### Builder changes
- Increased meshlet max vertices/triangles from 64v/64t to 255v/128t
(meshoptimizer won't allow 256v sadly). This gives us a much greater
percentage of meshlets with max triangle count (128). Still not perfect,
we still end up with some tiny <=10 triangle meshlets that never really
get simplified, but it's progress.
- Removed the error target limit. Now we allow meshoptimizer to simplify
as much as possible. No reason to cap this out, as the cluster culling
code will choose a good LOD level anyways. Again leads to higher quality
LOD trees.
- After some discussion and consulting the Nanite slides again, changed
meshlet group error from _adding_ the max child's error to the group
error, to doing `group_error = max(group_error, max_child_error)`. Error
is already cumulative between LODs as the edges we're collapsing during
simplification get longer each time.
- Bumped the 65% simplification threshold to allow up to 95% of the
original geometry (e.g. accept simplification as valid even if we only
simplified 5% of the triangles). This gives us closer to
log2(initial_meshlet_count) LOD levels, and fewer meshlet roots in the
DAG.
Still more work to be done in the future here. Maybe trying METIS for
meshlet building instead of meshoptimizer.
Using ~8 clusters per group instead of ~4 might also make a big
difference. The Nanite slides say that they have 8-32 meshlets per
group, suggesting some kind of heuristic. Unfortunately meshopt's
compute_cluster_bounds won't work with large groups atm
(https://github.com/zeux/meshoptimizer/discussions/750#discussioncomment-10562641)
so hard to test.
Based on discussion from
https://github.com/bevyengine/bevy/discussions/14998,
https://github.com/zeux/meshoptimizer/discussions/750, and discord.
### Runtime changes
- cluster:triangle packed IDs are now stored 25:7 instead of 26:6 bits,
as max triangles per cluster are now 128 instead of 64
- Hardware raster now spawns 128 * 3 vertices instead of 64 * 3 vertices
to account for the new max triangles limit
- Hardware raster now outputs NaN triangles (0 / 0) instead of
zero-positioned triangles for extra vertex invocations over the cluster
triangle count. Shouldn't really be a difference idt, but I did it
anyways.
- Software raster now does 128 threads per workgroup instead of 64
threads. Each thread now loads, projects, and caches a vertex (vertices
0-127), and then if needed does so again (vertices 128-254). Each thread
then rasterizes one of 128 triangles.
- Fixed a bug with `needs_dispatch_remap`. I had the condition backwards
in my last PR, I probably committed it by accident after testing the
non-default code path on my GPU.
# Objective
A previous issue describes the same problem: #14248.
This particular link was seemingly missed by #14276.
## Solution
- Search repo for `bevyengine.org/learn/errors/#`
- Remove `#`
- Verify link goes to right place
# Objective
- Crate-level prelude modules, such as `bevy_ecs::prelude`, are plagued
with inconsistency! Let's fix it!
## Solution
Format all preludes based on the following rules:
1. All preludes should have brief documentation in the format of:
> The _name_ prelude.
>
> This includes the most common types in this crate, re-exported for
your convenience.
2. All documentation should be outer, not inner. (`///` instead of
`//!`.)
3. No prelude modules should be annotated with `#[doc(hidden)]`. (Items
within them may, though I'm not sure why this was done.)
## Testing
- I manually searched for the term `mod prelude` and updated all
occurrences by hand. 🫠
---------
Co-authored-by: Gino Valente <49806985+MrGVSV@users.noreply.github.com>
# Objective
- `CursorIcon` had derived `Reflect`, but it wasn't registered
## Solution
- Use `register_type` on it
- I also moved the cursor code to it's own plugin because there was
starting to be too much cursor code outside the cursor file.
## Testing
- window_settings example still works with the custom cursor
Commit 65252bb87 (Consistently use `PI` to specify angles in examples.
(#5825)) changed from using push_children to add_child without adjusting
the comment
# Objective
- Fix the comments to align with the code
Co-authored-by: Alexander Stein <alexander.stein@mailbox.org>
# Objective
#14957 added the `pick_rounded_rect` function to `bevy_ui` in the
`picking_backend` module, which is gated behind the `bevy_picking`
feature. This function is used in that module, as well as in the `focus`
module. The latter usage is not gated behind the `bevy_picking` feature,
causing a compile error when the feature is disabled.
## Solution
Move the `pick_rounded_rect` function out of the `picking_backend`
module, as it does not depend on anything defined in that module. I put
it in `lib.rs` but it could reasonably be moved somewhere else instead.
## Testing
Encountered this compile error in a project and confirmed that this
patch fixes it.
# Objective
Fixes#13479
This also fixes the gaps you can sometimes observe in outlines
(screenshot from main, not this PR):
<img width="636" alt="outline-gaps"
src="https://github.com/user-attachments/assets/c11dae24-20f5-4aea-8ffc-1894ad2a2b79">
The outline around the last item in each section has vertical gaps.
## Solution
Draw the outlines with corner radius using the existing border rendering
for uinodes. The outline radius is very simple to calculate. We just
take the computed border radius of the node, and if it's greater than
zero, add it to the distance from the edge of the node to the outer edge
of the node's outline.
---
## Showcase
<img width="634" alt="outlines-radius"
src="https://github.com/user-attachments/assets/1ecda26c-65c5-41ef-87e4-5d9171ddc3ae">
---------
Co-authored-by: Jan Hohenheim <jan@hohenheim.ch>
# Objective
Correctly order picking events. Resolves
https://github.com/bevyengine/bevy/issues/5984.
## Solution
Event ordering [very long standing
problem](https://github.com/aevyrie/bevy_mod_picking/issues/294) with
mod picking, stemming from two related issues. The first problem was
that `Pointer<T>` events of different types couldn't be ordered, but we
have already gotten around that in the upstream by switching to
observers. Since observers run in the order they are triggered, this
isn't an issue.
The second problem was that the underlying event streams that picking
uses to create it's pointer interaction events *also* lacked ordering,
and the systems that generated the points couldn't interleave events.
This PR fixes that by unifying the event streams and integrating the
various interaction systems.
The concrete changes are as follows:
+ `bevy_winit::WinitEvent` has been moved to `bevy_window::WindowEvent`.
This provides a unified (and more importantly, *ordered*) input stream
for both `bevy_window` and `bevy_input` events.
+ Replaces `InputMove` and `InputPress` with `PointerInput`, a new
unified input event which drives picking and interaction. This event is
built to have drop-in forward compatibility with [winit's upcoming
pointer abstraction](https://github.com/rust-windowing/winit/pull/3876).
I have added code to emulate it using the current winit input
abstractions, but this entire thing will be much more robust when it
lands.
+ Rolls `pointer_events` `send_click_and_drag_events` and
`send_drag_over_events` into a single system, which operates directly on
`PointerEvent` and triggers observers as output.
The PR also improves docs and takes the opportunity to
refactor/streamline the pointer event dispatch logic.
## Status & Testing
This PR is now feature complete and documented. While it is
theoretically possible to add unit tests for the ordering, building the
picking mocking for that will take a little while.
Feedback on the chosen ordering of events is within-scope.
## Migration Guide
For users switching from `bevy_mod_picking` to `bevy_picking`:
+ Instead of adding an `On<T>` component, use `.observe(|trigger:
Trigger<T>|)`. You may now apply multiple handlers to the same entity
using this command.
+ Pointer interaction events now have semi-deterministic ordering which
(more or less) aligns with the order of the raw input stream. Consult
the docs on `bevy_picking::event::pointer_events` for current
information. You may need to adjust your event handling logic
accordingly.
+ `PointerCancel` has been replaced with `Pointer<Cancled>`, which now
has the semantics of an OS touch pointer cancel event.
+ `InputMove` and `InputPress` have been merged into `PointerInput`. The
use remains exactly the same.
+ Picking interaction events are now only accessible through observers,
and no `EventReader`. This functionality may be re-implemented later.
For users of `bevy_winit`:
+ The event `bevy_winit::WinitEvent` has moved to
`bevy_window::WindowEvent`. If this was the only thing you depended on
`bevy_winit` for, you should switch your dependency to `bevy_window`.
+ `bevy_window` now depends on `bevy_input`. The dependencies of
`bevy_input` are a subset of the existing dependencies for `bevy_window`
so this should be non-breaking.
# Objective
Fixes#15032
## Solution
Reimplement support for the `flip_x` and `flip_y` fields.
This doesn't flip the border geometry, I'm not really sure whether that
is desirable or not.
Also fixes a bug that was causing the side and center slices to tile
incorrectly.
### Testing
```
cargo run --example ui_texture_slice_flip_and_tile
```
## Showcase
<img width="787" alt="nearest"
src="https://github.com/user-attachments/assets/bc044bae-1748-42ba-92b5-0500c87264f6">
With tiling need to use nearest filtering to avoid bleeding between the
slices.
---------
Co-authored-by: Jan Hohenheim <jan@hohenheim.ch>
Co-authored-by: Alice Cecile <alice.i.cecile@gmail.com>
# Objective
- follow of #14049 ,we could use it on our Parallel Iterator,this pr
also unified the used function in both regular iter and parallel
iterations.
## Performance
no performance regression for regular itertaion
3.5X faster in hybrid parallel iteraion,this number is far greater than
the benefits obtained in regular iteration(~1.81) because mutable
iterations on continuous memory can effectively reduce the cost of
mataining core cache coherence
# Objective
As discussed in https://github.com/bevyengine/bevy/issues/7386, system
order ambiguities within `DefaultPlugins` are a source of bugs in the
engine and badly pollute diagnostic output for users.
We should eliminate them!
This PR is an alternative to #15027: with all external ambiguities
silenced, this should be much less prone to merge conflicts and the test
output should be much easier for authors to understand.
Note that system order ambiguities are still permitted in the
`RenderApp`: these need a bit of thought in terms of how to test them,
and will be fairly involved to fix. While these aren't *good*, they'll
generally only cause graphical bugs, not logic ones.
## Solution
All remaining system order ambiguities have been resolved.
Review this PR commit-by-commit to see how each of these problems were
fixed.
## Testing
`cargo run --example ambiguity_detection` passes with no panics or
logging!
# Objective
- Fixes https://github.com/bevyengine/bevy/issues/14593.
## Solution
- Add `ViewportConversionError` and return it from viewport conversion
methods on Camera.
## Testing
- I successfully compiled and ran all changed examples.
## Migration Guide
The following methods on `Camera` now return a `Result` instead of an
`Option` so that they can provide more information about failures:
- `world_to_viewport`
- `world_to_viewport_with_depth`
- `viewport_to_world`
- `viewport_to_world_2d`
Call `.ok()` on the `Result` to turn it back into an `Option`, or handle
the `Result` directly.
---------
Co-authored-by: Lixou <82600264+DasLixou@users.noreply.github.com>
Co-authored-by: Alice Cecile <alice.i.cecile@gmail.com>
Co-authored-by: Zachary Harrold <zac@harrold.com.au>
# Objective
- Fixes#14966
## Solution
- Added a _Performance_ section to the documentation for
`LogPlugin::filter` explaining that long filter strings can degrade
performance and to instead rely on `LogPlugin::level` when possible.
## Testing
- CI passed locally.
# Objective
Replaced the existing instantiation of the 2D Circle in the 2d_shapes.rs
file with the `new` method.
- Ensures consistency in instantiating 2D primitive shapes in the
examples.
## Solution
Replaced the existing instantiation of the 2D Circle in the 2d_shapes.rs
file with the `new` method.
- Ensures consistency in instantiating 2D primitive shapes in the
examples.
## Testing
- None: Should be straight-forward enough to not warrant a test (I will
eat my words if I am wrong).
---
# Objective
- Currently we have the `ColorRange` trait to interpolate linearly
between two colors
- It would be cool to have:
1. linear interpolation between n colors where `n >= 1`
2. other kinds of interpolation
## Solution
1. Implement `ColorGradient` which takes `n >= 1` colors and linearly
interpolates between consecutive pairs of them
2. Implement `Curve` intergration for this `ColorGradient` which yields
a curve struct. After that we can apply all of the cool curve adaptors
like `.reparametrize()` and `.map()` to the gradient
## Testing
- Added doc tests
- Added tests
## Showcase
```rust
// let gradient = ColorGradient::new(vec![]).unwrap(); // panic! 💥
let gradient = ColorGradient::new([basic::RED, basic::LIME, basic::BLUE]).expect("non-empty");
let curve = gradient.to_curve();
let brighter_curve = curve.map(|c| c.mix(&basic::WHITE, 0.5));
```
---
Kind of related to
https://github.com/bevyengine/bevy/pull/14971#discussion_r1736337631
---------
Co-authored-by: Zachary Harrold <zac@harrold.com.au>
Co-authored-by: Matty <weatherleymatthew@gmail.com>
# Objective
Fixes https://github.com/bevyengine/bevy/issues/14183
## Solution
Reimplement the UI texture atlas slicer using a shader.
The problems with #14183 could be fixed more simply by hacking around
with the coordinates and scaling but that way is very fragile and might
get broken again the next time we make changes to the layout
calculations. A shader based solution is more robust, it's impossible
for gaps to appear between the image slices with these changes as we're
only drawing a single quad.
I've not tried any benchmarks yet but it should much more efficient as
well, in the worst cases even hundreds or thousands of times faster.
Maybe could have used the UiMaterialPipeline. I wrote the shader first
and used fat vertices and then realised it wouldn't work that way with a
UiMaterial. If it's rewritten it so it puts all the slice geometry in
uniform buffer, then it might work? Adding the uniform buffer would
probably make the shader more complicated though, so don't know if it's
even worth it. Instancing is another alternative.
## Testing
The examples are working and it seems to match the old API correctly but
I've not used the texture atlas slicing API for anything before, I
reviewed the PR but that was back in January.
Needs a review by someone who knows the rendering pipeline and wgsl
really well because I don't really have any idea what I'm doing.
# Objective
Minor cosmetic improvements and code cleanup for this example
## Solution
- Use the default font and shorter constructor where the the full font
is not needed
- Use 12px padding for instruction text to match other examples
- Clean up formatting of instruction text
- Fix one instance of a font handle not being reused
<details>
<summary>Before / After</summary>
*Question mark boxes are expected. We don't include a font with the
glyphs I'm testing the IME with*
<img width="1280" alt="Screenshot 2024-09-02 at 11 14 17 AM"
src="https://github.com/user-attachments/assets/e0a904a3-c01b-4c18-b71f-41fbc1d09169">
<img width="1280" alt="Screenshot 2024-09-02 at 11 13 14 AM"
src="https://github.com/user-attachments/assets/4b2a2d31-6b9f-4b9a-a245-4258a6aa9519">
</details>
## Testing
Tested all functionality of the example (macOS)
# Objective
Thought I had found all of these... noticed some `10px` in #15013 and
did another sweep.
Continuation of #8478, #13583.
## Solution
- Position example text (and other elements) 12px from the edge of the
screen
# Objective
The `BorderRadius` component shouldn't be required to draw borders for
nodes with sharp corners.
## Solution
Make `BorderRadius` optional in `extract_uinode_borders`'s UI node
query.
# Objective
The `Parent` component holds a reference to the parent entity of the
entity it is inserted onto. The `with_child` function erroneously
forgets to insert this component onto the child entity that it spawns,
causing buggy behaviour when the function is used instead of the other
child-spawning functions.
## Solution
Ensure `with_child` inserts the `Parent` component, the same as all the
other child-spawning functions.
## Testing
Checked before/after with a bevy_ui layout where this patch fixed buggy
behaviour I was seeing in parent/child UI nodes.
# Objective
- Fixes#14969
## Solution
- Added `Deserialize` to the list of reflected traits for `SmolStr`
## Testing
- CI passed locally.
---------
Co-authored-by: Alice Cecile <alice.i.cecile@gmail.com>
This commit adds support for *masks* to the animation graph. A mask is a
set of animation targets (bones) that neither a node nor its descendants
are allowed to animate. Animation targets can be assigned one or more
*mask group*s, which are specific to a single graph. If a node masks out
any mask group that an animation target belongs to, animation curves for
that target will be ignored during evaluation.
The canonical use case for masks is to support characters holding
objects. Typically, character animations will contain hand animations in
the case that the character's hand is empty. (For example, running
animations may close a character's fingers into a fist.) However, when
the character is holding an object, the animation must be altered so
that the hand grips the object.
Bevy currently has no convenient way to handle this. The only workaround
that I can see is to have entirely separate animation clips for
characters' hands and bodies and keep them in sync, which is burdensome
and doesn't match artists' expectations from other engines, which all
effectively have support for masks. However, with mask group support,
this task is simple. We assign each hand to a mask group and parent all
character animations to a node. When a character grasps an object in
hand, we position the fingers as appropriate and then enable the mask
group for that hand in that node. This allows the character's animations
to run normally, while the object remains correctly attached to the
hand.
Note that even with this PR, we won't have support for running separate
animations for a character's hand and the rest of the character. This is
because we're missing additive blending: there's no way to combine the
two masked animations together properly. I intend that to be a follow-up
PR.
The major engines all have support for masks, though the workflow varies
from engine to engine:
* Unity has support for masks [essentially as implemented here], though
with layers instead of a tree. However, when using the Mecanim
("Humanoid") feature, precise control over bones is lost in favor of
predefined muscle groups.
* Unreal has a feature named [*layered blend per bone*]. This allows for
separate blend weights for different bones, effectively achieving masks.
I believe that the combination of blend nodes and masks make Bevy's
animation graph as expressible as that of Unreal, once we have support
for additive blending, though you may have to use more nodes than you
would in Unreal. Moreover, separating out the concepts of "blend weight"
and "which bones this node applies to" seems like a cleaner design than
what Unreal has.
* Godot's `AnimationTree` has the notion of [*blend filters*], which are
essentially the same as masks as implemented in this PR.
Additionally, this patch fixes a bug with weight evaluation whereby
weights weren't properly propagated down to grandchildren, because the
weight evaluation for a node only checked its parent's weight, not its
evaluated weight. I considered submitting this as a separate PR, but
given that this PR refactors that code entirely to support masks and
weights under a unified "evaluated node" concept, I simply included the
fix here.
A new example, `animation_masks`, has been added. It demonstrates how to
toggle masks on and off for specific portions of a skin.
This is part of #14395, but I'm going to defer closing that issue until
we have additive blending.
[essentially as implemented here]:
https://docs.unity3d.com/560/Documentation/Manual/class-AvatarMask.html
[*layered blend per bone*]:
https://dev.epicgames.com/documentation/en-us/unreal-engine/using-layered-animations-in-unreal-engine
[*blend filters*]:
https://docs.godotengine.org/en/stable/tutorials/animation/animation_tree.html
## Migration Guide
* The serialized format of animation graphs has changed with the
addition of animation masks. To upgrade animation graph RON files, add
`mask` and `mask_groups` fields as appropriate. (They can be safely set
to zero.)
# Objective
`resolve_outlines_system` wasn't updated when multi-window support was
added and it always uses the size of the primary window when resolving
viewport coords, regardless of the layout's camera target.
Fixes#14945
## Solution
It's awkward to get the viewport size of the target for an individual
node without walking the tree or adding extra fields to `Node`, so I
removed `resolve_outlines_system` and instead the outline values are
updated in `ui_layout_system`.
---------
Co-authored-by: Alice Cecile <alice.i.cecile@gmail.com>
# Objective
Make the documentation for `SystemParamBuilder` nicer by combining the
tuple implementations into a single line of documentation.
## Solution
Use `#[doc(fake_variadic)]` for `SystemParamBuilder` tuple impls.
(This got missed originally because #14050 and #14703 were open at the
same time.)
Adds a new `Handle<Storage>` asset type that can be used as a render
asset, particularly for use with `AsBindGroup`.
Closes: #13658
# Objective
Allow users to create storage buffers in the main world without having
to access the `RenderDevice`. While this resource is technically
available, it's bad form to use in the main world and requires mixing
rendering details with main world code. Additionally, this makes storage
buffers easier to use with `AsBindGroup`, particularly in the following
scenarios:
- Sharing the same buffers between a compute stage and material shader.
We already have examples of this for storage textures (see game of life
example) and these changes allow a similar pattern to be used with
storage buffers.
- Preventing repeated gpu upload (see the previous easier to use `Vec`
`AsBindGroup` option).
- Allow initializing custom materials using `Default`. Previously, the
lack of a `Default` implement for the raw `wgpu::Buffer` type made
implementing a `AsBindGroup + Default` bound difficult in the presence
of buffers.
## Solution
Adds a new `Handle<Storage>` asset type that is prepared into a
`GpuStorageBuffer` render asset. This asset can either be initialized
with a `Vec<u8>` of properly aligned data or with a size hint. Users can
modify the underlying `wgpu::BufferDescriptor` to provide additional
usage flags.
## Migration Guide
The `AsBindGroup` `storage` attribute has been modified to reference the
new `Handle<Storage>` asset instead. Usages of Vec` should be converted
into assets instead.
---------
Co-authored-by: IceSentry <IceSentry@users.noreply.github.com>
# Objective
- Avoid cloning the `CosmicBuffer` every time you create a new text
measurement.
## Solution
- Inject a buffer query when calculating layout so existing buffers can
be reused.
## Testing
- I tested the `text`, `text_debug`, and `text_wrap_debug` examples.
- I did not do a performance test.
Fixes#14993 (maybe). Adds a system ordering constraint that was missed
in the refactor in #14833. The theory here is that the single threaded
forces a topology that causes the prepare system to run before
`prepare_windows` in a way that causes issues. For whatever reason, this
appears to be unlikely when multi-threading is enabled.
# Objective
* Fixes https://github.com/bevyengine/bevy/issues/14889
## Solution
Exposes `bevy_animation::{animatable, graph, transition}` to the world.
## Testing
- Did you test these changes? If so, how?
- These changes do not need testing, as they do not modify/add/remove
any functionality.
- ~~Are there any parts that need more testing?~~
- ~~How can other people (reviewers) test your changes? Is there
anything specific they need to know?~~
- ~~If relevant, what platforms did you test these changes on, and are
there any important ones you can't test?~~
---------
Co-authored-by: Alice Cecile <alice.i.cecile@gmail.com>
# Objective
The `reflect` module in `bevy_state` is gated behind the `bevy_reflect`
feature, but the type exports from that module in the crate prelude are
erroneously gated behind the `bevy_app` feature, causing a compile error
when the `bevy_reflect` feature is disabled, but the `bevy_app` feature
is enabled.
## Solution
Change the feature gate to `bevy_reflect`.
## Testing
- Discovered by depending on `bevy_state` with `default-features =
false, features = ["bevy_app"]`
- Tested by running `cargo check -p bevy_state --no-default-features
--features bevy_app`
# Objective
- Fixes#14974
## Solution
- Replace all* instances of `NonZero*` with `NonZero<*>`
## Testing
- CI passed locally.
---
## Notes
Within the `bevy_reflect` implementations for `std` types,
`impl_reflect_value!()` will continue to use the type aliases instead,
as it inappropriately parses the concrete type parameter as a generic
argument. If the `ZeroablePrimitive` trait was stable, or the macro
could be modified to accept a finite list of types, then we could fully
migrate.
