This is iflix's style guide for the Swift programming language. Feel free to fork it, propose changes, but most importantly read it!
The document uses the raywenderlich.com Swift styleguide as a template, for its thoroughness.
- Correctness
- Naming
- Code Organization
- Spacing
- Comments
- Classes and Structures
- Function Declarations
- Closure Expressions
- Types
- Functions vs Methods
- Memory Management
- Access Control
- Control Flow
- Golden Path
- Semicolons
- Parentheses
- Use of
self - Images
Consider warnings to be errors. This rule informs many stylistic decisions such as C-style for loops, or strings as selectors.
Use descriptive names with camel case for classes, methods, variables, etc. Type names (classes, structures, enumerations and protocols) should be capitalized, while method names and variables should start with a lower case letter.
Also see the "Naming" section in the Swift 3 API guidelines.
Preferred:
private let maximumWidgetCount = 100
class WidgetContainer {
var widgetButton: UIButton
let widgetHeightPercentage = 0.85
}Not Preferred:
let MAX_WIDGET_COUNT = 100
class app_widgetContainer {
var wBut: UIButton
let wHeightPct = 0.85
}Abbreviations and acronyms should generally be avoided. Following the API Design Guidelines, abbreviations and initialisms that appear in all uppercase should be uniformly uppercase or lowercase. Examples:
Preferred
let urlString: URLString
let userID: UserIDNot Preferred
let uRLString: UrlString
let userId: UserIdFor functions and init methods, prefer named parameters for all arguments unless the context is very clear. Include external parameter names if it makes function calls more readable.
func dateFromString(dateString: String) -> NSDate
func convertPointAt(column: Int, row: Int) -> CGPoint
func timedAction(afterDelay delay: NSTimeInterval, perform action: SKAction) -> SKAction!
// would be called like this:
dateFromString("2014-03-14")
convertPointAt(column: 42, row: 13)
timedAction(afterDelay: 1.0, perform: someOtherAction)For methods, follow the standard Apple convention of referring to the first parameter in the method name:
class Counter {
func combineWith(otherCounter: Counter, options: Dictionary?) { ... }
func incrementBy(amount: Int) { ... }
}Following Apple's API Design Guidelines, protocols names that describe what something is should be a noun. Examples: Collection, WidgetFactory. Protocols names that describe an ability should end in -ing, -able, or -ible. Examples: Equatable, Resizing.
Following Apple's API Design Guidelines for Swift 3, use lowerCamelCase for enumeration values.
enum Shape {
case rectangle
case square
case rightTriangle
case equilateralTriangle
}When referring to functions in prose (documentation, code comments, code reviews) include the required parameter names from the caller's perspective or _ for unnamed parameters. Examples:
Call
convertPointAt(column:row:)from your owninitimplementation.If you call
dateFromString(_:)make sure that you provide a string with the format "yyyy-MM-dd".If you call
timedAction(afterDelay:perform:)fromviewDidLoad()remember to provide an adjusted delay value and an action to perform.You shouldn't call the data source method
tableView(_:cellForRowAtIndexPath:)directly.
This is the same as the #selector syntax. When in doubt, look at how Xcode lists the method in the jump bar – our style here matches that.
Swift types are automatically namespaced by the module that contains them and you should not add a class prefix such as ABC. If two names from different modules collide you can disambiguate by prefixing the type name with the module name. However, only specify the module name when there is possibility for confusion which should be rare.
import SomeModule
let myClass = MyModule.UsefulClass()Selectors are Obj-C methods that act as handlers for many Cocoa and Cocoa Touch APIs. Prior to Swift 2.2, they were specified using type unsafe strings. This now causes a compiler warning. The "Fix it" button replaces these strings with the fully qualified type safe selector. Often, however, you can use context to shorten the expression. This is the preferred style.
Preferred:
let selector = #selector(viewDidLoad)Not Preferred:
let sel = #selector(ViewController.viewDidLoad)Generic type parameters should be descriptive, upper camel case names. When a type name doesn't have a meaningful relationship or role, use a traditional single uppercase letter such as T, U, or V.
Preferred:
struct Stack<Element> { ... }
func writeTo<Target: OutputStream>(inout target: Target)
func max<T: Comparable>(x: T, _ y: T) -> TNot Preferred:
struct Stack<T> { ... }
func writeTo<target: OutputStream>(inout t: target)
func max<Thing: Comparable>(x: Thing, _ y: Thing) -> ThingWhen moving extensions in a dedicated file name it ExtendedType+NewFunctionality.swift.
Preferred:
// String+Awesome.swift
extension String {
func makeAwesome() -> String { return "AWESOME" }
}Not Preferred:
// AwesomeString.swift
extension String {
func makeAwesome() -> String { return "AWESOME" }
}Use US English spelling to match Apple's API.
Preferred:
let color = "red"Not Preferred:
let colour = "red"When using abbreviations and initialisms, conform to the Swift API Design Guidelines and Apple's API when naming variables and types.
Preferred:
var isRepresentableAsASCII = true
var radarDetector: RadarScanner
var uniqueID: UUIDNot Preferred:
var isRepresentableAsAscii = true
var radarDetector: RADARScanner
var uniqueId: UUIDUse extensions to organize your code into logical blocks of functionality. Each extension not in a dedicated file should be set off with a // MARK: - comment to keep things well-organized. Refer to the extensions naming guide to know how to name files containing a single extension.
In particular, when adding protocol conformance to a model, prefer adding a separate extension for the protocol methods. This keeps the related methods grouped together with the protocol and can simplify instructions to add a protocol to a class with its associated methods.
Preferred:
class MyViewcontroller: UIViewController {
// class stuff here
}
// MARK: - UITableViewDataSource
extension MyViewcontroller: UITableViewDataSource {
// table view data source methods
}
// MARK: - UIScrollViewDelegate
extension MyViewcontroller: UIScrollViewDelegate {
// scroll view delegate methods
}Not Preferred:
class MyViewcontroller: UIViewController, UITableViewDataSource, UIScrollViewDelegate {
// all methods
}Since the compiler does not allow you to re-declare protocol conformance in a derived class, it is not always required to replicate the extension groups of the base class. This is especially true if the derived class is a terminal class and a small number of methods are being overriden. When to preserve the extension groups is left to the discretion of the author.
For UIKit view controllers, consider grouping lifecycle, custom accessors, and IBAction in separate class extensions.
Unused (dead) code, including Xcode template code and placeholder comments should be removed.
Methods whose implementation simply calls the super class should also be removed. This includes any empty/unused UIApplicationDelegate methods.
Not Preferred:
override func didReceiveMemoryWarning() {
super.didReceiveMemoryWarning()
// Dispose of any resources that can be recreated.
}
override func numberOfSectionsInTableView(tableView: UITableView) -> Int {
// #warning Incomplete implementation, return the number of sections
return 1
}
override func tableView(tableView: UITableView, numberOfRowsInSection section: Int) -> Int {
// #warning Incomplete implementation, return the number of rows
return Database.contacts.count
}Preferred:
override func tableView(tableView: UITableView, numberOfRowsInSection section: Int) -> Int {
return Database.contacts.count
}Also do not include Xcode generated comments header, that information can be taken out of git, and we don't have copyright concerns because the code is not open source.
Keep imports minimal. For example, don't import UIKit when importing Foundation will suffice.
-
Indent using 4 spaces rather than tabs to conserve space and help prevent line wrapping. Be sure to set this preference in Xcode and in the Project settings as shown below:
-
Method braces and other braces (
if/else/switch/whileetc.) always open on the same line as the statement but close on a new line. -
Except for simple one liner checks like
guard let strongSelf = self else { return }.
Preferred:
if user.isHappy {
// Do something
} else {
// Do something else
}Not Preferred:
if user.isHappy
{
// Do something
}
else {
// Do something else
}-
There should be exactly one blank line between methods to aid in visual clarity and organization. Whitespace within methods should separate functionality, but having too many sections in a method often means you should refactor into several methods.
-
Colons always have no space on the left and one space on the right. Exceptions are the ternary operator
? :and empty dictionary[:].
Preferred:
class TestDatabase: Database {
var data: [String: CGFloat] = ["A": 1.2, "B": 3.2]
}Not Preferred:
class TestDatabase : Database {
var data :[String:CGFloat] = ["A" : 1.2, "B":3.2]
}When they are needed, use comments to explain why a particular piece of code does something. Comments must be kept up-to-date or deleted.
Avoid block comments inline with code, as the code should be as self-documenting as possible. Exception: This does not apply to those comments used to generate documentation.
Remember, structs have value semantics. Use structs for things that do not have an identity. An array that contains [a, b, c] is really the same as another array that contains [a, b, c] and they are completely interchangeable. It doesn't matter whether you use the first array or the second, because they represent the exact same thing. That's why arrays are structs.
Classes have reference semantics. Use classes for things that do have an identity or a specific life cycle. You would model a person as a class because two person objects are two different things. Just because two people have the same name and birthdate, doesn't mean they are the same person. But the person's birthdate would be a struct because a date of 3 March 1950 is the same as any other date object for 3 March 1950. The date itself doesn't have an identity.
Sometimes, things should be structs but need to conform to AnyObject or are historically modeled as classes already (Date, Set). Try to follow these guidelines as closely as possible.
Here's an example of a well-styled class definition:
class Circle: Shape {
var x: Int, y: Int
var radius: Double
var diameter: Double {
get {
return radius * 2
}
set {
radius = newValue / 2
}
}
init(x: Int, y: Int, radius: Double) {
self.x = x
self.y = y
self.radius = radius
}
convenience init(x: Int, y: Int, diameter: Double) {
self.init(x: x, y: y, radius: diameter / 2)
}
func describe() -> String {
return "I am a circle at \(centerString()) with an area of \(computeArea())"
}
override func computeArea() -> Double {
return M_PI * radius * radius
}
private func centerString() -> String {
return "(\(x),\(y))"
}
}The example above demonstrates the following style guidelines:
- Specify types for properties, variables, constants, argument declarations and other statements with a space after the colon but not before, e.g.
x: Int, andCircle: Shape. - Define multiple variables and structures on a single line if they share a common purpose / context.
- Indent getter and setter definitions and property observers.
- Don't add modifiers such as
internalwhen they're already the default. Similarly, don't repeat the access modifier when overriding a method.
For conciseness, avoid using self since Swift does not require it to access an object's properties or invoke its methods.
Use self when required to differentiate between property names and arguments in initializers, and when referencing properties in closure expressions (as required by the compiler):
class BoardLocation {
let row: Int, column: Int
init(row: Int, column: Int) {
self.row = row
self.column = column
let closure = {
print(self.row)
}
}
}For conciseness, if a computed property is read-only, omit the get clause. The get clause is required only when a set clause is provided.
Preferred:
var diameter: Double {
return radius * 2
}Not Preferred:
var diameter: Double {
get {
return radius * 2
}
}Mark classes final only when inheritance should specifically not be allowed. In previous versions of Xcode, the standard compiler release mode did not perform Whole Module Optimization, and the final keyword added performance optimiziations by avoiding dynamic dispatch. Whole Module Optimization is now the default in Xcode, meaning the compiler automatically infers final for all internal, fileprivate and private classes. This means we can remove this boilerplate, and focus using this keyword on the cases where subclassing should be deliberately disabled.
When marking a class as final, you should also document why it was finalized. Example:
/// This class is marked as final because it's super unsafe to subclass it due to...
final class SuperUnsafeBox<T> {
let value: T
init(_ value: T) {
self.value = value
}
}Keep short function declarations on one line including the opening brace:
func reticulateSplines(spline: [Double]) -> Bool {
// reticulate code goes here
}If a function has declaration longer than the recommended line length (100) then put each parameter on a new line:
func reticulateSplines(
spline: [Double],
adjustmentFactor: Double,
translateConstant: Int,
comment: String
) -> Bool {Be sure to set this preference in Xcode and in the Project settings as shown below:
The rationale behind this convention is called zero one inifinity rule.
Use trailing closure syntax only if there's a single closure expression parameter at the end of the argument list. Give the closure parameters descriptive names.
Preferred:
UIView.animateWithDuration(1.0) {
self.myView.alpha = 0
}
UIView.animateWithDuration(1.0,
animations: {
self.myView.alpha = 0
},
completion: { finished in
self.myView.removeFromSuperview()
}
)Not Preferred:
UIView.animateWithDuration(1.0, animations: {
self.myView.alpha = 0
})
UIView.animateWithDuration(1.0,
animations: {
self.myView.alpha = 0
}) { f in
self.myView.removeFromSuperview()
}For single-expression closures where the context is clear, use implicit returns:
attendeeList.sort { a, b in
a > b
}Chained methods using trailing closures should be clear and easy to read in context. Decisions on spacing, line breaks, and when to use named versus anonymous arguments is left to the discretion of the author. Examples:
let value = numbers.map { $0 * 2 }.filter { $0 % 3 == 0 }.indexOf(90)
let value = numbers
.map {$0 * 2}
.filter {$0 > 50}
.map {$0 + 10}Always use Swift's native types when available. Swift offers bridging to Objective-C so you can still use the full set of methods as needed.
Preferred:
let width = 120.0 // Double
let widthString = (width as Number).stringValue // StringNot Preferred:
let width: Number = 120.0 // Number
let widthString: String = width.stringValue // StringIn Sprite Kit, Core Animation, Core Graphics code, use CGFloat if it makes the code more succinct by avoiding too many conversions.
Constants are defined using the let keyword, and variables with the var keyword. Always use let instead of var if the value of the variable will not change.
Tip: A good technique is to define everything using let and only change it to var if the compiler complains!
You can define constants on a type rather than an instance of that type using type properties. To declare a type property as a constant simply use static let. Type properties declared in this way are generally preferred over global constants because they are easier to distinguish from instance properties. Example:
Preferred:
enum Math {
static let e = 2.718281828459045235360287
static let pi = 3.141592653589793238462643
}
radius * Math.pi * 2 // circumferenceNote: The advantage of using a case-less enumeration is that it can't accidentally be instantiated and works as a pure namespace.
Not Preferred:
let e = 2.718281828459045235360287 // pollutes global namespace
let pi = 3.141592653589793238462643
radius * pi * 2 // is pi instance data or a global constant?Static methods and type properties work similarly to global functions and global variables and should be used sparingly. They are useful when functionality is scoped to a particular type or when Objective-C interoperability is required.
Declare variables and function return types as optional with ? where a nil value is acceptable.
Use implicitly unwrapped types declared with ! only for IBOutlet view properties.
In most cases don't force unwrap.
The only exception is when the occurence of a nil value would be due to a developer error. For example if we create a URL from a static string, the initializer is failable, but it would only result in a nil value if there was a typo in the string.
Therefore it's ok to force-unwrap the result in this case and crash immediately if we've made a mistake:
let url: URL = URL(string: "https://www.iflix.com")!In such cases, you can read the ! character as an assertion. For more context, read this article
Another example is when the compiler throws a warning regarding the coercion of IBOutlets in LayoutConstraints.
In this case force unwrap is accepted too.
See https://github.com/iflix/apple-ios/issues/4911.
In all other cases, when accessing an optional value, use optional chaining if the value is only accessed once or if there are many optionals in the chain:
self.textContainer?.textLabel?.setNeedsDisplay()Use optional binding when it's more convenient to unwrap once and perform multiple operations:
if let textContainer = self.textContainer {
// do many things with textContainer
}When naming optional variables and properties, avoid naming them like optionalString or maybeView since their optional-ness is already in the type declaration.
For optional binding, shadow the original name when appropriate rather than using names like unwrappedView or actualLabel.
Preferred:
var subview: UIView?
var volume: Double?
// later on...
if let subview = subview, volume = volume {
// do something with unwrapped subview and volume
}Not Preferred:
var optionalSubview: UIView?
var volume: Double?
if let unwrappedSubview = optionalSubview {
if let realVolume = volume {
// do something with unwrappedSubview and realVolume
}
}Use the native Swift struct initializers rather than the legacy CGGeometry constructors.
Preferred:
let bounds = CGRect(x: 40, y: 20, width: 120, height: 80)
let centerPoint = CGPoint(x: 96, y: 42)Not Preferred:
let bounds = CGRectMake(40, 20, 120, 80)
let centerPoint = CGPointMake(96, 42)Prefer the struct-scope constants CGRect.infinite, CGRect.null, etc. over global constants CGRectInfinite, CGRectNull, etc. For existing variables, you can use the shorter .zero.
Consider using lazy initialization for finer grain control over object lifetime. This is especially true for UIViewController that loads views lazily. You can either use a closure that is immediately called { }() or call a private factory method. Example:
lazy var locationManager: CLLocationManager = self.makeLocationManager()
private func makeLocationManager() -> CLLocationManager {
let manager = CLLocationManager()
manager.desiredAccuracy = kCLLocationAccuracyBest
manager.delegate = self
manager.requestAlwaysAuthorization()
return manager
}Notes:
[unowned self]is not required here. A retain cycle is not created.- Location manager has a side-effect for popping up UI to ask the user for permission so fine grain control makes sense here.
Prefer compact code and let the compiler infer the type for constants or variables of single instances. Type inference is also appropriate for small (non-empty) arrays and dictionaries. When required, specify the specific type such as CGFloat or Int16.
Preferred:
let message = "Click the button"
let currentBounds = computeViewBounds()
var names = ["Mic", "Sam", "Christine"]
let maximumWidth: CGFloat = 106.5Not Preferred:
let message: String = "Click the button"
let currentBounds: CGRect = computeViewBounds()
let names = [String]()For empty arrays and dictionaries, use type annotation. (For an array or dictionary assigned to a large, multi-line literal, use type annotation.)
Preferred:
var names: [String] = []
var lookup: [String: Int] = [:]Not Preferred:
var names = [String]()
var lookup = [String: Int]()NOTE: Following this guideline means picking descriptive names is even more important than before.
Prefer the shortcut versions of type declarations over the full generics syntax.
Preferred:
var deviceModels: [String]
var employees: [Int: String]
var faxNumber: Int?Not Preferred:
var deviceModels: Array<String>
var employees: Dictionary<Int, String>
var faxNumber: Optional<Int>Free functions, which aren't attached to a class or type, should be used sparingly. When possible, prefer to use a method instead of a free function. This aids in readability and discoverability.
Free functions are most appropriate when they aren't associated with any particular type or instance.
Preferred
let sorted = items.mergeSort() // easily discoverable
rocket.launch() // clearly acts on the modelNot Preferred
let sorted = mergeSort(items) // hard to discover
launch(&rocket)Free Function Exceptions
let tuples = zip(a, b) // feels natural as a free function (symmetry)
let value = max(x,y,z) // another free function that feels naturalCode (even non-production, tutorial demo code) should not create reference cycles. Analyze your object graph and prevent strong cycles with weak and unowned references. Alternatively, use value types (struct, enum) to prevent cycles altogether.
Extend object lifetime using the [weak self] and guard let strongSelf = self else { return } idiom. [weak self] is preferred to [unowned self] where it is not immediately obvious that self outlives the closure. Explicitly extending lifetime is preferred to optional unwrapping.
Preferred
resource.request().onComplete { [weak self] response in
guard let strongSelf = self else { return }
let model = strongSelf.updateModel(response)
strongSelf.updateUI(model)
}Not Preferred
// might crash if self is released before response returns
resource.request().onComplete { [unowned self] response in
let model = self.updateModel(response)
self.updateUI(model)
}Not Preferred
// deallocate could happen between updating the model and updating UI
resource.request().onComplete { [weak self] response in
let model = self?.updateModel(response)
self?.updateUI(model)
}Always use the strictest access control possible.
This rule applies to private vs fileprivate as well. When choosing between the two, always use private, unless the code wouldn't compile otherwise, which should only happen when the intended behaviour is "non visible to the outside world, but visible in extensions or other types defined in the same file".
Use private as the leading property specifier. The only things that should come before access control are the static specifier or attributes such as @IBAction and @IBOutlet.
Preferred:
class TimeMachine {
private dynamic lazy var fluxCapacitor = FluxCapacitor()
}Not Preferred:
class TimeMachine {
lazy dynamic private var fluxCapacitor = FluxCapacitor()
}When writing negative conditions, avoid using ! because it's easy to miss it.
Preferred:
if foo == false {
// do something
}Not Preferred:
if !foo {
// do something
}Code using conditionals should be as straightforward to read as possible, avoiding negative conditions makes it easier to read.
Preferred:
if hungry {
eat()
} else {
sleep()
}Not Preferred:
if hungry == false {
sleep()
} else {
eat()
}Preferred:
for _ in 0..<3 {
print("Hello three times")
}
for (index, person) in attendeeList.enumerate() {
print("\(person) is at position #\(index)")
}
for index in 0.stride(to: items.count, by: 2) {
print(index)
}
for index in (0...3).reverse() {
print(index)
}Not Preferred:
var i = 0
while i < 3 {
print("Hello three times")
i += 1
}
var i = 0
while i < attendeeList.count {
let person = attendeeList[i]
print("\(person) is at position #\(i)")
i += 1
}When coding with conditionals, the left hand margin of the code should be the "golden" or "happy" path. That is, don't nest if statements. Multiple return statements are OK. The guard statement is built for this.
Preferred:
func computeFFT(context: Context?, inputData: InputData?) throws -> Frequencies {
guard let context = context else { throw FFTError.noContext }
guard let inputData = inputData else { throw FFTError.noInputData }
// use context and input to compute the frequencies
return frequencies
}Not Preferred:
func computeFFT(context: Context?, inputData: InputData?) throws -> Frequencies {
if let context = context {
if let inputData = inputData {
// use context and input to compute the frequencies
return frequencies
}
else {
throw FFTError.noInputData
}
}
else {
throw FFTError.noContext
}
}When multiple optionals are unwrapped either with guard or if let, minimize nesting by using the compound version when possible. Example:
Preferred:
guard let number1 = number1, number2 = number2, number3 = number3 else { fatalError("impossible") }
// do something with numbersNot Preferred:
if let number1 = number1 {
if let number2 = number2 {
if let number3 = number3 {
// do something with numbers
}
else {
fatalError("impossible")
}
}
else {
fatalError("impossible")
}
}
else {
fatalError("impossible")
}Guard statements are required to exit in some way. Generally, this should be simple one line statement such as return, throw, break, continue, and fatalError(). Large code blocks should be avoided. If cleanup code is required for multiple exit points, consider using a defer block to avoid cleanup code duplication.
Swift does not require a semicolon after each statement in your code. They are only required if you wish to combine multiple statements on a single line.
Do not write multiple statements on a single line separated with semicolons.
The only exception to this rule is the for-conditional-increment construct, which requires semicolons. However, alternative for-in constructs should be used where possible.
Preferred:
let swift = "not a scripting language"Not Preferred:
let swift = "not a scripting language";NOTE: Swift is very different from JavaScript, where omitting semicolons is generally considered unsafe
Parentheses around conditionals are not required and should be omitted.
Preferred:
if name == "Hello" {
print("World")
}Not Preferred:
if (name == "Hello") {
print("World")
}Don't use self, unless your code can't compile without it.
Use self only when required by the compiler (in @escaping closures, or in initializers to disambiguate properties from arguments).
Don't use #imageLiteral(resourceName:).
It might be nice to see the image in the code (when it works),
but it hides the name of the image file making it harder to read the code.
Use UIImage(named:) instead.