Accelerators missing from dialogs in KDE 4.3

[David Jarvie]

On Saturday 25 July 2009 22:10:37 Albert Astals Cid wrote:
> A Dissabte, 25 de juliol de 2009, Tom Albers va escriure:
> > Op Saturday 25 July 2009 18:31 schreef u:
> > > A Dimarts, 7 de juliol de 2009, Stephan Kulow va escriure:
> > > > Am Montag 29 Juni 2009 schrieb Albert Astals Cid:
> > > > > A Dilluns, 29 de juny de 2009, Andreas Pakulat va escriure:
> > > > > > On 29.06.09 12:19:18, David Faure wrote:
> > > > > > > On Friday 05 June 2009, Andreas Pakulat wrote:
> > > > > > > > Apart from that this is actually not the purpose (and never
> > > > > > > > has been) of KAcceleratorManager - at least not in the way it
> > > > > > > > used right now. The purpose was to use this during
> > > > > > > > development to easily find good accelerator keys and then
> > > > > > > > assign them in code/ui files.
> > > > > > >
> > > > > > > Did coolo confirm that? Because it doesn't match what I
> > > > > > > remember from kde3 times. I distinctly remember that the goal
> > > > > > > was to have it done automatically when no accels were specified
> > > > > > > by the app developer explicitely.
> > > > > >
> > > > > > Well, that was what I gathered from the commits and source code,
> > > > > > when it was broken some time ago and fixed by Albert. There's a
> > > > > > comment in the code that activates the auto-assignment that
> > > > > > suggests it wasn't meant to be used during runtime, but during
> > > > > > development to find good accelerators.
> > > > > >
> > > > > > Can't find the commits easily at the moment and neither remember
> > > > > > the code file where the comment was.
> > > > >
> > > > > Stephan, can you clarify if the goal for KAcceleratorManager
> > > > > automatic accelerator assignment was:
> > > > >  a) have it done automatically when no accels were specified by the
> > > > > app developer explicitely and so should be always enabled
> > > > >  b) have a translators/developers tool but KAcceleratorManager
> > > > > automatic accelerator assignment should be generally disabled
> > > >
> > > > Hi Albert,
> > > >
> > > > The assignment should be done automatically as history proved that
> > > > it's impossible for a developer and even less possible for a
> > > > translator to pick accelerators that do not clash all the time. So my
> > > > goal was to remove the accelerators from the code in general and only
> > > > leave those that are really essential to the user, e.g. always have
> > > > them on standard actions and assign the free letters by an algorithm.
> > >
> > > Ok, so we can see here that Coolo's intention was to remove
> > > accelerators altogether from code and let KAcceleratorManager manage
> > > everything.
> > >
> > > From the i18n point of view not having accelerators in strings to
> > > translate is not good as non latin1 languages like Japanese do things
> > > like
> > >
> > > msgid "&Source:"
> > > msgstr "???(&S):"
> > >
> > > And without the & marker they can't know if they have to add one.
> > >
> > > So having automatic accelerator assignment is good for developers and
> > > latin1 users but not for the broad i18n community.
> > >
> > > So I see three possibilities here:
> > >  a) Always use automatic accelerator assignment and ignore the i18n
> > > problem b) Try to add as much accelerators in code as possible but
> > > leave the automatic accelerator enabled so that it somehow works if the
> > > developer is not careful enough
> > >  c) Disable the automatic accelerator altogether and force developers
> > > to put correct accelerators
> > >
> > > I'm on b) because c) is asking for the impossible.
> >
> > a) is no option.
> >
> > I like to think: do c), especially when that means we gain even more
> > speed by it.
> >
> > Some other remarks:
> > When we do b) can we somehow ( in debugging mode), make the automatic
> > accelarators different as the manually assigned ones? This way I, as
> > developer, know I need to fix it, and face it, it is a great junior job
> > for a lot of people. Can we make a kaboutdata setting to disable it? This
> > way apps that have completed can maybe get that tiny bit of extra speed?
>
> The problem with c) is that it's impossible to achive, one thing the
> automatic accelerator does is trying to remove duplicate accelerators, so
> if you get two accelerators with A, it tries to find a new non used one for
> one of them. This is something the developer can try to do but he will
> fail, we have programs that embed KParts arbitrarily so it's actually most
> of the times impossible to make sure you are going to get them right.
>
> That's why i'd suggest going with b), that is, not changing anything,
> execpt maybe the documentation that says that KAcceleratorManager is just
> for debugging.

It's time this was decided definitively. I stripped out many of the 
accelerators from strings in KAlarm for KDE4, believing that automatic 
assignment was now the correct way to do it. Reinserting accelerators into 
strings will mess up translations again, so should existing strings be left 
without hard-coded accelerators, or should they be edited to insert them?
 
-- 
David Jarvie.
KDE developer.
KAlarm author and maintainer.
http://www.astrojar.org.uk/kalarm