Difference between these two NSString methods

问题:

So I just got asked this at an interview today and after some googling am still unable to figure out the answer (in fact I couldn't even find any code at all which used the [NSString string] method).

What is the difference between

1. NSString *someString = [NSString string];
2. NSString *someString = [[NSString alloc] init];

Now my initial thoughts were that [NSString string] would return an object which would be autoreleased whereas using alloc and init would return an object which has been retained. However it seems that this answer was incorrect.

I've looked at the NSString class reference in the apple docs but all it says is

Returns an empty string.

+ (id)string 

Return Value
An empty string.

Could somebody explain to me exactly what the difference between these two are?

回答1:

Now my initial thoughts were that [NSString string] would return an object which would be autoreleased

Technically, it’s a placeholder string that is constant, i.e., it lives throughout the entire program execution, never being released. It’s not an autoreleased string. Conceptually, and this is what I’d focus as an interviewer, it’s a string (an empty string) that is not owned by the caller, hence the caller shouldn’t release it.

whereas using alloc and init would return an object which has been retained

Technically, it’s a placeholder string that is constant, i.e., it lives throughout the entire program execution. In fact, it’s the same object as the one above, and it is not retained. Conceptually, and this is what I’d focus as an interviewer, it’s a string (an empty string) that is owned by the caller, hence the caller is responsible for releasing it when it’s not needed any longer.

回答2:

Was that your response, and did you ask why your answer was incorrect? I ask because your assumption is mostly correct (at a higher level).

It's not exactly 'retained' when returned from alloc+init, it is an object you hold one reference to, and should balance with a release or autorelease. For the convenience constructor (+[NSString string]), you are returned an object which you hold zero references to, but one which you can expect to live until the current autorelease pool is popped unless you send it an explicit retain (assuming MRC or ARC, since it is tagged iOS).

At the lower level, you could make some guesses, but I wouldn't expect that question in many objc interviews (unless you told them you were mid or senior level). Basically, it is implementation defined, but both forms could return the same static, constant NSString (that may have been what the interviewer was looking for). To illustrate:

@implementation NSString

static NSString * const EmptyNSString = @"";

- (id)init
{
    self = [super init];
    [self release];
    return EmptyNSString;
}

+ (id)string
{
    return EmptyNSString;
}

...

Again, that's implementation defined, but an obvious optimization. As well, that optimization makes physically subclassing concrete immutable types (NSString) difficult for mutable variants (NSMutableString) in some cases.

回答3:

The correct answer is that

NSString *someString = [NSString string];

gives you an empty string that you do not own and that you must not release (according to the memory management rules)

whereas

NSString *someString = [[NSString alloc] init];

gives you an empty string you do own and that you must release (according to the memory management rules).

Without poking into the implementation, you can't say anything else about those two strings. You can't say that they are autoreleased, because they might not be and you can't say what the retain count will be.

In actual fact, you'll probably get (in both cases) the same pointer to a constant object of some NSString subclass, probably with a retain count of UINT_MAX which is used by the run time as a flag to disable normal retain release behaviour for constant strings. I haven't actually tried the above because nobody except the maintainers of the Objective-C SDK needs to care.

回答4:

You don't often see

NSString *someString = [NSString string]; 

because it's the same as

NSString *someString = @""; 

which is shorter. It's usually used to create an empty NSMutableString

NSMutableString* s = [NSMutableString string];

回答5:

The only thing I can imagine is that:

1. Won't allocate memory since it is not made with alloc. It is a constant (an empty string) made by the system and doesn't need to be released.

2. You allocate the memory for the NSString yourself which means you have to keep track if the NSString still 'lives' or not when you are done with it, and thus need to release it.