iOS Memory Management in Under 6 Minutes
What you should know about ARC in old good Objective-C
In the last couple of days, I was searching more on memory management in iOS. It was a kind of a hard topic for me to understand in Swift. Therefore, I decided to go a little deeper in old good Objective-C — as I’m learning it already, to understand memory management more and get an in-depth understanding of that topic. Like, how was is it before the ARC? What happened after it? And what is the difference between ARC and Garbage Collector?… etc.
When you start working with objects, you will notice an interesting thing.
Objects are more demanding on memory than primitive types like integers and floats. This remains true whether these objects are from your own classes or they are objects you make from classes in any of the Apple frameworks.
Luckily we don’t have to worry about working manually with memory addresses and manually allocating or deallocating areas of memory (which is still true for some languages). Instead in Objective-C, we use something called reference counting.
What does reference counting mean?
When your application starts, an area of memory became available for your objects. So you create an object and an area of memory is claimed for that object. Now, the variable that holds the object is a pointer to that object. A pointer to an area of memory to be exact. What happens under the hood is that when the object was created it’s been given what’s called a retain count or a reference count and it represents the number of owners for a particular object. Therefore, we can imagine it to be one, as you see on the image below:
The variable that called var is a pointer to a block of memory. So, in your program, you work with that pointer and call methods on it as long as you want. But when you arrive at the end of the code block. This variable is no longer reachable or accessible by any part of the program. The retain count goes again to zero because there’s nothing pointing to that block of memory now. And when the retain count become zero the runtime engine says, here’s a block of memory that no one cares about. Therefore, it will release this block of memory and make it available to other objects to use.
What might go wrong?
The only issue here can be in situations when you are creating objects and passing them around from one place to another. So it’s not clear that the pointer is still in scope or not. And up until 2011 — when ARC (Automatic Reference Counting) was added, you had to write statements on when you were finished using that object.
Before ARC was added to Objective-C we would have to do something like this
MyClass *myObject = [[MyClass alloc] init];
[myObject myMethod]; // call methods
... // doing some stuff with the object
[myObject release]; // releasing the object
We’d write a little bit of code to create an object. It would be created. We’d call methods of that object. But at some point, we would have to actually explicitly call a
release statement. And that would be what was responsible for taking a retain count down.
This is not a problem with a small number of objects. But when you had hundreds or thousands of objects being created, manipulated, used as parameters, passed from object to object… you had to keep track of them all. If you were passing an object from one area of your program to another, you might not even be sure if you could release it yet. Or if some other part of the program would take care of releasing it or might even release before you were done with it. So you could also write what was called, retain calls to that object. But you would still need to match any retain call up with an additional release call.
Basically, before ARC, you had to envision every possible scenario-logic that your application will go through to make sure that all of your objects’ lifetime is managed correctly. Not that easy.
Luckily, when using ARC you are no longer need to use
retain calls. But it’s important to understand the dangers of incorrectly writing this code.
One of the problems you could have is you might release too soon. You’d create an object, have a pointer to that area of memory, you’d call the methods, at some point you’d release it. But if you still had a valid pointer, there’s nothing that would actually stop you writing another line of code that tried to use it. This would be called a dangling pointer. The pointer existed, but what it was pointing to was no longer valid in memory and this could cause a crash.
The flip side of this is you might not release at all, you might create an object. Start calling methods of that object and then just allow the pointer to drop out of scope and disappear. But you never released the object so you just started claiming more and more memory, what was referred to as a memory leak.
So having to write a lot of this code was certainly very prone to errors and issues. Now you might wonder if we’re using this new ARC feature why do I even mention the existence of things like release and retain calls. Because the idea of release and retain calls have not disappeared the language still does reference counting.
The language hasn’t changed. The difference is you simply don’t need to write the retain and release and autorelease calls because the compiler does. ARC is taking the fact that compilers have gotten so good, that any time you build your project, the compiler, in this case, llvm, which is what Xcode is using behind the scenes, is able to determine all the possible paths for your code. And it basically goes through your code synthesizing the write, retain, release, and autorelease calls that you would need.
What the compiler’s doing is effectively writing the same code you would write yourself if you were really, really, really good at writing memory management code.
ARC doesn’t physically change your source code files, but this is effectively what the compiler is doing when you compile a project, and you’re using ARC.
Difference between Garbage Collector and ARC
ARC is a very different effect than having a garbage collector. Languages that use garbage collection are often what is referred to as non-deterministic. It means that you can’t tell exactly when objects are being reclaimed because it's being managed by the runtime by an external process. ARC allows us to be completely deterministic. The code controls when these objects are released. It’s just the code to release them has been written by the compiler, not by you. In fact, not only do you not write these calls when you’re using ARC, you can’t write these calls. If you try and do even the simple release call you’re going to get an error.
You need to be aware of the idea of retain or release. It is still what’s going on under the hood, but, certainly if you haven’t had to do it be thankful that you don’t have to do this anymore.
I’d love to hear more from you on that topic, and what do you think of going back to a language like Objective-C in order to understand a topic like memory management better? Have you ever done that before?