By books first explains the problem (using ARC) about "unvalid" references, like in this example:
NSDate* date1=[[NSDate alloc]init];
NSDate* date2=date1;
[date1 release];
NSLog(#"%#",date2); // bad access
So I have understood the retain/release mechanism: in this case the instruction would be:
date2=[date1 retain];
But when it talks about strong/weak references, it sounds like a contradiction to me:
"By default, references are strong. If you assign an object to a strong reference, ARC assumes that you want that object to stick around and retains it implicitly"
Isn't this a contradiction of what has said before?
date2 is strong by default, so it should implicitly retain date1 and there wouldn't be a bad access exception.
Of course I have misunderstood something, could someone explain this better to me?
Strong is the default because it's typically what you want, but using ARC, the compiler analyzes how long the object's life needs to be and releases the memory at the appropriate time. For example:
- (void)someMethod
{
NSDate* date = [[NSDate alloc] init]; // date is __strong by default
NSLog(#"The date: %#", date); // date still contains the object created above
// Sometime before this point, the object date pointed to is released by the compiler
}
Weak references only keep the object around while it has one or more other strong references to it. As soon as the last strong reference is broken, the object is released by the compiler and the weak object reference (the variable) is changed to nil by the runtime. This makes weak variables almost useless in the local scope like the example above. For example:
- (void)someMethod
{
__weak NSDate* date = [[NSDate alloc] init]; // The date created is released before it's ever assigned to date
// because date is __weak and the newly created date has no
// other __strong references
NSLog(#"The date: %#", date); // This always prints (null) since date is __weak
}
To see an example of a weak and strong variable working together in the local scope (this would have only severely limited usefulness and is really shown here only to demonstrate weak variable references):
- (void)someMethod
{
NSDate* date = [[NSDate alloc] init]; // date stays around because it's __strong
__weak NSDate* weakDate = date;
// Here, the dates will be the same, the second pointer (the object) will be the same
// and will remain retained, and the first pointer (the object reference) will be different
NSLog(#"Date(%p/%p): %#", &date, date, date);
NSLog(#"Weak Date(%p/%p): %#", &weakDate, weakDate, weakDate);
// This breaks the strong link to the created object and the compiler will now
// free the memory. This will also make the runtime zero-out the weak variable
date = nil;
NSLog(#"Date: %#", date); // prints (null) as expected
NSLog(#"Weak Date: %#", weakDate); // also prints (null) since it was weak and there were no more strong references to the original object
}
The key mistake is in colluding manual-retain-release behavior and ARC as being identical. It isn't. Under ARC, object assignments are both atomic and, in and of themselves, expressions that determine object lifecycle.
Take your example and remove the retain. You end up with this:
NSDate* date1=[[NSDate alloc]init];
NSDate* date2=date1;
NSLog(#"%#",date2);
Makes perfect sense under ARC; there is no manual release screwing things up. I.e. it automatically just works.
Better yet, because the compiler does the flow control analysis behind the scenes, there is no extra retains or releases needed. The code will literally be something like:
NSDate* date1; // date1 initialized to nil -- or not given that it is never read before...
date1 = [[NSDate alloc]init]; // ... this assignment (of a +1 retain count obj)
NSDate* date2=date1; // retain count unchanged
NSLog(#"%#",date2); // retain count unchanged
.... compiler emits equivalent to [date1 release] ...
Since the compiler won't emit that release until after the last use of date1 or date2, there can never be a dangling pointer.
Related
As we know, using strong self within a block can lead to retain cycles and memory leak. Is the common practice to use weak self in a block, or is it better to assign the weak self to strong within the block and then use it as such so the weak self is not released during block execution? Does it matter since weak self will be zero-ed out anyway?
Due to the volatile nature of weak variables, you should use them with care. If you are using weak variables in a multithreading environment, it is considered good practice to assign the weak variable to a strong one and check for nil before using. This will ensure that the object will not be released in the middle of your method, causing unexpected results.
Consider the following case:
__weak id var;
//...
if(var != nil)
{
//var was released here on another thread and there are not more retaining references.
[anotherObj performActionWithAnObjThatMustNotBeNil:var]; //<- You may crash here.
}
The compiler can be configured to throw a warning on a consecutive access of a weak variable.
On the other hand, if your use is in the main thread, and all calls to the object are on the main thread, this problem is moot, since the object will either be released before the block call or after, thus it being safe to access the weak variable directly.
There are two possible questions here that are easy to get confused:
Is it possible for a __weak reference to become nil in the middle of a method?
id __strong strongObject = ...;
id __weak weakObject = strongObject;
dispatch_async(dispatch_get_main_queue(), ^{
[weakObject method1]; // if weakObject is non-nil here
[weakObject method2]; // can it become non-nil here?
});
Yes! Xcode will even warn you about it.
Is it possible for self to become nil in the middle of a method if the method is called on a __weak lvalue as below?
id __strong strongObject = ...;
id __weak weakObject = strongObject;
dispatch_async(dispatch_get_main_queue(), ^{
// is it possible for weakObject to be deallocated
// while methodUsingSelf is being called?
[weakObject methodUsingSelf];
});
- (void)methodUsingSelf {
NSLog(#"%#", self); // Could this be non-nil
NSLog(#"%#", self); // while this is nil?
}
No! Joe Groff, of the Swift team at Apple, said so:
self is guaranteed kept alive by ObjC ARC while a method on self is
executing.
Clang's official ARC documentation covers this case in the Semantics/Reading subsection:
Reading occurs when performing a lvalue-to-rvalue conversion on an
object lvalue.
For __weak objects, the current pointee is retained and then released
at the end of the current full-expression. This must execute
atomically with respect to assignments and to the final release of the
pointee.
Thus, calling a method on a __weak variable, is roughly equivalent to the following Manual Retain/Release (MRR) code:
id retainedObject = ...;
id assignedObject = strongObject;
dispatch_async(dispatch_get_main_queue(), ^{
{
[assignedObject retain];
[assignedObject methodUsingSelf];
[assignedObject release];
}
});
Of course, in MRR, [assignedObject retain]; might crash because the object assignedObject points to might have been deallocated, so assignedObject might point to garbage. ARC doesn't have this problem because it zeroes weak references.
I think that even if using the weak will work and be retained as long as needed, assigning it to strong before using will make it more readable and "worries free"...:
__weak id weakThing = thing;
thing.someBlock = ^{
if (weakThing) {
id strongThing = weakThing;
strongThing doThisWithThat...
}
};
Compiler won't complain and it is safe and maybe not less importantly - easy to understand for John Doe who will try to read this code tomorrow....
You can continue to use the weak self. The only time you'd need to use strong self is if you are trying to access a self->ivar directly, instead of going through a property.
I bet that I could find the answer of this question from reading similar threads here or by googling, but I would like "hear" it first hand so to speak because it's an anomaly in my understanding.
So here's the thing, I have some code which a previous employee wrote and I see a lot of a certain type of construct which looks rather odd to me and I just want to clarify what is "right and wrong".
For example
- (void) setWwanServiceId: (NSString *) newValue {
[wwanServiceId autorelease];
wwanServiceId = [newValue copy];
}
Here wwanServiceIdis a NSString member of the class, and to me this seem like a strange way to do it. As far as I understand it would begin by putting an autorelease on the object, basically saying: "whenever this object seem to not be used, release it for me I don't really care" then the copy will up the retain count +1 on.... wwanServiceId? or newValue? I guess the first.
Then to make me all more confused let's just quickly run through the life-cycle of the wwanServiceId-string..
basically the value will be set if we receive a notification which then from the notification-handler method will call the above -setWwanServiceId: method. other than that it will only ever be accessed for reading, so we are safe to say that it will at any given point:
put autorelease on object
retain a new string copy
Then there is one more quirk to this and this is where I am getting rather suspicious, namely in the -dealloc method which looks like this:
- (void) dealloc {
[[self wwanServiceId] release];
[super dealloc];
}
So what happens there? it will afaik release the wwanServiceId and as I said the only time it's memory management is being touched (if I haven't missed anything but I feel pretty sure) is to put an autorelease and retain it.
So to summarize:
Is the idea here that he thought that since he always retain a new copy after putting autorelease he need to release it in the end.. or well it's the only thing I can think about. Or just felt it would be safe to do an extra release in the end just in case..?
Because as far as I understand it, if this setter is called one time it will put an autorelease (-1 in future), do a retain (+1) and when destructor is called it will do the "final release" (-1).
Any ideas or suggestions helping me understand (if in fact I am wrong and the memory handling is correct as is) would be appreciated.
You wrote:
Here wwanServiceIdis a NSString member of the class, and to me this seem like a strange way to do it. As far as I understand it would begin by putting an autorelease on the object, basically saying: "whenever this object seem to not be used, release it for me I don't really care" then the copy will up the retain count +1 on.... wwanServiceId? or newValue? I guess the first.
This seems to point to the source of your confusion. wwanServiceId is a variable which can contain a reference to an object of type NSString. Variables do not have reference counts only objects do.
The previous employee wrote:
- (void) setWwanServiceId: (NSString *) newValue {
[wwanServiceId autorelease];
wwanServiceId = [newValue copy];
}
The expression [wwanServiceId autorelease] means: read the reference stored in wwanServiceId and autorelease the object that reference refers to - let's call that object A. Important: This does not delete object A; it will be released as some later stage and if at that time there are no remaining references object A will be deleted.
The expression [newValue copy] means: read the reference stored in newValue, use that to locate the object (call it object B), make a copy of that object to produce a new object (call it object C), and return a reference to the new object. This new object is owned by the caller of copy, so there is no need to retain it.
Finally the assignment stores the reference to object C into wwanServiceId.
So there is a maximum of three distinct objects involved:
A: the original object referenced by wwanServiceId, this is autoreleased to remove the ownership of wwanServiceId.
B: the object referenced by newValue, this is left untouched
C: a newly created copy of B owned through wwanServiceId
Why "autorelease" in the code and "maximum of three distinct" above?
The method could be called with newValue referencing object A, e.g. as in:
[self setWwanServiceId:[self wwanServiceId]]
If this was to occur and (a) release was used instead of autorelease and (b) there was no other reference to object A then the release would delete object A and then when [newValue copy] was evaluated newValue would be referencing a deleted object... The use of autorelease in this case delays the deletion until after the copy.
So what the previous employee wrote is not in any way "wrong", but as some of the other answer suggest it may be an unusual style. Another way you see this written is:
- (void) setWwanServiceId: (NSString *) newValue
{
NSString *oldValue = wwanServiceId;
wwanServiceId = [newValue copy];
[oldValue release];
}
which also ensures any deletion occurs after the copy.
HTH.
To be short and helpful:
This is wrong:
- (void) setWwanServiceId: (NSString *) newValue {
[wwanServiceId autorelease];
wwanServiceId = [newValue copy];
}
This is right:
- (void) setWwanServiceId: (NSString *) newValue {
if (newValue != wwanServiceId) {
[wwanServiceId release];
wwanServiceId = [newValue copy];
}
}
To explain in short:
[wwanServiceId autorelease]; is an unnecessary sent message, because autoreleasing an object will reduce the retain count at some unknown point in the future. And in the next line you are wwanServiceId = [newValue copy]; instantly setting the instance variable. So, in your memory you now have a to be autoreleased object and a new object. One of them is too much The new object is, where the pointer of your IVar is pointing to. The old one is swimming in your Memory Pool with probably no reference to it :-)
Autorelease as few as possible or use ARC.
Oh: And in the dealloc method, please do not send the message like this:
[[self wwanServiceId] release];
Better like this:
[wwanServiceId release];
as Apple recommends to work directly with instance methods in init and dealloc methods, instead of using getters and setters there.
Debug it and have a look.
[wwanServiceId autorelease];
wwanServiceId has an address. This statement does not change it. It does decrement the retain count of this object though.
Thats it.
wwanServiceId = [newValue copy];
This statement creates a new object. The new object is a copy of newValue. Compare the addresses of the objects and you will see, that the address inn wwanServiceId will vary from the address of newValue and it will vary from the address that wwanServiceId did have just before the statement was executed.
The retain, that is implicit in copy will affect wwanServiceId, but it affects the new object, that was just creaed with copy. It does not affect the wwanServiceId object which was autoreleased during the statement before.
At some point after the execution of setWwanServiceId had finished, the old and autoreleased object will dissappear. (Assuming that the retain count is 0 now. If it is >0 because it is still retained for other reasons or just for error, then it will remain and potentially cause a leak.)
Once you understood that you will not question any more what is happening in the dealloc method. wwanServiceId is released. Meaning its retain count is reduced by 1. In the event that it is 0 then it will be deallocated automatically.
You could even autorelease it there too. The diffrerence to autorelease is basically that an autoreleased object is still around and available while the current method is being executed. Its release comes into effect at some later point in time.
But there is no reason for autoreleasing the object in dealloc.
In the example given there is not even a good reason to autorelease the object in the setter setWwanServiceId. You may well release the object directly in both methods.
Assuming wwanServiceId is the private ivar mapped by the getter wwanServiceId and setter setWwanServiceId, I think it is not correct to autorelease the ivar in the setter.
I would have coded the following:
- (void) setWwanServiceId: (NSString *) newValue {
if (newValue != wwanServiceId) {
[wwanServiceId release];
wwanServiceId = [newValue copy];
}
}
What I mean is: it is not necessary to give ownership of your var to the autorelease pool (which it may be drained at the end of the application). Simply release the ivar. If someone is using it, no problem, it will have a strong reference to it. Otherwise it will be deallocated.
I'm currently developing an iOS application which was started by another developer.
Usually, I make a property for every instance variable (assign for int, bool etc. / retain for all classes).
So in my projects, this line causes a leak:
myVar = [[NSString alloc] init]; (alloc/init +1, retain in setter +1, release in dealloc -1 => +1)
So I use:
NSString *tmpMyVar = [[NSString alloc] init];
[self setMyVar: tmpMyVar];
[tmpMyVar release];
Or:
NSString *tmpMyVar = [[[NSString alloc] init] autorelease];
[self setMyVar: tmpMyVar];
In this new project, the previous developer didn't use #property/#synthesize so I'm wondering what will be the result of the previous line of code in this context (it doesn't call setter I guess)? Memory Leak?
The previous developer releases variable in dealloc method, just like me.
Thank you very much!
Since it directly assigns the instance variable to the allocated object it's retain count is 1 (because, like you said, a setter isn't called).
And because it's released in dealloc, it's all balanced out. So no memory leaks.
So in my projects, this line causes a leak:
myVar = [[NSString alloc] init]; (alloc/init +1, retain in setter +1, release in dealloc -1 => +1)
No,it wouldn't even in your projects, because, as you pointed out, no setter is used.
Also, when using properties, it is the recommended way to access instance variables directly in the init method, instead of using setters.
To inspect for questionable memory-leaks like your example, also use the clang static analyzer or instrument's leak tool.
You need to look at the other developer's setter implementation. Make sure they release the existing value and retain the new value; something like:
- (void)setMyString:(NSString *)string
{
[string retain];
[_string release]; // ivar
_string = string;
}
The only advantage to implementing your own setter/getter methods is to do something (other than setting the ivar) when a value is set. If the methods don't do anything like this then why not change all implementations to #property/#synthensize?
If I have a piece of code that looks like this:
- (void)testSomething
{
__weak NSString *str = [[NSString alloc] initWithFormat:#"%#", [NSDate date]];
NSLog(#"%#", str);
}
the output will be (null) because there are no strong references to str and it will be immediately released after I allocate it. This makes sense and is spelled out in the Transitioning to ARC guide.
If my code looks like this:
- (void)testSomething
{
__weak NSString *str = [NSString stringWithFormat:#"%#", [NSDate date]];
NSLog(#"%#", str);
}
then it correctly prints out the current date. Obviously you would expect it to work in a non-ARC world, since str would be autoreleased and therefore valid until this method exits. However, in ARC-enabled code people generally consider the two forms (stringWithFormat & alloc/initWithFormat) to be equivalent.
So my question is whether code like the second example is guaranteed to work under ARC. That is, if I have a weak reference to an object that I get via what we would normally consider an autoreleasing convenience constructor, is it guaranteed to be safe to use that reference in the same scope I normally would have without ARC (i.e. until the method exits)?
The conventions of autoreleasing and allocing still apply in the world of ARC. The only difference is that ARC will insert extra retain/release calls to make it much harder to leak objects or access a dealloced object.
In this code:
__weak NSString *str = [[NSString alloc] initWithFormat:#"%#", [NSDate date]];
The only place the object is retained (or equivalent) is the alloc. ARC will automatically insert a release command, causing it to be immediately dealloced.
Meanwhile, in this code:
__weak NSString *str = [NSString stringWithFormat:#"%#", [NSDate date]];
By convention, the return value of a convenience constructor like this must be an autoreleased object*. That means the current autoreleasepool has retained the object and will not release it until the pool is drained. You are therefore all but guaranteed that this object will exist for at least the duration of your method - although you probably shouldn't rely on this behaviour.
(* or retained in some other way)
The lifetime of a local weak variable is not guaranteed at all. If the object that the variable points to is deallocated, the weak variable will point to nil afterwards.
If you have a weak reference to an object that you got via a method that does not return a retained object, it is not safe to assume that this object lives until the method exits. If you want to make sure that the object survives, use a strong reference.
Here is an example that shows that a non-retaining method's return value is not guaranteed to end up in the autorelease pool:
Create a new iOS project (Single View App using ARC and Storyboards)
Add this method to the AppDelegate.m:
+ (id)anObject
{
return [[NSObject alloc] init];
}
Replace -application:didFinishLaunchingWithOptions::
- (BOOL)application:(UIApplication *)application didFinishLaunchingWithOptions:(NSDictionary *)launchOptions
{
__weak id x = [AppDelegate anObject];
NSLog(#"%#", x);
return YES;
}
Important: Now set the Optimization level for Debug to -Os.
In this example, +[AppDelegate anObject] acts like a convenience constructor, but you will see (null) logged if you execute it on a device with -Os optimization. The reason for that is a nifty ARC optimization that prevents the overhead of adding the object to the autorelease pool.
You may have noticed that I switched to not using a library method like +[NSString stringWithFormat:]. These seem to always put objects in the autorelease pool, that may be for compatibility reasons.
Please see my comments in code:
-(id)initWithCoordinate:(CLLocationCoordinate2D)c title:(NSString *)t
{
[super init];
coordinate = c;
NSDate *today = [NSDate date];
NSDateFormatter *dateFormatter = [[NSDateFormatter alloc] init];
[dateFormatter setDateStyle:NSDateFormatterLongStyle];
NSString* formattedDate = [NSString stringWithFormat:#"%# %#",
[dateFormatter stringFromDate:today], t];
[self setTitle:formattedDate]; //Why does the app crash when I try and release formattedDate? I have after all passed its reference to the title property?
[dateFormatter release]; //I need to release the dateformatter because I have finished using it and I have not passed on a reference to it
return self;
}
Nope, looks fine. stringWithFormat returns an autoreleased object - you don't own it unless you retain it, which you haven't, so you mustn't release it. You allocated dateFormatter yourself, so you do own it and must release.
(See the object ownership documentation and 8 squillion very similar SO questions. This must be the number one objective-c problem that crops up here by a significant margin.)
Edit: Actually, looking at your comment, there is a very slightly subtler issue here, though the upshot is the same (and actually underlines the point about ownership).
When you pass the string to setTitle, it could retain the string itself, in which case a call to release here might not cause a crash immediately. You would still be over-releasing, however, and it would bite you in the end. It just would be more difficult to locate the cause of the problem further down the line.
As it happens, setTitle makes a copy rather than retaining yours. This is pretty common with NSString properties. One reason is that you might pass an NSMutableString instead, and then randomly change it at some later point that might screw up the receiver. Taking a private copy is safer. As a bonus it exposes your over-releasing right away and allows you to fix it easily.
In any case, the point stands: only ever release something you know you own. If you've just passed it to some other object, it's that object's responsibility to handle whether it owns it or not, not yours.
You appear to have some radical misconceptions about how retain and release work. You say:
"Why does the app crash when I try and release formattedDate? I have after all passed its reference to the title property"
"I need to release the dateformatter because … I have not passed on a reference to it"
Whether you pass a reference to something is irrelevant to whether you're responsible for releasing it. The memory management rules dictate whether you should retain or release something, and they have practically nothing to do with passing references. You should read that document, because it's very fundamental and not really all that hard. Once you understand it, all your questions about this sort of thing will vanish.
The nutshell version is, you own an object and are responsible for releasing it if you obtained it from a new alloc or copy method or if you retained it. In the case of the NSDateFormatter, you got it from alloc, so you must release it. In the case of the string, none of those things are true, so you must not release this object that you don't own.
Slightly unrelated, but...
Initializsers in ObjC are supposed to look like
- (id)init;
{
self = [super init];
if (self == nil)
return self;
...
return self;
}
This allows the super class to return a replacement. This is pretty advanced usage of ObjC, though, but it's a good habit to stick to this pattern.
More info at ADC: Allocating and Initializing Objects.