[Python-Dev] Re: gettext in the standard library

[M.-A. Lemburg]

Martin von Loewis wrote:
> 
> > Hmm, if your catalogs are encoded in UTF-8 and use non-ASCII
> > chars then the traditional API would have to raise encoding
> > errors
> 
> I don't know what you mean by "traditional" here. The gettext.gettext
> implementation in Barry's patch will return the UTF-8 encoded byte
> string, instead of raising encoding errors - no code conversion takes
> place.

True.
 
> > Perhaps the return value type of .gettext() should be given on
> > the .install() call: e.g. encoding='utf-8' would have .gettext()
> > return a string using UTF-8 while encoding='unicode' would have
> > it return Unicode objects.
> 
> No. You should have the option of either receiving byte strings, or
> Unicode strings. If you want byte strings, you should get the ones
> appearing in the catalog.

So you're all for the two different API version ? After some
more thinking, I think I agree. The reason is that the lookup
itself will have to be Unicode-aware too:

gettext.unigettext(u"Löschen") would have to convert u"Löschen"
to UTF-8, then look this up and convert the returned value
back to Unicode.

gettext.gettext(u"Löschen") will fail with ASCII default encoding.

-- 
Marc-Andre Lemburg
______________________________________________________________________
Business:                                      http://www.lemburg.com/
Python Pages:                           http://www.lemburg.com/python/