Guarantees in Swift for inexperienced persons

Sync vs async execution

Writing asynchronous code is likely one of the hardest a part of constructing an app.

What precisely is the distinction between a synchronous and an asynchronous execution? Nicely, I already defined this in my Dispatch framework tutorial, however here’s a fast recap. A synchoronousoperate normally blocks the present thread and returns some worth afterward. An asynchronousoperate will immediately return and passes the outcome worth right into a completion handler. You should utilize the GCD framework to carry out duties sync on async on a given queue. Let me present you a fast instance:

func aBlockingFunction() -> String {
    sleep(.random(in: 1...3))
    return "Hiya world!"
}

func syncMethod() -> String {
    return aBlockingFunction()
}

func asyncMethod(completion block: @escaping ((String) -> Void)) {
    DispatchQueue.world(qos: .background).async {
        block(aBlockingFunction())
    }
}

print(syncMethod())
print("sync technique returned")
asyncMethod { worth in
    print(worth)
}
print("async technique returned")

As you’ll be able to see the async technique runs solely on a background queue, the operate will not block the present thread. For this reason the async technique can return immediately, so you may at all times see the return output earlier than the final howdy output. The async technique’s completion block is saved for later execution, that is the rationale why is it attainable to call-back and return the string worth approach after the unique operate have returned.

What occurs should you do not use a distinct queue? The completion block will probably be executed on the present queue, so your operate will block it. It is going to be considerably async-like, however in actuality you are simply shifting the return worth right into a completion block.

func syncMethod() -> String {
    return "Hiya world!"
}

func fakeAsyncMethod(completion block: ((String) -> Void)) {
    block("Hiya world!")
}

print(syncMethod())
print("sync technique returned")
fakeAsyncMethod { worth in
    print(worth)
}
print("pretend async technique returned")

I do not actually need to concentrate on completion blocks on this article, that might be a standalone put up, however if you’re nonetheless having hassle with the concurrency mannequin or you do not perceive how duties and threading works, it’s best to learn perform a little analysis first.

Callback hell and the pyramid of doom

What is the drawback with async code? Or what’s the results of writing asynchronous code? The quick reply is that it’s a must to use completion blocks (callbacks) as a way to deal with future outcomes.

The lengthy reply is that managing callbacks sucks. You must watch out, as a result of in a block you’ll be able to simply create a retain-cycle, so it’s a must to go round your variables as weak or unowned references. Additionally if it’s a must to use a number of async strategies, that’ll be a ache within the donkey. Pattern time! 🐴

struct Todo: Codable {
    let id: Int
    let title: String
    let accomplished: Bool
}

let url = URL(string: "https://jsonplaceholder.typicode.com/todos")!

URLSession.shared.dataTask(with: url) { knowledge, response, error in
    if let error = error {
        fatalError("Community error: " + error.localizedDescription)
    }
    guard let response = response as? HTTPURLResponse else {
        fatalError("Not a HTTP response")
    }
    guard response.statusCode <= 200, response.statusCode > 300 else {
        fatalError("Invalid HTTP standing code")
    }
    guard let knowledge = knowledge else {
        fatalError("No HTTP knowledge")
    }

    do {
        let todos = strive JSONDecoder().decode([Todo].self, from: knowledge)
        print(todos)
    }
    catch {
        fatalError("JSON decoder error: " + error.localizedDescription)
    }
}.resume()

The snippet above is an easy async HTTP knowledge request. As you’ll be able to see there are many optionally available values concerned, plus it’s a must to do some JSON decoding if you wish to use your individual sorts. This is only one request, however what should you’d have to get some detailed information from the primary ingredient? Let’s write a helper! #no 🤫

func request(_ url: URL, completion: @escaping ((Information) -> Void)) {
    URLSession.shared.dataTask(with: url) { knowledge, response, error in
        if let error = error {
            fatalError("Community error: " + error.localizedDescription)
        }
        guard let response = response as? HTTPURLResponse else {
            fatalError("Not a HTTP response")
        }
        guard response.statusCode <= 200, response.statusCode > 300 else {
            fatalError("Invalid HTTP standing code")
        }
        guard let knowledge = knowledge else {
            fatalError("No HTTP knowledge")
        }
        completion(knowledge)
    }.resume()
}


let url = URL(string: "https://jsonplaceholder.typicode.com/todos")!
request(url) { knowledge in
    do {
        let todos = strive JSONDecoder().decode([Todo].self, from: knowledge)
        guard let first = todos.first else {
            return
        }
        let url = URL(string: "https://jsonplaceholder.typicode.com/todos/(first.id)")!
        request(url) { knowledge in
            do {
                let todo = strive JSONDecoder().decode(Todo.self, from: knowledge)
                print(todo)
            }
            catch {
                fatalError("JSON decoder error: " + error.localizedDescription)
            }
        }
    }
    catch {
        fatalError("JSON decoder error: " + error.localizedDescription)
    }
}

See? My drawback is that we’re slowly shifting down the rabbit gap. Now what if now we have a third request? Hell no! You must nest all the pieces one stage deeper once more, plus it’s a must to go across the essential variables eg. a weak or unowned view controller reference as a result of in some unspecified time in the future in time it’s a must to replace your complete UI based mostly on the end result. There should be a greater method to repair this. 🤔

Outcomes vs futures vs guarantees?

The outcome kind was launched in Swift 5 and it is extraordinarily good for eliminating the optionally available issue from the equation. This implies you do not have to cope with an optionally available knowledge, and an optionally available error kind, however your result’s both of them.

Futures are mainly representing a price sooner or later. The underlying worth could be for instance a outcome and it ought to have one of many following states:

• pending – no worth but, ready for it…
• fulfilled – success, now the outcome has a price
• rejected – failed with an error

By definition a futures should not be writeable by the end-user. Because of this builders shouldn’t be capable of create, fulfill or reject one. But when that is the case and we observe the principles, how can we make futures?

We promise them. You must create a promise, which is mainly a wrapper round a future that may be written (fulfilled, rejected) or reworked as you need. You do not write futures, you make guarantees. Nevertheless some frameworks lets you get again the longer term worth of a promise, however you should not be capable to write that future in any respect.

Sufficient principle, are you able to fall in love with guarantees? ❤️

Guarantees 101 – a newbie’s information

Let’s refactor the earlier instance by utilizing my promise framework!

extension URLSession {

    enum HTTPError: LocalizedError {
        case invalidResponse
        case invalidStatusCode
        case noData
    }

    func dataTask(url: URL) -> Promise<Information> {
        return Promise<Information> { [unowned self] fulfill, reject in
            self.dataTask(with: url) { knowledge, response, error in
                if let error = error {
                    reject(error)
                    return
                }
                guard let response = response as? HTTPURLResponse else {
                    reject(HTTPError.invalidResponse)
                    return
                }
                guard response.statusCode <= 200, response.statusCode > 300 else {
                    reject(HTTPError.invalidStatusCode)
                    return
                }
                guard let knowledge = knowledge else {
                    reject(HTTPError.noData)
                    return
                }
                fulfill(knowledge)
            }.resume()
        }
    }
}

enum TodoError: LocalizedError {
    case lacking
}

let url = URL(string: "https://jsonplaceholder.typicode.com/todos")!
URLSession.shared.dataTask(url: url)
.thenMap { knowledge in
    return strive JSONDecoder().decode([Todo].self, from: knowledge)
}
.thenMap { todos -> Todo in
    guard let first = todos.first else {
        throw TodoError.lacking
    }
    return first
}
.then { first in
    let url = URL(string: "https://jsonplaceholder.typicode.com/todos/(first.id)")!
    return URLSession.shared.dataTask(url: url)
}
.thenMap { knowledge in
    strive JSONDecoder().decode(Todo.self, from: knowledge)
}
.onSuccess { todo in
    print(todo)
}
.onFailure(queue: .principal) { error in
    print(error.localizedDescription)
}

What simply occurred right here? Nicely, I made kind of a promisified model of the info process technique carried out on the URLSession object as an extension. After all you’ll be able to return the HTTP outcome or simply the standing code plus the info should you want additional information from the community layer. You should utilize a brand new response knowledge mannequin or perhaps a tuple. 🤷‍♂️

Anyway, the extra fascinating half is the underside half of the supply. As you’ll be able to see I am calling the model new dataTask technique which returns a Promise<Information> object. As I discussed this earlier than a promise could be reworked. Or ought to I say: chained?

Chaining guarantees is the largest benefit over callbacks. The supply code is just not wanting like a pyramid anymore with loopy indentations and do-try-catch blocks, however extra like a sequence of actions. In each single step you’ll be able to remodel your earlier outcome worth into one thing else. If you’re accustomed to some practical paradigms, it may be very easy to know the next:

• thenMap is an easy map on a Promise
• then is mainly flatMap on a Promise
• onSuccess solely will get referred to as if all the pieces was fantastic within the chain
• onFailure solely will get referred to as if some error occurred within the chain
• at all times runs at all times whatever the consequence

If you wish to get the primary queue, you’ll be able to merely go it by a queue parameter, like I did it with the onFailure technique, however it works for each single ingredient within the chain. These capabilities above are simply the tip of the iceberg. You may also faucet into a sequence, validate the outcome, put a timeout on it or get better from a failed promise.

There may be additionally a Guarantees namespace for different helpful strategies, like zip, which is able to zipping collectively 2, 3 or 4 completely different sort of guarantees. Identical to the Guarantees.all technique the zip operate waits till each promise is being accomplished, then it offers you the results of all the guarantees in a single block.

Guarantees.all(guarantees)
.thenMap { arrayOfResults in
    
}

Guarantees.zip(promise1, promise2)
.thenMap { result1, result2 in
    
}

It is also price to say that there’s a first, delay, timeout, race, wait and a retry technique underneath the Guarantees namespace. Be at liberty to mess around with these as properly, generally they’re extremly helpful and highly effective too. 💪

There are solely two issues with guarantees

The primary situation is cancellation. You possibly can’t merely cancel a operating promise. It is doable, however it requires some superior or some say “hacky” methods.

The second is async / await. If you wish to know extra about it, it’s best to learn the concurrency manifesto by Chis Lattner, however since it is a newbie’s information, let’s simply say that these two key phrases can add some syntactic sugar to your code. You will not want the additional (then, thenMap, onSuccess, onFailure) traces anymore, this fashion you’ll be able to focus in your code. I actually hope that we’ll get one thing like this in Swift 6, so I can throw away my Promise library for good. Oh, by the best way, libraries…

Promise libraries price to examine

My promise implementation is way from good, however it’s a fairly easy one (~450 traces of code) and it serves me very well. This weblog put up by @khanlou helped me loads to know guarantees higher, it’s best to learn it too! 👍

There are many promise libraries on github, but when I had to select from them (as a substitute my very own implementation), I might undoubtedly go along with one of many following ones:

• PromiseKit – The preferred one
• Guarantees by Google – characteristic wealthy, fairly widespread as properly
• Promise by Khanlou – small, however based mostly on on the JavaScript Guarantees/A+ spec
• SwiftNIO – not an precise promise library, however it has a superbly written occasion loop based mostly promise implementation underneath the hood

Professional tip: do not attempt to make your individual Promise framework, as a result of multi-threading is extraordinarily onerous, and you do not need to fiddle with threads and locks.

Guarantees are actually addictive. When you begin utilizing them, you’ll be able to’t merely return and write async code with callbacks anymore. Make a promise immediately! 😅