Question

There are a few concepts about iPhone memory management that have got me confused, so I was hoping that someone could clarify these.

(a) Say that I am calling a function which returns an image, and I want to store that returned image in a variable so that I can assign it to various other variables without having to re-call the image generation function each time. At the moment I am doing the following:

UIImage *storedImage = [someFunction returnImage];

and because the storedImage is not alloc‘ed I am not releasing the variable.

However, should I be explicitly alloc‘ing and releasing the UIImage instead?

UIImage *storedImage = [[UIImage alloc] init];
storedImage = [someFunction returnImage];
...do stuff...
[storedImage release];

What is the implication of doing the direct assignment without alloc rather than alloc‘ing the variable and then assigning?

(b) In the init method for various classes I am setting up the instance variables. At the moment I am doing the following:

self.arrayVariable = [[NSMutableArray alloc] init];

However, I have seen others do the assignment this way:

theArrayVariable = [[NSMutableArray alloc] init];
self.arrayVariable = theArrayVariable;
[theArrayVariable release];

or

theArrayVariable = [[NSMutableArray alloc] init];
arrayVariable = theArrayVariable;
[theArrayVariable release];

…which is the same as above, but without the self.

What is the difference between the three methods and which is better to use?

Answer 1

Regarding returning objects from methods, you should always return an autoreleased object from any method which does not begin with the name alloc, new, or contains copy. This is defined in Apple’s object ownership policy, which states that you own any object that you create (“create” is defined as an object which you sent retain to, or used any of the aforementioned messages to retrieve an object), and that you are responsible for relinquishing ownership of that object by sending it the release or autorelease message.

The first method using self uses the property setter to set the instance variable to the argument (in this case whatever is on the RHS of the assignment).
This will do whatever you specified in your @property declaration (for example if you specified retain, the setter will retain the new value and release the old value).

The second method sets up a pointer to an NSMutableArray and passes it off to your property setter via self, which will most likely retain it, thereby bringing the reference count up to 2, since the object was previously alloc-ed, so you need to release it after this line to bring it back down to 1.

The third method will not work, because you are releasing an object with a reference count of 1 at the point of invoking release. How so you ask? Well, the first line sets up a pointer to an alloc-ed object, then directly assigns it to your instance variable, which will just point the ivar to the same object that theArrayVariable is pointing to. Then, that same object that theArrayVariable is pointing to gets sent the release method, which will effectively bring down the reference count of your ivar as well as the receiver, to 0. At this point both your instance variable and theArrayVariable will get deallocated.