Re: RFA: [base] GSObjCRuntime

[David Ayers]

Richard Frith-Macdonald wrote:

> On Tuesday, March 25, 2003, at 12:31  pm, David Ayers wrote:
>
> > GS_EXPORT IMP GSObjCGetMethod(Class class, SEL sel); /*Notice I 
> > changed the name again ;-) */
> > GS_EXPORT IMP GSObjCReplaceMethod(Class class, SEL sel, IMP imp);
>
> Sorry to be awkward but ... 

Please do so ;-)

> while I have no actual objection to you comitting what you've done,
> wouldn't it be even better to make the name of the function 
> GSObjCSetMethod rather than GSObjCReplaceMethod ... for consistency 
> with the ivar management functions, and to add functionality such 
> that, if the method does not already exist in the class, it is added 
> (and a null
> IMP returned for the old value of the method implementation).

This sounds like a good implementation for GSObjCSetMethod, but that's 
not what GSObjCReplaceMethod is currently doing.  I have no objections 
to implementing GSObjCSetMethod as you describe it, but this would be 
more than just replaceing the implementation pointer.  We would need to:
- find/create the corresponding struct objc_method_list
- augment it with an struct objc_method at the end,
- set SEL method_name,
- set const char* method_types /* signature*/
I think the hardest part would be extending an existing list.  It would 
be relativly easy if we added a new list with a single entry for each 
new method, but that feels very inefficient.

I'll leave this until it is needed but feel free to implement it ;-)

I think we should stick with GSObjCReplaceMethod for the implementation 
we have now.

Cheers,
Dave