Re: Obj-C Namespacing

[Helge Hess]

On 04.12.2003, at 22:41, Ziemowit Laski wrote:
> > > That's right.  I believe that adding namespaces to ObjC/ObjC++ 
> > > constructs would actually require adding namespace support to the 
> > > runtime, and runtime metadata. :-(
> > Where do you see problems?
> I can't think of anything too problematic, actually, aside from 
> actually agreeing to do it.

Hehe ;-)

>  One would have to come up with a way to "mangle" ObjC class/protocol 
> names so that they contain namespace qualifications and do not collide 
> with names constructed by hand.  (E.g., mangling 'org.gnustep.MyClass' 
> as 'org_gnustep_MyClass' would not do, although you're probably right 
> that leaving it as 'org.gnustep.MyClass' might just do the trick.)

Well, as you already point out doing like Java is just fine in this 
case ;-) The "." is not valid in an ObjC class or method name.

> > @namespace org.gnustep;
> >
> > @interface MyClass : com.apple.Foundation.NSObject
> > - (id)blub;
> > @end
> >
> > @end
> > ---snap---
> >
> > Should work and produce "org.gnustep.MyClass" as the class name in 
> > the class lookup table.
> > Do I miss anything?
>
> If you did, I can't see it.  :-(  My first inclination was to rely on 
> C++ namespace notation, but that would mean that only ObjC++ users 
> could benefit from the features.  Your notation could work with both 
> ObjC and ObjC++, and seems quite reasonable.

Oh, yes. I would in no way try to mix up C++ and Objective-C. Using 
them in a single source file is OK for me, but reusing C++ features in 
ObjC is just huhuhuhu. Keep them separated.
So yes, I suggest using new, Objective-C specific directives for 
Objective-C namespaces.

Anyway, I wonder whether someone is willing to implement that for some 
upcoming GCC. I would certainly appreciate that and backwards 
compatibility should be very good.

best regards,
  Helge
--
OpenGroupware.org       http://www.opengroupware.org/