Best way to deal with retain properties - objective-c

Question

If you have an object with a property which has a retain setter, which of these is best practice?
1
-(id)init {
if((self = [super init])) {
self->_retainingProperty = [[NSObject alloc] init];
}
return self;
}
2
-(id)init {
if((self = [super init])) {
self.retainingProperty = [[NSObject alloc] init];
[self.retainingProperty release];
}
return self;
}
3
-(id)init {
if((self = [super init])) {
NSObject *obj = [[NSObject alloc] init];
self.retainingProperty = obj;
[obj release];
}
return self;
}
All of these would be coupled with a release in dealloc
Perhaps there is another way I've missed.

Answer 1

I generally just do:
- (id ) init
{
self = [super init];
if ( self )
{
retainingProperty = [[NSObject alloc] init];
}
return self;
}
I would not suggest #2 or #3, unless you are aware they might invoke KVO stuff that you do not intend.

Answer 2

All of the above are fine and broadly equivalent. You can also access the member variable using just its name:
-(id)init {
if((self = [super init])) {
_retainingProperty = [[NSObject alloc] init];
}
return self;
}
Property setters are really just convenience methods to ensure the retain/release dance is done correctly, so if you are doing things correctly you can do without them.
If you've added some custom logic to the setter method, you may want to ensure it's called by always using the property setter syntax. Or alternatively, you may want to deliberately sidestep that logic in some instances, and therefore avoid using it some of the time. It's up to you - whatever works for your use case.

Answer 3

At least two other ways immediately spring to mind. There's direct ivar access:
- (id) init
{
if ( self = [super init] )
{
_retainingProperty = [[NSObject alloc] init];
}
return self;
}
Or you could use the autorelease pool:
- (id) init
{
if ( self = [super init] )
{
self._retainingProperty = [[[NSObject alloc] init] autorelease];
}
return self;
}
Depending on the class of the property, there may also be convenience methods equivalent to the latter.
Personally, I would pretty much always go with the direct ivar access. It is usually considered bad form to call property accessors in init, because they may have side effects that you wouldn't want to happen while the object is incompletely initialised.
Similar considerations apply in dealloc, btw: it's better to release your ivars directly rather than using the property accessors.