If a method is independent of instance state it should be declared as type method, preferably by using static keyword (or class in case when the method is designed to be overridden in a subclass).
This is to avoid methods giving a false impression that they are dependent on the instance state while in fact they are not.
Subclassing behavior should be designed with maximal hermetization principle in mind - choosing most restrictive access level possible.
Declare class final unless you explicitly design it to be inheritable. For non final class declare all instance members as final and type members as static except for those you explicitly design for overriding.
When designing a module prefer public for non final class unless you explicitly design it to be inheritable by external clients. For open class prefer public for members except for those you explicitly design for overriding by external clients.
Every point that can be overridden increases complexity, thus it should always be a conscious choice to add it.
Use precondition(_:_:file:line:) for internal sanity checks by default. If you do not want to impact performance of shipping code apply assert(_:_:file:line:) instead.
This is to avoid internal sanity checks being removed in optimized builds. Using preconditions to enforce valid data and state causes your application to terminate more predictably if an invalid state occurs, and helps make the problem easier to debug. Stopping execution as soon as an invalid state is detected also helps limit the damage caused by that invalid state.
Utilize fatalError(_:file:line:) in places where application's state and logic are not evaluated at all, for example: unimplemented method stubs.
Use guard and preconditionFailure(_:file:line:) instead of force unwrapping to clarify your intent.
guard let value = optionalValue else { preconditionFailure("reason why value must be present") }This is to avoid situations in which reason for force unwrapping is undocumented.
func didDeliver(parcel: Parcel) {
// noop
}This is to make clear function is empty by intention and not by mistake
Do not declare type if compiler is able to infer it.
let string = "a string"
let double = 1.0
let int = 1
let array = [1, 2, 3]
let dictionary = [1: "one", 2: "two"]This is to minimize clutter.
Keep type on the right hand side of assignment operator (this rule does not apply for closures, declare the type wherever it helps readability).
let height = CGFloat(44)
let duration = NSTimeInterval(0.25)
let view = UIView()
var array = [Int]()
var dictionary = [Int: String]()
let accumulator: (Int, Int) -> Int = { $0 + $1 }Since not all the types can be declared by a literal and type on the left hand side of assignment operator (e.g.
let height: CGFloat = 20), let's keep all the types on the right side.
Declare constants within appropriate type.
class PhysicalObject {
/// N * m^2 * kg^-2
static let gravitationalConstant = 6.667e-11
}This is to increase discoverability and avoid global namespace pollution.
Gather IBOutlets in one group above all the other properties and keep them private.
@IBOutlet private weak var view: UIView!Outlets are implementation detail and should be kept private. Gathering them in one group above all the other properties is old convention.
Access control modifiers should not be specified on extension level:
private extension MyType {
func method() {
// code using helper()
}
func helper() { ... }
}But on member level:
extension MyType {
fileprivate func method() {
// code using helper()
}
private func helper() { ... }
}This is to introduce distinction between members that should be visible from different types in the same file and those which should not. (
private extensionhas the same result asfileprivate extensionwhere all members end up beingfileprivate).
Don't not avoid double negation.
if !unavailable {
//...
}This is to avoid confusion.
Use "positive" names for variables and properties.
if available {
//...
}return available ? operationWhenAvailable() : nilThis is to decrease mental tax often caused by "negative" properties. It will also help keeping "main" application flow on the left hand side of colon when conditional operator
?:is used.
Do not use negative conditions with guard.
guard !array.isEmpty else { return }This is to restrain overusing of
guardstatement whereifwould suffice.if array.isEmpty { return }
Nesting if statements should be avoided.
func buy(soda id: SodaID, with money: [Coin]) throws -> (Soda, [Coin]) {
if let soda = soda(for: id) {
if value(of: money) >= soda.price {
if let change = change(from: money, minus: soda.price) {
return (soda, change)
}
else {
throw PurchaseError.noChange
}
}
else {
throw PurchaseError.insufficientFunds
}
}
else {
throw PurchaseError.outOfStock
}
}All error conditions should be handled at the beginning of the function leaving "the working code" on the first indentation level.
func buy(soda id: SodaID, with money: [Coin]) throws -> (Soda, [Coin]) {
guard let soda = soda(for: id) else {
throw PurchaseError.outOfStock
}
if value(of: money) < soda.price {
throw PurchaseError.insufficientFunds
}
guard let change = change(from: money, minus: soda.price) else {
throw PurchaseError.noChange
}
return (soda, change)
}It is easier to understand error conditions written in linear manner than as a nested structure.
It is common to misuse primitive type like in the function below. Any string can be passed as the argument. It is too general provided that application expects email to have certain format.
func showUser(withEmail: String) { }Define a new type to make the API more strict, provide validation and convenience methods/properties if needed.
struct Email: RawRepresentable {
let username: String
let host: String
var rawValue: String {
return username + "@" + host
}
init?(rawValue: String) {
// parsing and validation code
}
}
func showUser(with: Email) { }This is to increase source code reliability by building it on data you can trust.
If you are breaking reference cycle for an object bind it to a strong reference for the duration of the closure’s execution.
operation.onComplete { [weak self] result in
guard let self = self else { return }
let model = self.updateModel(result)
self.updateUI(model)
}This is to avoid deallocation of weakly referenced object during closure execution. Use
strongSelfinstead ofselffor local variable name until SR-6156 is fixed.
Do not access singleton instance directly via static method - Singleton.sharedInstance(). Decouple from it with dependency injection.
If a singleton is used directly there is no way to perform unit tests in isolation from it. Moreover, it may introduce unexpected shared state between unit tests if they are executed in one run.
Place shared helper methods in extension of dedicated separate protocol.
// MyProtocol.swift
protocol MyProtocol {
func method()
}
protocol MyProtocolBase { }
extension MyProtocolBase {
func helperMethod() { }
}// MyProtocolImplA.swift
class MyProtocolImplA: MyProtocol {
func method() {
// helperMethod() can be called here due to MyProtocolBase conformance
}
}
extension MyProtocolImplA: MyProtocolBase { }// MyProtocolImplB.swift
class MyProtocolImplB: MyProtocol {
func method() {
// helperMethod() can be called here due to MyProtocolBase conformance
}
}
extension MyProtocolImplB: MyProtocolBase { }This is to avoid exposing helper methods in the main protocol and thus polluting it.