Question

Working in C# and Java, I’ve seen basically one way everybody initializes singletons:

static obj _inst = null;

obj getInstance() {
  if (_inst == null) {
    _inst = new obj();
  }
  return _inst;
}

Now, when I move to Objective-C for the iPhone, whenever I see code samples, I see basically the same thing:

static obj _inst = nil;

+ (obj *) sharedObj {
  if (!_inst) {
    _inst = [[obj alloc] init];
  }
  return _inst;
}

There’s a class method +initialize that’s called on every class in the Objective-C runtime before it’s used. Is there any reason I couldn’t use that to create my singletons?

static obj _inst = nil;

+ (void) initialize {
  if (self == [obj class]) {
    _inst = [[obj alloc] init];
  }
}

+ (obj *) sharedObj {
  return _inst;
}

It works just fine when I try it in my code, and it gets rid of checking to see if it exists every time before it’s accessed. Is there any reason I shouldn’t create singletons this way?

Answer 1

As far as I’m concerned +initialize is the way to do it. Peter Hosey suggests a couple of other catches (inside -init and -allocWithZone:) to make sure you can’t instantiate more than one instance of the class, ever. Thus making it a true singleton class and not just a class with a pointer to a particular instance of itself within it.