Swift and Cocoa Essentials
Threads, Queues, and Concurrency
It can be challenging or even overwhelming to prepare for a job interview or project meeting. The number of subjects you need to be familiar with as a developer is staggering. Mastering the fundamentals of the platform you're developing software for should always be a key focus area. That is why a significant portion of the content published on Cocoacasts focuses on the fundamentals of Swift and Cocoa development.
This episode zooms in on threading and concurrency. It isn't uncommon for developers to avoid these subjects because they seem complex or more advanced. It's true that threading and concurrency aren't the easiest concepts to learn, but you need to understand them to build software that is performant and reliable.
What Is a Thread?
Let's start with a basic question many developers don't have a good answer for. What is a thread? I won't go into the low-level details of threads and concurrency. While it's undoubtedly interesting, it's less important to understand the nitty gritty details if you're a Swift or Cocoa developer.
I want to avoid that this episode turns into a theoretical discussion of threading and concurrency. Let's make it more tangible. Launch Xcode and create a new project by choosing the Single View App template from the iOS section.
Name the project Threads and click Create to create the project.
Run the application in the simulator or on a physical device. With the application running, pause the execution of the application's process by clicking the Pause button in the Debug Bar at the bottom.
Open the Debug Navigator on the left. The Debug Navigator shows us a list of the threads the application uses to perform its work.
An application isn't functional unless the code we wrote is executed. A thread is nothing more than a context in which commands are executed. It's a line or thread of execution through the application we created. The stack trace of a thread illustrates this. Each frame of the stack trace is a command that is being executed by the thread. You can learn more about threads and stack traces in Debugging Applications With Xcode.
A thread has many more properties, such as a unique identifier, a stack, and a collection of registers. Don't worry about these properties for now. They are not important for the rest of this discussion.
Every Cocoa application has a minimum of one thread, the main thread. The main thread is the thread on which the application starts its life. The topmost thread in the list of threads in the Debug Navigator is the main thread. We talk more about the main thread in the next episode.
Despite the simplicity of the project, the Debug Navigator shows us that the application uses multiple threads to carry out its tasks. We could say that the application is multithreaded since it uses more than one thread. Multithreading is a pattern to increase the performance of an application. By scheduling work on multiple threads, more work can be performed in parallel. Performing work in parallel is also known as concurrency. Multiple commands are executed in parallel or concurrently.
Multithreading and concurrency have gained in importance over the years because devices have transitioned from having a single processor with a single core to one or more processors with multiple cores. Taking advantage of this power is important, but it's also what makes software development complex and challenging. There are several options to manage that complexity. Grand Central Dispatch is one of them.
Grand Central Dispatch
Threading and concurrency are complex concepts, but there's also good news. As a Cocoa developer, you rarely directly interact with threads. Several years ago, Apple introduced Grand Central Dispatch, GCD for short, a technology that makes working with threads easier and more efficient. There's rarely a need to manually create and manage a thread. Grand Central Dispatch offers developers an API that facilitates the scheduling of tasks. Which thread is used to execute a task is handled by Grand Central Dispatch, not the developer.
As I mentioned earlier, the devices we develop applications for are powered by powerful processors and it’s important to take advantage of that power. An application needs to be performant and responsive. Achieving these goals can be challenging.
Grand Central Dispatch manages a collection of dispatch queues. They are usually referred to as queues. The work submitted to these dispatch queues is executed on a pool of threads. As a developer, you don't need to worry about which thread is used to execute a block of work and Grand Central Dispatch doesn't want you to know. It doesn't offer any guarantees. Which thread is used for a particular task is an implementation detail of Grand Central Dispatch. We take a closer look at Grand Central Dispatch in a later episode.
What's important to understand is that Grand Central Dispatch operates at the system level, which means that it has an accurate overview of which processes are running, which resources are available, and how to best schedule the work that needs to be done. It's a very powerful solution.
What Is a Queue?
I already mentioned that Grand Central Dispatch manages a collection of dispatch queues. As the name implies, a dispatch queue is a queue onto which work can be scheduled for execution.
There are two types of dispatch queues, serial dispatch queues and concurrent dispatch queues. It's easy to understand the difference. A serial queue executes the commands it's given in a predictable order. If you schedule two tasks on a serial dispatch queue, then the second task is executed after the first task has completed. The advantage is predictability. The disadvantage is decreased performance. The second task needs to wait until the first task has finished executing.
Concurrent queues are different in that they sacrifice predictability for performance. The tasks scheduled on a concurrent queue are executed in parallel or concurrently. If you schedule two tasks on a concurrent dispatch queue, then the second task can be scheduled before the first task has finished executing. This means that it's possible for the second task to finish executing before the first one.
Let me illustrate this with a simple example. I've created a simple application that downloads three images and displays each image in an image view. The application displays two buttons. The button on the left is labeled Serial and performs its work on a serial dispatch queue. The button on the right is labeled Concurrent and performs its work on a concurrent dispatch queue.
Before I illustrate the difference, let's have a look at the implementation of the
ViewController class. It keeps a reference to a serial dispatch queue and a concurrent dispatch queue. If the left button is tapped, the
download(using:) method is invoked and a reference to the serial dispatch queue is passed in as an argument. If the right button is tapped, the
download(using:) method is invoked and a reference to the concurrent dispatch queue is passed in as an argument. In this example, we explicitly create a serial and a concurrent dispatch queue. This is fine, but it's more common to ask Grand Central Dispatch for a reference to one of its dispatch queues. Even though we create a serial and a concurrent dispatch queue, they are managed by Grand Central Dispatch.
Most interesting is the implementation of the
download(using:) method. We reset the image views, start an activity indicator view for each image view, and schedule a block of work on the dispatch queue that is passed to the
download(using:) method. It isn't a problem if you're not familiar with the API of Grand Central Dispatch. The idea is quite simple.
The application creates a
Data instance using the
URL instance. The image data is used to create a
UIImage instance, which is assigned to the
image property of the image view. Downloading the image data and creating the
UIImage instance is done in the background.
The key difference is that a serial dispatch queue executes the work it's given serially or sequentially. This means that the image data for the second image is fetched after the image data of the first image has completed downloading. The image data of the third image is fetched after the image data of the second image has completed downloading. If that is true, then the first image should be displayed first, then the second image, and finally the third image. In other words, the images are downloaded and displayed sequentially.
This isn't true if we schedule the work on a concurrent dispatch queue. In that scenario, the tasks we dispatch to the concurrent queue are executed in parallel. It's possible that the application has finished downloading the image data of the second image before the image data of the first and third image have finished downloading. In that scenario, the second image is displayed before the first image even though the first task was scheduled before the second task.
That's enough theory for now. Let's run the application and find out what happens. If we tap the button on the left, the images are displayed in a predictable order, from top to bottom. If we tap the button on the right, the images are displayed in an unpredictable order. I deliberately made the file size of the third image smaller than that of the other two images to illustrate the difference. Even though the task to fetch the image data of the third image is scheduled last, the third image is displayed first.
Main Queue and Global Queues
We manually created a serial and a concurrent dispatch queue in the example. That's fine, but it's more common to ask Grand Central Dispatch for an existing dispatch queue. Every application has access to a main queue and several background queues.
As the name implies, the main queue is associated with the main thread. What does that mean? Work that is scheduled on the main queue is guaranteed to be executed on the main thread. This is useful if you need to make sure a task is guaranteed to be executed on the main thread, such as updating the user interface. We talk more about the main queue and the main thread in the next episode.
In the example, the application updates the
image property of the image views on the main thread by taking advantage of Grand Central Dispatch. The application asks Grand Central Dispatch for a reference to the main queue and passes it a closure, a block of work, in which the
image property of the image view is set.
Grand Central Dispatch also manages a number of concurrent queues that perform work on a pool of background threads. These dispatch queues are also known as global queues. Remember that a concurrent queue doesn't care about the order of execution. Performance is key at the cost of predictability.
I hope that this episode has demystified a few more advanced concepts, such as threads, queues, and concurrency. If you take the time to understand what these concepts entail, then it makes working with technologies like Grand Central Dispatch much easier. The next episode zooms in on the main thread. What is it and why is it important?