Chapter 9 Exercise 4 - Which method is best?

[MarkReid]

I've written my add: methods taking id object as an argument and returning an id object. However in one I assign result as an id variable and in the other I leave it as a class object.

-(id) add: (id) f
{
   // To add two fractions
   // a/b + c/d = ((a*d) + (b*c)) / (b * d)
   
   // result will store the result of the addition
   id result = [[Fraction alloc] init];
   int resultNum, resultDenom;
   
   resultNum = numerator * [f denominator] + denominator * [f numerator];
   resultDenom = denominator * [f denominator];
   
   [result setTo: resultNum over: resultDenom];
   [result reduce];
   
   return result;
   [result release];
}

AND

-(id) add: (id) f
{
   Complex *result = [[Complex alloc] init];
   
   [result setReal: real + [f real] andImaginary: imaginary + [f imaginary]];
   
   return result;
   [result release];
}

Can someone tell me which approach is best or does it not really matter?

I don't think I fully understand what
id result = [[Fraction alloc] init];
actually does so any clarification on that would be greatly appreciated too.

My understanding of it is that it performs the same function as Fraction *result but does not explicitly set the object type. As such it is dynamic typed and so it's checked at runtime to see if the methods I call on it work with it or not. Am I correct in thinking this method isn't preferred as errors may appear at runtime rather than compile time and I should use static typing when I know what type of object is being used?

Hopefully that makes sense, not sure it did to me. 

[MarkReid]

One more thing...

I'm releasing result at the end of the method to avoid memory leaks, is it correct to be doing this after I return result. At this point in the text is that the correct way to release result? Am I correct that I can't access it to release it within main and it is a "local" use of memory?

[esc]

One more thing...

I'm releasing result at the end of the method to avoid memory leaks, is it correct to be doing this after I return result. At this point in the text is that the correct way to release result? Am I correct that I can't access it to release it within main and it is a "local" use of memory?

Hi Mark,

Any statement after "return" won't be executed.  Secondly, since you are returning "result", if you "release result" before your method returns, you will essentially be returning nil from your method.

To avoid memory leak, the caller of your method can release the memory allocated, e.g.
fraction_result = [a_fraction add: b_fraction];
[fraction_result release];

But the preferred way, as noted on top of page 191, is that the method is responsible for releasing the memory by "autorelease", e.g.
id result = [[Fraction alloc] init];
[result autorelease];
The book explains this later.  At this point, chapter 9, suffice to have the caller of your method release the result.

Going back to your original post, "id" is a pointer to any type of object.  So yes, the compiler won't flag an incompatible message sent to an id object.  There is flexibility in using dynamic binding...program 9.2 is an example.

Hope this helps.

[MarkReid]

Here's my main from this program, at least the section that I have a question on...

   id dataValue1, dataValue2, result;

   Complex *cp1 = [[Complex alloc] init];
   Complex *cp2 = [[Complex alloc] init];
   Fraction *f1 = [[Fraction alloc] init];
   Fraction *f2 = [[Fraction alloc] init];
   [cp1 setReal:2.0 andImaginary: 3.0];
   [cp2 setReal:4.0 andImaginary: 2.0];

   [f1 setTo: 3 over: 5];
   [f2 setTo: 1 over: 2];
   
   dataValue1 = cp1;
   dataValue2 = cp2;
   
   result = [dataValue1 add: dataValue2];
   [result print];
   
   [cp1 release];
   [cp2 release];
   
   dataValue1 = f1;
   dataValue2 = f2;
   
   result = [dataValue1 add: dataValue2];
   [result print];
   
   [f1 release];
   [f2 release];
   [result release];

I've removed the release from the add: methods. Can you tell me if the following is correct or not please?

result is never alloc/inited in main. It is declared as an object of the id type. Would that be the equivilant to a statement like this if it wasn't an id type variable?

Fraction* result;

Meaning in that in both cases result is just a pointer to memory?

As the memory is alloc/inited within the add: method and that is assigned to result, that means the memory location used by the add method is the same location pointed to by the result variable? So that's why when I release result it clears the memory?

If that's the case am I correct in thinking that I should be releasing result after each print method is called. If not the second add: method call creates a second memory location for an object but the first is not able to be release because nothing now refers to it?

I am sorry there are so many questions, I really appreciate you helping me understand this better.

[skochan]

result is never alloc/inited in main. It is declared as an object of the id type. Would that be the equivilant to a statement like this if it wasn't an id type variable?

Fraction* result;

Meaning in that in both cases result is just a pointer to memory?

Yes, the fact that the id type doesn't use  the * I think adds to the confusion here.  (id is actually a typedef, which is discussed in Chapter 10.  Take a look at the example for the NumberObject typedef there).

As the memory is alloc/inited within the add: method and that is assigned to result, that means the memory location used by the add method is the same location pointed to by the result variable? So that's why when I release result it clears the memory?

Yes!   

Cheers,

Steve

[MarkReid]

Thank you Steve, that helps a lot.

For the last part of my original question, am I correct in thinking that I should be releasing result after each print method is called. If not the second add: method call creates a second memory location for an object but the first is not able to be release because nothing now refers to it?

[skochan]

Sorry, I missed that.  Yes, you are correct in the conclusion that it should be released.

Cheers,

Steve Kochan