Swift API Design Guidelines

Swift API Design Guidelines

Apply the Swift API Design Guidelines to naming, labels, documentation, and

call-site clarity. For mixed requests, handle the API-design portion here and

route language/type-system work to swift-language, concurrency to

swift-concurrency, and lint configuration to swiftlint.

Contents

• Argument Label Rules
• Side-Effect Naming
• Mutating and Nonmutating Pairs
• Documentation Comments
• Clarity and Naming
• Fluent Usage and Protocols
• General Conventions
• Common Mistakes
• Review Checklist
• References

Argument Label Rules

Argument labels determine how a call site reads. Apply the first matching row:

| Situation | Rule | Example |

|-----------|------|---------|

| First arg completes grammatical phrase | Omit label, merge words into base name | addSubview(y) |

| Value-preserving init conversion | Omit first label | Int64(someUInt32) |

| Arguments are indistinguishable peers | Omit all labels | min(x, y) |

| First arg completes prepositional phrase | Label with preposition | fade(from: red) |

| First two args form a single abstraction | Fold preposition into base name | moveTo(x: b, y: c) |

| Everything else | Label it | split(maxSplits: 2) |

Load Argument Labels and Parameters

when resolving abstraction boundaries, multiple prepositions, conversion

initializers, indistinguishable peers, parameter naming, or default arguments.

Side-Effect Naming

Use imperative verbs for operations with side effects, result-describing noun

or adjective phrases for operations without side effects, and assertion-style

names for Boolean APIs.

array.sort()
array.append(newElement)
let d = point.distance(to: origin)
line.isEmpty
set.contains(element)

Load Side Effects and Mutating Pairs

when reviewing extended pure/mutating examples or Boolean naming.

Mutating and Nonmutating Pairs

Name mutating/nonmutating pairs from the operation's natural description:

• For verb operations, use the imperative for mutation and a result-describing

participle for the copy: sort()/sorted() or append(_:)/appending(_:).

Prefer -ed; use -ing only when -ed is ungrammatical or describes the

direct object instead of the returned result.

• For noun operations, use the noun for the copy and form + noun for

mutation: union(_:) / formUnion(_:).

• Prefix factories that create new values with make.

Load the -ed/-ing Decision Tree

when the returned-result grammar is unclear. The same reference contains

expanded form-prefix, Boolean, and factory patterns.

Documentation Comments

Every public declaration must have a documentation comment.

Summary rules by declaration kind

| Declaration | Summary describes |

|-------------|-------------------|

| Function / method | What it does and what it returns |

| Subscript | What it accesses |

| Initializer | What it creates |

| Type / property / variable | What it is |

Write summaries as a single sentence fragment, beginning with a verb (for actions) or a noun phrase (for entities), ending in a period.

/// Returns the element at the specified index.
func element(at index: Int) -> Element { ... }

/// The number of elements in the collection.
var count: Int { ... }

/// Creates a new array with the given elements.
init(_ elements: some Sequence<Element>) { ... }

/// Accesses the element at the specified position.
subscript(index: Int) -> Element { ... }

Symbol markup

Use standard symbol markup after the summary when relevant:

• - Parameter name: for individual parameters
• - Parameters: block for multiple parameters
• - Returns: for the return value
• - Throws: for errors thrown
• - Complexity: for algorithmic complexity

/// Removes and returns the element at the specified position.
///
/// - Parameter index: The position of the element to remove.
/// - Returns: The removed element.
/// - Complexity: O(*n*), where *n* is the length of the collection.
mutating func remove(at index: Int) -> Element { ... }

O(1) complexity rule

Document the complexity of any computed property that is not O(1). Callers assume properties are O(1) by default. If a property does more than constant-time work, state the complexity explicitly.

/// The total weight of all items.
///
/// - Complexity: O(*n*), where *n* is the number of items.
var totalWeight: Double {
    items.reduce(0) { $0 + $1.weight }
}

For documentation patterns and examples, see references/conventions-and-special-rules.md.

Clarity and Naming

Clarity at the point of use is the most important goal. Every design decision serves the person reading a call site.

Clarity over brevity. Longer names are acceptable when they remove ambiguity. Do not abbreviate.

// GOOD
employees.remove(at: position)

// BAD — ambiguous: remove the element? remove at position?
employees.remove(position)

Include words needed to avoid ambiguity. If omitting a word makes the call site unclear, keep it.

// GOOD — "at" clarifies the argument's role
friends.remove(at: index)

// BAD — is "index" the element to remove or the position?
friends.remove(index)

Omit needless words. Do not repeat type information already available from the context.

// GOOD
allViews.remove(cancelButton)

// BAD — "Element" repeats the type
allViews.removeElement(cancelButton)

Name variables and parameters by role, not type. Use the entity's role in the current context, not its type name.

// GOOD — describes the role
var greeting: String
func add(_ observer: NSObject, for keyPath: String)

// BAD — names the type
var string: String
func add(_ object: NSObject, for string: String)

Compensate for weak type information. When a parameter type is Any, AnyObject, or a fundamental type like Int or String, add role-clarifying words to the name.

// GOOD — role is clear despite weak types
func addObserver(_ observer: NSObject, forKeyPath path: String)

// BAD — what does "string" mean here?
func add(_ object: NSObject, for string: String)

For extended naming examples and patterns, see references/naming-and-clarity.md.

Fluent Usage and Protocols

Call sites read as grammatical English. Prefer names that form grammatical phrases at the point of use.

// GOOD — reads fluently
x.insert(y, at: z)          // "x, insert y at z"
x.subviews.remove(at: i)    // "x's subviews, remove at i"
x.makeIterator()             // "x, make iterator"

// BAD — ungrammatical
x.insert(y, position: z)
x.subviews.remove(i)

Initializer first argument. The first argument to an initializer should not form a phrase continuing the type name.

// GOOD
let foreground = Color(red: 32, green: 64, blue: 128)

// BAD — "Color with red" reads awkwardly
let foreground = Color(havingRGBValuesRed: 32, green: 64, blue: 128)

Protocol naming conventions:

| Protocol describes | Naming pattern | Examples |

|--------------------|----------------|----------|

| What something is | Noun | Collection, IteratorProtocol |

| A capability | -able, -ible, or -ing suffix | Equatable, Hashable, Sendable |

General Conventions

Casing. Types and protocols use UpperCamelCase. Everything else uses lowerCamelCase. Acronyms that are commonly all-caps in American English appear uniformly upper- or lower-cased based on position.

var utf8Bytes: [UTF8.CodeUnit]
var isRepresentableAsASCII = true
var userSMTPServer: SMTPServer

Methods and properties over free functions. Prefer methods and properties. Use free functions only when:

• There is no obvious self — min(x, y)
• The function is an unconstrained generic — print(value)
• The function syntax is established domain notation — sin(x)

Default arguments over method families. Prefer a single method with default parameters over a family of methods that differ only in which parameters they accept. Place defaulted parameters at the end. Parameters with default values should always have argument labels — defaulted parameters are usually omitted at call sites, so their labels must be clear when they do appear.

// GOOD — labeled with defaults
func decode(_ data: Data, encoding: String.Encoding = .utf8) -> String?

// BAD — method family
func decode(_ data: Data) -> String?
func decode(_ data: Data, encoding: String.Encoding) -> String?

Overload safety. Methods may share a base name when they operate in different type domains or when their meaning is clear from context. Avoid return-type-only overloads that cause ambiguity at the call site.

For casing edge cases, overload patterns, and tuple/closure naming, see references/conventions-and-special-rules.md.

Common Mistakes

| Mistake | Correction |

|---|---|

| Ambiguous or missing labels | Make the call read grammatically, such as remove(at:). |

| Wrong mutating/nonmutating form | Use imperative verbs for mutation and a grammatical -ed/-ing or noun form for copies. |

| Names describe types or implementation | Name roles and semantic effects. |

| Public API lacks purpose or complexity docs | Add a concise summary and document non-O(1) properties. |

| form or factory prefixes are misapplied | Reserve form for noun operations; use make for factories. |

| Type information is repeated | Remove words already clear from the declaration and context. |

| Overloads differ only by return type | Add a semantic name or parameter distinction. |

| Tuple or closure components are positional | Label public components and closure parameters. |

Review Checklist

Argument Labels

• [ ] First argument follows the correct label rule (grammatical phrase, prepositional, conversion, or labeled)
• [ ] Prepositional labels do not incorrectly group independent arguments
• [ ] Value-preserving conversion initializers omit the first label
• [ ] All non-special-case arguments have labels

Naming Semantics

• [ ] Mutating methods use imperative verb form
• [ ] Nonmutating methods use -ed/-ing or noun form
• [ ] Mutating/nonmutating pairs follow the correct pattern (verb pair or noun/form-noun pair)
• [ ] Boolean properties read as assertions (isEmpty, isValid, contains)
• [ ] Variables and parameters are named by role, not type

Documentation

• [ ] Every public declaration has a doc comment
• [ ] Summaries are single sentence fragments ending in a period
• [ ] Summaries describe the correct thing per declaration kind (action, access, creation, entity)
• [ ] Non-O(1) computed properties document their complexity
• [ ] Parameters, return values, and thrown errors are documented with symbol markup

Conventions

• [ ] Types and protocols use UpperCamelCase; everything else uses lowerCamelCase
• [ ] Acronyms are uniformly cased based on position
• [ ] Default arguments are preferred over method families
• [ ] Overloads do not differ only in return type
• [ ] Protocol names follow the noun (is-a) or suffix (capability) convention

References

• Naming clarity, role-based naming, weak-type compensation, and terminology: references/naming-and-clarity.md
• Argument label edge cases, parameter naming, and default argument strategy: references/argument-labels-and-parameters.md
• Side-effect naming examples, -ed/-ing decision tree, form- prefix patterns, and factory methods: references/side-effects-and-mutating-pairs.md
• Casing edge cases, complexity documentation, overload safety, tuple/closure naming, and free function exceptions: references/conventions-and-special-rules.md