Re: [Gnustep-cvs] Fix NSCharacterSet caching bug

[David Ayers]

Richard Frith-Macdonald  wrote:

> *+* * Source/NSCharacterSet.m 
> *+* ([NSCharacterSet _bitmapForSet:number:]): Restore caching while
> *+* retaining the intended behavior of the previous change (is that
> *+* intended behavior correct?).
> *+* Perhaps should maintain a separate cache for each subclass ...
> *+* or should that be the responsibility of the subclass itsself?

Actually my intent was *not* to retain a cache for subclasses 
(especially mutable sets). If I want mutable set, then I generally mean 
to alter it (like inverting itn).  If the next caller get's the altered 
cached object, he's not getting what he asks for.  We can only cache 
immutable versions, and then each subclass must do it's own caching.  
(Maybe I should have added a comment.)  But I'll add an extra test.

Yet (except for the caching) I'm not sure what the real change is.  
Instead of:

        if ([self isSubclassOfClass: [NSCharacterSet class]])

we now have

        if (self != abstract && self != [set class])

where the [set class] is always initialized by:

        set = [NSCharacterSet characterSetWithBitmapRepresentation: data];

Yet if it's not the same class but it invokes this implementation must'n it be 
the a subclass?

So in that sense, your change didn't alter the behavior.
 

Cheers,
David