How to enforce __weak reference in parameters for protocol implementation - objective-c

I have a protocol for which I want the implementations to use __weak references for method arguments. I define my protocol as:
#protocol TestProtocol
-(void) op:(__weak NSObject*)x;
#end
Then I create an implementation:
-(void) op:(NSObject*)x
{
}
And x becomes an strong reference. If I move the __weak declaration to the protocol implementation, 'x' is a weak reference.
As the caller of the protocol already keeps a strong reference, I want to ensure that the implementations don't create another strong reference.

I'm a tad surprised the compiler doesn't complain about a declaration mismatch. And, sure enough, it doesn't. I filed radar #13730581 to see what the compiler folks say.
As #hypercrypt cites, ARC isn't going to retain or release the arguments to a method in the general case. One of the few cases where the performance vs. correctness sacrifice was large enough to warrant such a tradeoff. (Note that you'll see a storeStrong in the non-optimized case but no such thing in the -Os case Release build case).
Which means you are really defending against a strong reference being created against the object by implication of the code in the method body (and, as per comments, that said strong reference isn't torn down when an #throw goes "over" the frame).
There isn't much you can do about that -- if you pass x off to, say, NSMutableArray's addObject: method, it'll create a strong reference.
Thus, in the general case, there isn't really a means of preventing a strong reference to an object from being created as a side effect of variable usage.
Specific to NSException cleanup, it isn't worth worrying about. Exceptions should be fatal and the state of the app after an exception is thrown is undefined (unless you can ensure that no system code was in any way involved with the exception, which isn't really viable).
// code tossed in main.m to play w/compiler behavior
#protocol TestProtocol
-(void) op:(__weak NSObject*)x;
#end
#interface DogBowl:NSObject <TestProtocol>
#end
#implementation DogBowl
-(void) op:(NSObject*)x
{
}
#end

Related

Objective-C init quirks

In a class interface I define some ivar
#property (strong,nonatomic) id < Protocol > initEst; // Initial estimate
This compiles without a problem but when I run the program it crashes with EXC_BAD_ACCESS and [Object .cxx_destruct] indicated by the debugger as the reason.
What is going on?
It's all about the rules of ARC automatic memory management. An initializer has special rules for how it treats the returned value: it retains and returns. See https://clang.llvm.org/docs/AutomaticReferenceCounting.html#semantics-of-init.
Objective-C in general, and ARC in particular, have some pretty strict rules about what names of methods mean. initXXX means "this is an initializer". If this isn't an initializer, don't use the init prefix.
You could turn off ARC entirely and manage memory yourself, but it's easier just to obey the conventions, and it fits better in case of interaction with other languages (such as Swift).
I have tested this some more and there seems to be three conditions for this particular quirk to show up.
In my particular case the ivar's Protocol was also the same as that of the containing class. This seems to be an additional requirement for this problem to surface (refering here to my earlier answer that did not mention this).
So to elaborate on my earlier answer. If
initXXX is an ivar
of id type
that implements a Protocol that is the same as the containing class
then the Objective-C + ARC compiler will happily compile the code but be unable to execute it.
Here is a sample of the code I used to test
#interface Dog : NSObject < Animal >
#property (nonatomic,strong) id < Animal > initState;
#end
Something like this will cause problems simply because the name starts with init. Change the name and all problems disappear.
For reference, the runtime error this generates is
Dog object overreleased while already deallocating
This snippet is pretty abstract but this may bite you in places where you need to specify some initial condition and where it is natural to name some ivar initXxx but beware, if you use Objective-C you do not have that luxury nor will the compiler warn you that it is wrong.
The original error seemed memory allocation related and caused me to suspect the way I used the autoreleasepool but now I am fairly convinced this has nothing to do with the issue.

What will happen if we declare all properties weak?

For example, there is a class
#interface Person: NSObject {
#property (weak) NSString *str;
#end
#implementation
- (void) init {
str = #"XYZ";
}
#end
Is it safe to declare all properties to weak to avoid reference count
overhead?
If not safe then why it's not safe?
What will be the reference count of str?
Is it safe to declare all properties to weak to avoid reference count overhead?
This is problematic on two counts. First, weak doesn't avoid reference counting. It actually performs more bookkeeping than strong. See Lots of overhead for weak property? for a good discussion of that.
Next, this is wrong because weak and strong mean different things that cannot be interchanged. As MartinM notes, if you make your properties weak, they'll just disappear as soon as you assign them if nothing else is holding onto them.
If not safe then why it's not safe?
It is totally safe. It just won't work. Your program will have well defined behavior, and shouldn't crash due to marking all your properties weak. Most of your properties will just be nil. If you used unsafe_unretained instead of weak, then you would avoid some reference counting overhead, and your program would typically crash (because the objects would still be destroyed, the pointers would just become invalid).
What will be the reference count of str?
That is not a meaningful question, even though it feels like one. In this particular case the answer is 9,223,372,036,854,775,807 because that's always the retain count of a static string, because static strings cannot be destroyed and ignore retain/release calls. (That value is NSNotFound in ObjC. I believe this used to return 1,152,921,504,606,846,975, but my tests indicate it's changed.)
But in the general case, it could be anything at this point in time. And what it is at this point in time versus at the end of the event loop (when the autoreleasepool drains) could be dramatically different. Do not chase retain count values. They will always lie to you. Create strong references to things you care about. Create weak references only when you need them for their specialized purposes.
Retain counts are an internal implementation detail. Before ARC, they were misleading to the point of uselessness. Since ARC, they're like asking "what assembly instructions will this compile to." It's important, but it depends on everything else about the program and how it's compiled.
If you have a specific performance problem related to memory management, then there are techniques for dealing with that in a case-by-case basis, and StackOverflow can help you. But ARC is extremely good at optimizing memory management. Let it do its job.
You should know which properties should be weak and which strong. It is a common practice declaring for example all IBOutlets as weak because they are being retained by the view that comes from the storyboard, so it is not necessary to keep a strong reference in the viewcontroller.
Another example would be a delegate that should always be weak, or it will prevent the delegate (mostly another viewcontroller) from being deallocated because the delegate's holder is still alive.
In your case, str will be nil after init, because there are no other references to its content. You should actually get a compiler warning if you do this.
Take a look at this thread with more detailed explanation: Weak and strong property setter attributes in Objective-C

Why doesn't Xcode provide autocompletion for dot-notated properties on objects of type id<protocol>?

Given this protocol definition:
#protocol MyProtocol <NSObject>
#property (nonatomic, strong) NSString *someProperty;
#end
Why will Xcode gladly offer autocompletion for this statement:
id<MyProtocol> thing = [ThingManager currentThing];
[thing someProperty]; // Xcode offered autocompletion here
But it doesn't offer autocompletion when I try to access the same property using dot-notation:
id<MyProtocol> thing = [ThingManager currentThing];
thing.someProperty; // Xcode claimed there were
// "No completions" available
// after the period
Because id is a base type, Xcode and CLANG are uneasy about providing dot-syntax access against it because dot syntax is just syntactic sugar for a method call to an associated setter or getter in a normal object, but id has no defined method members. Looking at it from the C side of things, id is a typedef for a struct pointer that the compiler cannot see the members of, which means it cannot access them (never mind the fact that you would need to dereference id before dot-access would make any semantic sense).
Back to the Objective-C side of things, protocols don't actually add methods or properties to the classes that claim to implement them, rather they serve as a specifier to other classes that an object that conforms to a given protocol implements a series of methods. As for the method-syntax being completed, Xcode pools all of the given methods of all the files imported into a given .m file because, an object of type id can receive any message*
*of course, it can receive the message, but it'll still crash if it's unimplemented.
This is kind of a tangential answer, and a thought experiment.
But before that, I'll note that you could get your property autocomplete by skipping id, like this:
NSObject<MyProtocol> *thing;
thing.▮
But assuming you don't want the entire list of NSObject methods gumming up your completion, you could do something like
EmptyClass<MyProtocol> *thing = [ThingManager currentThing];
// completion list will be (close) to only the protocol props
thing.▮
EmptyClass serves a similar "OK, no promises!" role that id does, but autocomplete likes it. Here's EmptyClass:
NS_ROOT_CLASS
#interface EmptyClass
#end
#implementation EmptyClass
+ (void)initialize {} // required
#end
Mind you, the object in thing is not actually rooted on EmptyClass (can't be), so this is high fakery. However, it
vastly underpromises what thing can actually do.
doesn't (can't!) instantiate an EmptyClass object.
So why not? If you try really hard, you can cause problems like
EmptyClass *nooooo = [[NSClassFromString(#"EmptyClass") alloc] init];
which will immediately exception. But not really a tricky bug to avoid.
A gotcha wouldn't surprise me, but I don't know one right now. Please, leave a comment if you do.

Adding Member Variables in Objective C

First I have to do
#property (retain) aMember;
Then in implementation file I got to do
#synthesize aMember;
Then in dealloc, I got to do
self.aMember= nil; (or [aMember release])
That's 3 times writing what essentially is the same
Is there a way to speed this up?
I mean I can drag drop a control from a IB and xcode automatically generate those codes why I can't do that for more normal codes?
As someone coming from C# and managed languages for my day job I completely agree with you in questioning this 3 step process. In fact its almost crazy easy to create properties in C# in MS Visual Studio, but I digress.
Even though there are these 3 lines you have to write there is a huge amount of work going on under the covers for your.
Your declaration of the #property tells objective-c some important attributes (atomic, nonatomic, retain, copy, etc) in how to deal with your property when it is set by users of your class. When you think about this, these attributes (without you writing any code) are; helping you create thread safe code, handling references to objects so you don't have to worry about them disappearing on you, or copying values so you have your own copy of an object. The #property is also important since it is declared in your header file (typically). This give other developers an idea of the properties of your class and some small hints as to how objects they pass into those properties will be handled during its lifetime.
The #synthesize is also doing quite a bit of work by creating the getters and setters for that property, that also handle all sorts of memory management for you. You don't need to worry about releasing the old references and correctly referencing the new object. This alone to me is a great feature, especially when you are new to objective-c and it is easy to forget to deal with memory management at every turn. The #synthesize just does it for you and you don't have to write all the get and set code yourself.
The dealloc call is just life in a non-memory managed environment. While it adds additional steps, I appreciate the benefits that explicit memory management allows in a constrained environment such as the phone.
So all 3 steps are required, are different and when you think about it actually do quite a bit of work for you under the covers.
Unfortunately, that's how it is (for now). Apple had recently toyed with allowing Clang to implicitly synthesize your properties, which would have reduced your work to:
#interface Blah : NSObject
#property (retain) Blorg *blorg;
#end
#implementation Blah
- (void)dealloc {
[blorg release];
[super dealloc];
}
#end
When you didn't want an instance variable to be synthesized, you'd just explicitly put #dynamic blorg in your implementation. But this feature was removed due to some unforeseen complications, despite mostly positive reactions from developers.
So, I think it's safe to expect that Apple's still working on this. But for now, you do need to explicitly synthesize.
A few other notes:
If you are using garbage collection, you don't need to implement -dealloc: just make sure to do any last-minute cleanup in -finalize (such as notification unregistration).
You could also avoid the -dealloc bit by wrapping your instance variable in a C++ class which performs memory management during construction and destruction: #property prop_wrapper<Blorg> blorg; would work. Then, when your object is destroyed, ~prop_wrapper() would be called on your object. I've done this, and it works, but I recommend against it, since it doesn't play nice with KVO and KVC.
You could iterate through the properties of an object, and release those that are annotated with copy or retain. Then, in -dealloc, you'd have something like [self releaseProperties]. I've also done this, but I also recommend against it, since it can cause subtle problems which may result in inexplicable crashes if you're not careful.
To actually add a member variable in objective-c you don't need to do any of that.
What you're doing in those 3 steps is:
Declare properties for a member variable. (In your case you are indicating that you want the property setter to 'retain' the object that it sets your member variable to)
Declare the property getters and setters in a default way for your property.
Release the object that your property is retaining.
IF you only wanted to declare a member variable, all you had to do was declare it inside your class:
#interface SomeClassObject : NSObject {
int someMemberVariable;
}
#end
That's 3 times writing what essentially is the same
No it isn't.
#property (retain) aMember;
The above line declares a property so that the compiler knows it is OK to send the messages -aMember and -setAMember: to objects of your class. It also tells the compiler (and developers) that the property is a retain property (i.e. the object you set the property to will be retained), that it is read/write and that it is atomic.
#synthesize aMember;
The above line tells the compiler that it should automatically generate the setter and getter methods for the declared property. You can leave that out but then you have to write your own setter and getter.
[aMember release]; // in dealloc
Is there to tell the runtime that when the object is being deallocated, it no longer needs to hold a reference to that instance variable. This is necessary because, when you use reference counting rather than garbage collection, the runtime does not automatically clean up unwanted objects.
Each of those lines does a different thing. So you are not doing the same thing three times.

Using instance variables with Modern Runtime

I have several years of experience in Obj-c and Cocoa, but am just now getting back into it and the advances of Obj-C 2.0 etc.
I'm trying to get my head around the modern runtime and declaring properties, etc. One thing that confuses me a bit is the ability in the modern runtime to have the iVars created implicitly. And of course this implies that in your code you should always be using self.property to access the value.
However, in init* and dealloc(assuming you're not using GC) methods we should be using the iVar directly (in the current runtime).
So questions are:
Should we use property accessors in init* and dealloc with Modern Runtime?
If so, why is this different? Is it just because the compiler can't see the iVar?
If I need to override an accessor, can I still access that iVar that will be defined at runtime or do I have to define an actual iVar that the runtime will then use?
Again, if I can access the synthesized iVar, why can't I continue to do this for the init* and dealloc methods?
I read the docs several times, but they seemed a bit vague about all of this and I want to be sure that I understand it well in order to decide how I want to continue coding.
Hope that my questions are clear.
Quick summary of testing:
If you don't declare the ivar in legacy, compiler is completely unhappy
If you use #ifndef __OBJC2__ around ivar in legacy compiler is happy and you can use both ivar directly and as property
In modern runtime, you can leave the ivar undefined and access as property
In modern runtime, trying to access ivar directly without declaration gives error during compile
#private declaration of ivar, of course, allows direct access to ivar, in both legacy and modern
Doesn't really give a clean way to go forward right now does it?
In the current (OS X 10.5/GCC 4.0.1) compiler, you cannot directly access the runtime-synthesized ivars. Greg Parker, one of the OS X runtime engineers put it this way on the cocoa-dev list (March 12, 2009):
You can't in the current compiler. A
future compiler should fix that. Use
explicit #private ivars in the
meantime. An #private ivar should not
be considered part of the contract -
that's what #private means, enforced
by compiler warnings and linker
errors.
And why isn't there a way to
explicitly declare instance variables
in the .m file for the new runtime?
Three reasons: (1) there are some
non-trivial design details to work
out, (2) compiler-engineer-hours are
limited, and (3) #private ivars are
generally good enough.
So, for now you must use dot-notation to access properties, even in init and dealloc. This goes against the best practice of using ivars directly in these cases, but there's no way around it. I find that the ease of using runtime-synthesized ivars (and the performance benefits) outweigh this in most cases. Where you do need to access the ivar directly, you can use a #private ivar as Greg Parker suggests (there's nothing that prevents you from mixing explicitly declared and runtime-synthesized ivars).
Update With OS X 10.6, the 64-bit runtime does allow direct access to the synthesized ivars via self->ivar.
Since instance variables themselves can only be synthesized in the modern runtime (and must be declared in the #interface under 32-bit or pre-Leopard), it's safest / most portable to also declare the ivar
Should we use property accessors in init* and dealloc with Modern Runtime?
My rule of thumb is "possibly" for -init*, and "usually not" for -dealloc.
When initializing an object, you want to make sure to properly copy/retain values for ivars. Unless the property's setter has some side effect that makes it inappropriate for initialization, definitely reuse the abstraction the property provides.
When deallocating an object, you want to release any ivar objects, but not store new ones. An easy way to do this is to set the property to nil (myObject.myIvar = nil), which basically calls [myObject setMyIvar:nil]. Since messages to nil are ignored, there is no danger in this. However, it's overkill when [myIvar release]; is usually all you need. In general, don't use the property (or directly, the setter) in situations where deallocation should behave differently than setting the variable.
I can understand eJames' argument against using property accessors in init/dealloc at all, but the flipside is that if you change the property behavior (for example, change from retain to copy, or just assign without retaining) and don't use it in init, or vice versa, the behavior can get out of sync too. If initializing and modifying an ivar should act the same, use the property accessor for both.
If so, why is this different? Is it just because the compiler can't see the ivar?
The modern runtime deals with class size and layout more intelligently, which is why you can change the layout of ivars without having to recompile subclasses. It is also able to infer the name and type of the ivar you want from the name and type of the corresponding property. The Objective-C 2.0 Runtime Programming Guide has more info, but again, I don't know how deeply the details explained there.
If I need to override an accessor, can I still access that iVar that will be defined at runtime or do I have to define an actual iVar that the runtime will then use?
I haven't tested this, but I believe you're allowed to access the named ivar in code, since it actually does have to be created. I'm not sure whether the compiler will complain, but I would guess that since it will let you synthesize the ivar without complaining, it is also smart enough to know about the synthesized ivar and let you refer to it by name.
Again, if I can access the synthesized iVar, why can't I continue to do this for the init* and dealloc methods?
You should be able to access the property and/or ivar anytime after the instance has been allocated.
There is another SO question with similar information, but it isn't quite a duplicate.
The bottom line, from the Objective-C 2.0 documentation, and quoted from Mark Bessey's answer is as follows:
There are differences in the behavior that depend on the runtime (see also “Runtime Differences”):
For the legacy runtimes, instance variables must already be declared in the #interface block. If an instance variable of the same name and compatible type as the property exists, it is used—otherwise, you get a compiler error.
For the modern runtimes, instance variables are synthesized as needed. If an instance variable of the same name already exists, it is used.
My understanding is as follows:
You should not use property accessors in init* and dealloc methods, for the same reasons that you should not use them in the legacy runtime: It leaves you open to potential errors if you later override the property methods, and end up doing something that shouldn't be done in init* or dealloc.
You should be able to both synthesize the ivar and override the property methods as follows:
#interface SomeClass
{
}
#property (assign) int someProperty;
#end
#implementation SomeClass
#synthesize someProperty; // this will synthesize the ivar
- (int)someProperty { NSLog(#"getter"); return someProperty; }
- (void)setSomeProperty:(int)newValue
{
NSLog(#"setter");
someProperty = newValue;
}
#end
Which leads me to think that you would be able to access the synthesized ivar in your init* and dealloc methods as well. The only gotcha I could think of is that the #synthesize line may have to come before the definitions of your init* and dealloc methods in the source file.
In the end, since having the ivars declared in the interface still works, that is still your safest bet.
I am running into the same problem. The way I am working around not being able to access the synthesized instance variables is the following:
public header
#interface MyObject:NSObject {
}
#property (retain) id instanceVar;
#property (retain) id customizedVar;
#end
private header / implementation
#interface MyObject()
#property (retain) id storedCustomizedVar;
#end
#implementation MyObject
#synthesize instanceVar, storedCustomizedVar;
#dynamic customizedVar;
- customizedVar {
if(!self.storedCustomizedVar) {
id newCustomizedVar;
//... do something
self.storedCustomizedVar= newCustomizedVar;
}
return self.storedCustomizedVar;
}
- (void) setCustomizedVar:aVar {
self.storedCustomizedVar=aVar;
}
#end
It's not that elegant, but at least it keeps my public header file clean.
If you use KVO you need to define customizedVar as dependent key of storedCustomizedVar.
I'm relatively new to Obj-C (but not to programming) and have also been confused by this topic.
The aspect that worries me is that it seems to be relatively easy to inadvertently use the iVar instead of the property. For example writing:
myProp = someObject;
instead of
self.myProp = someObject;
Admittedly this is "user" error, but it's still seems quite easy to do accidentally in some code, and for a retained or atomic property it could presumably lead to problems.
Ideally I'd prefer to be able to get the runtime to apply some pattern to the property name when generating any iVar. E.g. always prefix them with "_".
In practice at the moment I'm doing this manually - explicitly declaring my ivars, and deliberately giving them different names from the properties. I use an old-style 'm' prefix, so if my property is "myProp", my iVar will be "mMyProp". Then I use #synthesize myProp = mMyProp to associate the two.
This is a bit clumsy I admit, and a bit of extra typing, but it seems worth it to me to be able to disambiguate a little bit more clearly in the code. Of course I can still get it wrong and type mMyProp = someObject, but I'm hoping that the 'm' prefix will alert me to my error.
It would feel much nicer if I could just declare the property and let the compiler/runtime do the rest, but when I have lots of code my gut instinct tells me that I'll make mistakes that way if I still have to follow manual rules for init/dealloc.
Of course there are also plenty of other things I can also do wrong...