/swiftui-expert-skill

SwiftUI View Structure Reference

Table of Contents

• View Structure Principles
• View File Structure (optional readability suggestion)
• Struct or Method / Computed Property?
• Extract Subviews, Not Computed Properties
• @ViewBuilder
• SDK 27 @ContentBuilder
• Keep View Body Simple and Avoid High-Cost Operations
• Keep View init Cheap
• Single-Child Group
• When to Extract Subviews
• Container View Pattern
• Utilize Lazy Containers for Large Data Sets
• ZStack vs overlay/background
• Compositing Group Before Clipping
• Split State-Driven Parts into Custom View Types
• Reusable Styling with ViewModifier
• Skeleton Loading with Redacted Views
• AnyView
• UIViewRepresentable Essentials
• Troubleshooting
• Summary Checklist

View Structure Principles

SwiftUI's diffing algorithm compares view hierarchies to determine what needs updating. Proper view composition directly impacts performance.

View File Structure (optional readability suggestion)

Property ordering has no effect on correctness or performance, so treat this as a personal/team readability preference rather than a rule. Some developers find a consistent order easier to scan — for example, environment, then state, then passed-in properties, then init, body, and helper subviews. Adopt it only if your team wants the consistency; never reorder existing code solely to match it.

struct ContentView: View {
    // MARK: - Environment Properties
    @Environment(\.colorScheme) var colorScheme

    // MARK: - State Properties
    @Binding var isToggled: Bool
    @State private var viewModel: SomeViewModel

    // MARK: - Private Properties
    private let title: String = "SwiftUI Guide"

    // MARK: - Initializer (if needed)
    init(isToggled: Binding<Bool>) {
        self._isToggled = isToggled
    }

    // MARK: - Body
    var body: some View {
        VStack {
            header
            content
        }
    }

    // MARK: - Computed Subviews
    private var header: some View {
        Text(title).font(.largeTitle).padding()
    }

    private var content: some View {
        VStack {
            Text("Counter: \(counter)")
        }
    }
}

Struct or Method / Computed Property?

If a View is intended to be reusable across multiple screens, encapsulate it within a separate struct. If its usage is confined to a single context, it can be declared as a function or computed property within the containing View.

However, if a view maintains state using @State, @Binding, @ObservedObject, @Environment, @StateObject, or similar wrappers, it should generally be a separate struct.

• For simple, static views: a computed property is acceptable.
• For views requiring parameters: a method is more appropriate, but only when those parameters are stable. If parameters change per-call (e.g. inside a ForEach where each call receives a different item), prefer a separate struct so SwiftUI can diff inputs and skip body evaluation.
• For reusable, stateful, or logically independent UI sections: prefer a dedicated struct.

struct ContentView: View {
    var titleView: some View {
        Text("Hello from Property")
            .font(.largeTitle)
            .foregroundStyle(.blue)
    }

    func messageView(text: String, color: Color) -> some View {
        Text(text)
            .font(.title)
            .foregroundStyle(color)
            .padding()
    }

    var body: some View {
        VStack {
            titleView
            messageView(text: "Hello from Method", color: .red)
        }
    }
}

For conditional modifier composition and AnyShapeStyle, consult references/modifier-patterns.md.

Extract Subviews, Not Computed Properties

A view is SwiftUI's unit of invalidation. When an input changes, SwiftUI re-runs the body of the smallest enclosing view type that depends on it — every conditional, modifier chain, and string interpolation in that body, even if only one leaf actually depends on what changed. A computed property or @ViewBuilder helper is inlined into the parent's body, so it shares the parent's invalidation boundary and does not reduce update cost; it only reorganizes the code. A separate View type with narrow inputs becomes its own boundary and re-runs only when its own inputs change.

This is why "split your body for readability" is also a performance tool — but only when you split into real View types, not computed properties.

The Problem with @ViewBuilder Functions

When you use @ViewBuilder functions or computed properties for complex views, the entire function re-executes on every parent state change:

// BAD - re-executes complexSection() on every tap
struct ParentView: View {
    @State private var count = 0

    var body: some View {
        VStack {
            Button("Tap: \(count)") { count += 1 }
            complexSection()  // Re-executes every tap!
        }
    }

    @ViewBuilder
    func complexSection() -> some View {
        // Complex views that re-execute unnecessarily
        ForEach(0..<100) { i in
            HStack {
                Image(systemName: "star")
                Text("Item \(i)")
                Spacer()
                Text("Detail")
            }
        }
    }
}

The Solution: Separate Structs

Extract to separate struct views. SwiftUI can skip their body when inputs don't change:

// GOOD - ComplexSection body SKIPPED when its inputs don't change
struct ParentView: View {
    @State private var count = 0

    var body: some View {
        VStack {
            Button("Tap: \(count)") { count += 1 }
            ComplexSection()  // Body skipped during re-evaluation
        }
    }
}

struct ComplexSection: View {
    var body: some View {
        ForEach(0..<100) { i in
            HStack {
                Image(systemName: "star")
                Text("Item \(i)")
                Spacer()
                Text("Detail")
            }
        }
    }
}

Why This Works

1. SwiftUI compares the ComplexSection struct (which has no properties)
2. Since nothing changed, SwiftUI skips calling ComplexSection.body
3. The complex view code never executes unnecessarily

Multi-section detail views

The most common place this rule gets dropped is a detail screen with several distinct sections — header + gallery + description + reviews, header + ingredients + steps, hero + specs + related. The tempting shape is one big view with private var header: some View, private var gallery: some View, and so on. That shape shares one invalidation boundary, so a change that affects one section re-evaluates all of them. Factor each named section into its own View type that takes only the fields it renders, and keep the parent thin — it just composes the sections.

// PREFER: each section is its own type with narrow inputs.
struct ProductDetailView: View {
    let product: Product

    var body: some View {
        ScrollView {
            VStack(alignment: .leading, spacing: 24) {
                ProductHeader(name: product.name, price: product.price)
                ProductGallery(images: product.imageURLs)
                ProductDescription(text: product.descriptionText)
                ProductReviews(average: product.averageStars, count: product.reviewCount)
            }
            .padding()
        }
    }
}

This generalizes to every *DetailView: one View type per section, narrow inputs each, a thin parent that composes them. Small @ViewBuilder fragments reused two or three times within the same body are still fine — the rule targets factoring done for organization or to manage body length, where a real View type does the right thing.

@ViewBuilder

Use @ViewBuilder functions for small, simple sections (a few views, no expensive computation) that don't affect performance. They work particularly well for static content that doesn't depend on any @State or @Binding, since SwiftUI won't need to diff them independently. Extract to a separate struct when the section is complex, depends on state, or needs to be skipped during re-evaluation.

The @ViewBuilder attribute is only required when a function or computed property returns multiple different views conditionally, for example through if or switch:

@ViewBuilder
private var conditionalView: some View {
    if isExpanded {
        VStack {
            Text("Expanded View")
            Image(systemName: "star")
        }
    } else {
        Text("Collapsed View")
    }
}

If every branch returns the same concrete type, @ViewBuilder is unnecessary:

var conditionalText: some View {
    if Bool.random() {
        Text("Hello")
    } else {
        Text("World")
    }
}

Prefer @ViewBuilder when:

• there is conditional branching between multiple view types
• extracting a separate struct would not provide meaningful separation

SDK 27 @ContentBuilder

SDK 27 unifies many SwiftUI result builders under @ContentBuilder. Block contents are no longer constrained to View, so previously compiling source can become ambiguous. Choose the narrow fix matching the diagnostic. Do not broadly rewrite working builders or rename unrelated types.

• For ambiguous ShapeStyle.opacity or blendMode passed directly to overlay or background, select the builder overload:

  Rectangle().overlay {
      Color.blue.opacity(0.3).blendMode(.overlay)
  }

• Fully qualify a shadowed SwiftUI type, such as SwiftUI.Color.clear.
• Avoid spelling concrete TupleView or TupleContent generic structures; prefer opaque some View. If a concrete SDK 27 type is unavoidable, the builder now produces TupleContent. When the deployment target is below OS 27 and a concrete TupleView constraint is unavoidable, construct TupleView((...)) explicitly inside the builder as a back-deployment fallback.
• If an empty nested builder is ambiguous, provide EmptyContent() or EmptyView(). This can occur with MapKit in the dependency graph even when the file does not import MapKit, if member-import visibility is disabled. It also occurs with a conditional-compilation branch that becomes empty.
• For deeply branching Charts content that times out only when back-deployed (typically around 10+ if/else if or switch branches), extract the branches into an @ChartContentBuilder helper.

Keep View Body Simple and Avoid High-Cost Operations

Refrain from performing complex operations within the body of your view. Instead of passing a ready-to-use sequence with filtering, mapping, or sorting directly into ForEach, prepare the sequence outside the body.

// Avoid such things ...
var body: some View {
    List {
        ForEach(model.values.filter { $0 > 0 }, id: \.self) {
            Text(String($0))
                .padding()
        }
    }
}

Prefer already-prepared values. init is not a one-time filter — it reruns whenever the parent re-evaluates, so treat it as a constant-time copy of inputs. Cache derived collections on the model (or in @State updated from .onChange) and pass the prepared sequence in:

struct FilteredListView: View {
    let filteredValues: [Int]

    var body: some View {
        List {
            ForEach(filteredValues, id: \.self) { value in
                Text(String(value))
                    .padding()
            }
        }
    }
}

The reason this matters is that the system can call body multiple times during a single layout phase. Complex body computation makes those calls more expensive than necessary.

General guidance:

• avoid filtering, sorting, and mapping inline in body
• avoid constructing expensive formatters in body
• avoid heavy branching in large view trees
• move data preparation into the model layer or dedicated helpers; do not filter in init

Keep View init Cheap

A view's init runs every time the parent re-evaluates its body, which can be many times per second for views inside List, LazyVStack, scroll containers, or animated parents. Treat init as a constant-time copy of inputs into stored properties. Don't decode JSON, build a DateFormatter, touch the file system, or allocate large structures there — that work repeats on every parent body pass even when the inputs are identical.

// AVOID: decoding and formatting on every init
init(rawJSON: Data, date: Date) {
    self.summary = try! JSONDecoder().decode(WeatherSummary.self, from: rawJSON)
    let formatter = DateFormatter()
    formatter.dateStyle = .medium
    self.formattedDate = formatter.string(from: date)
}

// PREFER: take already-prepared values; format lazily in body
let summary: WeatherSummary
let date: Date

var body: some View {
    VStack {
        Text(summary.headline)
        Text(date, format: .dateTime.day().month().year())  // cached, locale-aware
    }
}

If a derived value genuinely needs to be computed once and kept, store it on an @State-owned @Observable model or compute it asynchronously in .task — not in init. init is not a one-time setup hook; it runs as often as the parent's body does.

Single-Child Group

Group { SomeView() } — a Group with exactly one concrete child — wraps the view in an extra Group<SomeView> type for no visual benefit. Every modifier chained after it must be type-checked against that wrapper, adding avoidable type-checking overhead in long chains. Drop the Group and chain modifiers directly on the child.

// AVOID: single concrete child wrapped in Group
Group { Text(status) }
    .padding(.horizontal, 8)
    .background(.thinMaterial, in: Capsule())

// PREFER: chain directly on the child
Text(status)
    .padding(.horizontal, 8)
    .background(.thinMaterial, in: Capsule())

The rule is specifically about one concrete view. A Group whose content is a ForEach, multiple sibling views, or an if/else (which produces _ConditionalContent) is doing real work — applying a shared modifier across siblings or both branches — and is fine.

When to Extract Subviews

Extract complex views into separate subviews when:

• The view has multiple logical sections or responsibilities
• The view contains reusable components
• The view body becomes difficult to read or understand
• You need to isolate state changes for performance
• The view is becoming large (keep views small for better performance)
• The section may evolve independently over time

Container View Pattern

Avoid Closure-Based Content

Closures can't be compared, causing unnecessary re-renders:

// BAD - closure prevents SwiftUI from skipping updates
struct MyContainer<Content: View>: View {
    let content: () -> Content

    var body: some View {
        VStack {
            Text("Header")
            content()  // Always called, can't compare closures
        }
    }
}

// Usage forces re-render on every parent update
MyContainer {
    ExpensiveView()
}

Use @ViewBuilder Property Instead

// GOOD - view can be compared
struct MyContainer<Content: View>: View {
    @ViewBuilder let content: Content

    var body: some View {
        VStack {
            Text("Header")
            content  // SwiftUI can compare and skip if unchanged
        }
    }
}

// Usage - SwiftUI can diff ExpensiveView
MyContainer {
    ExpensiveView()
}

Utilize Lazy Containers for Large Data Sets

When displaying extensive lists or grids, prefer LazyVStack, LazyHStack, LazyVGrid, or LazyHGrid. These containers load views only when they appear on the screen, reducing memory usage and improving performance.

struct ContentView: View {
    let items = Array(0..<1000)

    var body: some View {
        ScrollView {
            LazyVStack {
                ForEach(items, id: \.self) { item in
                    Text("Item \(item)")
                }
            }
        }
    }
}

Prefer lazy containers when:

• rendering large collections
• row views are non-trivial
• memory usage matters
• the content is inside ScrollView

ZStack vs overlay/background

Use ZStack to compose multiple peer views that should be layered together and jointly define layout.

Prefer overlay / background when you’re decorating a primary view.
Not primarily because they don’t affect layout size, but because they express intent and improve readability: the view being modified remains the clear layout anchor.

A key difference is size proposal behavior:

• In overlay / background, the child view implicitly adopts the size proposed to the parent when it doesn’t define its own size, making decorative attachments feel natural and predictable.
• In ZStack, each child participates independently in layout, and no implicit size inheritance exists. This makes it better suited for peer composition, but less intuitive for simple decoration.

Use ZStack (or another container) when the “decoration” must explicitly participate in layout sizing—for example, when reserving space, extending tappable/visible bounds, or preventing overlap with neighboring views.

Examples

// GOOD - decoration via overlay (layout anchored to button)
Button("Continue") { }
    .overlay(alignment: .trailing) {
        Image(systemName: "lock.fill").padding(.trailing, 8)
    }

// BAD - ZStack when overlay suffices (layout no longer anchored to button)
ZStack(alignment: .trailing) {
    Button("Continue") { }
    Image(systemName: "lock.fill").padding(.trailing, 8)
}

// GOOD - background shape takes parent size
HStack(spacing: 12) { Text("Inbox"); Text("Next") }
    .background { Capsule().strokeBorder(.blue, lineWidth: 2) }

Compositing Group Before Clipping

Always add .compositingGroup() before .clipShape() when clipping layered views (.overlay or .background). Without it, each layer is antialiased separately and then composited. Where antialiased edges overlap — typically at rounded corners — you get visible color fringes (semi-transparent pixels of different colors blending together).

let shape = RoundedRectangle(cornerRadius: 16)

// BAD - each layer antialiased separately, producing color fringes at corners
Color.red
    .overlay(.white, in: shape)
    .clipShape(shape)
    .frame(width: 200, height: 150)

// GOOD - layers composited first, antialiasing applied once during clipping
Color.red
    .overlay(.white, in: .rect)
    .compositingGroup()
    .clipShape(shape)
    .frame(width: 200, height: 150)

.compositingGroup() forces all child layers to be rendered into a single offscreen buffer before the clip is applied. This means antialiasing only happens once — on the final composited result — eliminating the fringe artifacts.

Split State-Driven Parts into Custom View Types

Large views often depend on multiple independent state sources. If a single view body depends on all of them, then any state change can cause the entire body to re-evaluate.

struct BigAndComplicatedView: View {
    @State private var counter = 0
    @State private var isToggled = false
    @StateObject private var viewModel = SomeViewModel()

    let title = "Big and Complicated View"

    var body: some View {
        VStack {
            Text(title)
                .font(.largeTitle)

            Text("Counter: \(counter)")
                .font(.title)

            Toggle("Enable Feature", isOn: $isToggled)
                .padding()

            Button("Increment Counter") {
                counter += 1
            }

            Text("ViewModel Data: \(viewModel.data)")
                .padding()

            Button("Fetch Data") {
                viewModel.fetchData()
            }
        }
    }
}

Better: Split Into Smaller Components

struct BigAndComplicatedView: View {
    @State private var counter = 0
    @State private var isToggled = false
    @StateObject private var viewModel = SomeViewModel()

    var body: some View {
        VStack {
            titleView
            CounterView(counter: $counter)
            ToggleView(isToggled: $isToggled)
            ViewModelDataView(data: viewModel.data) {
                viewModel.updateData()
            }
            .equatable()
        }
    }

    private var titleView: some View {
        Text("Big and Complicated View")
            .font(.largeTitle)
    }
}

Why this is better:

• changing counter only affects CounterView
• toggling only affects ToggleView
• updating the model data only affects ViewModelDataView

Notes on Equatable

Using Equatable for a view is not a universal best practice, but it can be useful in targeted cases where:

• the input is small and well-defined
• the comparison logic is meaningful
• you want to reduce unnecessary body evaluation for a specific subtree

Do not use Equatable as a blanket optimization technique.

Reusable Styling with ViewModifier

Extract repeated modifier combinations into a ViewModifier struct. Expose via a View extension for autocompletion:

private struct CardStyle: ViewModifier {
    func body(content: Content) -> some View {
        content
            .padding()
            .background(Color(.secondarySystemBackground))
            .clipShape(.rect(cornerRadius: 12))
    }
}

extension View {
    func cardStyle() -> some View {
        modifier(CardStyle())
    }
}

Custom ButtonStyle

Use the ButtonStyle protocol for reusable button designs. Use PrimitiveButtonStyle only when you need custom interaction handling (e.g., simultaneous gestures):

struct PrimaryButtonStyle: ButtonStyle {
    func makeBody(configuration: Configuration) -> some View {
        configuration.label
            .bold()
            .foregroundStyle(.white)
            .padding(.horizontal, 16)
            .padding(.vertical, 8)
            .background(Color.accentColor)
            .clipShape(Capsule())
            .scaleEffect(configuration.isPressed ? 0.95 : 1)
            .animation(.smooth, value: configuration.isPressed)
    }
}

Discoverability with Static Member Lookup

Make custom styles and modifiers discoverable via leading-dot syntax:

extension ButtonStyle where Self == PrimaryButtonStyle {
    static var primary: PrimaryButtonStyle { .init() }
}

// Usage: .buttonStyle(.primary)

This pattern works for any SwiftUI style protocol (ButtonStyle, ListStyle, ToggleStyle, etc.).

Skeleton Loading with Redacted Views

Use .redacted(reason: .placeholder) to show skeleton views while data loads. Use .unredacted() to opt out specific views:

VStack(alignment: .leading) {
    Text(article?.title ?? String(repeating: "X", count: 20))
        .font(.headline)
    Text(article?.author ?? String(repeating: "X", count: 12))
        .font(.subheadline)
    Text("SwiftLee")
        .font(.caption)
        .unredacted()
}
.redacted(reason: article == nil ? .placeholder : [])

Apply .redacted on a container to redact all children at once.

AnyView

AnyView is type erasure. SwiftUI uses structural identity based on type information to determine when views should be updated.

private var nameView: some View {
    if isEditable {
        TextField("Your name", text: $name)
    } else {
        Text(name)
    }
}

Avoid patterns like:

private var nameView: some View {
    if isEditable {
        return AnyView(TextField("Your name", text: $name))
    } else {
        return AnyView(Text(name))
    }
}

Because AnyView erases type information, SwiftUI loses some optimization opportunities. Prefer @ViewBuilder or conditional branches with concrete view types.

Use AnyView only when type erasure is truly necessary for API design.

UIViewRepresentable Essentials

When bridging UIKit views into SwiftUI:

• makeUIView(context:) is called once to create the UIKit view
• updateUIView(_:context:) is called on every SwiftUI redraw to sync state
• The representable struct itself is recreated on every redraw -- avoid heavy work in its init
• Use a Coordinator for delegate callbacks and two-way communication

struct MapView: UIViewRepresentable {
    let coordinate: CLLocationCoordinate2D

    func makeUIView(context: Context) -> MKMapView {
        let map = MKMapView()
        map.delegate = context.coordinator
        return map
    }

    func updateUIView(_ map: MKMapView, context: Context) {
        map.setCenter(coordinate, animated: true)
    }

    func makeCoordinator() -> Coordinator { Coordinator() }

    class Coordinator: NSObject, MKMapViewDelegate { }
}

Troubleshooting

Debug SwiftUI Renderings

See references/performance-patterns.md (item #8) for the _printChanges() vs _logChanges() comparison and the @self/@identity output meaning. The snippets below show the call sites.

If it is needed to debug render cycles and read console output you can leverage the _printChanges() or _logChanges() methods on View. These methods print information about when the view is being evaluated and what changes are triggering updates. This can be very helpful when your view body is called multiple times and you want to know why.

struct ContentView: View {
    @State private var counter: Int = 99

    init() {
        print(Self.self, #function)
    }

    var body: some View {
        let _ = Self._printChanges()

        VStack {
            Text("Counter: \(counter)")
            Button {
                counter += 1
            } label: {
                Text("Counter +1")
            }
            .buttonStyle(.borderedProminent)
        }
        .padding()
    }
}

As an alternative to Self._printChanges(), you can use _logChanges()

struct ContentView: View {
    @State private var counter: Int = 99

    var body: some View {
        let _ = Self._logChanges()

        VStack {
            Text("Counter: \(counter)")
            Button {
                counter += 1
            } label: {
                Text("Counter +1")
            }
            .buttonStyle(.borderedProminent)
        }
        .padding()
    }
}

Use these tools only for debugging and remove them from production code.

Handling "The Compiler Is Unable to Type-Check This Expression in Reasonable Time"

If you encounter:

The compiler is unable to type-check this expression in reasonable time; try breaking up the expression into distinct sub-expressions

it is often caused by overly complex view structures or expressions.

Ways to fix it:

• break large expressions into smaller computed values
• extract subviews
• split long modifier chains
• simplify nested generics and builders
• avoid huge inline closures

Summary Checklist

• Extract complex views into separate subviews, not computed properties
• Keep views small for readability and performance
• Use @ViewBuilder only where it actually adds value
• Avoid heavy filtering, mapping, sorting, or formatter creation inside body or init
• Keep view init cheap (constant-time input copy; it runs on every parent body pass)
• Resolve SDK 27 @ContentBuilder ambiguity with the matching narrow fix
• Avoid single-child Group { OneView() } (chain modifiers directly on the child)
• Use lazy containers for large data sets
• Container views use @ViewBuilder let content: Content
• Prefer overlay / background for decoration and ZStack for peer composition
• .compositingGroup() before .clipShape() on layered views to avoid antialiasing fringes
• Split state-heavy areas into smaller view types
• Extract repeated styling into ViewModifier or ButtonStyle
• Expose reusable styles via static member lookup when it improves discoverability
• Use .redacted(reason: .placeholder) for loading skeletons
• Avoid AnyView unless type erasure is truly needed
• In UIViewRepresentable, keep heavy work out of struct init
• Use _printChanges() / _logChanges() to debug rendering behavior
• Break up overly complex expressions when the compiler struggles