Question

I’ve seen this particular implementation of the singleton pattern everywhere:

+ (CargoBay *)sharedManager {
   static CargoBay *_sharedManager = nil;
   static dispatch_once_t onceToken;
   dispatch_once(&onceToken, ^{
      _sharedManager = [[CargoBay alloc] init];
   });
   return _sharedManager;
}

and it seems to be accepted as good practice (this one in particular is from CargoBay).

The only part I don’t understand is the first line static CargoBay *_sharedManager = nil;.

Why are you setting that static variable to nil?

Answer 1

It’s just a matter of readability, convention and practice. It’s not really needed, because:

One. Its value won’t ever be checked. In older singleton implementations there used to be the famous

+ (id)sharedInstance
{
    static SomeClass *shared = nil;
    if (shared == nil)
        shared = [[SomeClass alloc] init];

    return shared;
}

code – for this method to work, the backing variable has to be initialized to nil, since if it wasn’t nil for the first time, it would falsely omit the alloc-init in the if part and return a junk pointer. However, with the GCD solution, the nil-check is not anymore needed – GCD handles the ‘execute this code only once’ pragma.

Two. But nevertheless: static variables are implicitly initialized to zero. So even if you just write static id shared; it will initially be nil.

Three. Why this might be good practice? Because, despite the first two reasons I mentioned, it’s still more readable to let the reader of the source code know that something is explicitly initialized to zero. Or there may even exist some non-conforming implementations where static variables are not properly autoinitialized, and then this action shall be taken.