Our top goals are clarity, consistency and brevity, in that order. Other references:
- 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
- References
Strive to make your code compile without warnings.
Descriptive and consistent naming makes software easier to read and understand. Use the Swift naming conventions described in the API Design Guidelines. Some key takeaways include:
- striving for clarity at the call site
- prioritizing clarity over brevity
- using camel case (not snake case)
- using uppercase for types (and protocols), lowercase for everything else
- including all needed words while omitting needless words
- using names based on roles, not types
- sometimes compensating for weak type information
- striving for fluent usage
- beginning factory methods with
make
- naming methods for their side effects
- verb methods follow the -ed, -ing rule for the non-mutating version
- noun methods follow the formX rule for the mutating version
- boolean types should read like assertions
- protocols that describe what something is should read as nouns
- protocols that describe a capability should end in -able or -ible
- using terms that don't surprise experts or confuse beginners
- generally avoiding abbreviations
- using precedent for names
- preferring methods and properties to free functions
- casing acronyms and initialisms uniformly up or down
- giving the same base name to methods that share the same meaning
- avoiding overloads on return type
- choosing good parameter names that serve as documentation
- labeling closure and tuple parameters
- taking advantage of default parameters
Swift types are automatically namespaced by the module that contains them and you should not add a class prefix such as ZP. 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()
When creating custom delegate methods, an unnamed first parameter should be the delegate source. (UIKit contains numerous examples of this).
Preferred:
func namePickerView(_ namePickerView: NamePickerView, didSelectName name: String)
func namePickerViewShouldReload(_ namePickerView: NamePickerView) -> Bool
Not Preferred:
func didSelectName(namePicker: NamePickerViewController, name: String)
func namePickerShouldReload() -> Bool
Use compiler inferred context to write shorter, clear code. (Also see Type Inference.)
Preferred:
let selector = #selector(viewDidLoad)
view.backgroundColor = .red
let toView = context.view(forKey: .to)
let view = UIView(frame: .zero)
Not Preferred:
let selector = #selector(ViewController.viewDidLoad)
view.backgroundColor = UIColor.red
let toView = context.view(forKey: UITransitionContextViewKey.to)
let view = UIView(frame: CGRect.zero)
On the other hand, when declaring class properties, always specify the type. It makes it easier to read and faster to compile.
Preferred:
let selector = #selector(viewDidLoad)
view.backgroundColor = .red
let toView = context.view(forKey: .to)
let view = UIView(frame: .zero)
Not Preferred:
let selector = #selector(ViewController.viewDidLoad)
view.backgroundColor = UIColor.red
let toView = context.view(forKey: UITransitionContextViewKey.to)
let view = UIView(frame: CGRect.zero)
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 Person {
var firstName: String = ""
var lastName: String = ""
}
Not Preferred:
struct Person {
var firstName = ""
var lastName = ""
}
Use US English spelling to match Apple's API.
Preferred:
let color = "red"
Not Preferred:
let colour = "red"
Use extensions to organize your code into logical blocks of functionality. Each extension should be set off with a // MARK: -
comment to keep things well-organized.
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
}
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. This includes any empty/unused UIApplicationDelegate methods.
Preferred:
override func tableView(_ tableView: UITableView, numberOfRowsInSection section: Int) -> Int {
return Database.contacts.count
}
Not Preferred:
override func didReceiveMemoryWarning() {
super.didReceiveMemoryWarning()
// Dispose of any resources that can be recreated.
}
override func numberOfSections(in 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
}
Keep imports minimal. For example, don't import UIKit
when importing Foundation
will suffice.
- Indent using tabs
- Method braces and other braces (
if
/else
/switch
/while
etc.) always open on the same line as the statement but close on a new line. - Tip: You can re-indent by selecting some code (or โA to select all) and then Control-I (or Editor\Structure\Re-Indent in the menu). Some of the Xcode template code will have 4-space tabs hard coded, so this is a good way to fix that.
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
? :
, empty dictionary[:]
and#selector
syntax for unnamed parameters(_:)
.
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]
}
-
Long lines should be wrapped at around 70 characters. A hard limit is intentionally not specified.
-
Remove trailing whitespaces at the ends of lines.
-
Add a single newline character at the end of each file.
When they are needed, use comments to explain why a particular piece of code does something, especially for public methods. 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.
Here's an example of a well-styled class definition:
class Circle: Shape {
var x: Int
var 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)
}
override func area() -> Double {
return Double.pi * radius * radius
}
}
extension Circle: CustomStringConvertible {
var description: String {
return "center = \(centerString) area = \(area())"
}
private var 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
. - Indent getter and setter definitions and property observers. Keep the code in the same line of the getter and setter definitions if there's only one line of code.
- Don't add modifiers such as
internal
when they're already the default. Similarly, don't repeat the access modifier when overriding a method. - Organize extra functionality (e.g. printing) in extensions.
- Hide non-shared, implementation details such as
centerString
inside the extension usingprivate
access control.
For conciseness, avoid using self
since Swift does not require it to access an object's properties or invoke its methods.
Use self only when required by the compiler (in @escaping
closures, or in initializers to disambiguate properties from arguments). In other words, if it compiles without self
then omit it.
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
}
}
Keep short function declarations on one line including the opening brace:
func reticulateSplines(spline: [Double]) -> Bool {
// reticulate code goes here
}
For functions with long signatures, add line breaks at appropriate points and add an extra indent on subsequent lines:
func reticulateSplines(spline: [Double], adjustmentFactor: Double,
translateConstant: Int, comment: String) -> Bool {
// reticulate code goes here
}
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.animate(withDuration: 1.0) {
self.myView.alpha = 0
}
UIView.animate(withDuration: 1.0, animations: {
self.myView.alpha = 0
}, completion: { finished in
self.myView.removeFromSuperview()
})
Not Preferred:
UIView.animate(withDuration: 1.0, animations: {
self.myView.alpha = 0
})
UIView.animate(withDuration: 1.0, animations: {
self.myView.alpha = 0
}) { f in
self.myView.removeFromSuperview()
}
For single-expression closures where the context is clear, use implicit returns and keep it in the same line:
attendeeList.sort { $0 > $1 }
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 developer. Examples:
let value = numbers.map { $0 * 2 }.filter { $0 % 3 == 0 }.index(of: 90)
let value = numbers
.map { $0 * 2 }
.filter { $0 > 50 }
.map { $0 + 10 }
Always include exactly one white space after and before curly braces in closures.
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 NSNumber).stringValue // String
Not Preferred:
let width: NSNumber = 120.0 // NSNumber
let widthString: NSString = width.stringValue // NSString
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 on 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 root2 = 1.41421356237309504880168872
}
let hypotenuse = side * Math.root2
Note: 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 root2 = 1.41421356237309504880168872
let hypotenuse = side * root2 // what is root2?
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 instance variables that you know will be initialized later before use, such as subviews that will be set up in viewDidLoad
.
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, let 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
}
}
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 = 106.5
Not Preferred:
let message: String = "Click the button"
let currentBounds: CGRect = computeViewBounds()
let names = [String]()
Property declarations are an exception to this rule and should have type annotation.
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.mergeSorted() // easily discoverable
rocket.launch() // acts on the model
Not 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 natural
Code should not create reference cycles. Analyze your object graph and prevent strong cycles with weak
and unowned
references.
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)
}
Using private
and fileprivate
appropriately adds clarity and promotes encapsulation. Prefer private
to fileprivate
when possible. Using extensions may require you to use fileprivate
.
Only explicitly use open
, public
, and internal
when you require a full access control specification.
Use access control as the leading property specifier. The only things that should come before access control are the static
specifier or attributes such as @IBAction
, @IBOutlet
and @discardableResult
.
Preferred:
private let message = "Great Scott!"
class TimeMachine {
fileprivate dynamic lazy var fluxCapacitor = FluxCapacitor()
}
Not Preferred:
fileprivate let message = "Great Scott!"
class TimeMachine {
lazy dynamic fileprivate var fluxCapacitor = FluxCapacitor()
}
Prefer the for-in
style of for
loop over the while-condition-increment
style.
Preferred:
for _ in 0..<3 {
print("Hello three times")
}
for (index, person) in attendeeList.enumerated() {
print("\(person) is at position #\(index)")
}
for index in stride(from: 0, to: items.count, by: 2) {
print(index)
}
for index in (0...3).reversed() {
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,
let number2 = number2,
let number3 = number3 else {
fatalError("impossible")
}
// do something with numbers
Not 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.
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")
}
In larger expressions, optional parentheses can sometimes make code read more clearly.
Preferred:
let playerMark = (player == current ? "X" : "O")