Skip to content

Latest commit

 

History

History
756 lines (590 loc) · 20.7 KB

README.rst

File metadata and controls

756 lines (590 loc) · 20.7 KB

SwiftUI for React Developers

This is a cheat sheet that helps you React developers to quickly start with SwiftUI.

Note

I assume that you are familiar with React Hooks. For the transformation from Class Components to Hooks, I highly recommend you to visit Thinking in React Hooks, which is a great visualized explanation.

One of the core parts of these declarative UI frameworks is its DSL syntax, both of them do provide the special inline syntax for building the content. For React, that calls JSX and need to be transpiled by Babel (with plugins) or tsc. For SwiftUI, it's a built-in syntax in Swift 5.1 called Function Builders.

In React:

const Hello = () => {
  return (
    <div>
      <p>Hello</p>
      <p>React is awesome!</p>
    </div>
  );
};

In SwiftUI:

struct Hello: View {
    var body: some View {
        VStack {
            Text("Hello")
            Text("SwiftUI is awesome!")
        }
    }
}

As you can see, Swift's syntax feels more natural and JSX seems to be more exotic. Actually, Web developers should be more familiar with JSX, after all, it's just like HTML.

Most of components render different contents depend on what input is given to it. That is what props comes to play.

In React:

const Hello = ({name}) => {
  return <p>Hello, {name}!</p>;
};

In SwiftUI:

struct Hello: View {
    let name: String

    var body: some View {
        Text("Hello, \(name)!")
    }
}

Almost the same in semantic!

Structure of the contents can be dynamic, the most common patterns are conditional and list.

In React:

const UserList = ({ users }) => {
  if (!users.length) {
    return <p>No users</p>;
  }

  return (
    <ul>
      {users.map(e => (
        <li key={e.id}>{e.username}</li>
      ))}
    </ul>
  );
}

In SwiftUI:

struct UserList: View {
    let users: [User]

    var body: some View {
        Group {
            if users.isEmpty {
                Text("No users")
            } else {
                VStack {
                    ForEach(users, id: \.id) {
                        Text("\($0.username)")
                    }
                }
            }
        }
    }
}

SwiftUI has built-in ForEach element, you don't need to manually map the data array to views, so you can have a much neater code.

In React:

const Hello = () => {
  const clickHandler = useCallback(e => {
    console.log('Yay, the button is clicked!');
  }, []);
  return <button onClick={clickHandler}>Click Me</button>;
};

In SwiftUI:

struct Hello: View {
    var body: some View {
        Button("Click Me") {
            print("Yay, the button is clicked!")
        }
    }
}

SwiftUI looks cleaner because there is no useCallback meme. In JavaScript, if you create a function inside another function (let's say foo), the former always has a different reference every time foo is called. That means, the component receives the function as a prop will be rerendered every time.

In consideration of performance, React provided useCallback. It takes a value as dependency, and will return the same reference if the dependency is not changed.

In SwiftUI, Apple have not provided such mechanism, and developers can just take no account of that.

Sometimes, a component may retain some internal state even it's get updated by new props. Or it need to update itself without the props changed. State was born for this mission.

The example combines all the things we've talked above. Let's create a simple counter.

In React:

const Counter = ({ initialValue }) => {
  const [counter, setCounter] = useState(initialValue);
  const increaseCounter = useCallback(() => {
    setCounter(counter + 1);
  }, [counter]);

  return (
    <div>
      <p>{counter}</p>
      <button onClick={increaseCounter}>Increase</button>
    </div>
  );
};

In SwiftUI:

struct Counter: View {
    let initialValue: Int

    @State
    var counter: Int

    init(initialValue: Int) {
        self.initialValue = initialValue
        _counter = State(initialValue: initialValue)
    }

    var body: some View {
        VStack {
            Text("\(counter)")
            Button("Increase") {
                self.counter += 1
            }
        }
    }
}

It seems to be a little complicated, let's decompose them into pieces.

The counter has a internal state: counter, and it's initial value is from the input props. In SwiftUI, a state is declared with @State property wrapper. I'll explain that later but now, you could just consider it as a special mark.

The real counter value is wrapped in the _counter member variable (which has type of State<Int>), and we can use the input prop initialValue to initialize it.

We trigger an update by directly setting the counter value. This is not just an assignment, instead, this will cause some logic inside State to take effect and notify the SwiftUI framework to update our view. SwiftUI packed the xxx and setXXX functions into this little syntactic sugar to simplify our code.

How can we perform some side-effects when the component is updated? In React, we have useEffect:

const Hello = ({ greeting, name }) => {
  useEffect(() => {
    console.log(`Hey, ${name}!`);
  }, [name]);

  useEffect(() => {
    console.log('Something changed!');
  });

  return <p>{greeting}, {name}!</p>;
};

In SwiftUI:

func uniqueId() -> some Equatable {
    return UUID().uuidString  // Maybe not so unique?
}

struct Hello: View {
    let greeting: String
    let name: String

    var body: some View {
        Text("\(greeting), \(name)!")
            .onChange(of: name) { name in
                print("Hey, \(name)!")
            }
            .onChange(of: uniqueId()) { _ in
                print("Something changed!")
            }
    }
}

In SwiftUI, we have neither hook functions nor lifecycle functions, but we have modifiers! Every view type has a lot of modifier functions attached to it.

onChange behaves just like useEffect, the action closure is called every time the value changes and the first time the receiver view renders. But we must pass a value, if you need perform something whenever something changed, you can use a trick:

Create a function that returns an unique object every time it gets called. You can use UUID, global incrementing integer and even timestamps!

In React:

const Hello = () => {
  useEffect(() => {
    console.log('I\'m just mounted!');
    return () => {
      console.log('I\'m just unmounted!');
    };
  }, []);

  return <p>Hello</p>;
};

In SwiftUI:

struct Hello: View {
    var body: some View {
        Text("Hello")
            .onAppear {
                print("I'm just mounted!")
            }
            .onDisappear {
                print("I'm just unmounted!")
            }
    }
}

It's that easy.

Components can have some internal state that will not trigger view update when it is changed. In React, we have ref:

In React:

const Hello = () => {
  const timerId = useRef(-1);
  useEffect(() => {
    timerId.current = setInterval(() => {
      console.log('Tick!');
    }, 1000);
    return () => {
      clearInterval(timerId.current);
    };
  });

  return <p>Hello</p>;
};

In SwiftUI:

struct Hello: View {
    private class Refs: ObservableObject {
        var timer: Timer?
    }

    @StateObject
    private var refs = Refs()

    var body: some View {
        Text("Hello")
            .onAppear {
                refs.timer =
                    Timer.scheduledTimer(withTimeInterval: 1,
                                        repeats: true) { _ in
                        print("Tick!")
                    }
            }
            .onDisappear {
                refs.timer?.invalidate()
            }
    }
}

And we've got two approaches:

struct Hello: View {
    @State
    private var timer: Timer? = nil

    var body: some View {
        Text("Hello")
            .onAppear {
                self.timer =
                    Timer.scheduledTimer(withTimeInterval: 1,
                                        repeats: true) { _ in
                        print("Tick!")
                    }
            }
            .onDisappear {
                self.timer?.invalidate()
            }
    }
}

You may wonder why setting the state will not lead to view updates. SwiftUI is pretty clever to handle the state, it uses a technique called Dependency Tracking. If you are familiar with Vue.js or MobX, you may understand it immediately. That's say, if we never access the state's value in the view's building process (which not includes onAppear calls), that state will be unbound and can be updated freely without causing view updates.

Accessing the native DOM object is an advanced but essential feature for Web frontend development.

In React:

const Hello = () => {
  const pEl = useRef();
  useEffect(() => {
    pEl.current.innerHTML = '<b>Hello</b>, world!';
  }, []);

  return <p ref={pEl}></p>;
};

In SwiftUI, we apparently don't have DOM, but for native applications, View is a common concept. We can bridge native views to SwiftUI and gain control of them by the way.

First, let's bridge an existed UIView to SwiftUI:

struct MapView: UIViewRepresentable {
    let mapType: MKMapType
    let ref: RefBox<MKMapView>

    typealias UIViewType = MKMapView

    func makeUIView(context: Context) -> MKMapView {
        return MKMapView(frame: .zero)
    }

    func updateUIView(_ uiView: MKMapView, context: Context) {
        uiView.mapType = mapType
        ref.current = uiView
    }
}

Every time we modified the input props, the updateUIView gets called, we can update our UIView there. To export the UIView instance to the outer, we declare a ref prop, and set it's current property to the view instance whenever the updateUIView gets called.

Now we can manipulate the native view in our SwiftUI views:

struct Hello: View {
    @State
    var mapType = MKMapType.standard

    @StateObject
    var mapViewRef = RefBox<MKMapView>()

    var body: some View {
        VStack {
            MapView(mapType: mapType, ref: mapViewRef)
            Picker("Map Type", selection: $mapType) {
                Text("Standard").tag(MKMapType.standard)
                Text("Satellite").tag(MKMapType.satellite)
                Text("Hybrid").tag(MKMapType.hybrid)
            }
            .pickerStyle(SegmentedPickerStyle())
        }
        .onAppear {
            if let mapView = self.mapViewRef.current {
                mapView.setRegion(.init(center: .init(latitude: 34, longitude: 108),
                                        span: MKCoordinateSpan(latitudeDelta: 50,
                                                               longitudeDelta: 60)),
                                  animated: true)
            }
        }
    }
}

Note that, we'd better encapsulate all the manipulations of native views to a dedicated SwiftUI view. It's not a good practice to manipulate native objects everywhere, as well as in React.

Passing data between the components can be hard, especially when you travel through the hierachy. And Context to the rescue!

Let's look at an example in React:

const UserContext = createContext({});

const UserInfo = () => {
  const { username, logout } = useContext(UserContext);
  if (!username) {
    return <p>Welcome, please login.</p>;
  }
  return (
    <p>
      Hello, {username}.
      <button onClick={logout}>Logout</button>
    </p>
  );
}

const Panel = () => {
  return (
    <div>
      <UserInfo />
      <UserInfo />
    </div>
  );
}

const App = () => {
  const [username, setUsername] = useState('cyan');
  const logout = useCallback(() => {
    setUsername(null);
  }, [setUsername]);
  return (
    <UserContext.Provider value={{ username, logout }}>
      <Panel />
      <Panel />
    </UserContext.Provider>
  );
}

Even if the <UserInfo> is at a very deep position, we can use context to grab the data we need through the tree. And also, contexts are often used by components to communicate with each other.

In SwiftUI:

class UserContext: ObservableObject {
    @Published
    var username: String?

    init(username: String?) {
        self.username = username
    }

    func logout() {
        self.username = nil
    }
}

struct UserInfo: View {
    @EnvironmentObject
    var userContext: UserContext

    var body: some View {
        Group {
            if userContext.username == nil {
                Text("Welcome, please login.")
            } else {
                HStack {
                    Text("Hello, \(userContext.username!).")
                    Button("Logout") {
                        self.userContext.logout()
                    }
                }
            }
        }
    }
}

struct Panel: View {
    var body: some View {
        VStack {
            UserInfo()
            UserInfo()
        }
    }
}

struct App: View {
    @StateObject
    var userContext = UserContext(username: "cyan")

    var body: some View {
        VStack {
            Panel()
            Panel()
        }
        .environmentObject(userContext)
    }
}

Contexts are provided by environmentObject modifier and can be retrieved via @EnvironmentObject property wrapper. And in SwiftUI, context objects can use to update views. We don't need to wrap some functions that modifies the provider into the context objects. Context objects are ObservableObject, so they can notify all the consumers automatically when they are changed.

Another interesting fact is that the contexts are identified by the type of context objects, thus we don't need to maintain the context objects globally.

In SwiftUI, the View objects are different from the React.Component objects. Actually, there is no React.Component equivalent in SwiftUI. View objects are stateless themselves, they are just like Widget objects in Flutter, which are used to describe the configuration of views.

That means, if you want attach some state to the view, you must mark it using @State. Any other member variables are transient and live shorter than the view. After all, View objects are created and destroyed frequently during the building process, but meanwhile views may keep stable.

To explain this question, you should know what is property wrapper before. This proposal describe that in detail: [SE-0258] Property Wrappers.

Before the View is mounted, SwiftUI will use type metadata to find out all the State fields (backends of the properties marked with @State), and add them to a DynamicPropertyBuffer sequentially, we call this process as "registration".

The buffer is aware of the view's lifecycle. When a new View object is created, SwiftUI enumerates the State fields, and get its corresponding previous value from the buffer. These fields are identified by their storage index in container struct, pretty like how Hook works in React.

In this way, even though the View objects are recreated frequently, as long as the view is not unmounted, the state will be kept.

As we mention earlier, SwiftUI use Function Builders as DSL to let us build contents. There is also a draft proposal about it: Function builders (draft proposal).

Let's first take a look at how JSX is transpiled to JavaScript. We have this:

const UserInfo = ({ users }) => {
  if (!users.length) {
    return <p>No users</p>;
  }

  return (
    <div>
      <p>Great!</p>
      <p>We have {users.length} users!</p>
    </div>
  );
}

And this is the output from Babel with react preset:

const UserInfo = ({
  users
}) => {
  if (!users.length) {
    return /*#__PURE__*/React.createElement("p", null, "No users");
  }

  return /*#__PURE__*/React.createElement("div", null,
    /*#__PURE__*/React.createElement("p", null, "Great!"),
    /*#__PURE__*/React.createElement("p", null, "We have ", users.length, " users!")
  );
};

Most of the structure is identical, and the HTML tags are transformed to React.createElement calls. That makes sense, the function doesn't produce component instances, instead, it produces elements. Elements describe how to configure components or DOM elements.

Now, let's back to SwiftUI. There is the same example:

struct UserInfo: View {
    let users: [User]

    var body: some View {
        Group {
            if users.isEmpty {
                Text("No users")
            } else {
                VStack {
                    Text("Great!")
                    Text("We have \(users.count) users!")
                }
            }
        }
    }
}

And this is the actual code represented by it:

struct UserInfo: View {
    let users: [User]

    var body: some View {
        let v: _ConditionalContent<Text, VStack<TupleView<(Text, Text)>>>
        if users.isEmpty {
            v = ViewBuilder.buildEither(first: Text("No users"))
        } else {
            v = ViewBuilder.buildEither(second: VStack {
                return ViewBuilder.buildBlock(
                    Text("Great!"),
                    Text("We have \(users.count) users!")
                )
            })
        }
        return v
    }
}

Voila! All the dynamic structures are replaced by ViewBuilder method calls. In this way, we can use a complex type to represent the structure. Like if statement will be transformed to ViewBuilder.buildEither call, and its return value contains the information of both if block and else block.

ViewBuilder.buildBlock is used to represent a child element that contains multiple views.

With function builders, you can even create your own DSLs. And this year in WWDC20, Apple released more features based on function builders, like WidgetKit and SwiftUI App Structure.

All views in SwiftUI are like PureComponent in React by default. That means, all the member variables (props) will be used to evaluate the equality, of course it's shallow comparison.

What if you want to customize the update strategy? If you take a look at the declaration of View protocol, you will notice this subtle thing:

extension View where Self : Equatable {

    /// Prevents the view from updating its child view when its new value is the
    /// same as its old value.
    @inlinable public func equatable() -> EquatableView<Self>
}

SwiftUI provides an EquatableView to let you achieve that. All you need to do is make your view type conform Equatable and implement the == function. Then wrap it into EquatableView at the call-site.