# Objective
- The previous implementation of automatically inserting sync points did
not consider explicitly added sync points. This created additional sync
points. For example:
```
A-B
C-D-E
```
If `A` and `B` needed a sync point, and `D` was an `ApplyDeferred`, an
additional sync point would be generated between `A` and `B`.
```
A-D2-B
C-D -E
```
This can result in the following system ordering:
```
A-D2-(B-C)-D-E
```
Where only `B` and `C` run in parallel. If we could reuse `D` as the
sync point, we would get the following ordering:
```
(A-C)-D-(B-E)
```
Now we have two more opportunities for parallelism!
## Solution
- In the first pass, we:
- Compute the number of sync points before each node
- This was already happening but now we consider `ApplyDeferred` nodes
as creating a sync point.
- Pick an arbitrary explicit `ApplyDeferred` node for each "sync point
index" that we can (some required sync points may be missing!)
- In the second pass, we:
- For each edge, if two nodes have a different number of sync points
before them then there must be a sync point between them.
- Look for an explicit `ApplyDeferred`. If one exists, use it as the
sync point.
- Otherwise, generate a new sync point.
I believe this should also gracefully handle changes to the
`ScheduleGraph`. Since automatically inserted sync points are inserted
as systems, they won't look any different to explicit sync points, so
they are also candidates for "reusing" sync points.
One thing this solution does not handle is "deduping" sync points. If
you add 10 sync points explicitly, there will be at least 10 sync
points. You could keep track of all the sync points at the same
"distance" and then hack apart the graph to dedup those, but that could
be a follow-up step (and it's more complicated since you have to worry
about transferring edges between nodes).
## Testing
- Added a test to test the feature.
- The existing tests from all our crates still pass.
## Showcase
- Automatically inserted sync points can now reuse explicitly inserted
`ApplyDeferred` systems! Previously, Bevy would add new sync points
between systems, ignoring the explicitly added sync points. This would
reduce parallelism of systems in some situations. Now, the parallelism
has been improved!
# Objective
- Closes#12944.
## Solution
- Load `R8G8B8` textures by transcoding to an rgba format since `wgpu`
does not support texture formats with 3 channels.
- Switch to erroring out instead of panicking on an invalid dds file.
---
## Changelog
### Added
- DDS Textures with the `R8G8B8` format are now supported. They require
an additional conversion step, so using `R8G8B8A8` or a similar format
is preferable for texture loading performance.
# Objective
- Fixes#17960
## Solution
- Followed the [edition upgrade
guide](https://doc.rust-lang.org/edition-guide/editions/transitioning-an-existing-project-to-a-new-edition.html)
## Testing
- CI
---
## Summary of Changes
### Documentation Indentation
When using lists in documentation, proper indentation is now linted for.
This means subsequent lines within the same list item must start at the
same indentation level as the item.
```rust
/* Valid */
/// - Item 1
/// Run-on sentence.
/// - Item 2
struct Foo;
/* Invalid */
/// - Item 1
/// Run-on sentence.
/// - Item 2
struct Foo;
```
### Implicit `!` to `()` Conversion
`!` (the never return type, returned by `panic!`, etc.) no longer
implicitly converts to `()`. This is particularly painful for systems
with `todo!` or `panic!` statements, as they will no longer be functions
returning `()` (or `Result<()>`), making them invalid systems for
functions like `add_systems`. The ideal fix would be to accept functions
returning `!` (or rather, _not_ returning), but this is blocked on the
[stabilisation of the `!` type
itself](https://doc.rust-lang.org/std/primitive.never.html), which is
not done.
The "simple" fix would be to add an explicit `-> ()` to system
signatures (e.g., `|| { todo!() }` becomes `|| -> () { todo!() }`).
However, this is _also_ banned, as there is an existing lint which (IMO,
incorrectly) marks this as an unnecessary annotation.
So, the "fix" (read: workaround) is to put these kinds of `|| -> ! { ...
}` closuers into variables and give the variable an explicit type (e.g.,
`fn()`).
```rust
// Valid
let system: fn() = || todo!("Not implemented yet!");
app.add_systems(..., system);
// Invalid
app.add_systems(..., || todo!("Not implemented yet!"));
```
### Temporary Variable Lifetimes
The order in which temporary variables are dropped has changed. The
simple fix here is _usually_ to just assign temporaries to a named
variable before use.
### `gen` is a keyword
We can no longer use the name `gen` as it is reserved for a future
generator syntax. This involved replacing uses of the name `gen` with
`r#gen` (the raw-identifier syntax).
### Formatting has changed
Use statements have had the order of imports changed, causing a
substantial +/-3,000 diff when applied. For now, I have opted-out of
this change by amending `rustfmt.toml`
```toml
style_edition = "2021"
```
This preserves the original formatting for now, reducing the size of
this PR. It would be a simple followup to update this to 2024 and run
`cargo fmt`.
### New `use<>` Opt-Out Syntax
Lifetimes are now implicitly included in RPIT types. There was a handful
of instances where it needed to be added to satisfy the borrow checker,
but there may be more cases where it _should_ be added to avoid
breakages in user code.
### `MyUnitStruct { .. }` is an invalid pattern
Previously, you could match against unit structs (and unit enum
variants) with a `{ .. }` destructuring. This is no longer valid.
### Pretty much every use of `ref` and `mut` are gone
Pattern binding has changed to the point where these terms are largely
unused now. They still serve a purpose, but it is far more niche now.
### `iter::repeat(...).take(...)` is bad
New lint recommends using the more explicit `iter::repeat_n(..., ...)`
instead.
## Migration Guide
The lifetimes of functions using return-position impl-trait (RPIT) are
likely _more_ conservative than they had been previously. If you
encounter lifetime issues with such a function, please create an issue
to investigate the addition of `+ use<...>`.
## Notes
- Check the individual commits for a clearer breakdown for what
_actually_ changed.
---------
Co-authored-by: François Mockers <francois.mockers@vleue.com>
# Objective
`Eq`/`PartialEq` are currently implemented for `MeshMaterial{2|3}d` only
through the derive macro. Since we don't have perfect derive yet, the
impls are only present for `M: Eq` and `M: PartialEq`. On the other
hand, I want to be able to compare material components for my toy
reactivity project.
## Solution
Switch to manual `Eq`/`PartialEq` impl.
## Testing
Boy I hope this didn't break anything!
# Objective
Fixes#17488
## Solution
The world update logic happened in the the `about_to_wait` winit window
callback, but this is is not correct as (1) the winit documentation
states that the callback should not be used for that purpose and (2) the
callback is not fired when the window is resized or being dragged.
However, that callback was used in #11245 to fix an iOS bug (which
caused the regression). The solution presented here is a workaround
until the event loop code can be re-written.
## Testing
I confirmed that the `eased_motion` example continued to be animated
when dragging or resizing the window.
https://github.com/user-attachments/assets/ffaf0abf-4cd7-479b-83e9-e1850aaf3513
- Remove references to the short-lived `CommandError` type.
- Add a sentence to the explanation of error handlers.
- Clean up spacing/linebreaks.
- Use `where` notation for command-related trait `impl`s to make the big
ones easier to parse.
Fixes#17856.
## Migration Guide
- `EventWriter::send` has been renamed to `EventWriter::write`.
- `EventWriter::send_batch` has been renamed to
`EventWriter::write_batch`.
- `EventWriter::send_default` has been renamed to
`EventWriter::write_default`.
---------
Co-authored-by: François Mockers <mockersf@gmail.com>
# Objective
- Make transform propagation faster.
## Solution
- Work sharing worker threads
- Parallel tree traversal excluding leaves
- Second cache friendly wide pass over all leaves
- 3-10x faster than main
## Testing
- Tracy
- Caldera hotel is showing 3-7x faster on my M4 Max. Timing for bevy's
existing transform system shifts wildly run to run, so I don't know that
I would advertise a particular number. But this implementation is faster
in a... statistically significant way.
---------
Co-authored-by: Alice Cecile <alice.i.cecile@gmail.com>
Co-authored-by: François Mockers <mockersf@gmail.com>
PR #17898 regressed this, causing much of #17970. This commit fixes the
issue by freeing and reallocating materials in the
`MaterialBindGroupAllocator` on change. Note that more efficiency is
possible, but I opted for the simple approach because (1) we should fix
this bug ASAP; (2) I'd like #17965 to land first, because that unlocks
the biggest potential optimization, which is not recreating the bind
group if it isn't necessary to do so.
We might not be able to prepare a material on the first frame we
encounter a mesh using it for various reasons, including that the
material hasn't been loaded yet or that preparing the material is
exceeding the per-frame cap on number of bytes to load. When this
happens, we currently try to find the material in the
`MaterialBindGroupAllocator`, fail, and then fall back to group 0, slot
0, the default `MaterialBindGroupId`, which is obviously incorrect.
Worse, we then fail to dirty the mesh and reextract it when we *do*
finish preparing the material, so the mesh will continue to be rendered
with an incorrect material.
This patch fixes both problems. In `collect_meshes_for_gpu_building`, if
we fail to find a mesh's material in the `MeshBindGroupAllocator`, then
we detect that case, bail out, and add it to a list,
`MeshesToReextractNextFrame`. On subsequent frames, we process all the
meshes in `MeshesToReextractNextFrame` as though they were changed. This
ensures both that we don't render a mesh if its material hasn't been
loaded and that we start rendering the mesh once its material does load.
This was first noticed in the intermittent Pixel Eagle failures in the
`testbed_3d` patch in #17898, although the problem has actually existed
for some time. I believe it just so happened that the changes to the
allocator in that PR caused the problem to appear more commonly than it
did before.
This patch fixes two bugs in the new non-bindless material allocator
that landed in PR #17898:
1. A debug assertion to prevent double frees had been flipped: we
checked to see whether the slot was empty before freeing, while we
should have checked to see whether the slot was full.
2. The non-bindless allocator returned `None` when querying a slab that
hadn't been prepared yet instead of returning a handle to that slab.
This resulted in a 1-frame delay when modifying materials. In the
`animated_material` example, this resulted in the meshes never showing
up at all, because that example changes every material every frame.
Together with #17979, this patch locally fixes the problems with
`animated_material` on macOS that were reported in #17970.
# Objective
Fixes#8615
## Solution
Bevy currently interprets 1x1 dds textures as 1-dimensional. I think it
might be more common for game engines to assume two dimensions in this
ambiguous case. [citation needed]
I reworked the dimension choosing logic to only use 1d if there's a
dimension > 1, and assume 2d otherwise. I kept the assumption that
compressed textures are probably 2d.
## Testing
Modified `sprite.rs` to use `Tex_0012_0.dds` from the linked issue.
---------
Co-authored-by: Alice Cecile <alice.i.cecile@gmail.com>
# Objective
Fix#17924
## Solution
Use fully qualified syntax (`usize::from` rather than `.into()`).
## Testing
Ran a build for the platform specified in the issue.
---------
Co-authored-by: Gino Valente <49806985+MrGVSV@users.noreply.github.com>
Fixes#17883
# Objective + Solution
When doing normal scene root entity despawns (which are notably now
recursive), do not despawn instanced entities that are no longer in the
hierarchy.
(I would not classify this as a bug, but rather a behavior change)
## Migration Guide
If you previously relied on scene entities no longer in the hierarchy
being despawned when the scene root is despawned , use
`SceneSpawner::despawn_instance()` instead.
# Objective
Calling `define_label!` in a `no_std` 3rd party crate currently requires
the user to import `Box` themselves due to a non-fully-specified
reference to `Box`.
## Solution
Add a fully specified path for `Box` in the one location necessary, to
match all of the other cases.
Two-phase occlusion culling can be helpful for shadow maps just as it
can for a prepass, in order to reduce vertex and alpha mask fragment
shading overhead. This patch implements occlusion culling for shadow
maps from directional lights, when the `OcclusionCulling` component is
present on the entities containing the lights. Shadow maps from point
lights are deferred to a follow-up patch. Much of this patch involves
expanding the hierarchical Z-buffer to cover shadow maps in addition to
standard view depth buffers.
The `scene_viewer` example has been updated to add `OcclusionCulling` to
the directional light that it creates.
This improved the performance of the rend3 sci-fi test scene when
enabling shadows.
Currently, Bevy's implementation of bindless resources is rather
unusual: every binding in an object that implements `AsBindGroup` (most
commonly, a material) becomes its own separate binding array in the
shader. This is inefficient for two reasons:
1. If multiple materials reference the same texture or other resource,
the reference to that resource will be duplicated many times. This
increases `wgpu` validation overhead.
2. It creates many unused binding array slots. This increases `wgpu` and
driver overhead and makes it easier to hit limits on APIs that `wgpu`
currently imposes tight resource limits on, like Metal.
This PR fixes these issues by switching Bevy to use the standard
approach in GPU-driven renderers, in which resources are de-duplicated
and passed as global arrays, one for each type of resource.
Along the way, this patch introduces per-platform resource limits and
bumps them from 16 resources per binding array to 64 resources per bind
group on Metal and 2048 resources per bind group on other platforms.
(Note that the number of resources per *binding array* isn't the same as
the number of resources per *bind group*; as it currently stands, if all
the PBR features are turned on, Bevy could pack as many as 496 resources
into a single slab.) The limits have been increased because `wgpu` now
has universal support for partially-bound binding arrays, which mean
that we no longer need to fill the binding arrays with fallback
resources on Direct3D 12. The `#[bindless(LIMIT)]` declaration when
deriving `AsBindGroup` can now simply be written `#[bindless]` in order
to have Bevy choose a default limit size for the current platform.
Custom limits are still available with the new
`#[bindless(limit(LIMIT))]` syntax: e.g. `#[bindless(limit(8))]`.
The material bind group allocator has been completely rewritten. Now
there are two allocators: one for bindless materials and one for
non-bindless materials. The new non-bindless material allocator simply
maintains a 1:1 mapping from material to bind group. The new bindless
material allocator maintains a list of slabs and allocates materials
into slabs on a first-fit basis. This unfortunately makes its
performance O(number of resources per object * number of slabs), but the
number of slabs is likely to be low, and it's planned to become even
lower in the future with `wgpu` improvements. Resources are
de-duplicated with in a slab and reference counted. So, for instance, if
multiple materials refer to the same texture, that texture will exist
only once in the appropriate binding array.
To support these new features, this patch adds the concept of a
*bindless descriptor* to the `AsBindGroup` trait. The bindless
descriptor allows the material bind group allocator to probe the layout
of the material, now that an array of `BindGroupLayoutEntry` records is
insufficient to describe the group. The `#[derive(AsBindGroup)]` has
been heavily modified to support the new features. The most important
user-facing change to that macro is that the struct-level `uniform`
attribute, `#[uniform(BINDING_NUMBER, StandardMaterial)]`, now reads
`#[uniform(BINDLESS_INDEX, MATERIAL_UNIFORM_TYPE,
binding_array(BINDING_NUMBER)]`, allowing the material to specify the
binding number for the binding array that holds the uniform data.
To make this patch simpler, I removed support for bindless
`ExtendedMaterial`s, as well as field-level bindless uniform and storage
buffers. I intend to add back support for these as a follow-up. Because
they aren't in any released Bevy version yet, I figured this was OK.
Finally, this patch updates `StandardMaterial` for the new bindless
changes. Generally, code throughout the PBR shaders that looked like
`base_color_texture[slot]` now looks like
`bindless_2d_textures[material_indices[slot].base_color_texture]`.
This patch fixes a system hang that I experienced on the [Caldera test]
when running with `caldera --random-materials --texture-count 100`. The
time per frame is around 19.75 ms, down from 154.2 ms in Bevy 0.14: a
7.8× speedup.
[Caldera test]: https://github.com/DGriffin91/bevy_caldera_scene
Deferred rendering currently doesn't support occlusion culling. This PR
implements it in a straightforward way, mirroring what we already do for
the non-deferred pipeline.
On the rend3 sci-fi base test scene, this resulted in roughly a 2×
speedup when applied on top of my other patches. For that scene, it was
useful to add another option, `--add-light`, which forces the addition
of a shadow-casting light, to the scene viewer, which I included in this
patch.
This commit restructures the multidrawable batch set builder for better
performance in various ways:
* The bin traversal is optimized to make the best use of the CPU cache.
* The inner loop that iterates over the bins, which is the hottest part
of `batch_and_prepare_binned_render_phase`, has been shrunk as small as
possible.
* Where possible, multiple elements are added to or reserved from GPU
buffers as a batch instead of one at a time.
* Methods that LLVM wasn't inlining have been marked `#[inline]` where
doing so would unlock optimizations.
This code has also been refactored to avoid duplication between the
logic for indexed and non-indexed meshes via the introduction of a
`MultidrawableBatchSetPreparer` object.
Together, this improved the `batch_and_prepare_binned_render_phase` time
on Caldera by approximately 2×.
Eventually, we should optimize the batchable-but-not-multidrawable and
unbatchable logic as well, but these meshes are much rarer, so in the
interests of keeping this patch relatively small I opted to leave those
to a follow-up.
# Objective
Fix incorrect mesh culling where objects (particularly directional
shadows) were being incorrectly culled during the early preprocessing
phase. The issue manifested specifically on Apple M1 GPUs but not on
newer devices like the M4. The bug was in the
`view_frustum_intersects_obb` function, where including the w component
(plane distance) in the dot product calculations led to false positive
culling results. This caused objects to be incorrectly culled before
shadow casting could begin.
## Issue Details
The problem of missing shadows is reproducible on Apple M1 GPUs as of
this commit (bisected):
```
00722b8d0 Make indirect drawing opt-out instead of opt-in, enabling multidraw by default. (#16757)
```
and as recent as this commit:
```
c818c9214 Add option to animate materials in many_cubes (#17927)
```
- The frustum culling calculation incorrectly included the w component
(plane distance) when transforming basis vectors
- The relative radius calculation should only consider directional
transformation (xyz), not positional information (w)
- This caused false positive culling specifically on M1 devices likely
due to different device-specific floating-point behavior
- When objects were incorrectly culled, `early_instance_count` never
incremented, leading to missing geometry in the shadow pass
## Testing
- Tested on M1 and M4 devices to verify the fix
- Verified shadows and geometry render correctly on both platforms
- Confirmed the solution matches the existing Rust implementation's
behavior for calculating the relative radius:
c818c92143/crates/bevy_render/src/primitives/mod.rs (L77-L87)
- The fix resolves a mathematical error in the frustum culling
calculation while maintaining correct culling behavior across all
platforms.
---
## Showcase
`c818c9214`
<img width="1284" alt="c818c9214"
src="https://github.com/user-attachments/assets/fe1c7ea9-b13d-422e-b12d-f1cd74475213"
/>
`mate-h/frustum-cull-fix`
<img width="1283" alt="frustum-cull-fix"
src="https://github.com/user-attachments/assets/8a9ccb2a-64b6-4d5e-a17d-ac4798da5b51"
/>
# Objective
Make checked vs unchecked shaders configurable
Fixes#17786
## Solution
Added `ValidateShaders` enum to `Shader` and added
`create_and_validate_shader_module` to `RenderDevice`
## Testing
I tested the shader examples locally and they all worked. I'd like to
write a few tests to verify but am unsure how to start.
---------
Co-authored-by: Alice Cecile <alice.i.cecile@gmail.com>
The `check_visibility` system currently follows this algorithm:
1. Store all meshes that were visible last frame in the
`PreviousVisibleMeshes` set.
2. Determine which meshes are visible. For each such visible mesh,
remove it from `PreviousVisibleMeshes`.
3. Mark all meshes that remain in `PreviousVisibleMeshes` as invisible.
This algorithm would be correct if the `check_visibility` were the only
system that marked meshes visible. However, it's not: the shadow-related
systems `check_dir_light_mesh_visibility` and
`check_point_light_mesh_visibility` can as well. This results in the
following sequence of events for meshes that are in a shadow map but
*not* visible from a camera:
A. `check_visibility` runs, finds that no camera contains these meshes,
and marks them hidden, which sets the changed flag.
B. `check_dir_light_mesh_visibility` and/or
`check_point_light_mesh_visibility` run, discover that these meshes
are visible in the shadow map, and marks them as visible, again
setting the `ViewVisibility` changed flag.
C. During the extraction phase, the mesh extraction system sees that
`ViewVisibility` is changed and re-extracts the mesh.
This is inefficient and results in needless work during rendering.
This patch fixes the issue in two ways:
* The `check_dir_light_mesh_visibility` and
`check_point_light_mesh_visibility` systems now remove meshes that they
discover from `PreviousVisibleMeshes`.
* Step (3) above has been moved from `check_visibility` to a separate
system, `mark_newly_hidden_entities_invisible`. This system runs after
all visibility-determining systems, ensuring that
`PreviousVisibleMeshes` contains only those meshes that truly became
invisible on this frame.
This fix dramatically improves the performance of [the Caldera
benchmark], when combined with several other patches I've submitted.
[the Caldera benchmark]:
https://github.com/DGriffin91/bevy_caldera_scene
PR #17688 broke motion vector computation, and therefore motion blur,
because it enabled retention of `MeshInputUniform`s, and
`MeshInputUniform`s contain the indices of the previous frame's
transform and the previous frame's skinned mesh joint matrices. On frame
N, if a `MeshInputUniform` is retained on GPU from the previous frame,
the `previous_input_index` and `previous_skin_index` would refer to the
indices for frame N - 2, not the index for frame N - 1.
This patch fixes the problems. It solves these issues in two different
ways, one for transforms and one for skins:
1. To fix transforms, this patch supplies the *frame index* to the
shader as part of the view uniforms, and specifies which frame index
each mesh's previous transform refers to. So, in the situation described
above, the frame index would be N, the previous frame index would be N -
1, and the `previous_input_frame_number` would be N - 2. The shader can
now detect this situation and infer that the mesh has been retained, and
can therefore conclude that the mesh's transform hasn't changed.
2. To fix skins, this patch replaces the explicit `previous_skin_index`
with an invariant that the index of the joints for the current frame and
the index of the joints for the previous frame are the same. This means
that the `MeshInputUniform` never has to be updated even if the skin is
animated. The downside is that we have to copy joint matrices from the
previous frame's buffer to the current frame's buffer in
`extract_skins`.
The rationale behind (2) is that we currently have no mechanism to
detect when joints that affect a skin have been updated, short of
comparing all the transforms and setting a flag for
`extract_meshes_for_gpu_building` to consume, which would regress
performance as we want `extract_skins` and
`extract_meshes_for_gpu_building` to be able to run in parallel.
To test this change, use `cargo run --example motion_blur`.
Currently, the specialized pipeline cache maps a (view entity, mesh
entity) tuple to the retained pipeline for that entity. This causes two
problems:
1. Using the view entity is incorrect, because the view entity isn't
stable from frame to frame.
2. Switching the view entity to a `RetainedViewEntity`, which is
necessary for correctness, significantly regresses performance of
`specialize_material_meshes` and `specialize_shadows` because of the
loss of the fast `EntityHash`.
This patch fixes both problems by switching to a *two-level* hash table.
The outer level of the table maps each `RetainedViewEntity` to an inner
table, which maps each `MainEntity` to its pipeline ID and change tick.
Because we loop over views first and, within that loop, loop over
entities visible from that view, we hoist the slow lookup of the view
entity out of the inner entity loop.
Additionally, this patch fixes a bug whereby pipeline IDs were leaked
when removing the view. We still have a problem with leaking pipeline
IDs for deleted entities, but that won't be fixed until the specialized
pipeline cache is retained.
This patch improves performance of the [Caldera benchmark] from 7.8×
faster than 0.14 to 9.0× faster than 0.14, when applied on top of the
global binding arrays PR, #17898.
[Caldera benchmark]: https://github.com/DGriffin91/bevy_caldera_scene
Currently, Bevy rebuilds the buffer containing all the transforms for
joints every frame, during the extraction phase. This is inefficient in
cases in which many skins are present in the scene and their joints
don't move, such as the Caldera test scene.
To address this problem, this commit switches skin extraction to use a
set of retained GPU buffers with allocations managed by the offset
allocator. I use fine-grained change detection in order to determine
which skins need updating. Note that the granularity is on the level of
an entire skin, not individual joints. Using the change detection at
that level would yield poor performance in common cases in which an
entire skin is animated at once. Also, this patch yields additional
performance from the fact that changing joint transforms no longer
requires the skinned mesh to be re-extracted.
Note that this optimization can be a double-edged sword. In
`many_foxes`, fine-grained change detection regressed the performance of
`extract_skins` by 3.4x. This is because every joint is updated every
frame in that example, so change detection is pointless and is pure
overhead. Because the `many_foxes` workload is actually representative
of animated scenes, this patch includes a heuristic that disables
fine-grained change detection if the number of transformed entities in
the frame exceeds a certain fraction of the total number of joints.
Currently, this threshold is set to 25%. Note that this is a crude
heuristic, because it doesn't distinguish between the number of
transformed *joints* and the number of transformed *entities*; however,
it should be good enough to yield the optimum code path most of the
time.
Finally, this patch fixes a bug whereby skinned meshes are actually
being incorrectly retained if the buffer offsets of the joints of those
skinned meshes changes from frame to frame. To fix this without
retaining skins, we would have to re-extract every skinned mesh every
frame. Doing this was a significant regression on Caldera. With this PR,
by contrast, mesh joints stay at the same buffer offset, so we don't
have to update the `MeshInputUniform` containing the buffer offset every
frame. This also makes PR #17717 easier to implement, because that PR
uses the buffer offset from the previous frame, and the logic for
calculating that is simplified if the previous frame's buffer offset is
guaranteed to be identical to that of the current frame.
On Caldera, this patch reduces the time spent in `extract_skins` from
1.79 ms to near zero. On `many_foxes`, this patch regresses the
performance of `extract_skins` by approximately 10%-25%, depending on
the number of foxes. This has only a small impact on frame rate.
The GPU can fill out many of the fields in `IndirectParametersMetadata`
using information it already has:
* `early_instance_count` and `late_instance_count` are always
initialized to zero.
* `mesh_index` is already present in the work item buffer as the
`input_index` of the first work item in each batch.
This patch moves these fields to a separate buffer, the *GPU indirect
parameters metadata* buffer. That way, it avoids having to write them on
CPU during `batch_and_prepare_binned_render_phase`. This effectively
reduces the number of bits that that function must write per mesh from
160 to 64 (in addition to the 64 bits per mesh *instance*).
Additionally, this PR refactors `UntypedPhaseIndirectParametersBuffers`
to add another layer, `MeshClassIndirectParametersBuffers`, which allows
abstracting over the buffers corresponding indexed and non-indexed
meshes. This patch doesn't make much use of this abstraction, but
forthcoming patches will, and it's overall a cleaner approach.
This didn't seem to have much of an effect by itself on
`batch_and_prepare_binned_render_phase` time, but subsequent PRs
dependent on this PR yield roughly a 2× speedup.
# Objective
- #17787 removed sweeping of binned render phases from 2D by accident
due to them not using the `BinnedRenderPhasePlugin`.
- Fixes#17885
## Solution
- Schedule `sweep_old_entities` in `QueueSweep` like
`BinnedRenderPhasePlugin` does, but for 2D where that plugin is not
used.
## Testing
Tested with the modified `shader_defs` example in #17885 .
Fixes#17290.
<details>
<summary>Compilation errors before fix</summary>
`cargo clippy --tests --all-features --package bevy_image`:
```rust
error[E0061]: this function takes 7 arguments but 6 arguments were supplied
--> crates/bevy_core_pipeline/src/tonemapping/mod.rs:451:5
|
451 | Image::from_buffer(
| ^^^^^^^^^^^^^^^^^^
...
454 | bytes,
| ----- argument #1 of type `std::string::String` is missing
|
note: associated function defined here
--> /Users/josiahnelson/Desktop/Programming/Rust/bevy/crates/bevy_image/src/image.rs:930:12
|
930 | pub fn from_buffer(
| ^^^^^^^^^^^
help: provide the argument
|
451 | Image::from_buffer(/* std::string::String */, bytes, image_type, CompressedImageFormats::NONE, false, image_sampler, RenderAssetUsages::RENDER_WORLD)
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
```
`cargo clippy --tests --all-features --package bevy_gltf`:
```rust
error[E0560]: struct `bevy_pbr::StandardMaterial` has no field named `specular_channel`
--> crates/bevy_gltf/src/loader.rs:1343:13
|
1343 | specular_channel: specular.specular_channel,
| ^^^^^^^^^^^^^^^^ `bevy_pbr::StandardMaterial` does not have this field
|
= note: available fields are: `emissive_exposure_weight`, `diffuse_transmission`, `diffuse_transmission_channel`, `diffuse_transmission_texture`, `flip_normal_map_y` ... and 9 others
error[E0560]: struct `bevy_pbr::StandardMaterial` has no field named `specular_texture`
--> crates/bevy_gltf/src/loader.rs:1345:13
|
1345 | specular_texture: specular.specular_texture,
| ^^^^^^^^^^^^^^^^ `bevy_pbr::StandardMaterial` does not have this field
|
= note: available fields are: `emissive_exposure_weight`, `diffuse_transmission`, `diffuse_transmission_channel`, `diffuse_transmission_texture`, `flip_normal_map_y` ... and 9 others
error[E0560]: struct `bevy_pbr::StandardMaterial` has no field named `specular_tint_channel`
--> crates/bevy_gltf/src/loader.rs:1351:13
|
1351 | specular_tint_channel: specular.specular_color_channel,
| ^^^^^^^^^^^^^^^^^^^^^ `bevy_pbr::StandardMaterial` does not have this field
|
= note: available fields are: `emissive_exposure_weight`, `diffuse_transmission`, `diffuse_transmission_channel`, `diffuse_transmission_texture`, `flip_normal_map_y` ... and 9 others
error[E0560]: struct `bevy_pbr::StandardMaterial` has no field named `specular_tint_texture`
--> crates/bevy_gltf/src/loader.rs:1353:13
|
1353 | specular_tint_texture: specular.specular_color_texture,
| ^^^^^^^^^^^^^^^^^^^^^ `bevy_pbr::StandardMaterial` does not have this field
|
= note: available fields are: `emissive_exposure_weight`, `diffuse_transmission`, `diffuse_transmission_channel`, `diffuse_transmission_texture`, `flip_normal_map_y` ... and 9 others
```
</details>
# Objective
Fixes#17022
## Solution
Only enable `bevy_gltf/dds` if `bevy_gltf` is already enabled.
## Testing
Tested with empty project
```toml
[dependencies]
bevy = { version = "0.16.0-dev", path = "../bevy", default-features = false, features = [
"dds",
] }
```
### Before
```
cargo tree --depth 1 -i bevy_gltf
bevy_gltf v0.16.0-dev (/Users/robparrett/src/bevy/crates/bevy_gltf)
└── bevy_internal v0.16.0-dev (/Users/robparrett/src/bevy/crates/bevy_internal)
```
### After
```
cargo tree --depth 1 -i bevy_gltf
warning: nothing to print.
To find dependencies that require specific target platforms, try to use option `--target all` first, and then narrow your search scope accordingly.
```
# Context
Renaming `Parent` to `ChildOf` in #17247 has been contentious. While
those users concerns are valid (especially around legibility of code
IMO!), @cart [has
decided](https://discord.com/channels/691052431525675048/749335865876021248/1340434322833932430)
to stick with the new name.
> In general this conversation is unsurprising to me, as it played out
essentially the same way when I asked for opinions in my PR. There are
strong opinions on both sides. Everyone is right in their own way.
>
> I chose ChildOf for the following reasons:
>
> 1. I think it derives naturally from the system we have built, the
concepts we have chosen, and how we generally name the types that
implement a trait in Rust. This is the name of the type implementing
Relationship. We are adding that Relationship component to a given
entity (whether it "is" the relationship or "has" the relationship is
kind of immaterial ... we are naming the relationship that it "is" or
"has"). What is the name of the relationship that a child has to its
parent? It is a "child" of the parent of course!
> 2. In general the non-parent/child relationships I've seen in the wild
generally benefit from (or need to) use the naming convention in (1)
(aka calling the Relationship the name of the relationship the entity
has). Many relationships don't have an equivalent to the Parent/Child
name concept.
> 3. I do think we could get away with using (1) for pretty much
everything else and special casing Parent/Children. But by embracing the
naming convention, we help establish that this is in fact a pattern, and
we help prime people to think about these things in a consistent way.
Consistency and predictability is a generally desirable property. And
for something as divisive and polarizing as relationship naming, I think
drawing a hard line in the sand is to the benefit of the community as a
whole.
> 4. I believe the fact that we dont see as much of the XOf naming style
elsewhere is to our benefit. When people see things in that style, they
are primed to think of them as relationships (after some exposure to
Bevy and the ecosystem). I consider this a useful hint.
> 5. Most of the practical confusion from using ChildOf seems to be from
calling the value of the target field we read from the relationship
child_of. The name of the target field should be parent (we could even
consider renaming child_of.0 to child_of.parent for clarity). I suspect
that existing Bevy users renaming their existing code will feel the most
friction here, as this requires a reframing. Imo it is natural and
expected to receive pushback from these users hitting this case.
## Objective
The new documentation doesn't do a particularly good job at quickly
explaining the meaning of each component or how to work with them;
making a tricky migration more painful and slowing down new users as
they learn about some of the most fundamental types in Bevy.
## Solution
1. Clearly explain what each component does in the very first line,
assuming no background knowledge. This is the first relationships that
99% of users will encounter, so explaining that they are relationships
is unhelpful as an introduction.
2. Add doc aliases for the rejected `IsParent`/`IsChild`/`Parent` names,
to improve autocomplete and doc searching.
3. Do some assorted docs cleanup while we're here.
---------
Co-authored-by: Eagster <79881080+ElliottjPierce@users.noreply.github.com>
## Objective
There's no general error for when an entity doesn't exist, and some
methods are going to need one when they get Resultified. The closest
thing is `EntityFetchError`, but that error has a slightly more specific
purpose.
## Solution
- Added `EntityDoesNotExistError`.
- Contains `Entity` and `EntityDoesNotExistDetails`.
- Changed `EntityFetchError` and `QueryEntityError`:
- Changed `NoSuchEntity` variant to wrap `EntityDoesNotExistError` and
renamed the variant to `EntityDoesNotExist`.
- Renamed `EntityFetchError` to `EntityMutableFetchError` to make its
purpose clearer.
- Renamed `TryDespawnError` to `EntityDespawnError` to make it more
general.
- Changed `World::inspect_entity` to return `Result<[ok],
EntityDoesNotExistError>` instead of panicking.
- Changed `World::get_entity` and `WorldEntityFetch::fetch_ref` to
return `Result<[ok], EntityDoesNotExistError>` instead of `Result<[ok],
Entity>`.
- Changed `UnsafeWorldCell::get_entity` to return
`Result<UnsafeEntityCell, EntityDoesNotExistError>` instead of
`Option<UnsafeEntityCell>`.
## Migration Guide
- `World::inspect_entity` now returns `Result<impl Iterator<Item =
&ComponentInfo>, EntityDoesNotExistError>` instead of `impl
Iterator<Item = &ComponentInfo>`.
- `World::get_entity` now returns `EntityDoesNotExistError` as an error
instead of `Entity`. You can still access the entity's ID through the
error's `entity` field.
- `UnsafeWorldCell::get_entity` now returns `Result<UnsafeEntityCell,
EntityDoesNotExistError>` instead of `Option<UnsafeEntityCell>`.
Appending to these vectors is performance-critical in
`batch_and_prepare_binned_render_phase`, so `RawBufferVec`, which
doesn't have the overhead of `encase`, is more appropriate.
The `collect_buffers_for_phase` system tries to reuse these buffers, but
its efforts are stymied by the fact that
`clear_batched_gpu_instance_buffers` clears the containing hash table
and therefore frees the buffers. This patch makes
`clear_batched_gpu_instance_buffers` stop doing that so that the
allocations can be reused.
# Objective
Simplify the API surface by removing duplicated functionality between
`Query` and `QueryState`.
Reduce the amount of `unsafe` code required in `QueryState`.
This is a follow-up to #15858.
## Solution
Move implementations of `Query` methods from `QueryState` to `Query`.
Instead of the original methods being on `QueryState`, with `Query`
methods calling them by passing the individual parameters, the original
methods are now on `Query`, with `QueryState` methods calling them by
constructing a `Query`.
This also adds two `_inner` methods that were missed in #15858:
`iter_many_unique_inner` and `single_inner`.
One goal here is to be able to deprecate and eventually remove many of
the methods on `QueryState`, reducing the overall API surface. (I
expected to do that in this PR, but this change was large enough on its
own!) Now that the `QueryState` methods each consist of a simple
expression like `self.query(world).get_inner(entity)`, a future PR can
deprecate some or all of them with simple migration instructions.
The other goal is to reduce the amount of `unsafe` code. The current
implementation of a read-only method like `QueryState::get` directly
calls the `unsafe fn get_unchecked_manual` and needs to repeat the proof
that `&World` has enough access. With this change, `QueryState::get` is
entirely safe code, with the proof that `&World` has enough access done
by the `query()` method and shared across all read-only operations.
## Future Work
The next step will be to mark the `QueryState` methods as
`#[deprecated]` and migrate callers to the methods on `Query`.
# Objective
Support accessing resources using reflection when using
`FilteredResources` in a dynamic system. This is similar to how
components can be queried using reflection when using
`FilteredEntityRef|Mut`.
## Solution
Change `ReflectResource` from taking `&World` and `&mut World` to taking
`impl Into<FilteredResources>` and `impl Into<FilteredResourcesMut>`,
similar to how `ReflectComponent` takes `impl Into<FilteredEntityRef>`
and `impl Into<FilteredEntityMut>`. There are `From` impls that ensure
code passing `&World` and `&mut World` continues to work as before.
## Migration Guide
If you are manually creating a `ReflectComponentFns` struct, the
`reflect` function now takes `FilteredResources` instead `&World`, and
there is a new `reflect_mut` function that takes `FilteredResourcesMut`.
# Objective
Add reference to reported position space in picking backend docs.
Fixes#17844
## Solution
Add explanatory docs to the implementation notes of each picking
backend.
## Testing
`cargo r -p ci -- doc-check` & `cargo r -p ci -- lints`
# Objective
It is impossible to register a type with `TypeRegistry::register` if the
type is unnameable (in the current scope).
## Solution
Add `TypeRegistry::register_by_val` which mirrors std's `size_of_val`
and friends.
## Testing
There's a doc test (unrelated but there seem to be some pre-existing
broken doc links in `bevy_reflect`).
There was nonsense code in `batch_and_prepare_sorted_render_phase` that
created temporary buffers to add objects to instead of using the correct
ones. I think this was debug code. This commit removes that code in
favor of writing to the actual buffers.
Closes#17846.
Co-authored-by: Alice Cecile <alice.i.cecile@gmail.com>
# Objective
`bevy_assets` has long been unapproachable for contributors and users.
More and better documentation would help that.
We're gradually moving towards globally denying missing docs (#3492)!
However, writing all of the hundreds of missing doc strings in a single
go will be miserable to review.
## Solution
Remove the allow for missing docs temporarily, and then pick some easy
missing doc warnings largely at random to tackle.
Stop when the change set is starting to feel intimidating.
The `output_index` field is only used in direct mode, and the
`indirect_parameters_index` field is only used in indirect mode.
Consequently, we can combine them into a single field, reducing the size
of `PreprocessWorkItem`, which
`batch_and_prepare_{binned,sorted}_render_phase` must construct every
frame for every mesh instance, from 96 bits to 64 bits.
# Objective
Continuation of #16547.
We do not yet have parallel versions of `par_iter_many` and
`par_iter_many_unique`. It is currently very painful to try and use
parallel iteration over entity lists. Even if a list is not long, each
operation might still be very expensive, and worth parallelizing.
Plus, it has been requested several times!
## Solution
Once again, we implement what we lack!
These parallel iterators collect their input entity list into a
`Vec`/`UniqueEntityVec`, then chunk that over the available threads,
inspired by the original `par_iter`.
Since no order guarantee is given to the caller, we could sort the input
list according to `EntityLocation`, but that would likely only be worth
it for very large entity lists.
There is some duplication which could likely be improved, but I'd like
to leave that for a follow-up.
## Testing
The doc tests on `for_each_init` of `QueryParManyIter` and
`QueryParManyUniqueIter`.
# Objective
While surveying the state of documentation for bevy_assets, I noticed a
few minor issues.
## Solution
Revise the docs to focus on clear explanations of core ideas and
cross-linking related objects.
# Objective
Update typos, fix new typos.
1.29.6 was just released to fix an
[issue](https://github.com/crate-ci/typos/issues/1228) where January's
corrections were not included in the binaries for the last release.
Reminder: typos can be tossed in the monthly [non-critical corrections
issue](https://github.com/crate-ci/typos/issues/1221).
## Solution
I chose to allow `implementors`, because a good argument seems to be
being made [here](https://github.com/crate-ci/typos/issues/1226) and
there is now a PR to address that.
## Discussion
Should I exclude `bevy_mikktspace`?
At one point I think we had an informal policy of "don't mess with
mikktspace until https://github.com/bevyengine/bevy/pull/9050 is merged"
but it doesn't seem like that is likely to be merged any time soon.
I think these particular corrections in mikktspace are fine because
- The same typo mistake seems to have been fixed in that PR
- The entire file containing these corrections was deleted in that PR
## Typo of the Month
correspindong -> corresponding
# Objective
Updates the now inaccurate position docs
Fixes#17832
## Solution
From
`The position of the intersection in the world, if the data is available
from the backend.`
To
`The position reported by the backend, if the data is available.
Position data may be in any space (e.g. World space, Screen space, Local
space), specified by the backend providing it.`
## Testing
uhh reading :)
Currently, invocations of `batch_and_prepare_binned_render_phase` and
`batch_and_prepare_sorted_render_phase` can't run in parallel because
they write to scene-global GPU buffers. After PR #17698,
`batch_and_prepare_binned_render_phase` started accounting for the
lion's share of the CPU time, causing us to be strongly CPU bound on
scenes like Caldera when occlusion culling was on (because of the
overhead of batching for the Z-prepass). Although I eventually plan to
optimize `batch_and_prepare_binned_render_phase`, we can obtain
significant wins now by parallelizing that system across phases.
This commit splits all GPU buffers that
`batch_and_prepare_binned_render_phase` and
`batch_and_prepare_sorted_render_phase` touches into separate buffers
for each phase so that the scheduler will run those phases in parallel.
At the end of batch preparation, we gather the render phases up into a
single resource with a new *collection* phase. Because we already run
mesh preprocessing separately for each phase in order to make occlusion
culling work, this is actually a cleaner separation. For example, mesh
output indices (the unique ID that identifies each mesh instance on GPU)
are now guaranteed to be sequential starting from 0, which will simplify
the forthcoming work to remove them in favor of the compute dispatch ID.
On Caldera, this brings the frame time down to approximately 9.1 ms with
occlusion culling on.
# Objective
Fix unsoundness introduced by #15858. `QueryLens::query()` would hand
out a `Query` with the full `'w` lifetime, and the new `_inner` methods
would let the results outlive the `Query`. This could be used to create
aliasing mutable references, like
```rust
fn bad<'w>(mut lens: QueryLens<'w, EntityMut>, entity: Entity) {
let one: EntityMut<'w> = lens.query().get_inner(entity).unwrap();
let two: EntityMut<'w> = lens.query().get_inner(entity).unwrap();
assert!(one.entity() == two.entity());
}
```
Fixes#17693
## Solution
Restrict the `'world` lifetime in the `Query` returned by
`QueryLens::query()` to `'_`, the lifetime of the borrow of the
`QueryLens`.
The model here is that `Query<'w, 's, D, F>` and `QueryLens<'w, D, F>`
have permission to access their components for the lifetime `'w`. So
going from `&'a mut QueryLens<'w>` to `Query<'w, 'a>` would borrow the
permission only for the `'a` lifetime, but incorrectly give it out for
the full `'w` lifetime.
To handle any cases where users were calling `get_inner()` or
`iter_inner()` on the `Query` and expecting the full `'w` lifetime, we
introduce a new `QueryLens::query_inner()` method. This is only valid
for `ReadOnlyQueryData`, so it may safely hand out a copy of the
permission for the full `'w` lifetime. Since `get_inner()` and
`iter_inner()` were only valid on `ReadOnlyQueryData` prior to #15858,
that should cover any uses that relied on the longer lifetime.
## Migration Guide
Users of `QueryLens::query()` who were calling `get_inner()` or
`iter_inner()` will need to replace the call with
`QueryLens::query_inner()`.
Conceptually, bins are ordered hash maps. We currently implement these
as a list of keys with an associated hash map. But we already have a
data type that implements ordered hash maps directly: `IndexMap`. This
patch switches Bevy to use `IndexMap`s for bins. Because we're memory
bound, this doesn't affect performance much, but it is cleaner.
# Objective
Related to #17784. The ticket is actually about just getting rid of
`Entity{Ref,Mut}Except` in favor of `FilteredEntity{Ref,Mut}`, but I got
told the unification of Entity types is a bigger endeavor that has been
going on for a while now (as the "Pointing Fingers" working group) and I
should just add the functions I actually need in the meantime.
## Solution
This PR adds all of the functions necessary to access components by
TypeId or ComponentId instead of static types.
## Testing
> Did you test these changes? If so, how?
Haven't tested it yet, but the changes are mostly copy/paste from other
implementations in the same file, since there is a lot of duplicated
functionality there.
## Not a Migration Guide
There shouldn't be any breaking changes, it's just a few new functions
on existing types.
I had to shuffle around the lifetimes in `From<&EntityMutExcept<'a, B>>
for EntityRefExcept<'a, B>` (originally it was `From<&'a
EntityMutExcept<'_, B>> for EntityRefExcept<'_, B>`) to make the borrow
checker happy, but I don't think that this should have an impact on user
code (correct me if I'm wrong).
* Use texture atomics rather than buffer atomics for the visbuffer
(haven't tested perf on a raster-heavy scene yet)
* Unfortunately to clear the visbuffer we now need a compute pass to
clear it. Using wgpu's clear_texture function internally uses a buffer
-> image copy that's insanely expensive. Ideally it should be using
vkCmdClearColorImage, which I've opened an issue for
https://github.com/gfx-rs/wgpu/issues/7090. For now we'll have to stick
with a custom compute pass and all the extra code that brings.
* Faster resolve depth pass by discarding 0 depth pixels instead of
redundantly writing zero (2x faster for big depth textures like shadow
views)
# Objective
- Wgpu has some expensive code it injects into shaders to avoid the
possibility of things like infinite loops. Generally our shaders are
written by users who won't do this, so it just makes our shaders perform
worse.
## Solution
- Turn off the checks.
- We could try to conditionally keep them, but that complicates the code
and 99.9% of users won't want this.
## Migration Guide
- Bevy no longer turns on wgpu's runtime safety checks
https://docs.rs/wgpu/latest/wgpu/struct.ShaderRuntimeChecks.html. If you
were using Bevy with untrusted shaders, please file an issue.
---------
Co-authored-by: Alice Cecile <alice.i.cecile@gmail.com>
# Objective
Continuation of #17589 and #16547.
Slices have several methods that return iterators which themselves yield
slices, which we have not yet implemented.
An example use is `par_iter_many` style logic.
## Solution
Their implementation is rather straightforward, we simply delegate all
impls to `[T]`.
The resulting iterator types need their own wrappers in the form of
`UniqueEntitySliceIter` and `UniqueEntitySliceIterMut`.
We also add three free functions that cast slices of entity slices to
slices of `UniqueEntitySlice`.
These three should be sufficient, though infinite nesting is achievable
with a trait (like `TrustedEntityBorrow` works over infinite reference
nesting), should the need ever arise.
## Objective
Get rid of a redundant Cargo feature flag.
## Solution
Use the built-in `target_abi = "sim"` instead of a custom Cargo feature
flag, which is set for the iOS (and visionOS and tvOS) simulator. This
has been stable since Rust 1.78.
In the future, some of this may become redundant if Wgpu implements
proper supper for the iOS Simulator:
https://github.com/gfx-rs/wgpu/issues/7057
CC @mockersf who implemented [the original
fix](https://github.com/bevyengine/bevy/pull/10178).
## Testing
- Open mobile example in Xcode.
- Launch the simulator.
- See that no errors are emitted.
- Remove the code cfg-guarded behind `target_abi = "sim"`.
- See that an error now happens.
(I haven't actually performed these steps on the latest `main`, because
I'm hitting an unrelated error (EDIT: It was
https://github.com/bevyengine/bevy/pull/17637). But tested it on
0.15.0).
---
## Migration Guide
> If you're using a project that builds upon the mobile example, remove
the `ios_simulator` feature from your `Cargo.toml` (Bevy now handles
this internally).
Currently, we look up each `MeshInputUniform` index in a hash table that
maps the main entity ID to the index every frame. This is inefficient,
cache unfriendly, and unnecessary, as the `MeshInputUniform` index for
an entity remains the same from frame to frame (even if the input
uniform changes). This commit changes the `IndexSet` in the `RenderBin`
to an `IndexMap` that maps the `MainEntity` to `MeshInputUniformIndex`
(a new type that this patch adds for more type safety).
On Caldera with parallel `batch_and_prepare_binned_render_phase`, this
patch improves that function from 3.18 ms to 2.42 ms, a 31% speedup.
Currently, when a mesh slab overflows, we recreate the allocator and
reinsert all the meshes that were in it in an arbitrary order. This can
result in the meshes moving around. Before `MeshInputUniform`s were
retained, this was slow but harmless, because the `MeshInputUniform`
that contained the positions of the vertex and index data in the slab
would be recreated every frame. However, with mesh retention, there's no
guarantee that the `MeshInputUniform`, which could be cached from the
previous frame, will reflect the new position of the mesh data within
the buffer if that buffer happened to grow. This manifested itself as
seeming mesh data corruption when adding many meshes dynamically to the
scene.
There are three possible ways that I could have fixed this that I can
see:
1. Invalidate and rebuild all the `MeshInputUniform`s belonging to all
meshes in a slab when that mesh grows.
2. Introduce a second layer of indirection so that the
`MeshInputUniform` points to a *mesh allocation table* that contains the
current locations of the data of each mesh.
3. Avoid moving meshes when reallocating the buffer.
To be efficient, option (1) would require scanning meshes to see if
their positions changed, a la
`mark_meshes_as_changed_if_their_materials_changed`. Option (2) would
add more runtime indirection and would require additional bookkeeping on
the part of the allocator.
Therefore, this PR chooses option (3), which was remarkably simple to
implement. The key is that the offset allocator happens to allocate
addresses from low addresses to high addresses. So all we have to do is
to *conceptually* allocate the full 512 MiB mesh slab as far as the
offset allocator is concerned, and grow the underlying backing store
from 1 MiB to 512 MiB as needed. In other words, the allocator now
allocates *virtual* GPU memory, and the actual backing slab resizes to
fit the virtual memory. This ensures that the location of mesh data
remains constant for the lifetime of the mesh asset, and we can remove
the code that reinserts meshes one by one when the slab grows in favor
of a single buffer copy.
Closes#17766.
# Objective
- Fixes#17797
## Solution
- `mesh` in `bevy_pbr::mesh_bindings` is behind a `ifndef
MESHLET_MESH_MATERIAL_PASS`. also gate `get_tag` which uses this `mesh`
## Testing
- Run the meshlet example
# Objective
- In #17743, attention was raised to the fact that we supported an
unusual kind of step easing function. The author of the fix kindly
provided some links to standards used in CSS. It would be desirable to
support generally agreed upon standards so this PR here tries to
implement an extra configuration option of the step easing function
- Resolve#17744
## Solution
- Introduce `StepConfig`
- `StepConfig` can configure both the number of steps and the jumping
behavior of the function
- `StepConfig` replaces the raw `usize` parameter of the
`EasingFunction::Steps(usize)` construct.
- `StepConfig`s default jumping behavior is `end`, so in that way it
follows #17743
## Testing
- I added a new test per `JumpAt` jumping behavior. These tests
replicate the visuals that can be found at
https://developer.mozilla.org/en-US/docs/Web/CSS/easing-function/steps#description
## Migration Guide
- `EasingFunction::Steps` now uses a `StepConfig` instead of a raw
`usize`. You can replicate the previous behavior by replaceing
`EasingFunction::Steps(10)` with
`EasingFunction::Steps(StepConfig::new(10))`.
---------
Co-authored-by: François Mockers <francois.mockers@vleue.com>
Co-authored-by: Alice Cecile <alice.i.cecile@gmail.com>
## Objective
There's no need for the `span_index` and `color` variables in
`extract_text_shadows` and `extract_text_sections` and we can remove one
of the span index comparisons since text colors are only set per
section.
## Testing
<img width="454" alt="trace"
src="https://github.com/user-attachments/assets/3109d1df-0817-46c2-9889-0459ac93a42c"
/>
This commit builds on top of the work done in #16589 and #17051, by
adding support for fallible observer systems.
As with the previous work, the actual results of the observer system are
suppressed for now, but the intention is to provide a way to handle
errors in a global way.
Until then, you can use a `PipeSystem` to manually handle results.
---------
Signed-off-by: Jean Mertz <git@jeanmertz.com>
## What problem does this solve or what need does it fill?
There are some situations
(https://github.com/bevyengine/bevy/issues/13735) where the ticks that
are present inside `Ref` are incorrect, for example if `Ref` is created
outside of a `SystemParam`.
I still want to use `Ref` because it has convenient `is_added` and
`is_changed` methods.
My current solution is to build my own `Ref` by copy-pasting most the
bevy code to do that via something like
```rust
/// This method is necessary because there is no easy way to
pub(crate) fn get_ref<C: Component>(
world: &World,
entity: Entity,
last_run: Tick,
this_run: Tick,
) -> Ref<C> {
unsafe {
let component_id = world
.components()
.get_id(TypeId::of::<C>())
.unwrap_unchecked();
let world = world.as_unsafe_world_cell_readonly();
let entity_cell = world.get_entity(entity).unwrap_unchecked();
get_component_and_ticks(
world,
component_id,
C::STORAGE_TYPE,
entity,
entity_cell.location(),
)
.map(|(value, cells, _caller)| {
Ref::new(
value.deref::<C>(),
cells.added.deref(),
cells.changed.deref(),
last_run,
this_run,
#[cfg(feature = "track_location")]
_caller.deref(),
)
})
.unwrap_unchecked()
}
}
// Utility function to return
#[inline]
unsafe fn get_component_and_ticks(
world: UnsafeWorldCell<'_>,
component_id: ComponentId,
storage_type: StorageType,
entity: Entity,
location: EntityLocation,
) -> Option<(Ptr<'_>, TickCells<'_>, MaybeUnsafeCellLocation<'_>)> {
match storage_type {
StorageType::Table => {
let table = unsafe { world.storages().tables.get(location.table_id) }?;
// SAFETY: archetypes only store valid table_rows and caller ensure aliasing rules
Some((
table.get_component(component_id, location.table_row)?,
TickCells {
added: table
.get_added_tick(component_id, location.table_row)
.unwrap_unchecked(),
changed: table
.get_changed_tick(component_id, location.table_row)
.unwrap_unchecked(),
},
#[cfg(feature = "track_location")]
table
.get_changed_by(component_id, location.table_row)
.unwrap_unchecked(),
#[cfg(not(feature = "track_location"))]
(),
))
}
StorageType::SparseSet => {
let storage = unsafe { world.storages() }.sparse_sets.get(component_id)?;
storage.get_with_ticks(entity)
}
}
}
```
It would be very convenient if instead bevy exposed a way to create a
`Ref` object with custom `last_run` and `this_run` ticks.
This PR does this by exposing a function to overwrite the `last_run` and
`this_run` ticks.
(Same with `Mut`)
I am ok with marking the method unsafe or risky if it's deemed to risky
for end-users.
# Objective
Add position reporting to `HitData` sent from the UI picking backend.
## Solution
Add the computed normalized relative cursor position to `hit_data`
alongside the `Entity`.
The position reported in `HitData` is normalized relative to the node,
with `(0.,0.,0.)` at the top left and `(1., 1., 0.)` in the bottom
right. Coordinates are relative to the entire node, not just the visible
region.
`HitData` needs a `Vec3` so I just extended with 0.0. I considered
inserting the `depth` here but thought it would be redundant.
I also considered putting the screen space position in the `normal`
field of `HitData`, but that would require renaming of the field or a
separate data structure.
## Testing
Tested with mouse on X11 with entities that have `Node` components.
---
## Showcase
```rs
// Get click position relative to node
fn hit_position(trigger: Trigger<Pointer<Click>>) {
let hit_pos = trigger.event.hit.position.expect("no position");
info!("{}", hit_pos);
}
```
---------
Co-authored-by: Alice Cecile <alice.i.cecile@gmail.com>
You can now configure error handlers for fallible systems. These can be
configured on several levels:
- Globally via `App::set_systems_error_handler`
- Per-schedule via `Schedule::set_error_handler`
- Per-system via a piped system (this is existing functionality)
The default handler of panicking on error keeps the same behavior as
before this commit.
The "fallible_systems" example demonstrates the new functionality.
This builds on top of #17731, #16589, #17051.
---------
Signed-off-by: Jean Mertz <git@jeanmertz.com>
This is a follow up fix for #17330 and fixes#17780.
There was a logic error in the ambiguity detection of
`cargo-manifest-proc-macros`.
`cargo-manifest-proc-macros` now has a test for this case to prevent the
issue in the future.
I also opted to hard fail if the `cargo-manifest-proc-macros` crate
fails. That way the error is more obvious and easier to fix and
diagnose.
## Testing
- The reproducer:
https://github.com/bevyengine/bevy_editor_prototypes/pull/178 works for
me using these fixes.
# Objective
Currently, default query filters, as added in #13120 / #17514 are
hardcoded to only use a single query filter.
This is limiting, as multiple distinct disabling components can serve
important distinct roles. I ran into this limitation when experimenting
with a workflow for prefabs, which don't represent the same state as "an
entity which is temporarily nonfunctional".
## Solution
1. Change `DefaultQueryFilters` to store a SmallVec of ComponentId,
rather than an Option.
2. Expose methods on `DefaultQueryFilters`, `World` and `App` to
actually configure this.
3. While we're here, improve the docs, write some tests, make use of
FromWorld and make some method names more descriptive.
## Follow-up
I'm not convinced that supporting sparse set disabling components is
useful, given the hit to iteration performance and runtime checks
incurred. That's disjoint from this PR though, so I'm not doing it here.
The existing warnings are fine for now.
## Testing
I've added both a doc test and an mid-level unit test to verify that
this works!
# Objective
Allow quick jump to definition of types of GlTFs labeled assets.
## Solution
Add links to the types refered on the docs of `GltfAssetLabel`
## Testing
Ran `cargo run -p ci`
# Objective
Since previously we only had the alpha channel available, we stored the
mean of the transmittance in the aerial view lut, resulting in a grayer
fog than should be expected.
## Solution
- Calculate transmittance to scene in `render_sky` with two samples from
the transmittance lut
- use dual-source blending to effectively have per-component alpha
blending
Currently, we *sweep*, or remove entities from bins when those entities
became invisible or changed phases, during `queue_material_meshes` and
similar phases. This, however, is wrong, because `queue_material_meshes`
executes once per material type, not once per phase. This could result
in sweeping bins multiple times per phase, which can corrupt the bins.
This commit fixes the issue by moving sweeping to a separate system that
runs after queuing.
This manifested itself as entities appearing and disappearing seemingly
at random.
Closes#17759.
---------
Co-authored-by: Robert Swain <robert.swain@gmail.com>
# Objective
Because of mesh preprocessing, users cannot rely on
`@builtin(instance_index)` in order to reference external data, as the
instance index is not stable, either from frame to frame or relative to
the total spawn order of mesh instances.
## Solution
Add a user supplied mesh index that can be used for referencing external
data when drawing instanced meshes.
Closes#13373
## Testing
Benchmarked `many_cubes` showing no difference in total frame time.
## Showcase
https://github.com/user-attachments/assets/80620147-aafc-4d9d-a8ee-e2149f7c8f3b
---------
Co-authored-by: IceSentry <IceSentry@users.noreply.github.com>
# Objective
- Expand the documentation for `EasingCurve`.
- I suspect this might have avoided the confusion in
https://github.com/bevyengine/bevy/pull/17711.
- Also add a shortcut for simple cases.
## Solution
- Added various examples and extra context.
- Implemented `Curve<T>` for `EaseFunction`.
- This means `EasingCurve::new(0.0, 1.0, EaseFunction::X)` can be
shortened to `EaseFunction::X`.
- In some cases this will be a minor performance improvement.
- Added test to confirm they're the same.
- ~~Added some benchmarks for bonus points.~~
## Side Notes
- I would have liked to rename `EaseFunction` to `EaseFn` for brevity,
but that would be a breaking change and maybe controversial.
- Also suspect `EasingCurve` should be `EaseCurve`, but say la vee.
- Benchmarks show that calling `EaseFunction::Smoothstep` is still
slower than calling `smoothstep` directly.
- I think this is because the compiler refuses to inline
`EaseFunction::eval`.
- I don't see any good solution - might need a whole different
interface.
## Testing
```sh
cargo test --package bevy_math
cargo doc --package bevy_math
./target/doc/bevy_math/curve/easing/struct.EasingCurve.html
cargo bench --package benches --bench math -- easing
```
# Objective
https://github.com/bevyengine/bevy/issues/17746
## Solution
- Change `Image.data` from being a `Vec<u8>` to a `Option<Vec<u8>>`
- Added functions to help with creating images
## Testing
- Did you test these changes? If so, how?
All current tests pass
Tested a variety of existing examples to make sure they don't crash
(they don't)
- If relevant, what platforms did you test these changes on, and are
there any important ones you can't test?
Linux x86 64-bit NixOS
---
## Migration Guide
Code that directly access `Image` data will now need to use unwrap or
handle the case where no data is provided.
Behaviour of new_fill slightly changed, but not in a way that is likely
to affect anything. It no longer panics and will fill the whole texture
instead of leaving black pixels if the data provided is not a nice
factor of the size of the image.
---------
Co-authored-by: IceSentry <IceSentry@users.noreply.github.com>
# Objective
Currently
[CosmicBuffer](https://docs.rs/bevy/latest/bevy/text/struct.CosmicBuffer.html)
is a public type with a public field that is not used or accessible in
any public API. Since it is prominently shown in the docs it is the
obvious place to start when trying to access `cosmic_string` features
such as for mapping between screen coordinates and positions in the
displayed text.
The only place `CosmicBuffer` is currently used is as a field of
`ComputedTextBlock`, where a comment explains why the field is private:
/// Buffer for managing text layout and creating [`TextLayoutInfo`].
///
/// This is private because buffer contents are always refreshed from
ECS state when writing glyphs to
/// `TextLayoutInfo`. If you want to control the buffer contents
manually or use the `cosmic-text`
/// editor, then you need to not use `TextLayout` and instead manually
implement the conversion to
/// `TextLayoutInfo`.
#[reflect(ignore)]
pub(crate) buffer: CosmicBuffer,
Unfortunately this comment does not appear in the docs, so a user
looking for a way to access `CosmicBuffer` will not find it unless they
check the source code.
Also there does not seem to be any alternative way to map between screen
coordinates and positions in the displayed text, which would be highly
useful for things like text edit widgets or tool tips. The reasons given
for making the field private only apply for mutable access, so
non-mutable access would presumably be fine.
## Solution
I added a getter to `ComputedTextBlock`, and added the explanation for
why there is no mutable access in the comment:
/// Accesses the underling buffer which can be used for `cosmic-text`
APIs such as accessing layout information
/// or calculating a cursor position.
///
/// Mutable access not offered because changes would be overwritten
during the automated layout calculation.
/// If you want to control the buffer contents manually or use the
`cosmic-text`
/// editor, then you need to not use `TextLayout` and instead manually
implement the conversion to
/// `TextLayoutInfo`.
pub fn get_buffer(&self) -> &CosmicBuffer {
&self.buffer
}
## Testing
I tested that the getter could be used to map from screen coordinates to
string positions by creating a rudimentary text edit widget and trying
it out.
## Alternatives
An alternative to making `CosmicBuffer` accessible would be to make the
type private so that no one wastes time looking for a way of accessing
it, and adding additional methods to `ComputedTextBlock` that make use
of the buffer as implementation detail and offer access to currently
inaccessible functionality.
---------
Co-authored-by: Rob Parrett <robparrett@gmail.com>
# Objective
- bevy_math fails to publish because of the self dev-dependency
- it's used to enable the `approx` feature in tests
## Solution
- Don't specify a version in the dev-dependency. dependencies without a
version are ignored by cargo when publishing
- Gate all the tests that depend on the `approx` feature so that it
doesn't fail to compile when not enabled
- Also gate an import that wasn't used without `bevy_reflect`
## Testing
- with at least cargo 1.84: `cargo package -p bevy_math`
- `cd target/package/bevy_math_* && cargo test`
# Objective
Restore the behavior of `Query::get_many` prior to #15858.
When passed duplicate `Entity`s, `get_many` is supposed to return
results for all of them, since read-only queries don't alias. However,
#15858 merged the implementation with `get_many_mut` and caused it to
return `QueryEntityError::AliasedMutability`.
## Solution
Introduce a new `Query::get_many_readonly` method that consumes the
`Query` like `get_many_inner`, but that is constrained to `D:
ReadOnlyQueryData` so that it can skip the aliasing check. Implement
`Query::get_many` in terms of that new method. Add a test, and a comment
explaining why it doesn't match the pattern of the other `&self`
methods.
This method returns `None` if `meta.location.archetype_id` is
`ArchetypeId::INVALID`.
`EntityLocation::INVALID.archetype_id` is `ArchetypeId::INVALID`.
Therefore this method cannot return `Some(EntityLocation::INVALID)`.
Linking to it in the docs is futile anyway as that constant is not
public.
# Objective
`bevy_picking` currently does not support scroll events.
## Solution
This pr adds a new event type for scroll, and updates the default input
system for mouse pointers to read and emit this event.
## Testing
- Did you test these changes? If so, how?
- 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?
I haven't tested these changes, if the reviewers can advise me how to do
so I'd appreciate it!
# Objective
- Allow users to configure volume using decibels by changing the
`Volume` type from newtyping an `f32` to an enum with `Linear` and
`Decibels` variants.
- Fixes#9507.
- Alternative reworked version of closed#9582.
## Solution
Compared to https://github.com/bevyengine/bevy/pull/9582, this PR has
the following main differences:
1. It uses the term "linear scale" instead of "amplitude" per
https://github.com/bevyengine/bevy/pull/9582/files#r1513529491.
2. Supports `ops` for doing `Volume` arithmetic. Can add two volumes,
e.g. to increase/decrease the current volume. Can multiply two volumes,
e.g. to get the “effective” volume of an audio source considering global
volume.
[requested and blessed on Discord]:
https://discord.com/channels/691052431525675048/749430447326625812/1318272597003341867
## Testing
- Ran `cargo run --example soundtrack`.
- Ran `cargo run --example audio_control`.
- Ran `cargo run --example spatial_audio_2d`.
- Ran `cargo run --example spatial_audio_3d`.
- Ran `cargo run --example pitch`.
- Ran `cargo run --example decodable`.
- Ran `cargo run --example audio`.
---
## Migration Guide
Audio volume can now be configured using decibel values, as well as
using linear scale values. To enable this, some types and functions in
`bevy_audio` have changed.
- `Volume` is now an enum with `Linear` and `Decibels` variants.
Before:
```rust
let v = Volume(1.0);
```
After:
```rust
let volume = Volume::Linear(1.0);
let volume = Volume::Decibels(0.0); // or now you can deal with decibels if you prefer
```
- `Volume::ZERO` has been renamed to the more semantically correct
`Volume::SILENT` because `Volume` now supports decibels and "zero
volume" in decibels actually means "normal volume".
- The `AudioSinkPlayback` trait's volume-related methods now deal with
`Volume` types rather than `f32`s. `AudioSinkPlayback::volume()` now
returns a `Volume` rather than an `f32`. `AudioSinkPlayback::set_volume`
now receives a `Volume` rather than an `f32`. This affects the
`AudioSink` and `SpatialAudioSink` implementations of the trait. The
previous `f32` values are equivalent to the volume converted to linear
scale so the `Volume:: Linear` variant should be used to migrate between
`f32`s and `Volume`.
- The `GlobalVolume::new` function now receives a `Volume` instead of an
`f32`.
---------
Co-authored-by: Zachary Harrold <zac@harrold.com.au>
# Objective
Eliminate the need to write `cfg(feature = "track_location")` every time
one uses an API that may use location tracking. It's verbose, and a
little intimidating. And it requires code outside of `bevy_ecs` that
wants to use location tracking needs to either unconditionally enable
the feature, or include conditional compilation of its own. It would be
good for users to be able to log locations when they are available
without needing to add feature flags to their own crates.
Reduce the number of cases where code compiles with the `track_location`
feature enabled, but not with it disabled, or vice versa. It can be hard
to remember to test it both ways!
Remove the need to store a `None` in `HookContext` when the
`track_location` feature is disabled.
## Solution
Create an `MaybeLocation<T>` type that contains a `T` if the
`track_location` feature is enabled, and is a ZST if it is not. The
overall API is similar to `Option`, but whether the value is `Some` or
`None` is set at compile time and is the same for all values.
Default `T` to `&'static Location<'static>`, since that is the most
common case.
Remove all `cfg(feature = "track_location")` blocks outside of the
implementation of that type, and instead call methods on it.
When `track_location` is disabled, `MaybeLocation` is a ZST and all
methods are `#[inline]` and empty, so they should be entirely removed by
the compiler. But the code will still be visible to the compiler and
checked, so if it compiles with the feature disabled then it should also
compile with it enabled, and vice versa.
## Open Questions
Where should these types live? I put them in `change_detection` because
that's where the existing `MaybeLocation` types were, but we now use
these outside of change detection.
While I believe that the compiler should be able to remove all of these
calls, I have not actually tested anything. If we want to take this
approach, what testing is required to ensure it doesn't impact
performance?
## Migration Guide
Methods like `Ref::changed_by()` that return a `&'static
Location<'static>` will now be available even when the `track_location`
feature is disabled, but they will return a new `MaybeLocation` type.
`MaybeLocation` wraps a `&'static Location<'static>` when the feature is
enabled, and is a ZST when the feature is disabled.
Existing code that needs a `&Location` can call `into_option().unwrap()`
to recover it. Many trait impls are forwarded, so if you only need
`Display` then no changes will be necessary.
If that code was conditionally compiled, you may instead want to use the
methods on `MaybeLocation` to remove the need for conditional
compilation.
Code that constructs a `Ref`, `Mut`, `Res`, or `ResMut` will now need to
provide location information unconditionally. If you are creating them
from existing Bevy types, you can obtain a `MaybeLocation` from methods
like `Table::get_changed_by_slice_for()` or
`ComponentSparseSet::get_with_ticks`. Otherwise, you will need to store
a `MaybeLocation` next to your data and use methods like `as_ref()` or
`as_mut()` to obtain wrapped references.
# Objective
Fixes#15417.
## Solution
- Remove the `labeled_assets` fields from `LoadedAsset` and
`ErasedLoadedAsset`.
- Created new structs `CompleteLoadedAsset` and
`CompleteErasedLoadedAsset` to hold the `labeled_subassets`.
- When a subasset is `LoadContext::finish`ed, it produces a
`CompleteLoadedAsset`.
- When a `CompleteLoadedAsset` is added to a `LoadContext` (as a
subasset), their `labeled_assets` are merged, reporting any overlaps.
One important detail to note: nested subassets with overlapping names
could in theory have been used in the past for the purposes of asset
preprocessing. Even though there was no way to access these "shadowed"
nested subassets, asset preprocessing does get access to these nested
subassets. This does not seem like a case we should support though. It
is confusing at best.
## Testing
- This is just a refactor.
---
## Migration Guide
- Most uses of `LoadedAsset` and `ErasedLoadedAsset` should be replaced
with `CompleteLoadedAsset` and `CompleteErasedLoadedAsset` respectively.
# Objective
https://github.com/bevyengine/bevy/pull/16966 tried to fix a bug where
`slot` wasn't passed to `parallaxed_uv` when not running under bindless,
but failed to account for meshlets. This surfaces on macOS where
bindless is disabled.
## Solution
Lift the slot variable out of the bindless condition so it's always
available.
# Objective
It's difficult to understand or make changes to the UI systems because
of how each system needs to individually track changes to scale factor,
windows and camera targets in local hashmaps, particularly for new
contributors. Any major change inevitably introduces new scale factor
bugs.
Instead of per-system resolution we can resolve the camera target info
for all UI nodes in a system at the start of `PostUpdate` and then store
it per-node in components that can be queried with change detection.
Fixes#17578Fixes#15143
## Solution
Store the UI render target's data locally per node in a component that
is updated in `PostUpdate` before any other UI systems run.
This component can be then be queried with change detection so that UI
systems no longer need to have knowledge of cameras and windows and
don't require fragile custom change detection solutions using local
hashmaps.
## Showcase
Compare `measure_text_system` from main (which has a bug the causes it
to use the wrong scale factor when a node's camera target changes):
```
pub fn measure_text_system(
mut scale_factors_buffer: Local<EntityHashMap<f32>>,
mut last_scale_factors: Local<EntityHashMap<f32>>,
fonts: Res<Assets<Font>>,
camera_query: Query<(Entity, &Camera)>,
default_ui_camera: DefaultUiCamera,
ui_scale: Res<UiScale>,
mut text_query: Query<
(
Entity,
Ref<TextLayout>,
&mut ContentSize,
&mut TextNodeFlags,
&mut ComputedTextBlock,
Option<&UiTargetCamera>,
),
With<Node>,
>,
mut text_reader: TextUiReader,
mut text_pipeline: ResMut<TextPipeline>,
mut font_system: ResMut<CosmicFontSystem>,
) {
scale_factors_buffer.clear();
let default_camera_entity = default_ui_camera.get();
for (entity, block, content_size, text_flags, computed, maybe_camera) in &mut text_query {
let Some(camera_entity) = maybe_camera
.map(UiTargetCamera::entity)
.or(default_camera_entity)
else {
continue;
};
let scale_factor = match scale_factors_buffer.entry(camera_entity) {
Entry::Occupied(entry) => *entry.get(),
Entry::Vacant(entry) => *entry.insert(
camera_query
.get(camera_entity)
.ok()
.and_then(|(_, c)| c.target_scaling_factor())
.unwrap_or(1.0)
* ui_scale.0,
),
};
if last_scale_factors.get(&camera_entity) != Some(&scale_factor)
|| computed.needs_rerender()
|| text_flags.needs_measure_fn
|| content_size.is_added()
{
create_text_measure(
entity,
&fonts,
scale_factor.into(),
text_reader.iter(entity),
block,
&mut text_pipeline,
content_size,
text_flags,
computed,
&mut font_system,
);
}
}
core::mem::swap(&mut *last_scale_factors, &mut *scale_factors_buffer);
}
```
with `measure_text_system` from this PR (which always uses the correct
scale factor):
```
pub fn measure_text_system(
fonts: Res<Assets<Font>>,
mut text_query: Query<
(
Entity,
Ref<TextLayout>,
&mut ContentSize,
&mut TextNodeFlags,
&mut ComputedTextBlock,
Ref<ComputedNodeTarget>,
),
With<Node>,
>,
mut text_reader: TextUiReader,
mut text_pipeline: ResMut<TextPipeline>,
mut font_system: ResMut<CosmicFontSystem>,
) {
for (entity, block, content_size, text_flags, computed, computed_target) in &mut text_query {
// Note: the ComputedTextBlock::needs_rerender bool is cleared in create_text_measure().
if computed_target.is_changed()
|| computed.needs_rerender()
|| text_flags.needs_measure_fn
|| content_size.is_added()
{
create_text_measure(
entity,
&fonts,
computed_target.scale_factor.into(),
text_reader.iter(entity),
block,
&mut text_pipeline,
content_size,
text_flags,
computed,
&mut font_system,
);
}
}
}
```
## Testing
I removed an alarming number of tests from the `layout` module but they
were mostly to do with the deleted camera synchronisation logic. The
remaining tests should all pass now.
The most relevant examples are `multiple_windows` and `split_screen`,
the behaviour of both should be unchanged from main.
---------
Co-authored-by: UkoeHB <37489173+UkoeHB@users.noreply.github.com>
Co-authored-by: Alice Cecile <alice.i.cecile@gmail.com>
## Objective
A major critique of Bevy at the moment is how boilerplatey it is to
compose (and read) entity hierarchies:
```rust
commands
.spawn(Foo)
.with_children(|p| {
p.spawn(Bar).with_children(|p| {
p.spawn(Baz);
});
p.spawn(Bar).with_children(|p| {
p.spawn(Baz);
});
});
```
There is also currently no good way to statically define and return an
entity hierarchy from a function. Instead, people often do this
"internally" with a Commands function that returns nothing, making it
impossible to spawn the hierarchy in other cases (direct World spawns,
ChildSpawner, etc).
Additionally, because this style of API results in creating the
hierarchy bits _after_ the initial spawn of a bundle, it causes ECS
archetype changes (and often expensive table moves).
Because children are initialized after the fact, we also can't count
them to pre-allocate space. This means each time a child inserts itself,
it has a high chance of overflowing the currently allocated capacity in
the `RelationshipTarget` collection, causing literal worst-case
reallocations.
We can do better!
## Solution
The Bundle trait has been extended to support an optional
`BundleEffect`. This is applied directly to World immediately _after_
the Bundle has fully inserted. Note that this is
[intentionally](https://github.com/bevyengine/bevy/discussions/16920)
_not done via a deferred Command_, which would require repeatedly
copying each remaining subtree of the hierarchy to a new command as we
walk down the tree (_not_ good performance).
This allows us to implement the new `SpawnRelated` trait for all
`RelationshipTarget` impls, which looks like this in practice:
```rust
world.spawn((
Foo,
Children::spawn((
Spawn((
Bar,
Children::spawn(Spawn(Baz)),
)),
Spawn((
Bar,
Children::spawn(Spawn(Baz)),
)),
))
))
```
`Children::spawn` returns `SpawnRelatedBundle<Children, L:
SpawnableList>`, which is a `Bundle` that inserts `Children`
(preallocated to the size of the `SpawnableList::size_hint()`).
`Spawn<B: Bundle>(pub B)` implements `SpawnableList` with a size of 1.
`SpawnableList` is also implemented for tuples of `SpawnableList` (same
general pattern as the Bundle impl).
There are currently three built-in `SpawnableList` implementations:
```rust
world.spawn((
Foo,
Children::spawn((
Spawn(Name::new("Child1")),
SpawnIter(["Child2", "Child3"].into_iter().map(Name::new),
SpawnWith(|parent: &mut ChildSpawner| {
parent.spawn(Name::new("Child4"));
parent.spawn(Name::new("Child5"));
})
)),
))
```
We get the benefits of "structured init", but we have nice flexibility
where it is required!
Some readers' first instinct might be to try to remove the need for the
`Spawn` wrapper. This is impossible in the Rust type system, as a tuple
of "child Bundles to be spawned" and a "tuple of Components to be added
via a single Bundle" is ambiguous in the Rust type system. There are two
ways to resolve that ambiguity:
1. By adding support for variadics to the Rust type system (removing the
need for nested bundles). This is out of scope for this PR :)
2. Using wrapper types to resolve the ambiguity (this is what I did in
this PR).
For the single-entity spawn cases, `Children::spawn_one` does also
exist, which removes the need for the wrapper:
```rust
world.spawn((
Foo,
Children::spawn_one(Bar),
))
```
## This works for all Relationships
This API isn't just for `Children` / `ChildOf` relationships. It works
for any relationship type, and they can be mixed and matched!
```rust
world.spawn((
Foo,
Observers::spawn((
Spawn(Observer::new(|trigger: Trigger<FuseLit>| {})),
Spawn(Observer::new(|trigger: Trigger<Exploded>| {})),
)),
OwnerOf::spawn(Spawn(Bar))
Children::spawn(Spawn(Baz))
))
```
## Macros
While `Spawn` is necessary to satisfy the type system, we _can_ remove
the need to express it via macros. The example above can be expressed
more succinctly using the new `children![X]` macro, which internally
produces `Children::spawn(Spawn(X))`:
```rust
world.spawn((
Foo,
children![
(
Bar,
children![Baz],
),
(
Bar,
children![Baz],
),
]
))
```
There is also a `related!` macro, which is a generic version of the
`children!` macro that supports any relationship type:
```rust
world.spawn((
Foo,
related!(Children[
(
Bar,
related!(Children[Baz]),
),
(
Bar,
related!(Children[Baz]),
),
])
))
```
## Returning Hierarchies from Functions
Thanks to these changes, the following pattern is now possible:
```rust
fn button(text: &str, color: Color) -> impl Bundle {
(
Node {
width: Val::Px(300.),
height: Val::Px(100.),
..default()
},
BackgroundColor(color),
children![
Text::new(text),
]
)
}
fn ui() -> impl Bundle {
(
Node {
width: Val::Percent(100.0),
height: Val::Percent(100.0),
..default(),
},
children![
button("hello", BLUE),
button("world", RED),
]
)
}
// spawn from a system
fn system(mut commands: Commands) {
commands.spawn(ui());
}
// spawn directly on World
world.spawn(ui());
```
## Additional Changes and Notes
* `Bundle::from_components` has been split out into
`BundleFromComponents::from_components`, enabling us to implement
`Bundle` for types that cannot be "taken" from the ECS (such as the new
`SpawnRelatedBundle`).
* The `NoBundleEffect` trait (which implements `BundleEffect`) is
implemented for empty tuples (and tuples of empty tuples), which allows
us to constrain APIs to only accept bundles that do not have effects.
This is critical because the current batch spawn APIs cannot efficiently
apply BundleEffects in their current form (as doing so in-place could
invalidate the cached raw pointers). We could consider allocating a
buffer of the effects to be applied later, but that does have
performance implications that could offset the balance and value of the
batched APIs (and would likely require some refactors to the underlying
code). I've decided to be conservative here. We can consider relaxing
that requirement on those APIs later, but that should be done in a
followup imo.
* I've ported a few examples to illustrate real-world usage. I think in
a followup we should port all examples to the `children!` form whenever
possible (and for cases that require things like SpawnIter, use the raw
APIs).
* Some may ask "why not use the `Relationship` to spawn (ex:
`ChildOf::spawn(Foo)`) instead of the `RelationshipTarget` (ex:
`Children::spawn(Spawn(Foo))`)?". That _would_ allow us to remove the
`Spawn` wrapper. I've explicitly chosen to disallow this pattern.
`Bundle::Effect` has the ability to create _significant_ weirdness.
Things in `Bundle` position look like components. For example
`world.spawn((Foo, ChildOf::spawn(Bar)))` _looks and reads_ like Foo is
a child of Bar. `ChildOf` is in Foo's "component position" but it is not
a component on Foo. This is a huge problem. Now that `Bundle::Effect`
exists, we should be _very_ principled about keeping the "weird and
unintuitive behavior" to a minimum. Things that read like components
_should be the components they appear to be".
## Remaining Work
* The macros are currently trivially implemented using macro_rules and
are currently limited to the max tuple length. They will require a
proc_macro implementation to work around the tuple length limit.
## Next Steps
* Port the remaining examples to use `children!` where possible and raw
`Spawn` / `SpawnIter` / `SpawnWith` where the flexibility of the raw API
is required.
## Migration Guide
Existing spawn patterns will continue to work as expected.
Manual Bundle implementations now require a `BundleEffect` associated
type. Exisiting bundles would have no bundle effect, so use `()`.
Additionally `Bundle::from_components` has been moved to the new
`BundleFromComponents` trait.
```rust
// Before
unsafe impl Bundle for X {
unsafe fn from_components<T, F>(ctx: &mut T, func: &mut F) -> Self {
}
/* remaining bundle impl here */
}
// After
unsafe impl Bundle for X {
type Effect = ();
/* remaining bundle impl here */
}
unsafe impl BundleFromComponents for X {
unsafe fn from_components<T, F>(ctx: &mut T, func: &mut F) -> Self {
}
}
```
---------
Co-authored-by: Alice Cecile <alice.i.cecile@gmail.com>
Co-authored-by: Gino Valente <49806985+MrGVSV@users.noreply.github.com>
Co-authored-by: Emerson Coskey <emerson@coskey.dev>
# Objective
Solves https://github.com/bevyengine/bevy/issues/17747.
## Solution
- Adds an example for creating a default value for Local.
## Testing
- Example code compiles and passes assertions.
This pr uses the `extern crate self as` trick to make proc macros behave
the same way inside and outside bevy.
# Objective
- Removes noise introduced by `crate as` in the whole bevy repo.
- Fixes#17004.
- Hardens proc macro path resolution.
## TODO
- [x] `BevyManifest` needs cleanup.
- [x] Cleanup remaining `crate as`.
- [x] Add proper integration tests to the ci.
## Notes
- `cargo-manifest-proc-macros` is written by me and based/inspired by
the old `BevyManifest` implementation and
[`bkchr/proc-macro-crate`](https://github.com/bkchr/proc-macro-crate).
- What do you think about the new integration test machinery I added to
the `ci`?
More and better integration tests can be added at a later stage.
The goal of these integration tests is to simulate an actual separate
crate that uses bevy. Ideally they would lightly touch all bevy crates.
## Testing
- Needs RA test
- Needs testing from other users
- Others need to run at least `cargo run -p ci integration-test` and
verify that they work.
---------
Co-authored-by: Alice Cecile <alice.i.cecile@gmail.com>
Didn't remove WgpuWrapper. Not sure if it's needed or not still.
## Testing
- Did you test these changes? If so, how? Example runner
- Are there any parts that need more testing? Web (portable atomics
thingy?), DXC.
## Migration Guide
- Bevy has upgraded to [wgpu
v24](https://github.com/gfx-rs/wgpu/blob/trunk/CHANGELOG.md#v2400-2025-01-15).
- When using the DirectX 12 rendering backend, the new priority system
for choosing a shader compiler is as follows:
- If the `WGPU_DX12_COMPILER` environment variable is set at runtime, it
is used
- Else if the new `statically-linked-dxc` feature is enabled, a custom
version of DXC will be statically linked into your app at compile time.
- Else Bevy will look in the app's working directory for
`dxcompiler.dll` and `dxil.dll` at runtime.
- Else if they are missing, Bevy will fall back to FXC (not recommended)
---------
Co-authored-by: Alice Cecile <alice.i.cecile@gmail.com>
Co-authored-by: IceSentry <c.giguere42@gmail.com>
Co-authored-by: François Mockers <francois.mockers@vleue.com>
# Objective
Fix#16477.
## Solution
- Remove temporary silence introduced in #16763
- bump version of `notify-debouncer-full` to remove transitive
dependency on `instant` crate.
# Objective
- publish script copy the license files to all subcrates, meaning that
all publish are dirty. this breaks git verification of crates
- the order and list of crates to publish is manually maintained,
leading to error. cargo 1.84 is more strict and the list is currently
wrong
## Solution
- duplicate all the licenses to all crates and remove the
`--allow-dirty` flag
- instead of a manual list of crates, get it from `cargo package
--workspace`
- remove the `--no-verify` flag to... verify more things?
# Objective
Things were breaking post-cs.
## Solution
`specialize_mesh_materials` must run after
`collect_meshes_for_gpu_building`. Therefore, its placement in the
`PrepareAssets` set didn't make sense (also more generally). To fix, we
put this class of system in ~`PrepareResources`~ `QueueMeshes`, although
it potentially could use a more descriptive location. We may want to
review the placement of `check_views_need_specialization` which is also
currently in `PrepareAssets`.
PR #17684 broke occlusion culling because it neglected to set the
indirect parameter offsets for the late mesh preprocessing stage if the
work item buffers were already set. This PR moves the update of those
values to a new function, `init_work_item_buffers`, which is
unconditionally called for every phase every frame.
Note that there's some complexity in order to handle the case in which
occlusion culling was enabled on one frame and disabled on the next, or
vice versa. This was necessary in order to make the occlusion culling
toggle in the `occlusion_culling` example work again.
Right now, we key the cached light change ticks off the `LightEntity`.
This uses the render world entity, which isn't stable between frames.
Thus in practice few shadows are retained from frame to frame. This PR
fixes the issue by keying off the `RetainedViewEntity` instead, which is
designed to be stable from frame to frame.
This PR makes Bevy keep entities in bins from frame to frame if they
haven't changed. This reduces the time spent in `queue_material_meshes`
and related functions to near zero for static geometry. This patch uses
the same change tick technique that #17567 uses to detect when meshes
have changed in such a way as to require re-binning.
In order to quickly find the relevant bin for an entity when that entity
has changed, we introduce a new type of cache, the *bin key cache*. This
cache stores a mapping from main world entity ID to cached bin key, as
well as the tick of the most recent change to the entity. As we iterate
through the visible entities in `queue_material_meshes`, we check the
cache to see whether the entity needs to be re-binned. If it doesn't,
then we mark it as clean in the `valid_cached_entity_bin_keys` bit set.
If it does, then we insert it into the correct bin, and then mark the
entity as clean. At the end, all entities not marked as clean are
removed from the bins.
This patch has a dramatic effect on the rendering performance of most
benchmarks, as it effectively eliminates `queue_material_meshes` from
the profile. Note, however, that it generally simultaneously regresses
`batch_and_prepare_binned_render_phase` by a bit (not by enough to
outweigh the win, however). I believe that's because, before this patch,
`queue_material_meshes` put the bins in the CPU cache for
`batch_and_prepare_binned_render_phase` to use, while with this patch,
`batch_and_prepare_binned_render_phase` must load the bins into the CPU
cache itself.
On Caldera, this reduces the time spent in `queue_material_meshes` from
5+ ms to 0.2ms-0.3ms. Note that benchmarking on that scene is very noisy
right now because of https://github.com/bevyengine/bevy/issues/17535.
# Objective
The docs of `EaseFunction` don't visualize the different functions,
requiring you to check out the Bevy repo and running the
`easing_function` example.
## Solution
- Add tool to generate suitable svg graphs. This only needs to be re-run
when adding new ease functions.
- works with all themes
- also add missing easing functions to example.
---
## Showcase
---------
Co-authored-by: François Mockers <mockersf@gmail.com>
Right now, meshes aren't grouped together based on the bindless texture
slab when drawing shadows. This manifests itself as flickering in
Bistro. I believe that there are two causes of this:
1. Alpha masked shadows may try to sample from the wrong texture,
causing the alpha mask to appear and disappear.
2. Objects may try to sample from the blank textures that we pad out the
bindless slabs with, causing them to vanish intermittently.
This commit fixes the issue by including the material bind group ID as
part of the shadow batch set key, just as we do for the prepass and main
pass.
# Objective
Fixes#17718
## Solution
Schedule `text_system` before `AssetEvents`.
I guess what was happening here is that glyphs weren't shown because
`text_system` was running before `AssetEevents` and so `prepare_uinodes`
never recieves the the asset modified event about the glyph texture
atlas image.
Fixes#17535
Bevy's approach to handling "entity mapping" during spawning and cloning
needs some work. The addition of
[Relations](https://github.com/bevyengine/bevy/pull/17398) both
[introduced a new "duplicate entities" bug when spawning scenes in the
scene system](#17535) and made the weaknesses of the current mapping
system exceedingly clear:
1. Entity mapping requires _a ton_ of boilerplate (implement or derive
VisitEntities and VisitEntitesMut, then register / reflect MapEntities).
Knowing the incantation is challenging and if you forget to do it in
part or in whole, spawning subtly breaks.
2. Entity mapping a spawned component in scenes incurs unnecessary
overhead: look up ReflectMapEntities, create a _brand new temporary
instance_ of the component using FromReflect, map the entities in that
instance, and then apply that on top of the actual component using
reflection. We can do much better.
Additionally, while our new [Entity cloning
system](https://github.com/bevyengine/bevy/pull/16132) is already pretty
great, it has some areas we can make better:
* It doesn't expose semantic info about the clone (ex: ignore or "clone
empty"), meaning we can't key off of that in places where it would be
useful, such as scene spawning. Rather than duplicating this info across
contexts, I think it makes more sense to add that info to the clone
system, especially given that we'd like to use cloning code in some of
our spawning scenarios.
* EntityCloner is currently built in a way that prioritizes a single
entity clone
* EntityCloner's recursive cloning is built to be done "inside out" in a
parallel context (queue commands that each have a clone of
EntityCloner). By making EntityCloner the orchestrator of the clone we
can remove internal arcs, improve the clarity of the code, make
EntityCloner mutable again, and simplify the builder code.
* EntityCloner does not currently take into account entity mapping. This
is necessary to do true "bullet proof" cloning, would allow us to unify
the per-component scene spawning and cloning UX, and ultimately would
allow us to use EntityCloner in place of raw reflection for scenes like
`Scene(World)` (which would give us a nice performance boost: fewer
archetype moves, less reflection overhead).
## Solution
### Improved Entity Mapping
First, components now have first-class "entity visiting and mapping"
behavior:
```rust
#[derive(Component, Reflect)]
#[reflect(Component)]
struct Inventory {
size: usize,
#[entities]
items: Vec<Entity>,
}
```
Any field with the `#[entities]` annotation will be viewable and
mappable when cloning and spawning scenes.
Compare that to what was required before!
```rust
#[derive(Component, Reflect, VisitEntities, VisitEntitiesMut)]
#[reflect(Component, MapEntities)]
struct Inventory {
#[visit_entities(ignore)]
size: usize,
items: Vec<Entity>,
}
```
Additionally, for relationships `#[entities]` is implied, meaning this
"just works" in scenes and cloning:
```rust
#[derive(Component, Reflect)]
#[relationship(relationship_target = Children)]
#[reflect(Component)]
struct ChildOf(pub Entity);
```
Note that Component _does not_ implement `VisitEntities` directly.
Instead, it has `Component::visit_entities` and
`Component::visit_entities_mut` methods. This is for a few reasons:
1. We cannot implement `VisitEntities for C: Component` because that
would conflict with our impl of VisitEntities for anything that
implements `IntoIterator<Item=Entity>`. Preserving that impl is more
important from a UX perspective.
2. We should not implement `Component: VisitEntities` VisitEntities in
the Component derive, as that would increase the burden of manual
Component trait implementors.
3. Making VisitEntitiesMut directly callable for components would make
it easy to invalidate invariants defined by a component author. By
putting it in the `Component` impl, we can make it harder to call
naturally / unavailable to autocomplete using `fn
visit_entities_mut(this: &mut Self, ...)`.
`ReflectComponent::apply_or_insert` is now
`ReflectComponent::apply_or_insert_mapped`. By moving mapping inside
this impl, we remove the need to go through the reflection system to do
entity mapping, meaning we no longer need to create a clone of the
target component, map the entities in that component, and patch those
values on top. This will make spawning mapped entities _much_ faster
(The default `Component::visit_entities_mut` impl is an inlined empty
function, so it will incur no overhead for unmapped entities).
### The Bug Fix
To solve #17535, spawning code now skips entities with the new
`ComponentCloneBehavior::Ignore` and
`ComponentCloneBehavior::RelationshipTarget` variants (note
RelationshipTarget is a temporary "workaround" variant that allows
scenes to skip these components. This is a temporary workaround that can
be removed as these cases should _really_ be using EntityCloner logic,
which should be done in a followup PR. When that is done,
`ComponentCloneBehavior::RelationshipTarget` can be merged into the
normal `ComponentCloneBehavior::Custom`).
### Improved Cloning
* `Option<ComponentCloneHandler>` has been replaced by
`ComponentCloneBehavior`, which encodes additional intent and context
(ex: `Default`, `Ignore`, `Custom`, `RelationshipTarget` (this last one
is temporary)).
* Global per-world entity cloning configuration has been removed. This
felt overly complicated, increased our API surface, and felt too
generic. Each clone context can have different requirements (ex: what a
user wants in a specific system, what a scene spawner wants, etc). I'd
prefer to see how far context-specific EntityCloners get us first.
* EntityCloner's internals have been reworked to remove Arcs and make it
mutable.
* EntityCloner is now directly stored on EntityClonerBuilder,
simplifying the code somewhat
* EntityCloner's "bundle scratch" pattern has been moved into the new
BundleScratch type, improving its usability and making it usable in
other contexts (such as future cross-world cloning code). Currently this
is still private, but with some higher level safe APIs it could be used
externally for making dynamic bundles
* EntityCloner's recursive cloning behavior has been "externalized". It
is now responsible for orchestrating recursive clones, meaning it no
longer needs to be sharable/clone-able across threads / read-only.
* EntityCloner now does entity mapping during clones, like scenes do.
This gives behavior parity and also makes it more generically useful.
* `RelatonshipTarget::RECURSIVE_SPAWN` is now
`RelationshipTarget::LINKED_SPAWN`, and this field is used when cloning
relationship targets to determine if cloning should happen recursively.
The new `LINKED_SPAWN` term was picked to make it more generically
applicable across spawning and cloning scenarios.
## Next Steps
* I think we should adapt EntityCloner to support cross world cloning. I
think this PR helps set the stage for that by making the internals
slightly more generalized. We could have a CrossWorldEntityCloner that
reuses a lot of this infrastructure.
* Once we support cross world cloning, we should use EntityCloner to
spawn `Scene(World)` scenes. This would yield significant performance
benefits (no archetype moves, less reflection overhead).
---------
Co-authored-by: eugineerd <70062110+eugineerd@users.noreply.github.com>
Co-authored-by: Alice Cecile <alice.i.cecile@gmail.com>
# Objective
This is a follow up to #9822, which automatically adds sync points
during the Schedule build process.
However, the implementation in #9822 feels very "special case" to me. As
the number of things we want to do with the `Schedule` grows, we need a
modularized way to manage those behaviors. For example, in one of my
current experiments I want to automatically add systems to apply GPU
pipeline barriers between systems accessing GPU resources.
For dynamic modifications of the schedule, we mostly need these
capabilities:
- Storing custom data on schedule edges
- Storing custom data on schedule nodes
- Modify the schedule graph whenever it builds
These should be enough to allows us to add "hooks" to the schedule build
process for various reasons.
cc @hymm
## Solution
This PR abstracts the process of schedule modification and created a new
trait, `ScheduleBuildPass`. Most of the logics in #9822 were moved to an
implementation of `ScheduleBuildPass`, `AutoInsertApplyDeferredPass`.
Whether a dependency edge should "ignore deferred" is now indicated by
the presence of a marker struct, `IgnoreDeferred`.
This PR has no externally visible effects. However, in a future PR I
propose to change the `before_ignore_deferred` and
`after_ignore_deferred` API into a more general form,
`before_with_options` and `after_with_options`.
```rs
schedule.add_systems(
system.before_with_options(another_system, IgnoreDeferred)
);
schedule.add_systems(
system.before_with_options(another_system, (
IgnoreDeferred,
AnyOtherOption {
key: value
}
))
);
schedule.add_systems(
system.before_with_options(another_system, ())
);
```
# Objective
- Make use of the new `weak_handle!` macro added in
https://github.com/bevyengine/bevy/pull/17384
## Solution
- Migrate bevy from `Handle::weak_from_u128` to the new `weak_handle!`
macro that takes a random UUID
- Deprecate `Handle::weak_from_u128`, since there are no remaining use
cases that can't also be addressed by constructing the type manually
## Testing
- `cargo run -p ci -- test`
---
## Migration Guide
Replace `Handle::weak_from_u128` with `weak_handle!` and a random UUID.
# Objective
There was a bug in the default `Relationship::on_insert` implementation
that caused it to not properly handle entities targeting themselves in
relationships. The relationship component was properly removed, but it
would go on to add itself to its own target component.
## Solution
Added a missing `return` and a couple of tests
(`self_relationship_fails` failed on its second assert prior to this
PR).
## Testing
See above.
# Objective
Progresses #17569. The end goal here is to synchronize component
registration. See the other PR for details for the motivation behind
that.
For this PR specifically, the objective is to decouple `Components` from
`Storages`. What components are registered etc should have nothing to do
with what Storages looks like. Storages should only care about what
entity archetypes have been spawned.
## Solution
Previously, this was used to create sparse sets for relevant components
when those components were registered. Now, we do that when the
component is inserted/spawned.
This PR proposes doing that in `BundleInfo::new`, but there may be a
better place.
## Testing
In theory, this shouldn't have changed any functionality, so no new
tests were created. I'm not aware of any examples that make heavy use of
sparse set components either.
## Migration Guide
- Remove storages from functions where it is no longer needed.
- Note that SparseSets are no longer present for all registered sparse
set components, only those that have been spawned.
---------
Co-authored-by: SpecificProtagonist <vincentjunge@posteo.net>
Co-authored-by: Chris Russell <8494645+chescock@users.noreply.github.com>
# Objective
- Fixes#17411
## Solution
- Deprecated `Component::register_component_hooks`
- Added individual methods for each hook which return `None` if the hook
is unused.
## Testing
- CI
---
## Migration Guide
`Component::register_component_hooks` is now deprecated and will be
removed in a future release. When implementing `Component` manually,
also implement the respective hook methods on `Component`.
```rust
// Before
impl Component for Foo {
// snip
fn register_component_hooks(hooks: &mut ComponentHooks) {
hooks.on_add(foo_on_add);
}
}
// After
impl Component for Foo {
// snip
fn on_add() -> Option<ComponentHook> {
Some(foo_on_add)
}
}
```
## Notes
I've chosen to deprecate `Component::register_component_hooks` rather
than outright remove it to ease the migration guide. While it is in a
state of deprecation, it must be used by
`Components::register_component_internal` to ensure users who haven't
migrated to the new hook definition scheme aren't left behind. For users
of the new scheme, a default implementation of
`Component::register_component_hooks` is provided which forwards the new
individual hook implementations.
Personally, I think this is a cleaner API to work with, and would allow
the documentation for hooks to exist on the respective `Component`
methods (e.g., documentation for `OnAdd` can exist on
`Component::on_add`). Ideally, `Component::on_add` would be the hook
itself rather than a getter for the hook, but it is the only way to
early-out for a no-op hook, which is important for performance.
## Migration Guide
`Component::register_component_hooks` has been deprecated. If you are
manually implementing the `Component` trait and registering hooks there,
use the individual methods such as `on_add` instead for increased
clarity.
# Objective
- A common bevy pattern is to pre-allocate a weak `Handle` with a
static, random ID and fill it during `Plugin::build` via
`load_internal_asset!`
- This requires generating a random 128-bit number that is interpreted
as a UUID. This is much less convenient than generating a UUID directly,
and also, strictly speaking, error prone, since it often results in an
invalid UUIDv4 – they have to follow the pattern
`xxxxxxxx-xxxx-4xxx-xxxx-xxxxxxxxxxxx`, where `x` is a random nibble (in
practice this doesn't matter, since the UUID is just interpreted as a
bag of bytes).
## Solution
- Add a `weak_handle!` macro that internally calls
[`uuid::uuid!`](https://docs.rs/uuid/1.12.0/uuid/macro.uuid.html) to
parse a UUID from a string literal.
- Now any random UUID generation tool can be used to generate an asset
ID, such as `uuidgen` or entering "uuid" in DuckDuckGo.
Previously:
```rust
const SHADER: Handle<Shader> = Handle::weak_from_u128(314685653797097581405914117016993910609);
```
After this PR:
```rust
const SHADER: Handle<Shader> = weak_handle!("1347c9b7-c46a-48e7-b7b8-023a354b7cac");
```
Note that I did not yet migrate any of the existing uses. I can do that
if desired, but want to have some feedback first to avoid wasted effort.
## Testing
Tested via the included doctest.
# Objective
Basic `TextShadow` support.
## Solution
New `TextShadow` component with `offset` and `color` fields. Just insert
it on a `Text` node to add a shadow.
New system `extract_text_shadows` handles rendering.
It's not "real" shadows just the text redrawn with an offset and a
different colour. Blur-radius support will need changes to the shaders
and be a lot more complicated, whereas this still looks okay and took a
couple of minutes to implement.
I added the `TextShadow` component to `bevy_ui` rather than `bevy_text`
because it only supports the UI atm.
We can add a `Text2d` version in a followup but getting the same effect
in `Text2d` is trivial even without official support.
---
## Showcase
<img width="122" alt="text_shadow"
src="https://github.com/user-attachments/assets/0333d167-c507-4262-b93b-b6d39e2cf3a4"
/>
<img width="136" alt="g"
src="https://github.com/user-attachments/assets/9b01d5d9-55c9-4af7-9360-a7b04f55944d"
/>
# Objective
Fixes#17662
## Solution
Moved `Item` and `fetch` from `WorldQuery` to `QueryData`, and adjusted
their implementations accordingly.
Currently, documentation related to `fetch` is written under
`WorldQuery`. It would be more appropriate to move it to the `QueryData`
documentation for clarity.
I am not very experienced with making contributions. If there are any
mistakes or areas for improvement, I would appreciate any suggestions
you may have.
## Migration Guide
The `WorldQuery::Item` type and `WorldQuery::fetch` method have been
moved to `QueryData`, as they were not useful for `QueryFilter` types.
---------
Co-authored-by: Chris Russell <8494645+chescock@users.noreply.github.com>
# Objective
The feature gates for the `UiChildren` and `UiRootNodes` system params
make the unconstructable `GhostNode` `PhantomData` trick redundant.
## Solution
Remove the `GhostNode::new` method and change `GhostNode` into a unit
struct.
## Testing
```cargo run --example ghost_nodes```
still works
# Objective
Simplify and expand the API for `QueryState`.
`QueryState` has a lot of methods that mirror those on `Query`. These
are then multiplied by variants that take `&World`, `&mut World`, and
`UnsafeWorldCell`. In addition, many of them have `_manual` variants
that take `&QueryState` and avoid calling `update_archetypes()`. Not all
of the combinations exist, however, so some operations are not possible.
## Solution
Introduce methods to get a `Query` from a `QueryState`. That will reduce
duplication between the types, and ensure that the full `Query` API is
always available for `QueryState`.
Introduce methods on `Query` that consume the query to return types with
the full `'w` lifetime. This avoids issues with borrowing where things
like `query_state.query(&world).get(entity)` don't work because they
borrow from the temporary `Query`.
Finally, implement `Copy` for read-only `Query`s. `get_inner` and
`iter_inner` currently take `&self`, so changing them to consume `self`
would be a breaking change. By making `Query: Copy`, they can consume a
copy of `self` and continue to work.
The consuming methods also let us simplify the implementation of methods
on `Query`, by doing `fn foo(&self) { self.as_readonly().foo_inner() }`
and `fn foo_mut(&mut self) { self.reborrow().foo_inner() }`. That
structure makes it more difficult to accidentally extend lifetimes,
since the safe `as_readonly()` and `reborrow()` methods shrink them
appropriately. The optimizer is able to see that they are both identity
functions and inline them, so there should be no performance cost.
Note that this change would conflict with #15848. If `QueryState` is
stored as a `Cow`, then the consuming methods cannot be implemented, and
`Copy` cannot be implemented.
## Future Work
The next step is to mark the methods on `QueryState` as `#[deprecated]`,
and move the implementations into `Query`.
## Migration Guide
`Query::to_readonly` has been renamed to `Query::as_readonly`.
# Cold Specialization
## Objective
An ongoing part of our quest to retain everything in the render world,
cold-specialization aims to cache pipeline specialization so that
pipeline IDs can be recomputed only when necessary, rather than every
frame. This approach reduces redundant work in stable scenes, while
still accommodating scenarios in which materials, views, or visibility
might change, as well as unlocking future optimization work like
retaining render bins.
## Solution
Queue systems are split into a specialization system and queue system,
the former of which only runs when necessary to compute a new pipeline
id. Pipelines are invalidated using a combination of change detection
and ECS ticks.
### The difficulty with change detection
Detecting “what changed” can be tricky because pipeline specialization
depends not only on the entity’s components (e.g., mesh, material, etc.)
but also on which view (camera) it is rendering in. In other words, the
cache key for a given pipeline id is a view entity/render entity pair.
As such, it's not sufficient simply to react to change detection in
order to specialize -- an entity could currently be out of view or could
be rendered in the future in camera that is currently disabled or hasn't
spawned yet.
### Why ticks?
Ticks allow us to ensure correctness by allowing us to compare the last
time a view or entity was updated compared to the cached pipeline id.
This ensures that even if an entity was out of view or has never been
seen in a given camera before we can still correctly determine whether
it needs to be re-specialized or not.
## Testing
TODO: Tested a bunch of different examples, need to test more.
## Migration Guide
TODO
- `AssetEvents` has been moved into the `PostUpdate` schedule.
---------
Co-authored-by: Patrick Walton <pcwalton@mimiga.net>
# Objective
Follow-up to #17549 and #16547.
A large part of `Vec`s usefulness is behind its ability to be sliced,
like sorting f.e., so we want the same to be possible for
`UniqueEntityVec`.
## Solution
Add a `UniqueEntitySlice` type. It is a wrapper around `[T]`, and itself
a DST.
Because `mem::swap` has a `Sized` bound, DSTs cannot be swapped, and we
can freely hand out mutable subslices without worrying about the
uniqueness invariant of the backing collection!
`UniqueEntityVec` and the relevant `UniqueEntityIter`s now have methods
and trait impls that return `UniqueEntitySlice`s.
`UniqueEntitySlice` itself can deref into normal slices, which means we
can avoid implementing the vast majority of immutable slice methods.
Most of the remaining methods:
- split a slice/collection in further unique subsections/slices
- reorder the slice: `sort`, `rotate_*`, `swap`
- construct/deconstruct/convert pointer-like types: `Box`, `Arc`, `Rc`,
`Cow`
- are comparison trait impls
As this PR is already larger than I'd like, we leave several things to
follow-ups:
- `UniqueEntityArray` and the related slice methods that would return it
- denoted by "chunk", "array_*" for iterators
- Methods that return iterators with `UniqueEntitySlice` as their item
- `windows`, `chunks` and `split` families
- All methods that are capable of actively mutating individual elements.
While they could be offered unsafely, subslicing makes their safety
contract weird enough to warrant its own discussion.
- `fill_with`, `swap_with_slice`, `iter_mut`, `split_first/last_mut`,
`select_nth_unstable_*`
Note that `Arc`, `Rc` and `Cow` are not fundamental types, so even if
they contain `UniqueEntitySlice`, we cannot write direct trait impls for
them.
On top of that, `Cow` is not a receiver (like `self: Arc<Self>` is) so
we cannot write inherent methods for it either.
We were calling `clear()` on the work item buffer table, which caused us
to deallocate all the CPU side buffers. This patch changes the logic to
instead just clear the buffers individually, but leave their backing
stores. This has two consequences:
1. To effectively retain work item buffers from frame to frame, we need
to key them off `RetainedViewEntity` values and not the render world
`Entity`, which is transient. This PR changes those buffers accordingly.
2. We need to clean up work item buffers that belong to views that went
away. Amusingly enough, we actually have a system,
`delete_old_work_item_buffers`, that tries to do this already, but it
wasn't doing anything because the `clear_batched_gpu_instance_buffers`
system already handled that. This patch actually makes the
`delete_old_work_item_buffers` system useful, by removing the clearing
behavior from `clear_batched_gpu_instance_buffers` and instead making
`delete_old_work_item_buffers` delete buffers corresponding to
nonexistent views.
On Bistro, this PR improves the performance of
`batch_and_prepare_binned_render_phase` from 61.2 us to 47.8 us, a 28%
speedup.
This patch fixes a bug whereby we're re-extracting every mesh every
frame. It's a regression from PR #17413. The code in question has
actually been in the tree with this bug for quite a while; it's that
just the code didn't actually run unless the renderer considered the
previous view transforms necessary. Occlusion culling expanded the set
of circumstances under which Bevy computes the previous view transforms,
causing this bug to appear more often.
This patch fixes the issue by checking to see if the previous transform
of a mesh actually differs from the current transform before copying the
current transform to the previous transform.
# Objective
- Fixes CI failure due to `uuid` 1.13 using the new version of
`getrandom` which requires using a new API to work on Wasm.
## Solution
- Based on [`uuid` 1.13 release
notes](https://github.com/uuid-rs/uuid/releases/tag/1.13.0) I've enabled
the `js` feature on `wasm32`. This will need to be revisited once #17499
is up for review
- Updated minimum `uuid` version to 1.13.1, which fixes a separate issue
with `target_feature = atomics` on `wasm`.
## Testing
- `cargo check --target wasm32-unknown-unknown`
# Objective
Fix text 2d. Fixes https://github.com/bevyengine/bevy/issues/17670
## Solution
Evidently there's a 1:N extraction going on here that requires using the
render entity rather than main entity.
## Testing
Text 2d example
Data for the other batches is only accessed by the GPU, not the CPU, so
it's a waste of time and memory to store information relating to those
other batches.
On Bistro, this reduces time spent in
`batch_and_prepare_binned_render_phase` from 85.9 us to 61.2 us, a 40%
speedup.
# Objective
While working on #17649, I found the docs for `WorldQuery` and the
related traits frustratingly vague.
## Solution
Clarify them and add some more tangible advice.
Also fix a copy-pasted typo in related comments.
---------
Co-authored-by: James O'Brien <james.obrien@drafly.net>
# Objective
Allow mapping `Mut` to another value while returning a custom error on
failure.
## Solution
Added `try_map_unchanged` to `Mut` which returns a `Result` instead of
`Option` .
# Objective
- When obtaining an axis from the transform and putting that into
`Transform::rotate_axis` or `Transform::rotate_axis_local`, floating
point errors could accumulate exponentially, resulting in denormalized
rotation.
- This is an alternative to and closes#17604, due to lack of consent
around this in the [discord
discussion](https://discord.com/channels/691052431525675048/1203087353850364004/1334232710658392227)
- Closes#16480
## Solution
- Add a warning of this issue and a recommendation to normalize to the
API docs.
- Add a runtime warning that checks for denormalized axis in debug mode,
with a reference to the API docs.
# Objective
Prevent unsound uses of `DeferredWorld` as a `SystemParam`. It is
currently unsound because it does not check for existing access, and
because it incorrectly registers filtered access.
## Solution
Have `DeferredWorld` panic if a previous parameter has conflicting
access.
Have `DeferredWorld` update `archetype_component_access` so that the
multi-threaded executor sees the access.
Fix `FilteredAccessSet::read_all()` and `write_all()` to correctly add a
`FilteredAccess` with no filter so that `Query` is able to detect the
conflicts.
Remove redundant `read_all()` call, since `write_all()` already declares
read access.
Remove unnecessary `set_has_deferred()` call, since `<DeferredWorld as
SystemParam>::apply_deferred()` does nothing. Previously we were
inserting unnecessary `apply_deferred` systems in the schedule.
## Testing
Added unit tests for systems where `DeferredWorld` conflicts with a
`Query` in the same system.
# Objective
- Most of the `*MeshBuilder` classes are not implementing `Reflect`
## Solution
- Implementing `Reflect` for all `*MeshBuilder` were is possible.
- Make sure all `*MeshBuilder` implements `Default`.
- Adding new `MeshBuildersPlugin` that registers all `*MeshBuilder`
types.
## Testing
- `cargo run -p ci`
- Tested some examples like `3d_scene` just in case something was
broken.
# Objective
- Fix the atmosphere LUT parameterization in the aerial -view and
sky-view LUTs
- Correct the light accumulation according to a ray-marched reference
- Avoid negative values of the sun disk illuminance when the sun disk is
below the horizon
## Solution
- Adding a Newton's method iteration to `fast_sqrt` function
- Switched to using `fast_acos_4` for better precision of the sun angle
towards the horizon (view mu angle = 0)
- Simplified the function for mapping to and from the Sky View UV
coordinates by removing an if statement and correctly apply the method
proposed by the [Hillarie
paper](https://sebh.github.io/publications/egsr2020.pdf) detailed in
section 5.3 and 5.4.
- Replaced the `ray_dir_ws.y` term with a shadow factor in the
`sample_sun_illuminance` function that correctly approximates the sun
disk occluded by the earth from any view point
## Testing
- Ran the atmosphere and SSAO examples to make sure the shaders still
compile and run as expected.
---
## Showcase
<img width="1151" alt="showcase-img"
src="https://github.com/user-attachments/assets/de875533-42bd-41f9-9fd0-d7cc57d6e51c"
/>
---------
Co-authored-by: Emerson Coskey <emerson@coskey.dev>
# Objective
The method `World::as_unsafe_world_cell_readonly` is used to create an
`UnsafeWorldCell` which is only allowed to access world data immutably.
This can be tricky to use, as the data that an `UnsafeWorldCell` is
allowed to access exists only in documentation (you could think of it as
a "doc-time abstraction" rather than a "compile-time" abstraction). It's
quite easy to forget where a particular instance came from and attempt
to use it for mutable access, leading to instant, silent undefined
behavior.
## Solution
Add a debug-mode only flag to `UnsafeWorldCell` which tracks whether or
not the instance can be used to access world data mutably. This should
catch basic improper usages of `as_unsafe_world_cell_readonly`.
## Future work
There are a few ways that you can bypass the runtime checks introduced
by this PR:
* Any world accesses done via `UnsafeWorldCell::storages` are completely
invisible to these runtime checks. Unfortunately, `storages` constitutes
most of the world accesses used in the engine itself, so this PR will
mostly benefit downstream users of bevy.
* It's possible to call `get_resource_by_id`, and then convert the
returned `Ptr` to a `PtrMut` by calling `assert_unique`. In the future
we'll probably want to add a debug-mode only flag to `Ptr` which tracks
whether or not it can be upgraded to a `PtrMut`. I didn't include this
change in this PR as those types are currently defined using macros
which makes it a bit tricky to modify their definitions.
* Any data accesses done through a mutable `UnsafeWorldCell` are
completely unchecked, meaning it's possible to unsoundly create multiple
mutable references to a single component, for example. In the future we
may want to store an `Access<>` set inside of the world's `Storages` to
add granular debug-mode runtime checks.
That said, I'd consider this PR to be a good first step towards adding
full runtime checks to `UnsafeWorldCell`.
## Testing
Added a few tests that basic invalid mutable world access result in a
panic.
---------
Co-authored-by: Joseph <21144246+JoJoJet@users.noreply.github.com>
Co-authored-by: Alice I Cecile <alice.i.cecile@gmail.com>
# Objective
Our
[`TextSpan`](https://docs.rs/bevy/latest/bevy/prelude/struct.TextSpan.html)
docs include a code example that does not actually "work." The code
silently does not render anything, and the `Text*Writer` helpers fail.
This seems to be by design, because we can't use `Text` or `Text2d` from
`bevy_ui` or `bevy_sprite` within docs in `bevy_text`. (Correct me if I
am wrong)
I have seen multiple users confused by these docs.
Also fixes#16794
## Solution
Remove the code example from `TextSpan`, and instead encourage users to
seek docs on `Text` or `Text2d`.
Add examples with nested `TextSpan`s in those areas.
# Objective
```
cargo test --package bevy_ecs --lib --all-features
```
fails to compile, with output like
> error[E0433]: failed to resolve: could not find `Serialize` in `serde`
> --> crates/bevy_ecs/src/entity/index_set.rs:14:69
> |
> 14 | #[cfg_attr(feature = "serialize", derive(serde::Deserialize,
serde::Serialize))]
> | ^^^^^^^^^ could not find `Serialize` in `serde`
> |
> note: found an item that was configured out
> -->
/home/alice/.cargo/registry/src/index.crates.io-6f17d22bba15001f/serde-1.0.217/src/lib.rs:343:37
> |
> 343 | pub use serde_derive::{Deserialize, Serialize};
> | ^^^^^^^^^
> note: the item is gated behind the `serde_derive` feature
> -->
/home/alice/.cargo/registry/src/index.crates.io-6f17d22bba15001f/serde-1.0.217/src/lib.rs:341:7
> |
> 341 | #[cfg(feature = "serde_derive")]
> | ^^^^^^^^^^^^^^^^^^^^^^^^
## Solution
Add the required feature flags and get bevy_ecs compiling standalone
corrctly.
## Testing
The command above now compiles succesfully. Note that several system
stepping tests are failing, and were not being tested in CI. That's a
different PR's problem though.
# Objective
Currently, `prepare_sprite_image_bind_group` spawns sprite batches onto
an individual representative entity of the batch. This poses significant
problems for multi-camera setups, since an entity may appear in multiple
phase instances.
## Solution
Instead, move batches into a resource that is keyed off the view and the
representative entity. Long term we should switch to mesh2d and use the
existing BinnedRenderPhase functionality rather than naively queueing
into transparent and doing our own ad-hoc batching logic.
Fixes#16867, #17351
## Testing
Tested repros in above issues.
# Objective
We have default query filters now, but there is no first-party marker
for entity disabling yet
Fixes#17458
## Solution
Add the marker, cool recursive features and/or potential hook changes
should be follow up work
## Testing
Added a unit test to check that the new marker is enabled by default
# Objective
Expose accessor functions to the `ObserverDescriptor`, so that users can
use the `Observer` component to inspect what the observer is watching.
This would be useful for me, I don't think there's any reason to hide
these.
# Objective
- Fix off by one error introduced in
https://github.com/bevyengine/bevy/pull/17540 causing:
```
Cursor image StrongHandle<Image>{ id: Index(AssetIndex { generation: 0, index: 3 }), path: Some(cursors/kenney_crosshairPack/Tilesheet/crosshairs_tilesheet_white.png) } is invalid: The specified hotspot (64, 64) is outside the image bounds (64x64).
```
- First PR commit and run shows the bug:
https://github.com/bevyengine/bevy/actions/runs/13009405866/job/36283507530?pr=17571
- Second PR commit fixes it.
## Solution
- Hotspot coordinates are 0-indexed, so we need to subtract 1 from the
width and height.
## Testing
- Fix the tests which included the off-by-one error in their expected
values.
- Consolidate the tests into a single test for brevity.
- Test round trip transform to ensure we can "undo" to get back to the
original value.
- Add a specific bounds test.
- Ran the example again and observed there are no more error logs:
`cargo run --example custom_cursor_image --features=custom_cursor`.
# Objective
- Make working with `bevy_time` more ergonomic in `no_std` environments.
Currently `bevy_time` expects the getter in environments where time
can't be obtained automatically via the instruction set or the standard
library to be of type `*mut fn() -> Duration`.
[`fn()`](https://doc.rust-lang.org/beta/std/primitive.fn.html) is
already a function pointer, so `*mut fn()` is a _pointer to a function
pointer_. This is harder to use and error prone since creating a pointer
out of something like `&mut fn() -> Duration` when the lifetime of the
reference isn't static will lead to an undefined behavior once the
reference is freed
## Solution
- Accept a `fn() -> Duration` instead
## Testing
- I made a whole game on the Playdate that relies on `bevy_time`
heavily, see:
[bevydate_time](1b4f02adcd/src/lib.rs (L510-L546))
for usage of the Instant's getter.
---
## Showcase
<details>
<summary>Click to view showcase</summary>
https://github.com/user-attachments/assets/f687847f-6b62-4322-95f3-c908ada3db30
</details>
## Migration Guide
This is a breaking change but it's not for people coming from Bevy v0.15
### Small thank you note
Thanks to my friend https://github.com/repnop for helping me understand
how to deal with function pointers in `unsafe` environments
Co-authored-by: Wesley Norris <repnop@repnop.dev>
# Objective
Follow-up to #17615.
Bevy's entity collection types like `EntityHashSet` no longer implement
serde's `Serialize` and `Deserialize` after becoming newtypes instead of
type aliases in #16912. This broke some types that support serde for me
in Avian.
I also missed creating const constructors for `EntityIndexMap` and
`EntityIndexSet` in #17615. Oops!
## Solution
Implement `Serialize` and `Deserialize` for Bevy's entity collection
types, and add const constructors for `EntityIndexMap` and
`EntityIndexSet`.
I didn't implement `ReflectSerialize` or `ReflectDeserialize` here,
because I had some trouble fixing the resulting errors, and they were
not implemented previously either.
# Objective
- Linking to a specific AsBindGroup attribute is hard because it doesn't
use any headers and all the docs is in a giant block
## Solution
- Make each attribute it's own sub-header so they can be easily linked
---
## Showcase
Here's what the rustdoc output looks like with this change
## Notes
I kept the bullet point so the text is still indented like before. Not
sure if we should keep that or not
# Objective
- `ValArithmeticError` contains a typo, and one of it's variants is not
used
## Solution
- Rename `NonEvaluateable::NonEvaluateable ` variant to
`NonEvaluateable::NonEvaluable`.
- Remove variant `ValArithmeticError:: NonIdenticalVariants`.
## Testing
- `cargo run -p ci`
---
## Migration Guide
- `ValArithmeticError::NonEvaluateable` has been renamed to
`NonEvaluateable::NonEvaluable`
- `ValArithmeticError::NonIdenticalVariants ` has been removed
# Objective
- We kind of missed out on implementing the `Ease` trait for some
objects like `Isometry2D` and `Isometry3D` even though it makes sense
and isn't that hard
- Fixes#17539
## Testing
- wrote some minimal tests
- ~~noticed that quat easing isn't working as expected yet~~ I just
confused degrees and radians once again 🙈
# Objective
For most UI node entities there's a 1-to-1 mapping from the entity to
its associated Taffy node. Root UI nodes are an exception though, their
corresponding Taffy node in the Taffy tree is also given a parent that
represents the viewport. These viewport Taffy nodes are not removed when
a root UI node is despawned.
Parenting of an existing root UI node with an associated viewport Taffy
node also results in the leak of the viewport node.
These tests fail if added to the `layout` module's tests on the main
branch:
```rust
#[test]
fn no_viewport_node_leak_on_root_despawned() {
let (mut world, mut ui_schedule) = setup_ui_test_world();
let ui_root_entity = world.spawn(Node::default()).id();
// The UI schedule synchronizes Bevy UI's internal `TaffyTree` with the
// main world's tree of `Node` entities.
ui_schedule.run(&mut world);
// Two taffy nodes are added to the internal `TaffyTree` for each root UI entity.
// An implicit taffy node representing the viewport and a taffy node corresponding to the
// root UI entity which is parented to the viewport taffy node.
assert_eq!(
world.resource_mut::<UiSurface>().taffy.total_node_count(),
2
);
world.despawn(ui_root_entity);
// The UI schedule removes both the taffy node corresponding to `ui_root_entity` and its
// parent viewport node.
ui_schedule.run(&mut world);
// Both taffy nodes should now be removed from the internal `TaffyTree`
assert_eq!(
world.resource_mut::<UiSurface>().taffy.total_node_count(),
0
);
}
#[test]
fn no_viewport_node_leak_on_parented_root() {
let (mut world, mut ui_schedule) = setup_ui_test_world();
let ui_root_entity_1 = world.spawn(Node::default()).id();
let ui_root_entity_2 = world.spawn(Node::default()).id();
ui_schedule.run(&mut world);
// There are two UI root entities. Each root taffy node is given it's own viewport node parent,
// so a total of four taffy nodes are added to the `TaffyTree` by the UI schedule.
assert_eq!(
world.resource_mut::<UiSurface>().taffy.total_node_count(),
4
);
// Parent `ui_root_entity_2` onto `ui_root_entity_1` so now only `ui_root_entity_1` is a
// UI root entity.
world
.entity_mut(ui_root_entity_1)
.add_child(ui_root_entity_2);
// Now there is only one root node so the second viewport node is removed by
// the UI schedule.
ui_schedule.run(&mut world);
// There is only one viewport node now, so the `TaffyTree` contains 3 nodes in total.
assert_eq!(
world.resource_mut::<UiSurface>().taffy.total_node_count(),
3
);
}
```
Fixes#17594
## Solution
Change the `UiSurface::entity_to_taffy` to map to `LayoutNode`s. A
`LayoutNode` has a `viewport_id: Option<taffy::NodeId>` field which is
the id of the corresponding implicit "viewport" node if the node is a
root UI node, otherwise it is `None`. When removing or parenting nodes
this field is checked and the implicit viewport node is removed if
present.
## Testing
There are two new tests in `bevy_ui::layout::tests` included with this
PR:
* `no_viewport_node_leak_on_root_despawned`
* `no_viewport_node_leak_on_parented_root`
# Objective
Fix this comment in `queue_sprites`:
```
// batch_range and dynamic_offset will be calculated in prepare_sprites.
```
`Transparent2d` no longer has a `dynamic_offset` field and the
`batch_range` is calculated in `prepare_sprite_image_bind_groups` now.
# Objective
Fixes#17561
## Solution
The anti-aliasing function used by the UI fragment shader is this:
```wgsl
fn antialias(distance: f32) -> f32 {
return saturate(0.5 - distance); // saturate clamps between 0 and 1
}
```
The returned value is multiplied with the alpha channel value to get the
anti-aliasing effect.
The `distance` is a signed distance value. A positive `distance` means
we are outside the shape we're drawing and a negative `distance` means
we are on the inside.
So with `distance` at `0` (on the edge of the shape):
```
antialias(0) = saturate(0.5 - 0) = saturate(0.5) = 0.5
```
but we want it to be `1` at this point, so the entire interior of the
shape is given a solid colour, and then decrease as the signed distance
increases.
So in this PR we change it to:
```wgsl
fn antialias(distance: f32) -> f32 {
return saturate(1. - distance);
}
```
Then:
```
antialias(-0.5) = saturate(1 - (-1)) = saturate(2) = 1
antialias(1) = saturate(1 - 0) = 1
antialias(0.5) = saturate(1 - 0.5) = 0.5
antialias(1) = saturate(1 - 1) = 0
```
as desired.
## Testing
```cargo run --example button```
On main:
<img width="400" alt="bleg" src="https://github.com/user-attachments/assets/314994cb-4529-479d-b179-18e5c25f75bc" />
With this PR:
<img width="400" alt="bbwhite" src="https://github.com/user-attachments/assets/072f481d-8b67-4fae-9a5f-765090d1713f" />
Modified the `button` example to draw a white background to make the bleeding more obvious.
The code added in #14343 seems to be trying to ensure that a `Handle`
for each glTF node exists by topologically sorting the directed graph of
glTF nodes containing edges from parent to child and from skin to joint.
Unfortunately, such a graph can contain cycles, as there's no guarantee
that joints are descendants of nodes with the skin. In particular, glTF
exported from Maya using the popular babylon.js export plugin create
skins attached to nodes that animate their parent nodes. This was
causing the topological sort code to enter an infinite loop.
Assuming that the intent of the topological sort is indeed to ensure
that `Handle`s exist for each glTF node before populating them, there's
a better mechanism for this: `LoadContext::get_label_handle`. This is
the documented way to obtain a handle for a node before populating it,
obviating the need for a topological sort. This patch replaces the
topological sort with a pre-pass that uses
`LoadContext::get_label_handle` to get handles for each `Node` before
populating them. This fixes the problem with Maya rigs, in addition to
making the code simpler and faster.
I realized there wasn't a test for this yet and figured it would be
trivial to add. Why not? Unless there was a test for this, and I just
missed it?
I appreciate the unique error message it gives and wanted to make sure
it doesn't get broken at some point. Or worse, endlessly recurse.
---------
Co-authored-by: Alice Cecile <alice.i.cecile@gmail.com>
# Objective
- Improve CI when testing rendering by having smarter testbeds
## Solution
- CI testing no longer need a config file and will run with a default
config if not found
- It is now possible to give a name to a screenshot instead of just a
frame number
- 2d and 3d testbeds are now driven from code
- a new system in testbed will watch for state changed
- on state changed, trigger a screenshot 100 frames after (so that the
scene has time to render) with the name of the scene
- when the screenshot is taken (`Captured` component has been removed),
switch scene
- this means less setup to run a testbed (no need for a config file),
screenshots have better names, and it's faster as we don't wait 100
frames for the screenshot to be taken
## Testing
- `cargo run --example testbed_2d --features bevy_ci_testing`
# Objective
- Also support `f16` values when getting and setting colors.
## Solution
- Use the `half` crate to work with `f16` until it's in stable Rust.
# Objective
#16912 turned `EntityHashMap` and `EntityHashSet` into proper newtypes
instead of type aliases. However, this removed the ability to create
these collections in const contexts; previously, you could use
`EntityHashSet::with_hasher(EntityHash)`, but it doesn't exist anymore.
## Solution
Make `EntityHashMap::new` and `EntityHashSet::new` const methods.
