I want to release the superclass of a class. And then from one of the subclass’s methods re-initialise it but with different values. Moveover, if A is the super class and B is the subclass. I want to release A from inside a B method calls. Something like this:

@interface A : NSObject { int value; }

• (A*) initWithValue:(int)aValue; @end

@interface B : A {

}

• (void) setValue:(int)aValue; @end

@implementation B

• (void) setValue:(int)aValue { [super release]; super = A alloc] initWithValue:aValue]; } @end

I am having problems using the ‘super’ keyword in this scheme. I cannot assign it a value.

The reason I want to do this is build a Mutable version (class B) of an existing class (class A). So I only want the subclass to have the -set: method. And I don’t want to make any changes to the existing Non-mutable class (class A).

This approach might be undesirable but it is possible?

Cheers,

– DJF

---

This does not make any sense whatsoever, I’m afraid. super is not a separate object. It is the same object as self. The only difference is that if super is the target of a message send, the dispatch is performed a bit differently. But it doesn’t make any sense to reassign super because it’s just self. Likewise releasing super is just releasing self.

If you are going to create this along this design, then you will probably need to make it a has-a relationship rather than an is-a relationship, so have A as an ivar instead of as your superclass.

---

Yes; going off of what the previous poster said, you might as well have A as an instance variable…but also, if A is something like [[NSNumber, you can cheat a bit by subclassing it and passing along messages to an actual value instance variable. –JediKnil

@interface MutableNumber : NSNumber { NSNumber *value; } // Implement forwardInvocation: and methodSignatureForSelector: // to forward messages to value

• (void)setIntegerValue:(int)newValue;
• (void)setDoubleValue:(double)newValue // etc…

---

What you are trying to do is actually accomplished with “self”, but only really possible in the init method:

@implementation B

• (void)initWithValue:(int)aValue { [self release];

  self = [[A alloc] initWithValue:aValue];

  return self; } @end

Doing this at any other point is potentially disastrous because other people could still have a reference to the original object, and there is no way to notify them that it has changed.

---

Hi,

Thanks for your response! Yes I knew it was a bit of a screwy question from start; it’s never a good idea to fight the language (or framework) one works with! The reason for doing this was one of lazyness, e.g. it was easier to create a ‘new super’ than it was to workout exactly what instance variables of the super class to update etc., in the end that’s what I did.

Thank you to the first poster for explaining about super and self I didn’t appreciate that they were that same object; the standard init methods make a lot more sense now!

Cheers,

–DJF