Re: Key-Value Observing implementation

[Gregory John Casamento]

Richard,

--- Richard Frith-Macdonald <address@hidden> wrote:

> On 2005-06-18 04:19:02 +0100 Gregory John Casamento
> <address@hidden> wrote:
> 
> > Sa¹o,
> > 
> > --- Sa¹o Kiselkov <address@hidden> wrote:
> >> Hi. I'm working on a CoreData implementation for GNUstep, but that depends
> 
> >> on
> >> the Key-Value Observing feature. Is somebody already working on it, or may
> I
> >> start it?
> > 
> > I may not be starting on this right away and I don't want to hold you up.  
> If
> > you are ready to start, please go ahead and implement whatever you feel you
> > need to.
> 
> Sounds like my partial implementation is the most advanced code we have at
> present then.
> I could commit it to CVS as a work in progress, or provide it in some other
> way if it's needed.

I agree.  We should start with your implementation, since it sounds like it's
furthest along.

> It implements all the trickier parts of KVO, but misses out all the boring
> details :-)
> 
> The implementation is not as described in the page I put on the wiki ...
> after I wrote that I came to the conclusion that the method using a proxy
> would be too slow and inefficient, so what I actually wrote was code to use
> the runtime to create an observer subclass for each class being observed, and
> change the isa ivar of any observed object to point to that subclass, which
> could handle the observatiuon process, but perform all other methods at the
> same speed as the original.

I think this is okay.  In fact, I believe that Apple does it more or less the
same way.

Later, GJC

Gregory John Casamento 
-- CEO/President Open Logic Corp. (A MD Corp.)
## Maintainer of Gorm (IB Equiv.) for GNUstep.