Skip to content

Latest commit

 

History

History
269 lines (224 loc) · 10.2 KB

README.md

File metadata and controls

269 lines (224 loc) · 10.2 KB

SwiftConcurrencyCheatSheet

Dispatch Queues

Dispatch queue is an object that manages the execution of tasks serially or concurrently on your app's main thread or on a background thread.

Serial & Concurrent Queues

SERIAL queues can only use one thread, which means that only one task can be completed at a time.

let serialQueue = DispatchQueue(label: "serial")

CONCURRENT queues can only use as many threads as the system has resources for.

let cuncurrentQueue = DispatchQueue(label: "cuncurrent", attributes: .concurrent)

ASYNCHRONOUS DOES NOT MEAN CONCURRENT

An ASYNCHRONOUS queue executes the task in another thread, while a SYNCHRONOUS queue just waits for the task to complete before executing the next one.

Main Queue

The MAIN queue is the dispatch queue associated with the main thread that's responsible for UI. You should NEVER execute something synchronously in the main queue unless it is related to the UI. Otherwise, it will freeze your app until your synchronous task is completed.

In addition, all DispatchQueue.main.sync calls should NEVER be called from the main thread, otherwise, it will cause deadlock.

DispatchQueue.main.async {
    DispatchQueue.main.sync {
        tableView.reloadData()
    }
}

Global Queue

The GLOBAL queue is the dispatch queue that executes tasks concurrently using threads from the global thread pool.

let urlToImage = "https://picsum.photos/seed/picsum/200/300" 
DispatchQueue.global(qos: .utility).async {
    guard let url = URL(string: urlToImage), 
    let data = try? Data(contentsOf: url),
    let image = UIImage(data: data) else { return } 
    
    DispatchQueue.main.async {
        imageView.image = image 
    } 
}

QOS

The quality of service allows you to determine how important a task is and how quickly it should be completed.

  • .userInteractive is used for highest priority tasks, such as animations, event handling, or updating your app's user interface.
  • .userInitiated is used for second priority tasks that provide immediate results for something the user is doing, such as email opening.
  • .utility is used for long-running tasks that may have a progress indicator.
  • .background is used for lowest priority tasks, such as server synchronization.
  • .default is used for tasks that perform active work on the user's behalf.
  • .unspecified should not be used directly unless you know what are you doing.

Dispatch Groups

Dispatch groups allow you to organize tasks into groups that can perform completion blocks after tasks are completed. For example, the code below will print Work One at first, Work Two after two seconds delay, and All work is done at the end.

let dispatchGroup = DispatchGroup()

DispatchQueue.global(qos: .background).async {
    print("Work One")

    dispatchGroup.enter()
    DispatchQueue.main.asyncAfter(deadline: .now() + 2) {
        print("Work Two")
        dispatchGroup.leave()
    }
}

dispatchGroup.notify(queue: DispatchQueue.main) {
    print("All work is done")
}

Remember that count of .enter() and .leave() calls must be equal otherwise .notify() completion block will never be executed or will be executed not in time.

Dispatch Semaphores

A dispatch semaphore is an efficient implementation of a traditional counting semaphore which is defined as a non-negative integer variable and what a dispatch semaphore does is limit the number of concurrent tasks performed at a time.

You increment a semaphore count by calling the .signal() method, and decrement a semaphore count or wait for a signal by calling .wait(). For example, the code below will print two task numbers every 2 seconds.

let semaphore = DispatchSemaphore(value: 2)

for index in 1 ..< 5 {
    DispatchQueue.global(qos: .utility).async {
        semaphore.wait()
        
        Thread.sleep(forTimeInterval: 2)
        print("Task number is \(index)")
        
        semaphore.signal()
    }
}

Dispatch Work Items

A dispatch work item encapsulates work to be performed on a dispatch queue or within a dispatch group. A dispatch work item has a cancel flag. If it is canceled before running, the dispatch queue won’t execute it and will skip it. If it is canceled during its execution, the cancel property return true and the dispatch queue will continue executing so there is no way to stop a dispatch work item after it is started. Also, a dispatch work item can perform a completion block after work is completed.

For example, the code below will execute the first task but the second task will never be executed.

let dispatchQueue = DispatchQueue.global()

let firstDispatchWorkItem = DispatchWorkItem {
    print("First Work Item Started")
    Thread.sleep(forTimeInterval: 2)

    print("First Work Item Completed")
}
let secondDispatchWorkItem = DispatchWorkItem {
    print("Second Work Item Started")
    Thread.sleep(forTimeInterval: 2)

    print("Second Work Item Completed")
}

dispatchQueue.async(execute: firstDispatchWorkItem)

firstDispatchWorkItem.notify(queue: DispatchQueue.main) {
    secondDispatchWorkItem.cancel()
    print(secondDispatchWorkItem.isCancelled ? "Second Work Item Canceled": "Second Work Item Is Ready For Execution")

    dispatchQueue.async(execute: secondDispatchWorkItem)
}

Remember that .notify() method will be executed no matter isCancelled property set to false or not.

Block Operations

The BlockOperation class is a concrete subclass of Operation that manages the concurrent execution of one or more blocks on the default global queue but there's a catch. Execution blocks will run concurrently, but BlockOperation itself is not actually concurrent, it's serial and we can see that by calling BlockOperation().isConcurrent so the calling thread will be stuck until all the execution blocks finish their work.

import Foundation

let blockOperation = BlockOperation()

for index in 0 ..< 4 {
    blockOperation.addExecutionBlock {
        Thread.sleep(forTimeInterval: TimeInterval(index))
        print("Thread slept for \(index) seconds")
    }
}
blockOperation.completionBlock = {
    print("Thread woke up")
}

blockOperation.start()

print("Block Operation is serial so this line will be executed after blockOperation ends.")

So, if you want your operation to be fully concurrent, you must implement the appropriate functionality in an Operation's subclass.

Operation Queues

Operation Queues regulate the execution of operations, same as Dispatch Queues. Take a look at the Block Operations section code reimplementation.

import Foundation

let queue = OperationQueue()
let blockOperation = BlockOperation()

for index in 0 ..< 4 {
    blockOperation.addExecutionBlock {
        Thread.sleep(forTimeInterval: TimeInterval(index))
        print("Thread slept for \(index) seconds")
    }
}

queue.addOperation(blockOperation)

print("Operation queue does not block the main thread because it's concurrent so that line will run first")

Remember that you can pause any operation queue by setting the isSuspended property to true.

Subclassing Operations

You can create reusable operations by subclassing Operation class. For example, the code below is the synchronous reimplementation of code used in the Block Operations section.

import Foundation

class CustomOperation: Operation {
    var executionBlocksCount: Int
    
    init(executionBlocksCount: Int) {
        self.executionBlocksCount = executionBlocksCount
        super.init()
    }
    override func main() {
        for index in 0 ..< executionBlocksCount + 1 {
            Thread.sleep(forTimeInterval: TimeInterval(index))
            print("Thread slept for \(index) seconds")
        }
    }
}
let customOperation = CustomOperation(executionBlocksCount: 2)
customOperation.completionBlock = {
    print("Custom Operation is Completed")
}
customOperation.start()

Note that this code is synchronous because in the for in loop we don't add any execution blocks.

Asynchronous Operations

You can create reusable asynchronous operations by creating and subclassing custom AsynchronousOperation class. AsynchronousOperation is a subclass of Operation class and it overrides isReady, isExecuting and isFinished properties that allow us to manually finish or delay the operation. For example, the code below will implement AsynchronousOperation and its subclass that will be used for image downloading.

class AsynchronousOperation: Operation {
    enum State: String {
        case ready
        case executing
        case finished
        
        var key: String { "is\(rawValue.capitalized)" }
    }
    var state = State.ready {
        willSet {
            willChangeValue(forKey: newValue.key)
            willChangeValue(forKey: state.key)
        } didSet {
            didChangeValue(forKey: oldValue.key)
            didChangeValue(forKey: state.key)
        }
    }
    final override public var isAsynchronous: Bool { true }
    override public var isReady: Bool { super.isReady && state == .ready }
    override public var isExecuting: Bool { state == .executing }
    override public var isFinished: Bool { state == .finished }
    
    override func cancel() { state = .finished }
    final override func start() {
        guard !isCancelled else {
            state = .finished
            return
        }
        
        main()
        state = .executing
    }
}
class ImageFromNetworkOperation: AsynchronousOperation {
    var urlToImage: String
    let completion: (UIImage) -> Void
    
    init(urlToImage: String, completion: @escaping (UIImage) -> Void) {
        self.urlToImage = urlToImage
        self.completion = completion
    }
    
    override func main() {
        DispatchQueue.global(qos: .utility).async {
            guard let url = URL(string: self.urlToImage),
                  let data = try? Data(contentsOf: url),
                  let image = UIImage(data: data) else { return }
            self.completion(image)
            self.state = .finished
        }
    }
}
let urlToImage = "https://picsum.photos/seed/picsum/200/300"
let imageOperation = ImageFromNetworkOperation(urlToImage: urlToImage) { (image) in
    DispatchQueue.main.async {
        imageView.image = image
    }
}
imageOperation.start()