Re: KVC

[Chris Farber]

Helge,

In Apple's documentation for setValue:forKey:, they have this under  
compatibility notes:

For backward binary compatibility with takeValue:forKey:'s behavior,  
a method whose name matches the pattern -_set<Key>: is also  
recognized in step 1.
Note that KVC accessor methods whose names start with underscores are  
deprecated as of Mac OS X v10.3.

For backward binary compatibility, unableToSetNilForKey: will be  
invoked instead of setNilValueForKey: in step 1, if the  
implementation of unableToSetNilForKey: in the receiver's class is  
not NSObject's.

The behavior described in step 2 is different from  
takeValue:forKey:'s, in which the instance variable search order is  
<key>, _<key>.

For backward binary compatibility with the behavior of  
takeValue:forKey:, handleTakeValue:forUnboundKey: will be invoked  
instead of valueForUndefinedKey: in step 3 if the implementation of  
handleTakeValue:forUnboundKey: in the receiver's class is not  
NSObject's.


Is this what you are referring to?

Chris

On May 29, 2007, at 5:31 AM, Helge Hess wrote:

> On May 29, 2007, at 11:17, Richard Frith-Macdonald wrote:
> > Hi Helge, It's not clear to me what the problem is with - 
> > setNilValueForKey:
> >
> > The Apple documentation at http://developer.apple.com/ 
> > documentation/Cocoa/Reference/Foundation/Protocols/ 
> > NSKeyValueCoding_Protocol/Reference/Reference.html#//apple_ref/occ/ 
> > instm/NSObject/setNilValueForKey:
> > says that 'The default implementation raises an  
> > NSInvalidArgumentException'
> >
> > Are you saying that the Apple implementation does not do what the  
> > documentation says?
>
> Well, I don't have the time to properly reproduce the exact issues,  
> but I think Cocoa has a Compatibility mode for classes which do  
> implement -takeValue:forKey:.
> Which also involves (overrides) -setNilValueForKey:.
>
> I assume the documentation is correct if you use new-style KVC, but  
> for old-style code (probably the far majority of existing code!) an  
> possibly undocumented Compat mode kicks in.
>
> Greets,
>   Helge
>
> PS: of course we should also update OGo to implement the new API,  
> but thats quite a few places to fix.
> --
> Helge Hess
> http://www.helgehess.eu/