[PyKDE] Feedback Needed on Possible tr() Changes

[Phil Thompson]

On Tuesday 27 May 2003 7:40 am, Jan L. Nauta wrote:
> > On Monday 26 May 2003 2:55 pm, Jan L. Nauta wrote:
> > > > This still doesn't reproduce the C++ behaviour. If somebody
> >
> > ships a class
> >
> > > > library, implemented in Python and including translation files,
> > > > then there is the possibility that an application translation will be
> > > > used instead of the
> > > > intended class library one.
> > >
> > > Well, it _almost_ reproduces the C++ behaviour. But also doesn't
> > > require any changes to existing (generated) Python code. And it even
> > > has the benefit that you can 'override' translations, which depending
> > > on your viewpoint can also be a disadvantage.
> > >
> > > I also don't agree with your counter example. If somebody ships
> >
> > translation
> >
> > > files they will be searched _before_ the application translation. The
> > > application translation is only used as a fall-back. In more
> > > complicated situations, you could use a list of translators. This
> > > allows
> >
> > you to control
> >
> > > the sequence used to search the translators.
> >
> > # Assume translations exist for LibClass context.
> > class LibClass(QObject):
> > 	def hello(self):
> > 		return self.tr("Hello")
> >
> > # Assume translation don't exist for AppClass context.
> > class AppClass(LibClass):
> > 	def welcome(self):
> > 		return self.tr("Hello")
> >
> > a = AppClass
> > a.hello()
> > a.welcome()
> >
> > With the current implementation a.hello() will wrongly fail to be
> > translated.
> > With your scheme a.welcome() will be wrongly translated (using
> > the LibClass
> > context). I think.
>
> You're correct ofcourse, that is what I called the 'override' translations.
> Although it's not the C++ way, I think it shouldn't be a problem in
> practice. If it happens you could always use
> qApp.translate("AppClass_Workaround", "Hello") as a workaround, in your own
> class. But normally, and I know it's just my perception, the translations
> should match. Because although the contexts don't match they are in the
> same hierarchy path, and therefor are related. So I think the method I
> proposed works better than the existing one, and doesn't require many
> changes to Sip/PyQt/PYUIC or existing code.

I'm coming to the conclusion that no option is good enough and so, for the 
moment, I'm inclined to...

- leave it as it is until I see if SIP v4 offers a better option
- properly document the problem and recommend using qApp.translate() to make 
sure the problem can be avoided and that will continue to work in the future
- change pyuic so that it uses qApp.translate().

Phil