How the relative dimension modifier interacts with stack views – Ole Begemann

I’ve another factor to say on the relative sizing view modifier from my earlier put up, Working with percentages in SwiftUI structure. I’m assuming you’ve learn that article. The next is nice to know if you wish to use the modifier in your personal code, however I hope you’ll additionally be taught some basic tidbits about SwiftUI’s structure algorithm for HStacks and VStacks.

Utilizing relative sizing inside a stack view

Let’s apply the relativeProposed modifier to one of many subviews of an HStack:

HStack(spacing: 10) {
    Colour.blue
        .relativeProposed(width: 0.5)
    Colour.inexperienced
    Colour.yellow
}
.border(.major)
.body(peak: 80)

What do you count on to occur right here? Will the blue view take up 50 % of the out there width? The reply isn’t any. In reality, the blue rectangle turns into narrower than the others:

It’s because the HStack solely proposes a proportion of its out there width to every of its kids. Right here, the stack proposes one third of the out there area to its first youngster, the relative sizing modifier. The modifier then halves this worth, leading to one sixth of the overall width (minus spacing) for the blue shade. The opposite two rectangles then grow to be wider than one third as a result of the primary youngster view didn’t burn up its full proposed width.

Order issues

Now let’s transfer the modifier to the inexperienced shade within the center:

HStack(spacing: 10) {
    Colour.blue
    Colour.inexperienced
        .relativeProposed(width: 0.5)
    Colour.yellow
}

Naively, I’d count on an equal end result: the inexperienced rectangle ought to grow to be 100 pt huge, and blue and yellow must be 250 pt every. However that’s not what occurs — the yellow view finally ends up being wider than the blue one:

I discovered this unintuitive at first, however it is sensible in case you perceive that the HStack processes its kids in sequence:

1. The HStack proposes one third of its out there area to the blue view: (620 – 20) / 3 = 200. The blue view accepts the proposal and turns into 200 pt huge.

2. Subsequent up is the relativeProposed modifier. The HStack divides the remaining area by the variety of remaining subviews and proposes that: 400 / 2 = 200. Our modifier halves this proposal and proposes 100 pt to the inexperienced view, which accepts it. The modifier in flip adopts the scale of its youngster and returns 100 pt to the HStack.

3. For the reason that second subview used much less area than proposed, the HStack now has 300 pt left over to suggest to its closing youngster, the yellow shade.

Necessary: the order wherein the stack lays out its subviews occurs to be from left to proper on this instance, however that’s not all the time the case. Basically, HStacks and VStacks first group their subviews by structure precedence (extra on that under), after which order the views inside every group by flexibility such that the least versatile views are laid out first. For extra on this, see How an HStack Lays out Its Kids by Chris Eidhof. The views in our instance are all equally versatile (all of them can grow to be any width between 0 and infinity), so the stack processes them of their “pure” order.

Leftover area isn’t redistributed

By now you could have the opportunity guess how the structure seems after we transfer our view modifier to the final youngster view:

HStack(spacing: 10) {
    Colour.blue
    Colour.inexperienced
    Colour.yellow
        .relativeProposed(width: 0.5)
}

• Blue and inexperienced every obtain one third of the out there width and grow to be 200 pt huge. No surprises there.

• When the HStack reaches the relativeProposed modifier, it has 200 pt left to distribute. Once more, the modifier and the yellow rectangle solely use half of this quantity.

The top result’s that the HStack finally ends up with 100 pt left over. The method stops right here — the HStack does not begin over in an try and discover a “higher” resolution. The stack makes itself simply large enough to comprise its subviews (= 520 pt incl. spacing) and studies that dimension to its dad or mum.

Structure precedence

We will use the layoutPriority view modifier to affect how stacks and different containers lay out their kids. Let’s give the subview with the relative sizing modifier the next structure precedence (the default precedence is 0):

HStack(spacing: 10) {
    Colour.blue
    Colour.inexperienced
    Colour.yellow
        .relativeProposed(width: 0.5)
        .layoutPriority(1)
}

This ends in a structure the place the yellow rectangle really takes up 50 % of the out there area:

Rationalization:

1. The HStack teams its kids by structure precedence after which processes every group in sequence, from highest to lowest precedence. Every group is proposed the complete remaining area.

2. The primary structure group solely incorporates a single view, our relative sizing modifier with the yellow shade. The HStack proposes the complete out there area (minus spacing) = 600 pt. Our modifier halves the proposal, leading to 300 pt for the yellow view.

3. There are 300 pt left over for the second structure group. These are distributed equally among the many two kids as a result of every subview accepts the proposed dimension.

Conclusion

The code I used to generate the photographs on this article is out there on GitHub. I solely checked out HStacks right here, however VStacks work in precisely the identical means for the vertical dimension.

SwiftUI’s structure algorithm all the time follows this fundamental sample of proposed sizes and responses. Every of the built-in “primitive” views (e.g. mounted and versatile frames, stacks, Textual content, Picture, Spacer, shapes, padding, background, overlay) has a well-defined (if not all the time well-documented) structure conduct that may be expressed as a operate (ProposedViewSize) -> CGSize. You’ll must be taught the conduct for view to work successfully with SwiftUI.

A concrete lesson I’m taking away from this evaluation: HStack and VStack don’t deal with structure as an optimization drawback that tries to seek out the optimum resolution for a set of constraints (autolayout model). Somewhat, they type their kids in a selected means after which do a single proposal-and-response go over them. If there’s area leftover on the finish, or if the out there area isn’t sufficient, then so be it.