Re: KVO Response for incomplete paths

FROM : Ken Thomases
DATE : Wed May 21 22:57:20 2008

On May 21, 2008, at 2:55 PM, Mike Abdullah wrote:

> On 21 May 2008, at 19:51, Gordon Apple wrote:
>

>> Not having access to the innards of KVO, I wonder how dymanic the
>> process is.  For example, my inspector bindings are through an object
>> controller which binds through MSApp.mainWindow which seems to 
>> successfully
>> switch its focus to a new main window with no problems.  This, to me,
>> indicates that the process operates dynamically according to the 
>> path,
>> rather than statically binding when it is set up.


Correct, except it's not looking up every element along the key path 
every time it needs to access it.  It only needs to update itself when 
one of the objects along the key path reports that the relevant 
property has been changed.


>> So my new question is whether or not I can successfully observe a
>> property that does not exist at registration time, and then have it 
>> observe
>> properly when the path is created.  E.g., register to observe
>> xxx.shadow.angle when shadow does not exist, then add the shadow 
>> object and
>> have xxx.shadow.angle be observed properly.

>
> As long as xxx returns nil for -valueForKey:@"shadow" initially, and 
> then puts out proper KVO notifications when a shadow comes into 
> being, this works just fine.


Search the Foundation Release Notes <http://developer.apple.com/releasenotes/Cocoa/Foundation.html#KVCKVO
> for "Bug Fix in KVO's Dependency Mechanism" for a caveat to this 
pre-Leopard.  Understanding the exact meaning of that caveat is 
something else entirely. ;P

Cheers,
Ken