Articles, podcasts and news about Swift development, by John Sundell.

Conditional conformances in Swift

Published on 01 Apr 2018
Discover page available: Generics

One of the most exciting aspects of Swift is how it continuously keeps evolving, adding more and more powerful features and - through the Evolution process - addressing feedback and requests from the developer community. This week, Swift 4.1 was publicly released as part of Xcode 9.3 - and despite being a minor release, it brings some really interesting new features.

One such feature is Conditional Conformances, which enables a type to conform to a protocol only under a certain set of constraints. Like many of the more advanced features of Swift, it may not seem like something super useful at first glance - but for certain tasks this new feature proves to be quite a powerful tool.

This week, let's take a look at how conditional conformances work - and how this new feature enables us to design code in a much more recursive fashion, making it more flexible while also reducing duplication.

The basics

Let's start with the basics - how to declare conditional conformance to a protocol. Let's say we are building a game that has multiple types that can be converted into a score (it can be levels, collectibles, enemies, etc). To handle all of these types in a uniform way, we define a ScoreConvertible protocol:

protocol ScoreConvertible {
    func computeScore() -> Int

One thing that is very common when working with protocols like the above is to want to deal with arrays of values. In this case, we'd like to easily be able to sum up the total score of all elements of arrays containing ScoreConvertible values. One way of enabling that is to add an extension on Array for when its Element type conforms to ScoreConvertible, like this:

extension Array where Element: ScoreConvertible {
    func computeScore() -> Int {
        return reduce(0) { result, element in
            result + element.computeScore()

The above works great for single dimensional arrays, such as when summarizing the total score of an array of Level objects:

let levels = [Level(id: "water-0"), Level(id: "water-1")]
let score = levels.computeScore()

However, as soon as we start dealing with more complex arrays (like if we use nested arrays to group our levels into worlds), we start running into problems. Since Array itself doesn't actually conform to ScoreConvertible we wouldn't be able to compute an array of arrays into a total score. We also wouldn't want all arrays to conform to ScoreConvertible, since that wouldn't make sense for variants like [String] or [UIView].

This is the core problem that the conditional conformances feature aims to solve. Now, in Swift 4.1, we can make Array conform to ScoreConvertible only if it contains ScoreConvertible elements, like this:

extension Array: ScoreConvertible where Element: ScoreConvertible {
    func computeScore() -> Int {
        return reduce(0) { result, element in
            result + element.computeScore()

Which in turn lets us compute the total score of any number of nested arrays containing ScoreConvertible types:

let worlds = [
    [Level(id: "water-0"), Level(id: "water-1")],
    [Level(id: "sand-0"), Level(id: "sand-1")],
    [Level(id: "lava-0"), Level(id: "lava-1")]

let totalScore = worlds.computeScore()

As we iterate on our code base, having this level of flexibility can be really sweet! 🍭

Recursive design

The big benefit of conditional conformances is that they allow us to design our code and systems in a more recursive fashion. By being able to nest types and collections, like in the example above, we are free to structure our objects and values in much more flexible ways.

One of the most clear benefits of such recursive design in the Swift standard library, is that collections containing Equatable types are now also equatable themselves. Similar to our ScoreConvertible example above, we are now free to check nested collections for equality without having to write any extra code:

func didLoadArticles(_ articles: [String : [Article]]) {
    // We can now compare nested collections containing Equatable
    // types simply by using the == or != operators.
    guard articles != currentArticles else {

    currentArticles = articles

While being able to do the above is pretty neat, it's also important to remember that such an equality check can hide complexity - since checking if two collections are equal is an O(n) operation.

Multipart requests

Now let's take a look at a bit more advanced example, in which we'll use a conditional conformance to create a nice API for handling multiple network requests. We'll start by defining a protocol for a Request, that can return a Result type containing any Response, like this:

protocol Request {
    associatedtype Response

    typealias Handler = (Result<Response>) -> Void

    func perform(then handler: @escaping Handler)

Let's say we're building an app for a magazine that lets our users read articles within different categories. To be able to load an array of articles for a given category, we define an ArticleRequest type that conforms to the above Request protocol:

struct ArticleRequest: Request {
    typealias Response = [Article]

    let dataLoader: DataLoader
    let category: Category

    func perform(then handler: @escaping Handler) {
        let endpoint = Endpoint.articles(category)

        dataLoader.load(from: endpoint) { result in
            // Here we decode a Result<Data> value to either
            // produce an error or an array of models.

Just like we wanted to be able to sum up the total score for multiple ScoreConvertible values in the earlier example, let's say we want to have an easy way to perform multiple request in a synchronized fashion. For example, we might want to load the articles for multiple categories at once, and get a dictionary containing all of the combined results back.

You might be able to guess where this is going 😉. By making Dictionary conditionally conform to Request when it contains values that themselves conform to Request, we can again solve this problem in a very nice, recursive way.

Like we took a look at in "A deep dive into Grand Central Dispatch in Swift", we'll use GCD's DispatchGroup to synchronize our group of requests and produce an aggregated result, like this:

extension Dictionary: Request where Value: Request {
    typealias Response = [Key : Value.Response]

    func perform(then handler: @escaping Handler) {
        var responses = [Key : Value.Response]()
        let group = DispatchGroup()

        for (key, request) in self {

            request.perform { response in
                switch response {
                case .success(let value):
                    responses[key] = value
                case .error(let error):

        group.notify(queue: .main) {

With the above extension in place, we can now easily create groups of requests simply by using a dictionary literal:

extension TopArticlesViewController {
    func loadArticles() {
        let requests: [Category : ArticleRequest] = [
            .news: ArticleRequest(dataLoader: dataLoader, category: .news),
            .sports: ArticleRequest(dataLoader: dataLoader, category: .sports)

        requests.perform { [weak self] result in
            switch result {
            case .success(let articles):
                for (category, articles) in articles {
                    self?.render(articles, in: category)
            case .error(let error):

We can now use a single, unified implementation for combining multiple requests, rather than having to write separate implementations for various combinations of requests and collections 👍.


Conditional conformances enable us to specialize generics in new ways, which in turn enables systems to be designed in a much more recursive way. While using this feature can also add complication, especially if these type of conformance extensions are scattered all over a code base, when used with moderation it can help us reduce code duplication and make our code more flexible.

Since this is a brand new Swift feature, I'm sure we'll return to conditional conformances again in the future, and take a look at different ways they can be used to solve other kinds of problems. For now, I encourage you to play around with this new feature, and experiment with how it might be able to make your own code more recursive & flexible as well.

What do you think? Is this a feature you're excited about, and do you already have a use case in mind for it? Let me know, along with any questions, comments or feedback you might have - on Twitter @johnsundell.

Thanks for reading! 🚀