Alright, it's Monday, and that means a #BevyEngine merge train!
Let's take a look at the 10 relevant PRs: https://github.com/bevyengine/bevy/pulls?q=is%3Aopen+is%3Apr+label%3AS-Ready-For-Final-Review+-label%3AS-Controversial+-label%3AS-Blocked+-label%3AS-Adopt-Me+
1. https://github.com/bevyengine/bevy/pull/4775
Parallelized transform propagation! Great example of targeted, profile-guided optimization based on the challenges seen in practice.
Initially testing looks good. Seeing modest performance improvements on large, scenes, and nothing noticeable on small scenes. Needs to be rebased to provide a fair comparison due to our other changes, but even still it's faster!
Gut feel on change detection worked out, but didn't seem worth it in more realistic cases.
bors r+
2. https://github.com/bevyengine/bevy/pull/5214
#RustLang's `ExactSizeIterator` trait didn't work how we expected. The size hint is for "remaining items", not total items.
That's been better documented upstream now (thanks!) https://github.com/rust-lang/rust/issues/98721.
Now, this PR fixes our mistake.
bors r+
Objective Fix #5149 Solution Instead of returning the total count of elements in the QueryIter in size_hint, we return the count of remaining elements. This Fixes #5149 even when #5148 gets merged....
3. https://github.com/bevyengine/bevy/pull/6524
When we were panicking on a thread, it was leaking upwards, and breaking other threads. This resulted in some *really* frustrating flaky tests in CI with #[should_panic] annotations.
It was likely even more damaging in actual use cases, as it prevents graceful failure.
This PR handles this more gracefully, by catching the unwind (and just dropping the error). Annoying that there's a small perf regression, but correct code is better than fast code.
bors r+
4. https://github.com/bevyengine/bevy/pull/6535
Ah, flexbox, my old nemesis. `Margin::Auto` is used consistently throughout our examples as a way to center items, but it's not very robust and is an old pre-flexbox #css hack.
Instead, actually set the alignment and justification of your parent.
Annoying that this is more verbose, but this is much clearer for beginners to learn from (and tweak). Time to rebase https://github.com/bevyengine/bevy/pull/5513...
bors r+
5. https://github.com/bevyengine/bevy/pull/6547
Another PR by James, another sizable performance win. This one splits apart our change detection ticks (u32s used to track what's been modified) into added / changed, rather than a joint struct.
This means that in the tight tight loops, we're iterating over less data, and get better cache efficiency. Huge wins for small components mutably iterated.
bors r+
6. https://github.com/bevyengine/bevy/pull/6664
Expose some more functionality from winit: passing cursor input through to the covered app.
With this + transparent windows I think that we're well overdue for Bonsai Buddy: Bevy Edition.
bors r+
7. https://github.com/bevyengine/bevy/pull/6672
Create a box from two corners! Perfectly useful. I agree with James's nit, but totally not worth blocking the PR on.
bors r+
8. https://github.com/bevyengine/bevy/pull/6692
Another helper method for Quaternion rotation! Should be useful when working with characters.
bors r+
9. https://github.com/bevyengine/bevy/pull/6676
Track the changed frames in CorePlugin, not RenderingPlugin, to be nicer to custom renderers, turn-based games and more. I'll actually use this myself in https://github.com/Leafwing-Studios/leafwing_input_playback
bors r+
Objective Latest Release, "bevy 0.9" move the FrameCount updater into RenderPlugin, it leads to user who only run app with Core/Minimal Plugin cannot get the right number of FrameCount, it always r...
10. https://github.com/bevyengine/bevy/pull/6699
When rendering, components are selectively extracted to the render world (only rendering-relevant data should be moved).
To do so, we use a nice `ExtractComponent` plugin. But when the value we want to extract, and the original component aren't the same thing, this results in wasted work and pointless boilerplate.
This PR uses the magic of associated types to speed that up.
Boo for no default associated types yet! Still, way better.
bors r+
Objective Allow more use cases where the user may benefit from both ExtractComponentPlugin and UniformComponentPlugin. Solution Add an associated type to ExtractComponent in order to allow specifyi...
That's all I have for you today: help us out by reviewing PRs: https://github.com/bevyengine/bevy/pulls?q=is%3Aopen+is%3Apr+-label%3AS-Ready-For-Final-Review+-draft%3A%3Atrue+-label%3AS-Needs-RFC+-reviewed-by%3A%40me+-author%3A%40me
Yes, that means you, dear reader. Your comments and approvals matter, and are what helps @bevy scale and move quickly as a volunteer #foss organization :)
Even if you're new to #rustlang or #gamedev, it's a great way to learn, and comments of "I don't understand this", "why are you making this change" and "needs more tests" really do improve PRs.
Yes! These days the progress and development momentum of #bevyengine is often not bottlenecked on maintainers, but rather on #codereview .
Reviewing PRs is a very effective way to help the project move forward, and anyone can do it. Don't let impostor syndrome stop you! Go review things and leave comments if you have anything to say. If you think you aren't compenent, just make that clear in your comments! Your review is still helpful!
Yes. It is up to the maintainers to make the judgement about when PRs have had enough review.
You can also ask questions in Discord, if you just wanna learn about something. But even on GH, it might lead to interesting discussion. I guess, use your judgement.
Alice I Cecile
Tim
IyesGames
Brian Kung