Working with percentages in SwiftUI structure – Ole Begemann

SwiftUI’s structure primitives usually don’t present relative sizing choices, e.g. “make this view 50 % of the width of its container”. Let’s construct our personal!

Use case: chat bubbles

Think about this chat dialog view for instance of what I wish to construct. The chat bubbles all the time stay 80 % as huge as their container because the view is resized:

Constructing a proportional sizing modifier

1. The Structure

We are able to construct our personal relative sizing modifier on prime of the Structure protocol. The structure multiplies its personal proposed dimension (which it receives from its dad or mum view) with the given elements for width and top. It then proposes this modified dimension to its solely subview. Right here’s the implementation (the complete code, together with the demo app, is on GitHub):

/// A customized structure that proposes a proportion of its
/// acquired proposed dimension to its subview.
///
/// - Precondition: should comprise precisely one subview.
fileprivate struct RelativeSizeLayout: Structure {
    var relativeWidth: Double
    var relativeHeight: Double

    func sizeThatFits(
        proposal: ProposedViewSize, 
        subviews: Subviews, 
        cache: inout ()
    ) -> CGSize {
        assert(subviews.depend == 1, "expects a single subview")
        let resizedProposal = ProposedViewSize(
            width: proposal.width.map { $0 * relativeWidth },
            top: proposal.top.map { $0 * relativeHeight }
        )
        return subviews[0].sizeThatFits(resizedProposal)
    }

    func placeSubviews(
        in bounds: CGRect, 
        proposal: ProposedViewSize, 
        subviews: Subviews, 
        cache: inout ()
    ) {
        assert(subviews.depend == 1, "expects a single subview")
        let resizedProposal = ProposedViewSize(
            width: proposal.width.map { $0 * relativeWidth },
            top: proposal.top.map { $0 * relativeHeight }
        )
        subviews[0].place(
            at: CGPoint(x: bounds.midX, y: bounds.midY), 
            anchor: .heart, 
            proposal: resizedProposal
        )
    }
}

Notes:

• I made the kind non-public as a result of I wish to management how it may be used. That is essential for sustaining the idea that the structure solely ever has a single subview (which makes the mathematics a lot easier).

• Proposed sizes in SwiftUI may be nil or infinity in both dimension. Our structure passes these particular values via unchanged (infinity instances a proportion continues to be infinity). I’ll talk about under what implications this has for customers of the structure.

2. The View extension

Subsequent, we’ll add an extension on View that makes use of the structure we simply wrote. This turns into our public API:

extension View {
    /// Proposes a proportion of its acquired proposed dimension to `self`.
    public func relativeProposed(width: Double = 1, top: Double = 1) -> some View {
        RelativeSizeLayout(relativeWidth: width, relativeHeight: top) {
            // Wrap content material view in a container to ensure the structure solely
            // receives a single subview. As a result of views are lists!
            VStack { // alternatively: `_UnaryViewAdaptor(self)`
                self
            }
        }
    }
}

Notes:

• I made a decision to go together with a verbose title, relativeProposed(width:top:), to make the semantics clear: we’re altering the proposed dimension for the subview, which gained’t all the time end in a special precise dimension. Extra on this under.

• We’re wrapping the subview (self within the code above) in a VStack. This may appear redundant, however it’s mandatory to ensure the structure solely receives a single factor in its subviews assortment. See Chris Eidhof’s SwiftUI Views are Lists for a proof.

Utilization

The structure code for a single chat bubble within the demo video above appears to be like like this:

let alignment: Alignment = message.sender == .me ? .trailing : .main
chatBubble
    .relativeProposed(width: 0.8)
    .body(maxWidth: .infinity, alignment: alignment)

The outermost versatile body with maxWidth: .infinity is liable for positioning the chat bubble with main or trailing alignment, relying on who’s talking.

You may even add one other body that limits the width to a most, say 400 factors:

let alignment: Alignment = message.sender == .me ? .trailing : .main
chatBubble
    .body(maxWidth: 400)
    .relativeProposed(width: 0.8)
    .body(maxWidth: .infinity, alignment: alignment)

Right here, our relative sizing modifier solely has an impact because the bubbles grow to be narrower than 400 factors. In a wider window the width-limiting body takes priority. I like how composable that is!

80 % gained’t all the time end in 80 %

Should you watch the debugging guides I’m drawing within the video above, you’ll discover that the relative sizing modifier by no means reviews a width higher than 400, even when the window is huge sufficient:

It is because our structure solely adjusts the proposed dimension for its subview however then accepts the subview’s precise dimension as its personal. Since SwiftUI views all the time select their very own dimension (which the dad or mum can’t override), the subview is free to disregard our proposal. On this instance, the structure’s subview is the body(maxWidth: 400) view, which units its personal width to the proposed width or 400, whichever is smaller.

Understanding the modifier’s conduct

Proposed dimension ≠ precise dimension

It’s essential to internalize that the modifier works on the premise of proposed sizes. This implies it relies on the cooperation of its subview to attain its objective: views that ignore their proposed dimension will likely be unaffected by our modifier. I don’t discover this notably problematic as a result of SwiftUI’s total structure system works like this. Finally, SwiftUI views all the time decide their very own dimension, so you may’t write a modifier that “does the proper factor” (no matter that’s) for an arbitrary subview hierarchy.

nil and infinity

I already talked about one other factor to pay attention to: if the dad or mum of the relative sizing modifier proposes nil or .infinity, the modifier will cross the proposal via unchanged. Once more, I don’t suppose that is notably unhealthy, however it’s one thing to pay attention to.

Proposing nil is SwiftUI’s means of telling a view to grow to be its best dimension (fixedSize does this). Would you ever wish to inform a view to grow to be, say, 50 % of its best width? I’m unsure. Perhaps it’d make sense for resizable photos and related views.

By the best way, you might modify the structure to do one thing like this:

1. If the proposal is nil or infinity, ahead it to the subview unchanged.
2. Take the reported dimension of the subview as the brand new foundation and apply the scaling elements to that dimension (this nonetheless breaks down if the kid returns infinity).
3. Now suggest the scaled dimension to the subview. The subview may reply with a special precise dimension.
4. Return this newest reported dimension as your individual dimension.

This technique of sending a number of proposals to little one views is known as probing. A number of built-in containers views do that too, e.g. VStack and HStack.

Nesting in different container views

The relative sizing modifier interacts in an fascinating means with stack views and different containers that distribute the obtainable area amongst their youngsters. I believed this was such an fascinating matter that I wrote a separate article about it: How the relative dimension modifier interacts with stack views.

The code

The entire code is out there in a Gist on GitHub.

Digression: Proportional sizing in early SwiftUI betas

The very first SwiftUI betas in 2019 did embrace proportional sizing modifiers, however they have been taken out earlier than the ultimate launch. Chris Eidhof preserved a duplicate of SwiftUI’s “header file” from that point that exhibits their API, together with fairly prolonged documentation.

I don’t know why these modifiers didn’t survive the beta part. The discharge notes from 2019 don’t give a cause:

The relativeWidth(_:), relativeHeight(_:), and relativeSize(width:top:) modifiers are deprecated. Use different modifiers like body(minWidth:idealWidth:maxWidth:minHeight:idealHeight:maxHeight:alignment:) as an alternative. (51494692)

I additionally don’t bear in mind how these modifiers labored. They most likely had considerably related semantics to my answer, however I can’t ensure. The doc feedback linked above sound simple (“Units the width of this view to the desired proportion of its dad or mum’s width.”), however they don’t point out the intricacies of the structure algorithm (proposals and responses) in any respect.

containerRelativeFrame