Re: KVC

[Chris Farber]

On Jun 6, 2007, at 4:13 AM, Helge Hess wrote:

> On Jun 6, 2007, at 24:14, Chris Farber wrote:
> > > produces an exception because the ß can't be converted to ASCII.  
> > > I recommend switching the conversion to UTF-8.
> > I don't believe this is an issue, as you cannot use unicode for  
> > variables or method names.
>
> I'm talking about KVC, not methods and ivars. Key-Value Coding is  
> defined in terms of
>
>   - (id)valueForKey:(NSString *)_key
>   - (void)setValue:(id)_value forKey:(NSString *)_key;
>
> Given your argumentation I could not use @ or - in keys just  
> because they are not allowed in vars/method-names.
>
> No restrictions apply to the key. KVC is not just another way to  
> access methods, its a concept on its own and in fact there are  
> plenty of places where those methods are mapped to something which  
> has nothing to do with methods. Eg one of the more complicated ones  
> is D2WContext which maps keys to a rule system in Direct2Web.
>
> > > Please remember that KVC is not just for accessing methods but  
> > > for arbitary key/value combinations.
> > Perhaps you are thinking of NSMutableDictionary?

There are two parts to KVC: the informal protocol, and the  
implementation. I made the mistake of thinking you were talking about  
the implementation for two reasons. In your example, you used  
takeValue:forKey:; if you had overridden takeValue:forKey:, then it  
would not be our implementation at fault, and if you had not I could  
only assume that you were trying to set an ivar or call a method,  
neither of which can contain unicode. I also had not been thinking  
about key paths…

> This is one of the simplest objects which modify the way KVC works.  
> EOGenericRecord is another one and a good example where unicode  
> plays a role (Given that eg PostgreSQL table, column names can be  
> arbitary strings).
>
> > KVC is only for setting properties via appropriately named methods  
> > or direct access to instance variables.
>
> Sorry??? Have you ever used WebObjects or EOF (or a clone of  
> those)? :-) The power of KVC is that its separate. Eg it allows you  
> to do trampolines like this:
>
>   @implementation MyComponent
>   - (id)labels {
>     return self->labelHandler;
>   }
>   @end
>
>   @implementation MyLabelHandler
>   - (id)valueForKey:(NSString *)_key {
>     return [[NSBundle bundle] labelForKey:_key];
>   }
>   @end
>
>   LocalizableString.strings:
>   WelcomeTitle = "Hello Friendly Visitor!!!";
>
> And then in the WO template:
>   <#WOString var:value="labels.WelcomeTitle" />

In this, WebObjects should be invoking valueForKeyPath:, and you are  
totally correct - we need to support unicode here as the objects  
along the path may or may not be using the default implementation of  
KVC. Thanks for pointing this out.

Chris

> Anyways, to make a long story short: if you want to be compatible  
> with Cocoa, you need to allow arbitary keys. This should hopefully  
> stop further argumentation ;-)
>
> Thanks,
>   Helge
>
> PS: its just a matter of time until UTF-8 is allowed in selectors.  
> Java already does this (which BTW also affects KVC in combination  
> with Java bridges ...)
> --
> Helge Hess
> http://www.helgehess.eu/
>
>
>
>
> _______________________________________________
> Discuss-gnustep mailing list
> Discuss-gnustep@gnu.org
> http://lists.gnu.org/mailman/listinfo/discuss-gnustep