Protocols

Protocols

A protocol defines a blueprint of methods, properties, and other requirements that suit a particular task or piece of functionality. The protocol can then be adopted by a class, structure, or enumeration to provide an actual implementation of those requirements. Any type that satisfies the requirements of a protocol is said to conform to that protocol.

In addition to specifying requirements that conforming types must implement, you can extend a protocol to implement some of these requirements or to implement additional functionality that conforming types can take advantage of.

Protocol Syntax

You define protocols in a very similar way to classes, structures, and enumerations:

  1. protocol SomeProtocol {
  2. // protocol definition goes here
  3. }

Custom types state that they adopt a particular protocol by placing the protocol’s name after the type’s name, separated by a colon, as part of their definition. Multiple protocols can be listed, and are separated by commas:

  1. struct SomeStructure: FirstProtocol, AnotherProtocol {
  2. // structure definition goes here
  3. }

If a class has a superclass, list the superclass name before any protocols it adopts, followed by a comma:

  1. class SomeClass: SomeSuperclass, FirstProtocol, AnotherProtocol {
  2. // class definition goes here
  3. }

Property Requirements

A protocol can require any conforming type to provide an instance property or type property with a particular name and type. The protocol doesn’t specify whether the property should be a stored property or a computed property—it only specifies the required property name and type. The protocol also specifies whether each property must be gettable or gettable and settable.

If a protocol requires a property to be gettable and settable, that property requirement can’t be fulfilled by a constant stored property or a read-only computed property. If the protocol only requires a property to be gettable, the requirement can be satisfied by any kind of property, and it’s valid for the property to be also settable if this is useful for your own code.

Property requirements are always declared as variable properties, prefixed with the var keyword. Gettable and settable properties are indicated by writing { get set } after their type declaration, and gettable properties are indicated by writing { get }.

  1. protocol SomeProtocol {
  2. var mustBeSettable: Int { get set }
  3. var doesNotNeedToBeSettable: Int { get }
  4. }

Always prefix type property requirements with the static keyword when you define them in a protocol. This rule pertains even though type property requirements can be prefixed with the class or static keyword when implemented by a class:

  1. protocol AnotherProtocol {
  2. static var someTypeProperty: Int { get set }
  3. }

Here’s an example of a protocol with a single instance property requirement:

  1. protocol FullyNamed {
  2. var fullName: String { get }
  3. }

The FullyNamed protocol requires a conforming type to provide a fully qualified name. The protocol doesn’t specify anything else about the nature of the conforming type—it only specifies that the type must be able to provide a full name for itself. The protocol states that any FullyNamed type must have a gettable instance property called fullName, which is of type String.

Here’s an example of a simple structure that adopts and conforms to the FullyNamed protocol:

  1. struct Person: FullyNamed {
  2. var fullName: String
  3. }
  4. let john = Person(fullName: "John Appleseed")
  5. // john.fullName is "John Appleseed"

This example defines a structure called Person, which represents a specific named person. It states that it adopts the FullyNamed protocol as part of the first line of its definition.

Each instance of Person has a single stored property called fullName, which is of type String. This matches the single requirement of the FullyNamed protocol, and means that Person has correctly conformed to the protocol. (Swift reports an error at compile-time if a protocol requirement is not fulfilled.)

Here’s a more complex class, which also adopts and conforms to the FullyNamed protocol:

  1. class Starship: FullyNamed {
  2. var prefix: String?
  3. var name: String
  4. init(name: String, prefix: String? = nil) {
  5. self.name = name
  6. self.prefix = prefix
  7. }
  8. var fullName: String {
  9. return (prefix != nil ? prefix! + " " : "") + name
  10. }
  11. }
  12. var ncc1701 = Starship(name: "Enterprise", prefix: "USS")
  13. // ncc1701.fullName is "USS Enterprise"

This class implements the fullName property requirement as a computed read-only property for a starship. Each Starship class instance stores a mandatory name and an optional prefix. The fullName property uses the prefix value if it exists, and prepends it to the beginning of name to create a full name for the starship.

Method Requirements

Protocols can require specific instance methods and type methods to be implemented by conforming types. These methods are written as part of the protocol’s definition in exactly the same way as for normal instance and type methods, but without curly braces or a method body. Variadic parameters are allowed, subject to the same rules as for normal methods. Default values, however, can’t be specified for method parameters within a protocol’s definition.

As with type property requirements, you always prefix type method requirements with the static keyword when they’re defined in a protocol. This is true even though type method requirements are prefixed with the class or static keyword when implemented by a class:

  1. protocol SomeProtocol {
  2. static func someTypeMethod()
  3. }

The following example defines a protocol with a single instance method requirement:

  1. protocol RandomNumberGenerator {
  2. func random() -> Double
  3. }

This protocol, RandomNumberGenerator, requires any conforming type to have an instance method called random, which returns a Double value whenever it’s called. Although it’s not specified as part of the protocol, it’s assumed that this value will be a number from 0.0 up to (but not including) 1.0.

The RandomNumberGenerator protocol doesn’t make any assumptions about how each random number will be generated—it simply requires the generator to provide a standard way to generate a new random number.

Here’s an implementation of a class that adopts and conforms to the RandomNumberGenerator protocol. This class implements a pseudorandom number generator algorithm known as a linear congruential generator:

  1. class LinearCongruentialGenerator: RandomNumberGenerator {
  2. var lastRandom = 42.0
  3. let m = 139968.0
  4. let a = 3877.0
  5. let c = 29573.0
  6. func random() -> Double {
  7. lastRandom = ((lastRandom * a + c)
  8. .truncatingRemainder(dividingBy:m))
  9. return lastRandom / m
  10. }
  11. }
  12. let generator = LinearCongruentialGenerator()
  13. print("Here's a random number: \(generator.random())")
  14. // Prints "Here's a random number: 0.3746499199817101"
  15. print("And another one: \(generator.random())")
  16. // Prints "And another one: 0.729023776863283"

Mutating Method Requirements

It’s sometimes necessary for a method to modify (or mutate) the instance it belongs to. For instance methods on value types (that is, structures and enumerations) you place the mutating keyword before a method’s func keyword to indicate that the method is allowed to modify the instance it belongs to and any properties of that instance. This process is described in Modifying Value Types from Within Instance Methods.

If you define a protocol instance method requirement that is intended to mutate instances of any type that adopts the protocol, mark the method with the mutating keyword as part of the protocol’s definition. This enables structures and enumerations to adopt the protocol and satisfy that method requirement.

Note

If you mark a protocol instance method requirement as mutating, you don’t need to write the mutating keyword when writing an implementation of that method for a class. The mutating keyword is only used by structures and enumerations.

The example below defines a protocol called Togglable, which defines a single instance method requirement called toggle. As its name suggests, the toggle() method is intended to toggle or invert the state of any conforming type, typically by modifying a property of that type.

The toggle() method is marked with the mutating keyword as part of the Togglable protocol definition, to indicate that the method is expected to mutate the state of a conforming instance when it’s called:

  1. protocol Togglable {
  2. mutating func toggle()
  3. }

If you implement the Togglable protocol for a structure or enumeration, that structure or enumeration can conform to the protocol by providing an implementation of the toggle() method that is also marked as mutating.

The example below defines an enumeration called OnOffSwitch. This enumeration toggles between two states, indicated by the enumeration cases on and off. The enumeration’s toggle implementation is marked as mutating, to match the Togglable protocol’s requirements:

  1. enum OnOffSwitch: Togglable {
  2. case off, on
  3. mutating func toggle() {
  4. switch self {
  5. case .off:
  6. self = .on
  7. case .on:
  8. self = .off
  9. }
  10. }
  11. }
  12. var lightSwitch = OnOffSwitch.off
  13. lightSwitch.toggle()
  14. // lightSwitch is now equal to .on

Initializer Requirements

Protocols can require specific initializers to be implemented by conforming types. You write these initializers as part of the protocol’s definition in exactly the same way as for normal initializers, but without curly braces or an initializer body:

  1. protocol SomeProtocol {
  2. init(someParameter: Int)
  3. }

Class Implementations of Protocol Initializer Requirements

You can implement a protocol initializer requirement on a conforming class as either a designated initializer or a convenience initializer. In both cases, you must mark the initializer implementation with the required modifier:

  1. class SomeClass: SomeProtocol {
  2. required init(someParameter: Int) {
  3. // initializer implementation goes here
  4. }
  5. }

The use of the required modifier ensures that you provide an explicit or inherited implementation of the initializer requirement on all subclasses of the conforming class, such that they also conform to the protocol.

For more information on required initializers, see Required Initializers.

Note

You don’t need to mark protocol initializer implementations with the required modifier on classes that are marked with the final modifier, because final classes can’t subclassed. For more about the final modifier, see Preventing Overrides.

If a subclass overrides a designated initializer from a superclass, and also implements a matching initializer requirement from a protocol, mark the initializer implementation with both the required and override modifiers:

  1. protocol SomeProtocol {
  2. init()
  3. }
  4. class SomeSuperClass {
  5. init() {
  6. // initializer implementation goes here
  7. }
  8. }
  9. class SomeSubClass: SomeSuperClass, SomeProtocol {
  10. // "required" from SomeProtocol conformance; "override" from SomeSuperClass
  11. required override init() {
  12. // initializer implementation goes here
  13. }
  14. }

Failable Initializer Requirements

Protocols can define failable initializer requirements for conforming types, as defined in Failable Initializers.

A failable initializer requirement can be satisfied by a failable or nonfailable initializer on a conforming type. A nonfailable initializer requirement can be satisfied by a nonfailable initializer or an implicitly unwrapped failable initializer.

Protocols as Types

Protocols don’t actually implement any functionality themselves. Nonetheless, you can use protocols as a fully fledged types in your code. Using a protocol as a type is sometimes called an existential type, which comes from the phrase “there exists a type T such that T conforms to the protocol”.

You can use a protocol in many places where other types are allowed, including:

  • As a parameter type or return type in a function, method, or initializer

  • As the type of a constant, variable, or property

  • As the type of items in an array, dictionary, or other container

Note

Because protocols are types, begin their names with a capital letter (such as FullyNamed and RandomNumberGenerator) to match the names of other types in Swift (such as Int, String, and Double).

Here’s an example of a protocol used as a type:

  1. class Dice {
  2. let sides: Int
  3. let generator: RandomNumberGenerator
  4. init(sides: Int, generator: RandomNumberGenerator) {
  5. self.sides = sides
  6. self.generator = generator
  7. }
  8. func roll() -> Int {
  9. return Int(generator.random() * Double(sides)) + 1
  10. }
  11. }

This example defines a new class called Dice, which represents an n-sided dice for use in a board game. Dice instances have an integer property called sides, which represents how many sides they have, and a property called generator, which provides a random number generator from which to create dice roll values.

The generator property is of type RandomNumberGenerator. Therefore, you can set it to an instance of any type that adopts the RandomNumberGenerator protocol. Nothing else is required of the instance you assign to this property, except that the instance must adopt the RandomNumberGenerator protocol. Because its type is RandomNumberGenerator, code inside the Dice class can only interact with generator in ways that apply to all generators that conform to this protocol. That means it can’t use any methods or properties that are defined by the underlying type of the generator. However, you can downcast from a protocol type to an underlying type in the same way you can downcast from a superclass to a subclass, as discussed in Downcasting.

Dice also has an initializer, to set up its initial state. This initializer has a parameter called generator, which is also of type RandomNumberGenerator. You can pass a value of any conforming type in to this parameter when initializing a new Dice instance.

Dice provides one instance method, roll, which returns an integer value between 1 and the number of sides on the dice. This method calls the generator’s random() method to create a new random number between 0.0 and 1.0, and uses this random number to create a dice roll value within the correct range. Because generator is known to adopt RandomNumberGenerator, it’s guaranteed to have a random() method to call.

Here’s how the Dice class can be used to create a six-sided dice with a LinearCongruentialGenerator instance as its random number generator:

  1. var d6 = Dice(sides: 6, generator: LinearCongruentialGenerator())
  2. for _ in 1...5 {
  3. print("Random dice roll is \(d6.roll())")
  4. }
  5. // Random dice roll is 3
  6. // Random dice roll is 5
  7. // Random dice roll is 4
  8. // Random dice roll is 5
  9. // Random dice roll is 4

Delegation

Delegation is a design pattern that enables a class or structure to hand off (or delegate) some of its responsibilities to an instance of another type. This design pattern is implemented by defining a protocol that encapsulates the delegated responsibilities, such that a conforming type (known as a delegate) is guaranteed to provide the functionality that has been delegated. Delegation can be used to respond to a particular action, or to retrieve data from an external source without needing to know the underlying type of that source.

The example below defines two protocols for use with dice-based board games:

  1. protocol DiceGame {
  2. var dice: Dice { get }
  3. func play()
  4. }
  5. protocol DiceGameDelegate: AnyObject {
  6. func gameDidStart(_ game: DiceGame)
  7. func game(_ game: DiceGame, didStartNewTurnWithDiceRoll diceRoll: Int)
  8. func gameDidEnd(_ game: DiceGame)
  9. }

The DiceGame protocol is a protocol that can be adopted by any game that involves dice.

The DiceGameDelegate protocol can be adopted to track the progress of a DiceGame. To prevent strong reference cycles, delegates are declared as weak references. For information about weak references, see Strong Reference Cycles Between Class Instances. Marking the protocol as class-only lets the SnakesAndLadders class later in this chapter declare that its delegate must use a weak reference. A class-only protocol is marked by its inheritance from AnyObject, as discussed in Class-Only Protocols.

Here’s a version of the Snakes and Ladders game originally introduced in Control Flow. This version is adapted to use a Dice instance for its dice-rolls; to adopt the DiceGame protocol; and to notify a DiceGameDelegate about its progress:

  1. class SnakesAndLadders: DiceGame {
  2. let finalSquare = 25
  3. let dice = Dice(sides: 6, generator: LinearCongruentialGenerator())
  4. var square = 0
  5. var board: [Int]
  6. init() {
  7. board = Array(repeating: 0, count: finalSquare + 1)
  8. board[03] = +08; board[06] = +11; board[09] = +09; board[10] = +02
  9. board[14] = -10; board[19] = -11; board[22] = -02; board[24] = -08
  10. }
  11. weak var delegate: DiceGameDelegate?
  12. func play() {
  13. square = 0
  14. delegate?.gameDidStart(self)
  15. gameLoop: while square != finalSquare {
  16. let diceRoll = dice.roll()
  17. delegate?.game(self, didStartNewTurnWithDiceRoll: diceRoll)
  18. switch square + diceRoll {
  19. case finalSquare:
  20. break gameLoop
  21. case let newSquare where newSquare > finalSquare:
  22. continue gameLoop
  23. default:
  24. square += diceRoll
  25. square += board[square]
  26. }
  27. }
  28. delegate?.gameDidEnd(self)
  29. }
  30. }

For a description of the Snakes and Ladders gameplay, see Break.

This version of the game is wrapped up as a class called SnakesAndLadders, which adopts the DiceGame protocol. It provides a gettable dice property and a play() method in order to conform to the protocol. (The dice property is declared as a constant property because it doesn’t need to change after initialization, and the protocol only requires that it must be gettable.)

The Snakes and Ladders game board setup takes place within the class’s init() initializer. All game logic is moved into the protocol’s play method, which uses the protocol’s required dice property to provide its dice roll values.

Note that the delegate property is defined as an optional DiceGameDelegate, because a delegate isn’t required in order to play the game. Because it’s of an optional type, the delegate property is automatically set to an initial value of nil. Thereafter, the game instantiator has the option to set the property to a suitable delegate. Because the DiceGameDelegate protocol is class-only, you can declare the delegate to be weak to prevent reference cycles.

DiceGameDelegate provides three methods for tracking the progress of a game. These three methods have been incorporated into the game logic within the play() method above, and are called when a new game starts, a new turn begins, or the game ends.

Because the delegate property is an optional DiceGameDelegate, the play() method uses optional chaining each time it calls a method on the delegate. If the delegate property is nil, these delegate calls fail gracefully and without error. If the delegate property is non-nil, the delegate methods are called, and are passed the SnakesAndLadders instance as a parameter.

This next example shows a class called DiceGameTracker, which adopts the DiceGameDelegate protocol:

  1. class DiceGameTracker: DiceGameDelegate {
  2. var numberOfTurns = 0
  3. func gameDidStart(_ game: DiceGame) {
  4. numberOfTurns = 0
  5. if game is SnakesAndLadders {
  6. print("Started a new game of Snakes and Ladders")
  7. }
  8. print("The game is using a \(game.dice.sides)-sided dice")
  9. }
  10. func game(_ game: DiceGame, didStartNewTurnWithDiceRoll diceRoll: Int) {
  11. numberOfTurns += 1
  12. print("Rolled a \(diceRoll)")
  13. }
  14. func gameDidEnd(_ game: DiceGame) {
  15. print("The game lasted for \(numberOfTurns) turns")
  16. }
  17. }

DiceGameTracker implements all three methods required by DiceGameDelegate. It uses these methods to keep track of the number of turns a game has taken. It resets a numberOfTurns property to zero when the game starts, increments it each time a new turn begins, and prints out the total number of turns once the game has ended.

The implementation of gameDidStart(_:) shown above uses the game parameter to print some introductory information about the game that is about to be played. The game parameter has a type of DiceGame, not SnakesAndLadders, and so gameDidStart(_:) can access and use only methods and properties that are implemented as part of the DiceGame protocol. However, the method is still able to use type casting to query the type of the underlying instance. In this example, it checks whether game is actually an instance of SnakesAndLadders behind the scenes, and prints an appropriate message if so.

The gameDidStart(_:) method also accesses the dice property of the passed game parameter. Because game is known to conform to the DiceGame protocol, it’s guaranteed to have a dice property, and so the gameDidStart(_:) method is able to access and print the dice’s sides property, regardless of what kind of game is being played.

Here’s how DiceGameTracker looks in action:

  1. let tracker = DiceGameTracker()
  2. let game = SnakesAndLadders()
  3. game.delegate = tracker
  4. game.play()
  5. // Started a new game of Snakes and Ladders
  6. // The game is using a 6-sided dice
  7. // Rolled a 3
  8. // Rolled a 5
  9. // Rolled a 4
  10. // Rolled a 5
  11. // The game lasted for 4 turns

Adding Protocol Conformance with an Extension

You can extend an existing type to adopt and conform to a new protocol, even if you don’t have access to the source code for the existing type. Extensions can add new properties, methods, and subscripts to an existing type, and are therefore able to add any requirements that a protocol may demand. For more about extensions, see Extensions.

Note

Existing instances of a type automatically adopt and conform to a protocol when that conformance is added to the instance’s type in an extension.

For example, this protocol, called TextRepresentable, can be implemented by any type that has a way to be represented as text. This might be a description of itself, or a text version of its current state:

  1. protocol TextRepresentable {
  2. var textualDescription: String { get }
  3. }

The Dice class from above can be extended to adopt and conform to TextRepresentable:

  1. extension Dice: TextRepresentable {
  2. var textualDescription: String {
  3. return "A \(sides)-sided dice"
  4. }
  5. }

This extension adopts the new protocol in exactly the same way as if Dice had provided it in its original implementation. The protocol name is provided after the type name, separated by a colon, and an implementation of all requirements of the protocol is provided within the extension’s curly braces.

Any Dice instance can now be treated as TextRepresentable:

  1. let d12 = Dice(sides: 12, generator: LinearCongruentialGenerator())
  2. print(d12.textualDescription)
  3. // Prints "A 12-sided dice"

Similarly, the SnakesAndLadders game class can be extended to adopt and conform to the TextRepresentable protocol:

  1. extension SnakesAndLadders: TextRepresentable {
  2. var textualDescription: String {
  3. return "A game of Snakes and Ladders with \(finalSquare) squares"
  4. }
  5. }
  6. print(game.textualDescription)
  7. // Prints "A game of Snakes and Ladders with 25 squares"

Conditionally Conforming to a Protocol

A generic type may be able to satisfy the requirements of a protocol only under certain conditions, such as when the type’s generic parameter conforms to the protocol. You can make a generic type conditionally conform to a protocol by listing constraints when extending the type. Write these constraints after the name of the protocol you’re adopting by writing a generic where clause. For more about generic where clauses, see Generic Where Clauses.

The following extension makes Array instances conform to the TextRepresentable protocol whenever they store elements of a type that conforms to TextRepresentable.

  1. extension Array: TextRepresentable where Element: TextRepresentable {
  2. var textualDescription: String {
  3. let itemsAsText = self.map { $0.textualDescription }
  4. return "[" + itemsAsText.joined(separator: ", ") + "]"
  5. }
  6. }
  7. let myDice = [d6, d12]
  8. print(myDice.textualDescription)
  9. // Prints "[A 6-sided dice, A 12-sided dice]"

Declaring Protocol Adoption with an Extension

If a type already conforms to all of the requirements of a protocol, but has not yet stated that it adopts that protocol, you can make it adopt the protocol with an empty extension:

  1. struct Hamster {
  2. var name: String
  3. var textualDescription: String {
  4. return "A hamster named \(name)"
  5. }
  6. }
  7. extension Hamster: TextRepresentable {}

Instances of Hamster can now be used wherever TextRepresentable is the required type:

  1. let simonTheHamster = Hamster(name: "Simon")
  2. let somethingTextRepresentable: TextRepresentable = simonTheHamster
  3. print(somethingTextRepresentable.textualDescription)
  4. // Prints "A hamster named Simon"

Note

Types don’t automatically adopt a protocol just by satisfying its requirements. They must always explicitly declare their adoption of the protocol.

Adopting a Protocol Using a Synthesized Implementation

Swift can automatically provide the protocol conformance for Equatable, Hashable, and Comparable in many simple cases. Using this synthesized implementation means you don’t have to write repetitive boilerplate code to implement the protocol requirements yourself.

Swift provides a synthesized implementation of Equatable for the following kinds of custom types:

  • Structures that have only stored properties that conform to the Equatable protocol

  • Enumerations that have only associated types that conform to the Equatable protocol

  • Enumerations that have no associated types

To receive a synthesized implementation of ==, declare conformance to Equatable in the file that contains the original declaration, without implementing an == operator yourself. The Equatable protocol provides a default implementation of !=.

The example below defines a Vector3D structure for a three-dimensional position vector (x, y, z), similar to the Vector2D structure. Because the x, y, and z properties are all of an Equatable type, Vector3D receives synthesized implementations of the equivalence operators.

  1. struct Vector3D: Equatable {
  2. var x = 0.0, y = 0.0, z = 0.0
  3. }
  4. let twoThreeFour = Vector3D(x: 2.0, y: 3.0, z: 4.0)
  5. let anotherTwoThreeFour = Vector3D(x: 2.0, y: 3.0, z: 4.0)
  6. if twoThreeFour == anotherTwoThreeFour {
  7. print("These two vectors are also equivalent.")
  8. }
  9. // Prints "These two vectors are also equivalent."

Swift provides a synthesized implementation of Hashable for the following kinds of custom types:

  • Structures that have only stored properties that conform to the Hashable protocol

  • Enumerations that have only associated types that conform to the Hashable protocol

  • Enumerations that have no associated types

To receive a synthesized implementation of hash(into:), declare conformance to Hashable in the file that contains the original declaration, without implementing a hash(into:) method yourself.

Swift provides a synthesized implementation of Comparable for enumerations that don’t have a raw value. If the enumeration has associated types, they must all conform to the Comparable protocol. To receive a synthesized implementation of <, declare conformance to Comparable in the file that contains the original enumeration declaration, without implementing a < operator yourself. The Comparable protocol’s default implementation of <=, >, and >= provides the remaining comparison operators.

The example below defines a SkillLevel enumeration with cases for beginners, intermediates, and experts. Experts are additionally ranked by the number of stars they have.

  1. enum SkillLevel: Comparable {
  2. case beginner
  3. case intermediate
  4. case expert(stars: Int)
  5. }
  6. var levels = [SkillLevel.intermediate, SkillLevel.beginner,
  7. SkillLevel.expert(stars: 5), SkillLevel.expert(stars: 3)]
  8. for level in levels.sorted() {
  9. print(level)
  10. }
  11. // Prints "beginner"
  12. // Prints "intermediate"
  13. // Prints "expert(stars: 3)"
  14. // Prints "expert(stars: 5)"

Collections of Protocol Types

A protocol can be used as the type to be stored in a collection such as an array or a dictionary, as mentioned in Protocols as Types. This example creates an array of TextRepresentable things:

  1. let things: [TextRepresentable] = [game, d12, simonTheHamster]

It’s now possible to iterate over the items in the array, and print each item’s textual description:

  1. for thing in things {
  2. print(thing.textualDescription)
  3. }
  4. // A game of Snakes and Ladders with 25 squares
  5. // A 12-sided dice
  6. // A hamster named Simon

Note that the thing constant is of type TextRepresentable. It’s not of type Dice, or DiceGame, or Hamster, even if the actual instance behind the scenes is of one of those types. Nonetheless, because it’s of type TextRepresentable, and anything that is TextRepresentable is known to have a textualDescription property, it’s safe to access thing.textualDescription each time through the loop.

Protocol Inheritance

A protocol can inherit one or more other protocols and can add further requirements on top of the requirements it inherits. The syntax for protocol inheritance is similar to the syntax for class inheritance, but with the option to list multiple inherited protocols, separated by commas:

  1. protocol InheritingProtocol: SomeProtocol, AnotherProtocol {
  2. // protocol definition goes here
  3. }

Here’s an example of a protocol that inherits the TextRepresentable protocol from above:

  1. protocol PrettyTextRepresentable: TextRepresentable {
  2. var prettyTextualDescription: String { get }
  3. }

This example defines a new protocol, PrettyTextRepresentable, which inherits from TextRepresentable. Anything that adopts PrettyTextRepresentable must satisfy all of the requirements enforced by TextRepresentable, plus the additional requirements enforced by PrettyTextRepresentable. In this example, PrettyTextRepresentable adds a single requirement to provide a gettable property called prettyTextualDescription that returns a String.

The SnakesAndLadders class can be extended to adopt and conform to PrettyTextRepresentable:

  1. extension SnakesAndLadders: PrettyTextRepresentable {
  2. var prettyTextualDescription: String {
  3. var output = textualDescription + ":\n"
  4. for index in 1...finalSquare {
  5. switch board[index] {
  6. case let ladder where ladder > 0:
  7. output += "▲ "
  8. case let snake where snake < 0:
  9. output += "▼ "
  10. default:
  11. output += "○ "
  12. }
  13. }
  14. return output
  15. }
  16. }

This extension states that it adopts the PrettyTextRepresentable protocol and provides an implementation of the prettyTextualDescription property for the SnakesAndLadders type. Anything that is PrettyTextRepresentable must also be TextRepresentable, and so the implementation of prettyTextualDescription starts by accessing the textualDescription property from the TextRepresentable protocol to begin an output string. It appends a colon and a line break, and uses this as the start of its pretty text representation. It then iterates through the array of board squares, and appends a geometric shape to represent the contents of each square:

  • If the square’s value is greater than 0, it’s the base of a ladder, and is represented by .

  • If the square’s value is less than 0, it’s the head of a snake, and is represented by .

  • Otherwise, the square’s value is 0, and it’s a “free” square, represented by .

The prettyTextualDescription property can now be used to print a pretty text description of any SnakesAndLadders instance:

  1. print(game.prettyTextualDescription)
  2. // A game of Snakes and Ladders with 25 squares:
  3. // ○ ○ ▲ ○ ○ ▲ ○ ○ ▲ ▲ ○ ○ ○ ▼ ○ ○ ○ ○ ▼ ○ ○ ▼ ○ ▼ ○

Class-Only Protocols

You can limit protocol adoption to class types (and not structures or enumerations) by adding the AnyObject protocol to a protocol’s inheritance list.

  1. protocol SomeClassOnlyProtocol: AnyObject, SomeInheritedProtocol {
  2. // class-only protocol definition goes here
  3. }

In the example above, SomeClassOnlyProtocol can only be adopted by class types. It’s a compile-time error to write a structure or enumeration definition that tries to adopt SomeClassOnlyProtocol.

Note

Use a class-only protocol when the behavior defined by that protocol’s requirements assumes or requires that a conforming type has reference semantics rather than value semantics. For more about reference and value semantics, see Structures and Enumerations Are Value Types and Classes Are Reference Types.

Protocol Composition

It can be useful to require a type to conform to multiple protocols at the same time. You can combine multiple protocols into a single requirement with a protocol composition. Protocol compositions behave as if you defined a temporary local protocol that has the combined requirements of all protocols in the composition. Protocol compositions don’t define any new protocol types.

Protocol compositions have the form SomeProtocol & AnotherProtocol. You can list as many protocols as you need, separating them with ampersands (&). In addition to its list of protocols, a protocol composition can also contain one class type, which you can use to specify a required superclass.

Here’s an example that combines two protocols called Named and Aged into a single protocol composition requirement on a function parameter:

  1. protocol Named {
  2. var name: String { get }
  3. }
  4. protocol Aged {
  5. var age: Int { get }
  6. }
  7. struct Person: Named, Aged {
  8. var name: String
  9. var age: Int
  10. }
  11. func wishHappyBirthday(to celebrator: Named & Aged) {
  12. print("Happy birthday, \(celebrator.name), you're \(celebrator.age)!")
  13. }
  14. let birthdayPerson = Person(name: "Malcolm", age: 21)
  15. wishHappyBirthday(to: birthdayPerson)
  16. // Prints "Happy birthday, Malcolm, you're 21!"

In this example, the Named protocol has a single requirement for a gettable String property called name. The Aged protocol has a single requirement for a gettable Int property called age. Both protocols are adopted by a structure called Person.

The example also defines a wishHappyBirthday(to:) function. The type of the celebrator parameter is Named & Aged, which means “any type that conforms to both the Named and Aged protocols.” It doesn’t matter which specific type is passed to the function, as long as it conforms to both of the required protocols.

The example then creates a new Person instance called birthdayPerson and passes this new instance to the wishHappyBirthday(to:) function. Because Person conforms to both protocols, this call is valid, and the wishHappyBirthday(to:) function can print its birthday greeting.

Here’s an example that combines the Named protocol from the previous example with a Location class:

  1. class Location {
  2. var latitude: Double
  3. var longitude: Double
  4. init(latitude: Double, longitude: Double) {
  5. self.latitude = latitude
  6. self.longitude = longitude
  7. }
  8. }
  9. class City: Location, Named {
  10. var name: String
  11. init(name: String, latitude: Double, longitude: Double) {
  12. self.name = name
  13. super.init(latitude: latitude, longitude: longitude)
  14. }
  15. }
  16. func beginConcert(in location: Location & Named) {
  17. print("Hello, \(location.name)!")
  18. }
  19. let seattle = City(name: "Seattle", latitude: 47.6, longitude: -122.3)
  20. beginConcert(in: seattle)
  21. // Prints "Hello, Seattle!"

The beginConcert(in:) function takes a parameter of type Location & Named, which means “any type that’s a subclass of Location and that conforms to the Named protocol.” In this case, City satisfies both requirements.

Passing birthdayPerson to the beginConcert(in:) function is invalid because Person isn’t a subclass of Location. Likewise, if you made a subclass of Location that didn’t conform to the Named protocol, calling beginConcert(in:) with an instance of that type is also invalid.

Checking for Protocol Conformance

You can use the is and as operators described in Type Casting to check for protocol conformance, and to cast to a specific protocol. Checking for and casting to a protocol follows exactly the same syntax as checking for and casting to a type:

  • The is operator returns true if an instance conforms to a protocol and returns false if it doesn’t.

  • The as? version of the downcast operator returns an optional value of the protocol’s type, and this value is nil if the instance doesn’t conform to that protocol.

  • The as! version of the downcast operator forces the downcast to the protocol type and triggers a runtime error if the downcast doesn’t succeed.

This example defines a protocol called HasArea, with a single property requirement of a gettable Double property called area:

  1. protocol HasArea {
  2. var area: Double { get }
  3. }

Here are two classes, Circle and Country, both of which conform to the HasArea protocol:

  1. class Circle: HasArea {
  2. let pi = 3.1415927
  3. var radius: Double
  4. var area: Double { return pi * radius * radius }
  5. init(radius: Double) { self.radius = radius }
  6. }
  7. class Country: HasArea {
  8. var area: Double
  9. init(area: Double) { self.area = area }
  10. }

The Circle class implements the area property requirement as a computed property, based on a stored radius property. The Country class implements the area requirement directly as a stored property. Both classes correctly conform to the HasArea protocol.

Here’s a class called Animal, which doesn’t conform to the HasArea protocol:

  1. class Animal {
  2. var legs: Int
  3. init(legs: Int) { self.legs = legs }
  4. }

The Circle, Country and Animal classes don’t have a shared base class. Nonetheless, they’re all classes, and so instances of all three types can be used to initialize an array that stores values of type AnyObject:

  1. let objects: [AnyObject] = [
  2. Circle(radius: 2.0),
  3. Country(area: 243_610),
  4. Animal(legs: 4)
  5. ]

The objects array is initialized with an array literal containing a Circle instance with a radius of 2 units; a Country instance initialized with the surface area of the United Kingdom in square kilometers; and an Animal instance with four legs.

The objects array can now be iterated, and each object in the array can be checked to see if it conforms to the HasArea protocol:

  1. for object in objects {
  2. if let objectWithArea = object as? HasArea {
  3. print("Area is \(objectWithArea.area)")
  4. } else {
  5. print("Something that doesn't have an area")
  6. }
  7. }
  8. // Area is 12.5663708
  9. // Area is 243610.0
  10. // Something that doesn't have an area

Whenever an object in the array conforms to the HasArea protocol, the optional value returned by the as? operator is unwrapped with optional binding into a constant called objectWithArea. The objectWithArea constant is known to be of type HasArea, and so its area property can be accessed and printed in a type-safe way.

Note that the underlying objects aren’t changed by the casting process. They continue to be a Circle, a Country and an Animal. However, at the point that they’re stored in the objectWithArea constant, they’re only known to be of type HasArea, and so only their area property can be accessed.

Optional Protocol Requirements

You can define optional requirements for protocols. These requirements don’t have to be implemented by types that conform to the protocol. Optional requirements are prefixed by the optional modifier as part of the protocol’s definition. Optional requirements are available so that you can write code that interoperates with Objective-C. Both the protocol and the optional requirement must be marked with the @objc attribute. Note that @objc protocols can be adopted only by classes that inherit from Objective-C classes or other @objc classes. They can’t be adopted by structures or enumerations.

When you use a method or property in an optional requirement, its type automatically becomes an optional. For example, a method of type (Int) -> String becomes ((Int) -> String)?. Note that the entire function type is wrapped in the optional, not the method’s return value.

An optional protocol requirement can be called with optional chaining, to account for the possibility that the requirement was not implemented by a type that conforms to the protocol. You check for an implementation of an optional method by writing a question mark after the name of the method when it’s called, such as someOptionalMethod?(someArgument). For information on optional chaining, see Optional Chaining.

The following example defines an integer-counting class called Counter, which uses an external data source to provide its increment amount. This data source is defined by the CounterDataSource protocol, which has two optional requirements:

  1. @objc protocol CounterDataSource {
  2. @objc optional func increment(forCount count: Int) -> Int
  3. @objc optional var fixedIncrement: Int { get }
  4. }

The CounterDataSource protocol defines an optional method requirement called increment(forCount:) and an optional property requirement called fixedIncrement. These requirements define two different ways for data sources to provide an appropriate increment amount for a Counter instance.

Note

Strictly speaking, you can write a custom class that conforms to CounterDataSource without implementing either protocol requirement. They’re both optional, after all. Although technically allowed, this wouldn’t make for a very good data source.

The Counter class, defined below, has an optional dataSource property of type CounterDataSource?:

  1. class Counter {
  2. var count = 0
  3. var dataSource: CounterDataSource?
  4. func increment() {
  5. if let amount = dataSource?.increment?(forCount: count) {
  6. count += amount
  7. } else if let amount = dataSource?.fixedIncrement {
  8. count += amount
  9. }
  10. }
  11. }

The Counter class stores its current value in a variable property called count. The Counter class also defines a method called increment, which increments the count property every time the method is called.

The increment() method first tries to retrieve an increment amount by looking for an implementation of the increment(forCount:) method on its data source. The increment() method uses optional chaining to try to call increment(forCount:), and passes the current count value as the method’s single argument.

Note that two levels of optional chaining are at play here. First, it’s possible that dataSource may be nil, and so dataSource has a question mark after its name to indicate that increment(forCount:) should be called only if dataSource isn’t nil. Second, even if dataSource does exist, there’s no guarantee that it implements increment(forCount:), because it’s an optional requirement. Here, the possibility that increment(forCount:) might not be implemented is also handled by optional chaining. The call to increment(forCount:) happens only if increment(forCount:) exists—that is, if it isn’t nil. This is why increment(forCount:) is also written with a question mark after its name.

Because the call to increment(forCount:) can fail for either of these two reasons, the call returns an optional Int value. This is true even though increment(forCount:) is defined as returning a non-optional Int value in the definition of CounterDataSource. Even though there are two optional chaining operations, one after another, the result is still wrapped in a single optional. For more information about using multiple optional chaining operations, see Linking Multiple Levels of Chaining.

After calling increment(forCount:), the optional Int that it returns is unwrapped into a constant called amount, using optional binding. If the optional Int does contain a value—that is, if the delegate and method both exist, and the method returned a value—the unwrapped amount is added onto the stored count property, and incrementation is complete.

If it’s not possible to retrieve a value from the increment(forCount:) method—either because dataSource is nil, or because the data source doesn’t implement increment(forCount:)—then the increment() method tries to retrieve a value from the data source’s fixedIncrement property instead. The fixedIncrement property is also an optional requirement, so its value is an optional Int value, even though fixedIncrement is defined as a non-optional Int property as part of the CounterDataSource protocol definition.

Here’s a simple CounterDataSource implementation where the data source returns a constant value of 3 every time it’s queried. It does this by implementing the optional fixedIncrement property requirement:

  1. class ThreeSource: NSObject, CounterDataSource {
  2. let fixedIncrement = 3
  3. }

You can use an instance of ThreeSource as the data source for a new Counter instance:

  1. var counter = Counter()
  2. counter.dataSource = ThreeSource()
  3. for _ in 1...4 {
  4. counter.increment()
  5. print(counter.count)
  6. }
  7. // 3
  8. // 6
  9. // 9
  10. // 12

The code above creates a new Counter instance; sets its data source to be a new ThreeSource instance; and calls the counter’s increment() method four times. As expected, the counter’s count property increases by three each time increment() is called.

Here’s a more complex data source called TowardsZeroSource, which makes a Counter instance count up or down towards zero from its current count value:

  1. class TowardsZeroSource: NSObject, CounterDataSource {
  2. func increment(forCount count: Int) -> Int {
  3. if count == 0 {
  4. return 0
  5. } else if count < 0 {
  6. return 1
  7. } else {
  8. return -1
  9. }
  10. }
  11. }

The TowardsZeroSource class implements the optional increment(forCount:) method from the CounterDataSource protocol and uses the count argument value to work out which direction to count in. If count is already zero, the method returns 0 to indicate that no further counting should take place.

You can use an instance of TowardsZeroSource with the existing Counter instance to count from -4 to zero. Once the counter reaches zero, no more counting takes place:

  1. counter.count = -4
  2. counter.dataSource = TowardsZeroSource()
  3. for _ in 1...5 {
  4. counter.increment()
  5. print(counter.count)
  6. }
  7. // -3
  8. // -2
  9. // -1
  10. // 0
  11. // 0

Protocol Extensions

Protocols can be extended to provide method, initializer, subscript, and computed property implementations to conforming types. This allows you to define behavior on protocols themselves, rather than in each type’s individual conformance or in a global function.

For example, the RandomNumberGenerator protocol can be extended to provide a randomBool() method, which uses the result of the required random() method to return a random Bool value:

  1. extension RandomNumberGenerator {
  2. func randomBool() -> Bool {
  3. return random() > 0.5
  4. }
  5. }

By creating an extension on the protocol, all conforming types automatically gain this method implementation without any additional modification.

  1. let generator = LinearCongruentialGenerator()
  2. print("Here's a random number: \(generator.random())")
  3. // Prints "Here's a random number: 0.3746499199817101"
  4. print("And here's a random Boolean: \(generator.randomBool())")
  5. // Prints "And here's a random Boolean: true"

Protocol extensions can add implementations to conforming types but can’t make a protocol extend or inherit from another protocol. Protocol inheritance is always specified in the protocol declaration itself.

Providing Default Implementations

You can use protocol extensions to provide a default implementation to any method or computed property requirement of that protocol. If a conforming type provides its own implementation of a required method or property, that implementation will be used instead of the one provided by the extension.

Note

Protocol requirements with default implementations provided by extensions are distinct from optional protocol requirements. Although conforming types don’t have to provide their own implementation of either, requirements with default implementations can be called without optional chaining.

For example, the PrettyTextRepresentable protocol, which inherits the TextRepresentable protocol can provide a default implementation of its required prettyTextualDescription property to simply return the result of accessing the textualDescription property:

  1. extension PrettyTextRepresentable {
  2. var prettyTextualDescription: String {
  3. return textualDescription
  4. }
  5. }

Adding Constraints to Protocol Extensions

When you define a protocol extension, you can specify constraints that conforming types must satisfy before the methods and properties of the extension are available. You write these constraints after the name of the protocol you’re extending by writing a generic where clause. For more about generic where clauses, see Generic Where Clauses.

For example, you can define an extension to the Collection protocol that applies to any collection whose elements conform to the Equatable protocol. By constraining a collection’s elements to the Equatable protocol, a part of the standard library, you can use the == and != operators to check for equality and inequality between two elements.

  1. extension Collection where Element: Equatable {
  2. func allEqual() -> Bool {
  3. for element in self {
  4. if element != self.first {
  5. return false
  6. }
  7. }
  8. return true
  9. }
  10. }

The allEqual() method returns true only if all the elements in the collection are equal.

Consider two arrays of integers, one where all the elements are the same, and one where they aren’t:

  1. let equalNumbers = [100, 100, 100, 100, 100]
  2. let differentNumbers = [100, 100, 200, 100, 200]

Because arrays conform to Collection and integers conform to Equatable, equalNumbers and differentNumbers can use the allEqual() method:

  1. print(equalNumbers.allEqual())
  2. // Prints "true"
  3. print(differentNumbers.allEqual())
  4. // Prints "false"

Note

If a conforming type satisfies the requirements for multiple constrained extensions that provide implementations for the same method or property, Swift uses the implementation corresponding to the most specialized constraints.