files isn't
used by the Python Gettext implementation.

Hanno