[Merged by Bors] - depend on dioxus(and bevy)-maintained fork of stretch (taffy)

[colepoirier]

Objective

DioxusLabs and Bevy have taken over maintaining what was our abandoned ui layout dependency stretch. Dioxus' fork has had a lot of work done on it by @alice-i-cecile, @Weibye , @jkelleyrtp, @mockersf, @HackerFoo, @TimJentzsch and a dozen other contributors and now is in much better shape than stretch was. The updated crate is called taffy and is available on github here (taffy on crates.io). The goal of this PR is to replace stretch v0.3.2 with taffy v0.1.0.

Solution

I changed the bevy_ui Cargo.toml to depend on taffy instead of stretch and fixed all the errors rustc complained about.

---

Changelog

Changed bevy_ui layout dependency from stretch to taffy (the maintained fork of stretch).

fixes #677

Migration Guide

The public api of taffy is different from that of stretch so please advise me on what to do here @alice-i-cecile.

[colepoirier]

Bevy UI Examples Screenshots

Here are the screenshots for each of the four bevy ui examples showing that the layout is broken by the changes in this PR.

text_debug

main

this PR

text

main

this PR

button

main

this PR

ui

main

this PR

[alice-i-cecile]

Those breaks are pretty rough :( We're going to need a much better test suite to fix this.

[rparrett]

It would be really helpful if Dioxus maintained a changelog. It seems like the "24 commits ahead" are:

• Jk/nested fix DioxusLabs/taffy#1
  • Renaming the project and related commits
  • Fix Percent height values, which were using parent width. vislyhq/stretch#88
  • Fix performance issue with deep nesting vislyhq/stretch#81
• fix re-computing layout moves children of non-zero positioned parent DioxusLabs/taffy#2

So just a couple of layout bug fixes and davier's partial perf fix so far.

Are bevy's examples relying on buggy layout behavior, or are these fixes introducing new bugs? Or has the migration to stretch2 gone awry in some way?

[colepoirier]

It would be really helpful if Dioxus maintained a changelog. It seems like the "24 commits ahead" are:

* Renaming the project

* [Jk/nested fix DioxusLabs/stretch#1](https://github.com/DioxusLabs/stretch/pull/1)

* [fix re-computing layout moves children of non-zero positioned parent DioxusLabs/stretch#2](https://github.com/DioxusLabs/stretch/pull/2)

So just a couple of layout bug fixes so far.

Are bevy's examples relying on buggy layout behavior, or are these fixes introducing new bugs? Or has the migration to stretch2 gone awry in some way?

All excellent points @rparrett!

[colepoirier]

Jk/nested fix DioxusLabs/stretch#1 includes 17 commits FYI

[rparrett]

Jk/nested fix DioxusLabs/stretch#1 includes 17 commits FYI

Thanks, updated summary in my comment.

[Sheepyhead]

I tested this PR with my most complicated piece of bevy_ui code; an inventory and equipment half-window. Here is what it looks like on Bevy main at time of writing:

And the following is what it looks like when I depend on this PR's changes:

The code has been extracted from a game I've been working on and is largely unchanged except for the sake of brevity and cutting out all game-specific mechanics, you can see the repo here https://github.com/Sheepyhead/bevy_ui_test

[mockersf]

Stretch used to give relative layout for nodes, now it's the absolute layout. I believe it's an issue on Stretch side and opened DioxusLabs/taffy#6, but it's not complicated to fix on Bevy side if they want to stick to the new behaviour.

[colepoirier]

Stretch used to give relative layout for nodes, now it's the absolute layout. I believe it's an issue on Stretch side and opened DioxusLabs/stretch#6, but it's not complicated to fix on Bevy side if they want to stick to the new behaviour.

Tested the bevy ui and button examples using François' PR fixing relative positioning and it looks like it fixes it.

ui

button

[Sheepyhead]

I can confirm that using Francois' latest changes on this PR completely fixes the layout of my complicated gui

[alice-i-cecile]

Excellent; I'll see about cutting a release of stretch2 ASAP then.

[colepoirier]

The two UI examples look correct using stretch2 v0.4.3.

ui

button

@mockersf Are there any further steps I need to take with this PR?

Edit: It looks like the check-bans CI is still failing on something to do with the lockfile that I don't understand. Do you know what is happening here and if I need to do something to resolve it?

[mockersf]

@mockersf Are there any further steps I need to take with this PR?

I think @alice-i-cecile intend to release the crate renamed soon, maybe wait for it

[mockersf]

Edit: It looks like the check-bans CI is still failing on something to do with the lockfile that I don't understand. Do you know what is happening here and if I need to do something to resolve it?

It seems the new crate has a few dependencies in multiple versions which we don't allow in Bevy. You can investigate on the dependencies if the duplicates are really needed, then open PRs on the dependencies to update if possible... It can take a lot of time

[alice-i-cecile]

You can investigate on the dependencies if the duplicates are really needed, then open PRs on the dependencies to update if possible... It can take a lot of time

I will happily accept any PRs to do this for sprawl. I'll bump the version of the dependencies where I can for the initial release of sprawl as well.

[Weibye]

It seems the new crate has a few dependencies in multiple versions which we don't allow in Bevy. You can investigate on the dependencies if the duplicates are really needed, then open PRs on the dependencies to update if possible... It can take a lot of time

Seems all dependency issues start with heapless (direct dependency in sprawl) and it's dependency tree

[colepoirier]

It seems the new crate has a few dependencies in multiple versions which we don't allow in Bevy. You can investigate on the dependencies if the duplicates are really needed, then open PRs on the dependencies to update if possible... It can take a lot of time

Seems all dependency issues start with heapless (direct dependency in sprawl) and it's dependency tree

I'm not sure how to fix this. Do you have any suggestions?

[alice-i-cecile]

@colepoirier can you make a PR to sprawl removing heapless? I don't think it's pulling its weight, and I would like to simplify that section of the code anyways.

[colepoirier]

@colepoirier can you make a PR to sprawl removing heapless? I don't think it's pulling its weight, and I would like to simplify that section of the code anyways.

@alice-i-cecile switched from stretch2 v0.4.3 to sprawl git main to check if that resolves the CI check-bans failures. Regardless I will attempt to remove heapless from sprawl and make a PR. Additionally, the sprawl v 0.1.0 release seems to be missing the vast majority of the contents of the library so will need a v 0.2.0 before we can take a dependency on it for bevy.

[alice-i-cecile]

@colepoirier the PR title and PR description need to be updated too.

[colepoirier]

@colepoirier the PR title and PR description need to be updated too.

I rewrote it thanks for the heads up! @alice-i-cecile what should I do about the removal of Direction from taffy in bevy_ui? What I did is just delete the code that used to be impl From<Direction> for stretch::style::Direction. Is that the right thing to do? Is Direction something that will need to be added back to bevy_ui after it is re-implemented in taffy?

[alice-i-cecile]

Direction in this sense refers to text direction. IMO it's out of scope for a layout library. Bevy should implement its own type when it needs this.

[alice-i-cecile]

That was about as painless as I could have hoped. @colepoirier have you verified that the UI examples all look okay with the newest changes?

[colepoirier]

That was about as painless as I could have hoped. @colepoirier have you verified that the UI examples all look okay with the newest changes?

Yes both look identical to the screenshots above for the bevy_ui on main.

ui

button

[colepoirier]

FYI I also removed the section for stretch in the deny.toml. I think that lies within the scope of this PR but I'd like confirmation. Please let me know if that is correct.

[TimJentzsch]

Direction in this sense refers to text direction. IMO it's out of scope for a layout library. Bevy should implement its own type when it needs this.

Since the flexbox algorithm depends on writing direction to determine start and end we will probably need to add a way to provide this again.

[mockersf]

just tried a complex case that was running at 45 fps with stretch, I get 160 fps with this PR 👍

[cart]

Looks good to me! The Bevy side was an easy review. Couldn't find any regressions. Did a taffy/stretch diff review and the changes look good to me.

I'm excited for our new future of collaborative UI layout dev!

[cart]

bors r+

[bors]

Pull request successfully merged into main.

Build succeeded:

• build-and-install-on-iOS
• build-android
• build (nightly, ubuntu-latest)
• build (stable, macos-latest)
• build (stable, ubuntu-latest)
• build (stable, windows-latest)
• build-wasm (nightly, ubuntu-latest)
• build-wasm (stable, ubuntu-latest)
• check-compiles
• check-doc
• check-missing-examples-in-docs
• check-unused-dependencies
• ci
• markdownlint
• miri
• run-examples
• run-examples-on-wasm
• run-examples-on-windows