Re: KVC

[Helge Hess]

On Jun 6, 2007, at 20:55, Chris Farber wrote:
> The other thread about KVC and unicode reminded me of this email.  
> Are you talking about the behavior of valueForKeyPath: here?

Not sure what you mean. Is there anything missing in the quoted  
email? Its not related to pathes.

> > I just noticed a minor difference between Cocoa and gnustep-base  
> > KVC. Both support the new
> >   -setValue:forKey:
> > method.
> >
> > However, on Cocoa -takeValue:forKey: still seems to be called if  
> > the method isn't implemented (I suppose to ensure compatibility).  
> > On GNUstep it just fails (the unknown key exception).

Maybe if I re-explain it. In MacOS 10.4 (or 10.3?, don't remember)  
Apple changed the KVC API. The selector:

  takeValue:forKey:

got renamed to

  setValue:forKey:

( what a great idea ... ;-/ )

Since a lot of code (including OGo ;-) was written before MacOS  
10.3/4, the code does not override setValue:forKey: but  
takeValue:forKey: if it wants to modify the default KVC behaviour of  
an object.

Now GNUstep only seems to support the new API (setValue:forKey:) for  
code which changes the KVC behaviour. That is, our custom  
takeValue:forKey: implementations are not called anymore.
On Cocoa it seems to be different. If your class overrides - 
takeValue:forKey:, but not -setValue:forKey:, KVC will still call  
your overridden -takeValue:forKey: method.

Maybe I'm wrong. As Richard suggested it would be best to come up  
with a series of tests for the KVC behaviour in various setups.

Thanks,
  Helge
--
Helge Hess
http://www.helgehess.eu/