# Objective
- Fixes#7288
- Do not expose access directly to cursor position as it is the physical position, ignoring scale
## Solution
- Make cursor position private
- Expose getter/setter on the window to have access to the scale
# Objective
Fixes#6931
Continues #6954 by squashing `Msaa` to a flat enum
Helps out #7215
# Solution
```
pub enum Msaa {
Off = 1,
#[default]
Sample4 = 4,
}
```
# Changelog
- Modified
- `Msaa` is now enum
- Defaults to 4 samples
- Uses `.samples()` method to get the sample number as `u32`
# Migration Guide
```
let multi = Msaa { samples: 4 }
// is now
let multi = Msaa::Sample4
multi.samples
// is now
multi.samples()
```
Co-authored-by: Sjael <jakeobrien44@gmail.com>
After #6503, bevy_render uses the `send_blocking` method introduced in async-channel 1.7, but depended only on ^1.4.
I saw this after pulling main without running cargo update.
# Objective
- Fix minimum dependency version of async-channel
## Solution
- Bump async-channel version constraint to ^1.8, which is currently the latest version.
NOTE: Both bevy_ecs and bevy_tasks also depend on async-channel but they didn't use any newer features.
# Objective
Fixes#3184. Fixes#6640. Fixes#4798. Using `Query::par_for_each(_mut)` currently requires a `batch_size` parameter, which affects how it chunks up large archetypes and tables into smaller chunks to run in parallel. Tuning this value is difficult, as the performance characteristics entirely depends on the state of the `World` it's being run on. Typically, users will just use a flat constant and just tune it by hand until it performs well in some benchmarks. However, this is both error prone and risks overfitting the tuning on that benchmark.
This PR proposes a naive automatic batch-size computation based on the current state of the `World`.
## Background
`Query::par_for_each(_mut)` schedules a new Task for every archetype or table that it matches. Archetypes/tables larger than the batch size are chunked into smaller tasks. Assuming every entity matched by the query has an identical workload, this makes the worst case scenario involve using a batch size equal to the size of the largest matched archetype or table. Conversely, a batch size of `max {archetype, table} size / thread count * COUNT_PER_THREAD` is likely the sweetspot where the overhead of scheduling tasks is minimized, at least not without grouping small archetypes/tables together.
There is also likely a strict minimum batch size below which the overhead of scheduling these tasks is heavier than running the entire thing single-threaded.
## Solution
- [x] Remove the `batch_size` from `Query(State)::par_for_each` and friends.
- [x] Add a check to compute `batch_size = max {archeytpe/table} size / thread count * COUNT_PER_THREAD`
- [x] ~~Panic if thread count is 0.~~ Defer to `for_each` if the thread count is 1 or less.
- [x] Early return if there is no matched table/archetype.
- [x] Add override option for users have queries that strongly violate the initial assumption that all iterated entities have an equal workload.
---
## Changelog
Changed: `Query::par_for_each(_mut)` has been changed to `Query::par_iter(_mut)` and will now automatically try to produce a batch size for callers based on the current `World` state.
## Migration Guide
The `batch_size` parameter for `Query(State)::par_for_each(_mut)` has been removed. These calls will automatically compute a batch size for you. Remove these parameters from all calls to these functions.
Before:
```rust
fn parallel_system(query: Query<&MyComponent>) {
query.par_for_each(32, |comp| {
...
});
}
```
After:
```rust
fn parallel_system(query: Query<&MyComponent>) {
query.par_iter().for_each(|comp| {
...
});
}
```
Co-authored-by: Arnav Choubey <56453634+x-52@users.noreply.github.com>
Co-authored-by: Robert Swain <robert.swain@gmail.com>
Co-authored-by: François <mockersf@gmail.com>
Co-authored-by: Corey Farwell <coreyf@rwell.org>
Co-authored-by: Aevyrie <aevyrie@gmail.com>
## Problem
`extract_uinodes` checks if an image is loaded for nodes without images
## Solution
Move the image loading skip check so that it is only performed for nodes with a `UiImage` component.
# Objective
- Implement pipelined rendering
- Fixes#5082
- Fixes#4718
## User Facing Description
Bevy now implements piplelined rendering! Pipelined rendering allows the app logic and rendering logic to run on different threads leading to large gains in performance.
*tracy capture of many_foxes example*
To use pipelined rendering, you just need to add the `PipelinedRenderingPlugin`. If you're using `DefaultPlugins` then it will automatically be added for you on all platforms except wasm. Bevy does not currently support multithreading on wasm which is needed for this feature to work. If you aren't using `DefaultPlugins` you can add the plugin manually.
```rust
use bevy::prelude::*;
use bevy::render::pipelined_rendering::PipelinedRenderingPlugin;
fn main() {
App::new()
// whatever other plugins you need
.add_plugin(RenderPlugin)
// needs to be added after RenderPlugin
.add_plugin(PipelinedRenderingPlugin)
.run();
}
```
If for some reason pipelined rendering needs to be removed. You can also disable the plugin the normal way.
```rust
use bevy::prelude::*;
use bevy::render::pipelined_rendering::PipelinedRenderingPlugin;
fn main() {
App::new.add_plugins(DefaultPlugins.build().disable::<PipelinedRenderingPlugin>());
}
```
### A setup function was added to plugins
A optional plugin lifecycle function was added to the `Plugin trait`. This function is called after all plugins have been built, but before the app runner is called. This allows for some final setup to be done. In the case of pipelined rendering, the function removes the sub app from the main app and sends it to the render thread.
```rust
struct MyPlugin;
impl Plugin for MyPlugin {
fn build(&self, app: &mut App) {
}
// optional function
fn setup(&self, app: &mut App) {
// do some final setup before runner is called
}
}
```
### A Stage for Frame Pacing
In the `RenderExtractApp` there is a stage labelled `BeforeIoAfterRenderStart` that systems can be added to. The specific use case for this stage is for a frame pacing system that can delay the start of main app processing in render bound apps to reduce input latency i.e. "frame pacing". This is not currently built into bevy, but exists as `bevy`
```text
|-------------------------------------------------------------------|
| | BeforeIoAfterRenderStart | winit events | main schedule |
| extract |---------------------------------------------------------|
| | extract commands | rendering schedule |
|-------------------------------------------------------------------|
```
### Small API additions
* `Schedule::remove_stage`
* `App::insert_sub_app`
* `App::remove_sub_app`
* `TaskPool::scope_with_executor`
## Problems and Solutions
### Moving render app to another thread
Most of the hard bits for this were done with the render redo. This PR just sends the render app back and forth through channels which seems to work ok. I originally experimented with using a scope to run the render task. It was cuter, but that approach didn't allow render to start before i/o processing. So I switched to using channels. There is much complexity in the coordination that needs to be done, but it's worth it. By moving rendering during i/o processing the frame times should be much more consistent in render bound apps. See https://github.com/bevyengine/bevy/issues/4691.
### Unsoundness with Sending World with NonSend resources
Dropping !Send things on threads other than the thread they were spawned on is considered unsound. The render world doesn't have any nonsend resources. So if we tell the users to "pretty please don't spawn nonsend resource on the render world", we can avoid this problem.
More seriously there is this https://github.com/bevyengine/bevy/pull/6534 pr, which patches the unsoundness by aborting the app if a nonsend resource is dropped on the wrong thread. ~~That PR should probably be merged before this one.~~ For a longer term solution we have this discussion going https://github.com/bevyengine/bevy/discussions/6552.
### NonSend Systems in render world
The render world doesn't have any !Send resources, but it does have a non send system. While Window is Send, winit does have some API's that can only be accessed on the main thread. `prepare_windows` in the render schedule thus needs to be scheduled on the main thread. Currently we run nonsend systems by running them on the thread the TaskPool::scope runs on. When we move render to another thread this no longer works.
To fix this, a new `scope_with_executor` method was added that takes a optional `TheadExecutor` that can only be ticked on the thread it was initialized on. The render world then holds a `MainThreadExecutor` resource which can be passed to the scope in the parallel executor that it uses to spawn it's non send systems on.
### Scopes executors between render and main should not share tasks
Since the render world and the app world share the `ComputeTaskPool`. Because `scope` has executors for the ComputeTaskPool a system from the main world could run on the render thread or a render system could run on the main thread. This can cause performance problems because it can delay a stage from finishing. See https://github.com/bevyengine/bevy/pull/6503#issuecomment-1309791442 for more details.
To avoid this problem, `TaskPool::scope` has been changed to not tick the ComputeTaskPool when it's used by the parallel executor. In the future when we move closer to the 1 thread to 1 logical core model we may want to overprovide threads, because the render and main app threads don't do much when executing the schedule.
## Performance
My machine is Windows 11, AMD Ryzen 5600x, RX 6600
### Examples
#### This PR with pipelining vs Main
> Note that these were run on an older version of main and the performance profile has probably changed due to optimizations
Seeing a perf gain from 29% on many lights to 7% on many sprites.
<html>
<body>
<!--StartFragment--><google-sheets-html-origin>
| percent | | | Diff | | | Main | | | PR | |
-- | -- | -- | -- | -- | -- | -- | -- | -- | -- | -- | -- | --
tracy frame time | mean | median | sigma | mean | median | sigma | mean | median | sigma | mean | median | sigma
many foxes | 27.01% | 27.34% | -47.09% | 1.58 | 1.55 | -1.78 | 5.85 | 5.67 | 3.78 | 4.27 | 4.12 | 5.56
many lights | 29.35% | 29.94% | -10.84% | 3.02 | 3.03 | -0.57 | 10.29 | 10.12 | 5.26 | 7.27 | 7.09 | 5.83
many animated sprites | 13.97% | 15.69% | 14.20% | 3.79 | 4.17 | 1.41 | 27.12 | 26.57 | 9.93 | 23.33 | 22.4 | 8.52
3d scene | 25.79% | 26.78% | 7.46% | 0.49 | 0.49 | 0.15 | 1.9 | 1.83 | 2.01 | 1.41 | 1.34 | 1.86
many cubes | 11.97% | 11.28% | 14.51% | 1.93 | 1.78 | 1.31 | 16.13 | 15.78 | 9.03 | 14.2 | 14 | 7.72
many sprites | 7.14% | 9.42% | -85.42% | 1.72 | 2.23 | -6.15 | 24.09 | 23.68 | 7.2 | 22.37 | 21.45 | 13.35
<!--EndFragment-->
</body>
</html>
#### This PR with pipelining disabled vs Main
Mostly regressions here. I don't think this should be a problem as users that are disabling pipelined rendering are probably running single threaded and not using the parallel executor. The regression is probably mostly due to the switch to use `async_executor::run` instead of `try_tick` and also having one less thread to run systems on. I'll do a writeup on why switching to `run` causes regressions, so we can try to eventually fix it. Using try_tick causes issues when pipeline rendering is enable as seen [here](https://github.com/bevyengine/bevy/pull/6503#issuecomment-1380803518)
<html>
<body>
<!--StartFragment--><google-sheets-html-origin>
| percent | | | Diff | | | Main | | | PR no pipelining | |
-- | -- | -- | -- | -- | -- | -- | -- | -- | -- | -- | -- | --
tracy frame time | mean | median | sigma | mean | median | sigma | mean | median | sigma | mean | median | sigma
many foxes | -3.72% | -4.42% | -1.07% | -0.21 | -0.24 | -0.04 | 5.64 | 5.43 | 3.74 | 5.85 | 5.67 | 3.78
many lights | 0.29% | -0.30% | 4.75% | 0.03 | -0.03 | 0.25 | 10.29 | 10.12 | 5.26 | 10.26 | 10.15 | 5.01
many animated sprites | 0.22% | 1.81% | -2.72% | 0.06 | 0.48 | -0.27 | 27.12 | 26.57 | 9.93 | 27.06 | 26.09 | 10.2
3d scene | -15.79% | -14.75% | -31.34% | -0.3 | -0.27 | -0.63 | 1.9 | 1.83 | 2.01 | 2.2 | 2.1 | 2.64
many cubes | -2.85% | -3.30% | 0.00% | -0.46 | -0.52 | 0 | 16.13 | 15.78 | 9.03 | 16.59 | 16.3 | 9.03
many sprites | 2.49% | 2.41% | 0.69% | 0.6 | 0.57 | 0.05 | 24.09 | 23.68 | 7.2 | 23.49 | 23.11 | 7.15
<!--EndFragment-->
</body>
</html>
### Benchmarks
Mostly the same except empty_systems has got a touch slower. The maybe_pipelining+1 column has the compute task pool with an extra thread over default added. This is because pipelining loses one thread over main to execute systems on, since the main thread no longer runs normal systems.
<details>
<summary>Click Me</summary>
```text
group main maybe-pipelining+1
----- ------------------------- ------------------
busy_systems/01x_entities_03_systems 1.07 30.7±1.32µs ? ?/sec 1.00 28.6±1.35µs ? ?/sec
busy_systems/01x_entities_06_systems 1.10 52.1±1.10µs ? ?/sec 1.00 47.2±1.08µs ? ?/sec
busy_systems/01x_entities_09_systems 1.00 74.6±1.36µs ? ?/sec 1.00 75.0±1.93µs ? ?/sec
busy_systems/01x_entities_12_systems 1.03 100.6±6.68µs ? ?/sec 1.00 98.0±1.46µs ? ?/sec
busy_systems/01x_entities_15_systems 1.11 128.5±3.53µs ? ?/sec 1.00 115.5±1.02µs ? ?/sec
busy_systems/02x_entities_03_systems 1.16 50.4±2.56µs ? ?/sec 1.00 43.5±3.00µs ? ?/sec
busy_systems/02x_entities_06_systems 1.00 87.1±1.27µs ? ?/sec 1.05 91.5±7.15µs ? ?/sec
busy_systems/02x_entities_09_systems 1.04 139.9±6.37µs ? ?/sec 1.00 134.0±1.06µs ? ?/sec
busy_systems/02x_entities_12_systems 1.05 179.2±3.47µs ? ?/sec 1.00 170.1±3.17µs ? ?/sec
busy_systems/02x_entities_15_systems 1.01 219.6±3.75µs ? ?/sec 1.00 218.1±2.55µs ? ?/sec
busy_systems/03x_entities_03_systems 1.10 70.6±2.33µs ? ?/sec 1.00 64.3±0.69µs ? ?/sec
busy_systems/03x_entities_06_systems 1.02 130.2±3.11µs ? ?/sec 1.00 128.0±1.34µs ? ?/sec
busy_systems/03x_entities_09_systems 1.00 195.0±10.11µs ? ?/sec 1.00 194.8±1.41µs ? ?/sec
busy_systems/03x_entities_12_systems 1.01 261.7±4.05µs ? ?/sec 1.00 259.8±4.11µs ? ?/sec
busy_systems/03x_entities_15_systems 1.00 318.0±3.04µs ? ?/sec 1.06 338.3±20.25µs ? ?/sec
busy_systems/04x_entities_03_systems 1.00 82.9±0.63µs ? ?/sec 1.02 84.3±0.63µs ? ?/sec
busy_systems/04x_entities_06_systems 1.01 181.7±3.65µs ? ?/sec 1.00 179.8±1.76µs ? ?/sec
busy_systems/04x_entities_09_systems 1.04 265.0±4.68µs ? ?/sec 1.00 255.3±1.98µs ? ?/sec
busy_systems/04x_entities_12_systems 1.00 335.9±3.00µs ? ?/sec 1.05 352.6±15.84µs ? ?/sec
busy_systems/04x_entities_15_systems 1.00 418.6±10.26µs ? ?/sec 1.08 450.2±39.58µs ? ?/sec
busy_systems/05x_entities_03_systems 1.07 114.3±0.95µs ? ?/sec 1.00 106.9±1.52µs ? ?/sec
busy_systems/05x_entities_06_systems 1.08 229.8±2.90µs ? ?/sec 1.00 212.3±4.18µs ? ?/sec
busy_systems/05x_entities_09_systems 1.03 329.3±1.99µs ? ?/sec 1.00 319.2±2.43µs ? ?/sec
busy_systems/05x_entities_12_systems 1.06 454.7±6.77µs ? ?/sec 1.00 430.1±3.58µs ? ?/sec
busy_systems/05x_entities_15_systems 1.03 554.6±6.15µs ? ?/sec 1.00 538.4±23.87µs ? ?/sec
contrived/01x_entities_03_systems 1.00 14.0±0.15µs ? ?/sec 1.08 15.1±0.21µs ? ?/sec
contrived/01x_entities_06_systems 1.04 28.5±0.37µs ? ?/sec 1.00 27.4±0.44µs ? ?/sec
contrived/01x_entities_09_systems 1.00 41.5±4.38µs ? ?/sec 1.02 42.2±2.24µs ? ?/sec
contrived/01x_entities_12_systems 1.06 55.9±1.49µs ? ?/sec 1.00 52.6±1.36µs ? ?/sec
contrived/01x_entities_15_systems 1.02 68.0±2.00µs ? ?/sec 1.00 66.5±0.78µs ? ?/sec
contrived/02x_entities_03_systems 1.03 25.2±0.38µs ? ?/sec 1.00 24.6±0.52µs ? ?/sec
contrived/02x_entities_06_systems 1.00 46.3±0.49µs ? ?/sec 1.04 48.1±4.13µs ? ?/sec
contrived/02x_entities_09_systems 1.02 70.4±0.99µs ? ?/sec 1.00 68.8±1.04µs ? ?/sec
contrived/02x_entities_12_systems 1.06 96.8±1.49µs ? ?/sec 1.00 91.5±0.93µs ? ?/sec
contrived/02x_entities_15_systems 1.02 116.2±0.95µs ? ?/sec 1.00 114.2±1.42µs ? ?/sec
contrived/03x_entities_03_systems 1.00 33.2±0.38µs ? ?/sec 1.01 33.6±0.45µs ? ?/sec
contrived/03x_entities_06_systems 1.00 62.4±0.73µs ? ?/sec 1.01 63.3±1.05µs ? ?/sec
contrived/03x_entities_09_systems 1.02 96.4±0.85µs ? ?/sec 1.00 94.8±3.02µs ? ?/sec
contrived/03x_entities_12_systems 1.01 126.3±4.67µs ? ?/sec 1.00 125.6±2.27µs ? ?/sec
contrived/03x_entities_15_systems 1.03 160.2±9.37µs ? ?/sec 1.00 156.0±1.53µs ? ?/sec
contrived/04x_entities_03_systems 1.02 41.4±3.39µs ? ?/sec 1.00 40.5±0.52µs ? ?/sec
contrived/04x_entities_06_systems 1.00 78.9±1.61µs ? ?/sec 1.02 80.3±1.06µs ? ?/sec
contrived/04x_entities_09_systems 1.02 121.8±3.97µs ? ?/sec 1.00 119.2±1.46µs ? ?/sec
contrived/04x_entities_12_systems 1.00 157.8±1.48µs ? ?/sec 1.01 160.1±1.72µs ? ?/sec
contrived/04x_entities_15_systems 1.00 197.9±1.47µs ? ?/sec 1.08 214.2±34.61µs ? ?/sec
contrived/05x_entities_03_systems 1.00 49.1±0.33µs ? ?/sec 1.01 49.7±0.75µs ? ?/sec
contrived/05x_entities_06_systems 1.00 95.0±0.93µs ? ?/sec 1.00 94.6±0.94µs ? ?/sec
contrived/05x_entities_09_systems 1.01 143.2±1.68µs ? ?/sec 1.00 142.2±2.00µs ? ?/sec
contrived/05x_entities_12_systems 1.00 191.8±2.03µs ? ?/sec 1.01 192.7±7.88µs ? ?/sec
contrived/05x_entities_15_systems 1.02 239.7±3.71µs ? ?/sec 1.00 235.8±4.11µs ? ?/sec
empty_systems/000_systems 1.01 47.8±0.67ns ? ?/sec 1.00 47.5±2.02ns ? ?/sec
empty_systems/001_systems 1.00 1743.2±126.14ns ? ?/sec 1.01 1761.1±70.10ns ? ?/sec
empty_systems/002_systems 1.01 2.2±0.04µs ? ?/sec 1.00 2.2±0.02µs ? ?/sec
empty_systems/003_systems 1.02 2.7±0.09µs ? ?/sec 1.00 2.7±0.16µs ? ?/sec
empty_systems/004_systems 1.00 3.1±0.11µs ? ?/sec 1.00 3.1±0.24µs ? ?/sec
empty_systems/005_systems 1.00 3.5±0.05µs ? ?/sec 1.11 3.9±0.70µs ? ?/sec
empty_systems/010_systems 1.00 5.5±0.12µs ? ?/sec 1.03 5.7±0.17µs ? ?/sec
empty_systems/015_systems 1.00 7.9±0.19µs ? ?/sec 1.06 8.4±0.16µs ? ?/sec
empty_systems/020_systems 1.00 10.4±1.25µs ? ?/sec 1.02 10.6±0.18µs ? ?/sec
empty_systems/025_systems 1.00 12.4±0.39µs ? ?/sec 1.14 14.1±1.07µs ? ?/sec
empty_systems/030_systems 1.00 15.1±0.39µs ? ?/sec 1.05 15.8±0.62µs ? ?/sec
empty_systems/035_systems 1.00 16.9±0.47µs ? ?/sec 1.07 18.0±0.37µs ? ?/sec
empty_systems/040_systems 1.00 19.3±0.41µs ? ?/sec 1.05 20.3±0.39µs ? ?/sec
empty_systems/045_systems 1.00 22.4±1.67µs ? ?/sec 1.02 22.9±0.51µs ? ?/sec
empty_systems/050_systems 1.00 24.4±1.67µs ? ?/sec 1.01 24.7±0.40µs ? ?/sec
empty_systems/055_systems 1.05 28.6±5.27µs ? ?/sec 1.00 27.2±0.70µs ? ?/sec
empty_systems/060_systems 1.02 29.9±1.64µs ? ?/sec 1.00 29.3±0.66µs ? ?/sec
empty_systems/065_systems 1.02 32.7±3.15µs ? ?/sec 1.00 32.1±0.98µs ? ?/sec
empty_systems/070_systems 1.00 33.0±1.42µs ? ?/sec 1.03 34.1±1.44µs ? ?/sec
empty_systems/075_systems 1.00 34.8±0.89µs ? ?/sec 1.04 36.2±0.70µs ? ?/sec
empty_systems/080_systems 1.00 37.0±1.82µs ? ?/sec 1.05 38.7±1.37µs ? ?/sec
empty_systems/085_systems 1.00 38.7±0.76µs ? ?/sec 1.05 40.8±0.83µs ? ?/sec
empty_systems/090_systems 1.00 41.5±1.09µs ? ?/sec 1.04 43.2±0.82µs ? ?/sec
empty_systems/095_systems 1.00 43.6±1.10µs ? ?/sec 1.04 45.2±0.99µs ? ?/sec
empty_systems/100_systems 1.00 46.7±2.27µs ? ?/sec 1.03 48.1±1.25µs ? ?/sec
```
</details>
## Migration Guide
### App `runner` and SubApp `extract` functions are now required to be Send
This was changed to enable pipelined rendering. If this breaks your use case please report it as these new bounds might be able to be relaxed.
## ToDo
* [x] redo benchmarking
* [x] reinvestigate the perf of the try_tick -> run change for task pool scope
# Objective
- Add a configurable prepass
- A depth prepass is useful for various shader effects and to reduce overdraw. It can be expansive depending on the scene so it's important to be able to disable it if you don't need any effects that uses it or don't suffer from excessive overdraw.
- The goal is to eventually use it for things like TAA, Ambient Occlusion, SSR and various other techniques that can benefit from having a prepass.
## Solution
The prepass node is inserted before the main pass. It runs for each `Camera3d` with a prepass component (`DepthPrepass`, `NormalPrepass`). The presence of one of those components is used to determine which textures are generated in the prepass. When any prepass is enabled, the depth buffer generated will be used by the main pass to reduce overdraw.
The prepass runs for each `Material` created with the `MaterialPlugin::prepass_enabled` option set to `true`. You can overload the shader used by the prepass by using `Material::prepass_vertex_shader()` and/or `Material::prepass_fragment_shader()`. It will also use the `Material::specialize()` for more advanced use cases. It is enabled by default on all materials.
The prepass works on opaque materials and materials using an alpha mask. Transparent materials are ignored.
The `StandardMaterial` overloads the prepass fragment shader to support alpha mask and normal maps.
---
## Changelog
- Add a new `PrepassNode` that runs before the main pass
- Add a `PrepassPlugin` to extract/prepare/queue the necessary data
- Add a `DepthPrepass` and `NormalPrepass` component to control which textures will be created by the prepass and available in later passes.
- Add a new `prepass_enabled` flag to the `MaterialPlugin` that will control if a material uses the prepass or not.
- Add a new `prepass_enabled` flag to the `PbrPlugin` to control if the StandardMaterial uses the prepass. Currently defaults to false.
- Add `Material::prepass_vertex_shader()` and `Material::prepass_fragment_shader()` to control the prepass from the `Material`
## Notes
In bevy's sample 3d scene, the performance is actually worse when enabling the prepass, but on more complex scenes the performance is generally better. I would like more testing on this, but @DGriffin91 has reported a very noticeable improvements in some scenes.
The prepass is also used by @JMS55 for TAA and GTAO
discord thread: <https://discord.com/channels/691052431525675048/1011624228627419187>
This PR was built on top of the work of multiple people
Co-Authored-By: @superdump
Co-Authored-By: @robtfm
Co-Authored-By: @JMS55
Co-authored-by: Charles <IceSentry@users.noreply.github.com>
Co-authored-by: JMS55 <47158642+JMS55@users.noreply.github.com>
# Objective
- Safety comments for the `CommandQueue` type are quite sparse and very imprecise. Sometimes, they are right for the wrong reasons or use circular reasoning.
## Solution
- Document previously-implicit safety invariants.
- Rewrite safety comments to actually reflect the specific invariants of each operation.
- Use `OwningPtr` instead of raw pointers, to encode an invariant in the type system instead of via comments.
- Use typed pointer methods when possible to increase reliability.
---
## Changelog
+ Added the function `OwningPtr::read_unaligned`.
# Objective
Fix https://github.com/bevyengine/bevy/issues/4530
- Make it easier to open/close/modify windows by setting them up as `Entity`s with a `Window` component.
- Make multiple windows very simple to set up. (just add a `Window` component to an entity and it should open)
## Solution
- Move all properties of window descriptor to ~components~ a component.
- Replace `WindowId` with `Entity`.
- ~Use change detection for components to update backend rather than events/commands. (The `CursorMoved`/`WindowResized`/... events are kept for user convenience.~
Check each field individually to see what we need to update, events are still kept for user convenience.
---
## Changelog
- `WindowDescriptor` renamed to `Window`.
- Width/height consolidated into a `WindowResolution` component.
- Requesting maximization/minimization is done on the [`Window::state`] field.
- `WindowId` is now `Entity`.
## Migration Guide
- Replace `WindowDescriptor` with `Window`.
- Change `width` and `height` fields in a `WindowResolution`, either by doing
```rust
WindowResolution::new(width, height) // Explicitly
// or using From<_> for tuples for convenience
(1920., 1080.).into()
```
- Replace any `WindowCommand` code to just modify the `Window`'s fields directly and creating/closing windows is now by spawning/despawning an entity with a `Window` component like so:
```rust
let window = commands.spawn(Window { ... }).id(); // open window
commands.entity(window).despawn(); // close window
```
## Unresolved
- ~How do we tell when a window is minimized by a user?~
~Currently using the `Resize(0, 0)` as an indicator of minimization.~
No longer attempting to tell given how finnicky this was across platforms, now the user can only request that a window be maximized/minimized.
## Future work
- Move `exit_on_close` functionality out from windowing and into app(?)
- https://github.com/bevyengine/bevy/issues/5621
- https://github.com/bevyengine/bevy/issues/7099
- https://github.com/bevyengine/bevy/issues/7098
Co-authored-by: Carter Anderson <mcanders1@gmail.com>
# Objective
See:
- https://github.com/bevyengine/bevy/issues/7067#issuecomment-1381982285
- (This does not fully close that issue in my opinion.)
- https://discord.com/channels/691052431525675048/1063454009769340989
## Solution
This merge request adds documentation:
1. Alert users to the fact that `App::run()` might never return and code placed after it might never be executed.
2. Makes `winit::WinitSettings::return_from_run` discoverable.
3. Better explains why `winit::WinitSettings::return_from_run` is discouraged and better links to up-stream docs. on that topic.
4. Adds notes to the `app/return_after_run.rs` example which otherwise promotes a feature that carries caveats.
Furthermore, w.r.t `winit::WinitSettings::return_from_run`:
- Broken links to `winit` docs are fixed.
- Links now point to BOTH `EventLoop::run()` and `EventLoopExtRunReturn::run_return()` which are the salient up-stream pages and make more sense, taken together.
- Collateral damage: "Supported platforms" heading; disambiguation of "run" → `App::run()`; links.
## Future Work
I deliberately structured the "`run()` might not return" section under `App::run()` to allow for alternative patterns (e.g. `AppExit` event, `WindowClosed` event) to be listed or mentioned, beneath it, in the future.
# Objective
- Fixes#7260
## Solution
- #6649 used `init_non_send_resource` for `AudioOutput`, but this is before #6436 was merged.
- Use `init_resource` instead.
# Objective
Repeated calls to `init_non_send_resource` currently overwrite the old value because the wrong storage is being checked.
## Solution
Use the correct storage. Add some tests.
## Notes
Without the fix, the new test fails with
```
thread 'world::tests::init_non_send_resource_does_not_overwrite' panicked at 'assertion failed: `(left == right)`
left: `1`,
right: `0`', crates/bevy_ecs/src/world/mod.rs:2267:9
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
test world::tests::init_non_send_resource_does_not_overwrite ... FAILED
```
This was introduced by #7174 and it seems like a fairly straightforward oopsie.
# Objective
I was reading through the bevy_ecs code, trying to understand how everything works.
I was getting a bit confused when reading the doc comment for the `new_archetype` function; it looks like it doesn't create a new archetype but instead updates some internal state in the SystemParam to facility QueryIteration.
(I still couldn't find where a new archetype was actually created)
## Solution
- Adding a doc comment with a more correct explanation.
If it's deemed correct, I can also update the doc-comment for the other `new_archetype` calls
# Objective
Speed up the render phase of rendering. An extension of #6885.
`SystemState::get` increments the `World`'s change tick atomically every time it's called. This is notably more expensive than a unsynchronized increment, even without contention. It also updates the archetypes, even when there has been nothing to update when it's called repeatedly.
## Solution
Piggyback off of #6885. Split `SystemState::validate_world_and_update_archetypes` into `SystemState::validate_world` and `SystemState::update_archetypes`, and make the later `pub`. Then create safe variants of `SystemState::get_unchecked_manual` that still validate the `World` but do not update archetypes and do not increment the change tick using `World::read_change_tick` and `World::change_tick`. Update `RenderCommandState` to call `SystemState::update_archetypes` in `Draw::prepare` and `SystemState::get_manual` in `Draw::draw`.
## Performance
There's a slight perf benefit (~2%) for `main_opaque_pass_3d` on `many_foxes` (340.39 us -> 333.32 us)
## Alternatives
We can change `SystemState::get` to not increment the `World`'s change tick. Though this would still put updating the archetypes and an atomic read on the hot-path.
---
## Changelog
Added: `SystemState::get_manual`
Added: `SystemState::get_manual_mut`
Added: `SystemState::update_archetypes`
# Objective
Remove the `VerticalAlign` enum.
Text's alignment field should only affect the text's internal text alignment, not its position. The only way to control a `TextBundle`'s position and bounds should be through the manipulation of the constraints in the `Style` components of the nodes in the Bevy UI's layout tree.
`Text2dBundle` should have a separate `Anchor` component that sets its position relative to its transform.
Related issues: #676, #1490, #5502, #5513, #5834, #6717, #6724, #6741, #6748
## Changelog
* Changed `TextAlignment` into an enum with `Left`, `Center`, and `Right` variants.
* Removed the `HorizontalAlign` and `VerticalAlign` types.
* Added an `Anchor` component to `Text2dBundle`
* Added `Component` derive to `Anchor`
* Use `f32::INFINITY` instead of `f32::MAX` to represent unbounded text in Text2dBounds
## Migration Guide
The `alignment` field of `Text` now only affects the text's internal alignment.
### Change `TextAlignment` to TextAlignment` which is now an enum. Replace:
* `TextAlignment::TOP_LEFT`, `TextAlignment::CENTER_LEFT`, `TextAlignment::BOTTOM_LEFT` with `TextAlignment::Left`
* `TextAlignment::TOP_CENTER`, `TextAlignment::CENTER_LEFT`, `TextAlignment::BOTTOM_CENTER` with `TextAlignment::Center`
* `TextAlignment::TOP_RIGHT`, `TextAlignment::CENTER_RIGHT`, `TextAlignment::BOTTOM_RIGHT` with `TextAlignment::Right`
### Changes for `Text2dBundle`
`Text2dBundle` has a new field 'text_anchor' that takes an `Anchor` component that controls its position relative to its transform.
# Objective
- Enabling the `debug_asset_server` feature doesn't compile when using it with `load_internal_binary_asset!()`. The issue is because it assumes the loader takes an `&'static str` as a parameter, but binary assets loader expect `&'static [u8]`.
## Solution
- Add a generic type for the loader and use a different type in `load_internal_asset` and `load_internal_binary_asset`
# Objective
- Fixes#6361
- Fixes#6362
- Fixes#6364
## Solution
- Added an example for creating a custom `Decodable` type
- Clarified the documentation on `Decodable`
- Added an `AddAudioSource` trait and implemented it for `App`
Co-authored-by: dis-da-moe <84386186+dis-da-moe@users.noreply.github.com>
# Objective
Fix#4647. If any child is changed, or even reordered, `Changed<Children>` is true, which causes transform propagation to propagate changes to all siblings of a changed child, even if they don't need to be.
## Solution
As `Parent` and `Children` are updated in tandem in hierarchy commands after #4800. `Changed<Parent>` is true on the child when `Changed<Children>` is true on the parent. However, unlike checking children, checking `Changed<Parent>` is only localized to the current entity and will not force propagation to the siblings.
Also took the opportunity to change propagation to use `Query::iter_many` instead of repeated `Query::get` calls. Should cut a bit of the overhead out of propagation. This means we won't panic when there isn't a `Parent` on the child, just skip over it.
The tests from #4608 still pass, so the change detection here still works just fine under this approach.
# Objective
fix bloom when used on a camera with a viewport specified
## Solution
- pass viewport into the prefilter shader, and use it to read from the correct section of the original rendered screen
- don't apply viewport for the intermediate bloom passes, only for the final blend output
# Objective
- Fixes#3158
## Solution
- clear columns
My implementation of `clear_resources` do not remove the components itself but it clears the columns that keeps the resource data. I'm not sure if the issue meant to clear all resources, even the components and component ids (which I'm not sure if it's possible)
Co-authored-by: 2ne1ugly <47616772+2ne1ugly@users.noreply.github.com>
# Objective
The trait `ReadOnlySystemParam` is not implemented for `Option<NonSend<>>`, even though it should be.
Follow-up to #7243. This fixes another mistake made in #6919.
## Solution
Add the missing impl.
# Objective
The trait `ReadOnlySystemParam` is implemented for `NonSendMut`, when it should not be. This mistake was made in #6919.
## Solution
Remove the incorrect impl.
# Objective
Complete the first part of the migration detailed in bevyengine/rfcs#45.
## Solution
Add all the new stuff.
### TODO
- [x] Impl tuple methods.
- [x] Impl chaining.
- [x] Port ambiguity detection.
- [x] Write docs.
- [x] ~~Write more tests.~~(will do later)
- [ ] Write changelog and examples here?
- [x] ~~Replace `petgraph`.~~ (will do later)
Co-authored-by: james7132 <contact@jamessliu.com>
Co-authored-by: Michael Hsu <mike.hsu@gmail.com>
Co-authored-by: Mike Hsu <mike.hsu@gmail.com>
# Objective
- We rely on the construction of `EntityRef` to be valid elsewhere in unsafe code. This construction is not checked (for performance reasons), and thus this private method must be unsafe.
- Fixes#7218.
## Solution
- Make the method unsafe.
- Add safety docs.
- Improve safety docs slightly for the sibling `EntityMut::new`.
- Add debug asserts to start to verify these assumptions in debug mode.
## Context for reviewers
I attempted to verify the `EntityLocation` more thoroughly, but this turned out to be more work than expected. I've spun that off into #7221 as a result.
# Objective
Fixes#5859
## Solution
- Add `ClearChildren` and `ReplaceChildren` commands in the `crates/bevy_hierarchy/src/child_builder.rs`
---
## Changelog
- Added `ClearChildren` and `ReplaceChildren` struct
- Added `clear_children(&mut self) -> &mut Self` and `replace_children(&mut self, children: &[Entity]) -> &mut Self` function in `BuildChildren` trait
- Changed `PushChildren` `write` function body to a `push_children ` function to reused in `ReplaceChildren`
- Added `clear_children` function
- Added `push_and_replace_children_commands` and `push_and_clear_children_commands` test
Co-authored-by: ld000 <lidong9144@163.com>
Co-authored-by: lidong63 <lidong63@meituan.com>
# Objective
- Fixes a bug where `just_pressed` and `just_released` in `Input<GamepadButton>` might behave incorrectly due calling `clear` 3 times in a single frame through these three different systems: `gamepad_button_event_system`, `gamepad_axis_event_system` and `gamepad_connection_system` in any order
## Solution
- Call `clear` only once and before all the above three systems, i.e. in `gamepad_event_system`
## Additional Info
- Discussion in Discord: https://discord.com/channels/691052431525675048/768253008416342076/1064621963693273279
# Objective
The usages of the unsafe function `byte_add` are not properly documented.
Follow-up to #7151.
## Solution
Add safety comments to each call-site.
# Objective
Currently, the `AxisSettings::new` function is unusable due to
an implementation quirk. It only allows `AxisSettings` where
the bounds that are supposed to be positive are negative!
## Solution
- We fix the bound check
- We add a test to make sure the method is usable
Seems like the error slipped through because of the relatively
verbose code style. With all those `if/else`, very long names,
range syntax, the bound check is actually hard to spot. I first
refactored a lot of code, but I left out the refactor because the
fix should be integrated independently.
---
## Changelog
- Fix `AxisSettings::new` only accepting invalid bounds
# Objective
Add useful information about cursor position relative to a UI node. Fixes#7079.
## Solution
- Added a new `RelativeCursorPosition` component
---
## Changelog
- Added
- `RelativeCursorPosition`
- an example showcasing the new component
Co-authored-by: Dawid Piotrowski <41804418+Pietrek14@users.noreply.github.com>
# Objective
- Allow rendering queue systems to use a `Res<PipelineCache>` even for queueing up new rendering pipelines. This is part of unblocking parallel execution queue systems.
## Solution
- Make `PipelineCache` internally mutable w.r.t to queueing new pipelines. Pipelines are no longer immediately updated into the cache state, but rather queued into a Vec. The Vec of pending new pipelines is then later processed at the same time we actually create the queued pipelines on the GPU device.
---
## Changelog
`PipelineCache` no longer requires mutable access in order to queue render / compute pipelines.
## Migration Guide
* Most usages of `resource_mut::<PipelineCache>` and `ResMut<PipelineCache>` can be changed to `resource::<PipelineCache>` and `Res<PipelineCache>` as long as they don't use any methods requiring mutability - the only public method requiring it is `process_queue`.
# Objective
- The function `BlobVec::replace_unchecked` has informal use of safety comments.
- This function does strange things with `OwningPtr` in order to get around the borrow checker.
## Solution
- Put safety comments in front of each unsafe operation. Describe the specific invariants of each operation and how they apply here.
- Added a guard type `OnDrop`, which is used to simplify ownership transfer in case of a panic.
---
## Changelog
+ Added the guard type `bevy_utils::OnDrop`.
+ Added conversions from `Ptr`, `PtrMut`, and `OwningPtr` to `NonNull<u8>`.
# Objective
Fix#5248.
## Solution
Support `In<T>` parameters and allow returning arbitrary types in exclusive systems.
---
## Changelog
- Exclusive systems may now be used with system piping.
## Migration Guide
Exclusive systems (systems that access `&mut World`) now support system piping, so the `ExclusiveSystemParamFunction` trait now has generics for the `In`put and `Out`put types.
```rust
// Before
fn my_generic_system<T, Param>(system_function: T)
where T: ExclusiveSystemParamFunction<Param>
{ ... }
// After
fn my_generic_system<T, In, Out, Param>(system_function: T)
where T: ExclusiveSystemParamFunction<In, Out, Param>
{ ... }
```
# Objective
Pipelines can be customized by wrapping an existing pipeline in a newtype and adding custom logic to its implementation of `SpecializedMeshPipeline::specialize`. To make that easier, the wrapped pipeline type needs to implement `Clone`.
For example, the current non-cloneable pipelines require wrapper pipelines to pull apart the wrapped pipeline like this:
```rust
impl FromWorld for Wireframe2dPipeline {
fn from_world(world: &mut World) -> Self {
let p = &world.resource::<Material2dPipeline<ColorMaterial>>();
Self {
mesh2d_pipeline: p.mesh2d_pipeline.clone(),
material2d_layout: p.material2d_layout.clone(),
vertex_shader: p.vertex_shader.clone(),
fragment_shader: p.fragment_shader.clone(),
}
}
}
```
## Solution
Derive or implement `Clone` on all built-in pipeline types. This is easy to do since they mostly just contain cheaply clonable reference-counted types.
---
## Changelog
Implement `Clone` for all pipeline types.
# Objective
fix error with shadow shader's spotlight direction calculation when direction.y ~= 0
fixes#7152
## Solution
same as #6167: in shadows.wgsl, clamp 1-x^2-z^2 to >= 0 so that we can safely sqrt it
# Objective
There are some utility functions for actually working with `Storages` inside `entity_ref.rs` that are used both for `EntityRef/EntityMut` and `World`, with a `// TODO: move to Storages`.
This PR moves them to private methods on `World`, because that's the safest API boundary. On `Storages` you would need to ensure that you pass `Components` from the same world.
## Solution
- move get_component[_with_type], get_ticks[_with_type], get_component_and_ticks[_with_type] to `World` (still pub(crate))
- replace `pub use entity_ref::*;` with `pub use entity_ref::{EntityRef, EntityMut}` and qualified `entity_ref::get_mut[_by_id]` in `world.rs`
- add safety comments to a bunch of methods
# Objective
* `World::init_resource` and `World::get_resource_or_insert_with` are implemented naively, and as such they perform duplicate `TypeId -> ComponentId` lookups.
* `World::get_resource_or_insert_with` contains an additional duplicate `ComponentId -> ResourceData` lookup.
* This function also contains an unnecessary panic branch, which we rely on the optimizer to be able to remove.
## Solution
Implement the functions using engine-internal code, instead of combining high-level functions. This allows computed variables to persist across different branches, instead of being recomputed.
# Objective
It is often necessary to update an entity's parent
while keeping its GlobalTransform static. Currently
it is cumbersome and error-prone (two questions in
the discord `#help` channel in the past week)
- Part 2, resolves#5475
- Builds on: #7020.
## Solution
- Added the `BuildChildrenTransformExt` trait, it is part
of `bevy::prelude` and adds the following methods to `EntityCommands`:
- `set_parent_in_place`: Change the parent of an entity and
update its `Transform` in order to preserve its `GlobalTransform` after the parent change
- `remove_parent_in_place`: Remove an entity from a hierarchy,
while preserving its `GlobalTransform`.
---
## Changelog
- Added the `BuildChildrenTransformExt` trait, it is part
of `bevy::prelude` and adds the following methods to `EntityCommands`:
- `set_parent_in_place`: Change the parent of an entity and
update its `Transform` in order to preserve its `GlobalTransform` after the parent change
- `remove_parent_in_place`: Remove an entity from a hierarchy,
while preserving its `GlobalTransform`.
Co-authored-by: Nicola Papale <nicopap@users.noreply.github.com>
# Objective
- While building UI, it makes more sense for most nodes to have a `FocusPolicy` of `Pass`, so that user interaction can correctly bubble
- Only `ButtonBundle` blocks by default
This change means that for someone adding children to a button, it's not needed to change the focus policy of those children to `Pass` for the button to continue to work.
---
## Changelog
- `FocusPolicy` default has changed from `FocusPolicy::Block` to `FocusPolicy::Pass`
## Migration Guide
- `FocusPolicy` default has changed from `FocusPolicy::Block` to `FocusPolicy::Pass`
# Objective
The documentation of the bevy_render crate is still pretty incomplete.
This PR follows up on #6885 and improves the documentation of the `render_phase` module.
This module contains one of our most important rendering abstractions and the current documentation is pretty confusing. This PR tries to clarify what all of these pieces are for and how they work together to form bevy`s modular rendering logic.
## Solution
### Code Reformating
- I have moved the `rangefinder` into the `render_phase` module since it is only used there.
- I have moved the `PhaseItem` (and the `BatchedPhaseItem`) from `render_phase::draw` over to `render_phase::mod`. This does not change the public-facing API since they are reexported anyway, but this change makes the relation between `RenderPhase` and `PhaseItem` clear and easier to discover.
### Documentation
- revised all documentation in the `render_phase` module
- added a module-level explanation of how `RenderPhase`s, `RenderPass`es, `PhaseItem`s, `Draw` functions, and `RenderCommands` relate to each other and how they are used
---
## Changelog
- The `rangefinder` module has been moved into the `render_phase` module.
## Migration Guide
- The `rangefinder` module has been moved into the `render_phase` module.
```rust
//old
use bevy::render::rangefinder::*;
// new
use bevy::render::render_phase::rangefinder::*;
```
# Objective
Following #6681, both `TableRow` and `TableId` are now part of `EntityLocation`. However, the safety invariant on `EntityLocation` requires that all of the constituent fields are `repr(transprent)` or `repr(C)` and the bit pattern of all 1s must be valid. This is not true for `TableRow` and `TableId` currently.
## Solution
Mark `TableRow` and `TableId` to satisfy the safety requirement. Add safety comments on `ArchetypeId`, `ArchetypeRow`, `TableId` and `TableRow`.
# Objective
Improve safety testing when using `bevy_ptr` types. This is a follow-up to #7113.
## Solution
Add a debug-only assertion that pointers are aligned when casting to a concrete type. This should very quickly catch any unsoundness from unaligned pointers, even without miri. However, this can have a large negative perf impact on debug builds.
---
## Changelog
Added: `Ptr::deref` will now panic in debug builds if the pointer is not aligned.
Added: `PtrMut::deref_mut` will now panic in debug builds if the pointer is not aligned.
Added: `OwningPtr::read` will now panic in debug builds if the pointer is not aligned.
Added: `OwningPtr::drop_as` will now panic in debug builds if the pointer is not aligned.
# Objective
- It can be useful for third party crates to work independently on how bevy is imported
## Solution
- Expose an helper to get a subcrate path for macros
# Objective
- There is a warning when building in release:
```
warning: unused import: `bevy_ecs::system::Local`
--> crates/bevy_render/src/extract_resource.rs:5:5
|
5 | use bevy_ecs::system::Local;
| ^^^^^^^^^^^^^^^^^^^^^^^
|
= note: `#[warn(unused_imports)]` on by default
```
- It's used 59751d6e33/crates/bevy_render/src/extract_resource.rs (L47)
- Fix it
## Solution
- Gate the import
- repeat of #5320
# Objective
`MutUntyped` is a struct that stores a `PtrMut` alongside change tick metadata. Working with this type is cumbersome, and has few benefits over storing the pointer and change ticks separately.
Related: #6430 (title is out of date)
## Solution
Add a convenience method for transforming an untyped change detection pointer into its typed counterpart.
---
## Changelog
- Added the method `MutUntyped::with_type`.
As mentioned in https://github.com/bevyengine/bevy/pull/6530. It allows to not create a new constant and simply having it to show up in the documentation when someone is looking for "transparent" (case insensitive) in rustdoc search.
cc @alice-i-cecile
# Objective
Enums are now reflectable, but are not accessible via reflection paths.
This would allow us to do things like:
```rust
#[derive(Reflect)]
struct MyStruct {
data: MyEnum
}
#[derive(Reflect)]
struct MyEnum {
Foo(u32, u32),
Bar(bool)
}
let x = MyStruct {
data: MyEnum::Foo(123),
};
assert_eq!(*x.get_path::<u32>("data.1").unwrap(), 123);
```
## Solution
Added support for enums in reflection paths.
##### Note
This uses a simple approach of just getting the field with the given accessor. It does not do matching or anything else to ensure the enum is the intended variant. This means that the variant must be known ahead of time or matched outside the reflection path (i.e. path to variant, perform manual match, and continue pathing).
---
## Changelog
- Added support for enums in reflection paths
# Objective
There are times where we want to simply take an owned `dyn Reflect` and cast it to a type `T`.
Currently, this involves doing:
```rust
let value = value.take::<T>().unwrap_or_else(|value| {
T::from_reflect(&*value).unwrap_or_else(|| {
panic!(
"expected value of type {} to convert to type {}.",
value.type_name(),
std::any::type_name::<T>()
)
})
});
```
This is a common operation that could be easily be simplified.
## Solution
Add the `FromReflect::take_from_reflect` method. This first tries to `take` the value, calling `from_reflect` iff that fails.
```rust
let value = T::take_from_reflect(value).unwrap_or_else(|value| {
panic!(
"expected value of type {} to convert to type {}.",
value.type_name(),
std::any::type_name::<T>()
)
});
```
Based on suggestion from @soqb on [Discord](https://discord.com/channels/691052431525675048/1002362493634629796/1041046880316043374).
---
## Changelog
- Add `FromReflect::take_from_reflect` method
# Objective
- Fixes#7066
## Solution
- Split the ChangeDetection trait into ChangeDetection and ChangeDetectionMut
- Added Ref as equivalent to &T with change detection
---
## Changelog
- Support for Ref which allow inspecting change detection flags in an immutable way
## Migration Guide
- While bevy prelude includes both ChangeDetection and ChangeDetectionMut any code explicitly referencing ChangeDetection might need to be updated to ChangeDetectionMut or both. Specifically any reading logic requires ChangeDetection while writes requires ChangeDetectionMut.
use bevy_ecs::change_detection::DetectChanges -> use bevy_ecs::change_detection::{DetectChanges, DetectChangesMut}
- Previously Res had methods to access change detection `is_changed` and `is_added` those methods have been moved to the `DetectChanges` trait. If you are including bevy prelude you will have access to these types otherwise you will need to `use bevy_ecs::change_detection::DetectChanges` to continue using them.
# Objective
The types in the `bevy_ptr` accidentally did not document anything relating to alignment. This is unsound as many methods rely on the pointer being correctly aligned.
## Solution
This PR introduces new safety invariants on the `$ptr::new`, `$ptr::byte_offset` and `$ptr::byte_add` methods requiring them to keep the pointer aligned. This is consistent with the documentation of these pointer types which document them as being "type erased borrows".
As it was pointed out (by @JoJoJet in #7117) that working with unaligned pointers can be useful (for example our commands abstraction which does not try to align anything properly, see #7039) this PR also introduces a default type parameter to all the pointer types that specifies whether it has alignment requirements or not. I could not find any code in `bevy_ecs` that would need unaligned pointers right now so this is going unused.
---
## Changelog
- Correctly document alignment requirements on `bevy_ptr` types.
- Support variants of `bevy_ptr` types that do not require being correctly aligned for the pointee type.
## Migration Guide
- Safety invariants on `bevy_ptr` types' `new` `byte_add` and `byte_offset` methods have been changed. All callers should re-audit for soundness.
# Objective
- Spawn tasks from other threads onto an async executor, but limit those tasks to run on a specific thread.
- This is a continuation of trying to break up some of the changes in pipelined rendering.
- Eventually this will be used to allow `NonSend` systems to run on the main thread in pipelined rendering #6503 and also to solve #6552.
- For this specific PR this allows for us to store a thread executor in a thread local, rather than recreating a scope executor for every scope which should save on a little work.
## Solution
- We create a Executor that does a runtime check for what thread it's on before creating a !Send ticker. The ticker is the only way for the executor to make progress.
---
## Changelog
- create a ThreadExecutor that can only be ticked on one thread.
`Query`'s fields being `pub(crate)` means that the struct can be constructed via safe code from anywhere in `bevy_ecs` . This is Not Good since it is intended that all construction of this type goes through `Query::new` which is an `unsafe fn` letting various `Query` methods rely on those invariants holding even though they can be trivially bypassed.
This has no user facing impact
# Objective
It is possible to manually update `GlobalTransform`.
The engine actually assumes this is not possible.
For example, `propagate_transform` does not update children
of an `Entity` which **`GlobalTransform`** changed,
leading to unexpected behaviors.
A `GlobalTransform` set by the user may also be blindly
overwritten by the propagation system.
## Solution
- Remove `translation_mut`
- Explain to users that they shouldn't manually update the `GlobalTransform`
- Remove `global_vs_local.rs` example, since it misleads users
in believing that it is a valid use-case to manually update the
`GlobalTransform`
---
## Changelog
- Remove `GlobalTransform::translation_mut`
## Migration Guide
`GlobalTransform::translation_mut` has been removed without alternative,
if you were relying on this, update the `Transform` instead. If the given entity
had children or parent, you may need to remove its parent to make its transform
independent (in which case the new `Commands::set_parent_in_place` and
`Commands::remove_parent_in_place` may be of interest)
Bevy may add in the future a way to toggle transform propagation on
an entity basis.
# Objective
- Fix#7103.
- The issue is caused because I forgot to add a where clause to a generated struct in #7056.
## Solution
- Add the where clause.
`Query` relies on the `World` it stores being the same as the world used for creating the `QueryState` it stores. If they are not the same then everything is very unsound. This was not actually being checked anywhere, `Query::new` did not have a safety invariant or even an assertion that the `WorldId`'s are the same.
This shouldn't have any user facing impact unless we have really messed up in bevy and have unsoundness elsewhere (in which case we would now get a panic instead of being unsound).
# Objective
- In some cases, you need a `Mut<T>` pointer, but you only have a mutable reference to one. There is no easy way of converting `&'a mut Mut<'_, T>` -> `Mut<'a, T>` outside of the engine.
### Example (Before)
```rust
fn do_with_mut<T>(val: Mut<T>) { ... }
for x: Mut<T> in &mut query {
// The function expects a `Mut<T>`, so `x` gets moved here.
do_with_mut(x);
// Error: use of moved value.
do_a_thing(&x);
}
```
## Solution
- Add the function `reborrow`, which performs the mapping. This is analogous to `PtrMut::reborrow`.
### Example (After)
```rust
fn do_with_mut<T>(val: Mut<T>) { ... }
for x: Mut<T> in &mut query {
// We reborrow `x`, so the original does not get moved.
do_with_mut(x.reborrow());
// Works fine.
do_a_thing(&x);
}
```
---
## Changelog
- Added the method `reborrow` to `Mut`, `ResMut`, `NonSendMut`, and `MutUntyped`.
# Objective
This a follow-up to #6894, see https://github.com/bevyengine/bevy/pull/6894#discussion_r1045203113
The goal is to avoid cloning any string when getting a `&TypeRegistration` corresponding to a string which is being deserialized. As a bonus code duplication is also reduced.
## Solution
The manual deserialization of a string and lookup into the type registry has been moved into a separate `TypeRegistrationDeserializer` type, which implements `DeserializeSeed` with a `Visitor` that accepts any string with `visit_str`, even ones that may not live longer than that function call.
`BorrowedStr` has been removed since it's no longer used.
---
## Changelog
- The type `TypeRegistrationDeserializer` has been added, which simplifies getting a `&TypeRegistration` while deserializing a string.
# Objective
- Fix the name of function parameter name in docs
## Solution
- Change `f` to `sub_app_runner`
---
It confused me a bit when I was reading the docs in the autocomplete hint.
Hesitated about filing a PR since it's just a one single word change in the comment.
Is this the right process to change these docs?
# Objective
The type `Local<T>` unnecessarily has the bound `T: Sync` when the local is used in an exclusive system.
## Solution
Lift the bound.
---
## Changelog
Removed the bound `T: Sync` from `Local<T>` when used as an `ExclusiveSystemParam`.
# Objective
Fixes#3310. Fixes#6282. Fixes#6278. Fixes#3666.
## Solution
Split out `!Send` resources into `NonSendResources`. Add a `origin_thread_id` to all `!Send` Resources, check it on dropping `NonSendResourceData`, if there's a mismatch, panic. Moved all of the checks that `MainThreadValidator` would do into `NonSendResources` instead.
All `!Send` resources now individually track which thread they were inserted from. This is validated against for every access, mutation, and drop that could be done against the value.
A regression test using an altered version of the example from #3310 has been added.
This is a stopgap solution for the current status quo. A full solution may involve fully removing `!Send` resources/components from `World`, which will likely require a much more thorough design on how to handle the existing in-engine and ecosystem use cases.
This PR also introduces another breaking change:
```rust
use bevy_ecs::prelude::*;
#[derive(Resource)]
struct Resource(u32);
fn main() {
let mut world = World::new();
world.insert_resource(Resource(1));
world.insert_non_send_resource(Resource(2));
let res = world.get_resource_mut::<Resource>().unwrap();
assert_eq!(res.0, 2);
}
```
This code will run correctly on 0.9.1 but not with this PR, since NonSend resources and normal resources have become actual distinct concepts storage wise.
## Changelog
Changed: Fix soundness bug with `World: Send`. Dropping a `World` that contains a `!Send` resource on the wrong thread will now panic.
## Migration Guide
Normal resources and `NonSend` resources no longer share the same backing storage. If `R: Resource`, then `NonSend<R>` and `Res<R>` will return different instances from each other. If you are using both `Res<T>` and `NonSend<T>` (or their mutable variants), to fetch the same resources, it's strongly advised to use `Res<T>`.
# Objective
- Fixes#7061
## Solution
- Add and implement `insert` and `remove` methods for `List`.
---
## Changelog
- Added `insert` and `remove` methods to `List`.
- Changed the `push` and `pop` methods on `List` to have default implementations.
## Migration Guide
- Manual implementors of `List` need to implement the new methods `insert` and `remove` and
consider whether to use the new default implementation of `push` and `pop`.
Co-authored-by: radiish <thesethskigamer@gmail.com>
# Objective
Speed up the render phase for rendering.
## Solution
- Follow up #6988 and make the internals of atomic IDs `NonZeroU32`. This niches the `Option`s of the IDs in draw state, which reduces the size and branching behavior when evaluating for equality.
- Require `&RenderDevice` to get the device's `Limits` when initializing a `TrackedRenderPass` to preallocate the bind groups and vertex buffer state in `DrawState`, this removes the branch on needing to resize those `Vec`s.
## Performance
This produces a similar speed up akin to that of #6885. This shows an approximate 6% speed up in `main_opaque_pass_3d` on `many_foxes` (408.79 us -> 388us). This should be orthogonal to the gains seen there.
---
## Changelog
Added: `RenderContext::begin_tracked_render_pass`.
Changed: `TrackedRenderPass` now requires a `&RenderDevice` on construction.
Removed: `bevy_render::render_phase::DrawState`. It was not usable in any form outside of `bevy_render`.
## Migration Guide
TODO
# Objective
- This pulls out some of the changes to Plugin setup and sub apps from #6503 to make that PR easier to review.
- Separate the extract stage from running the sub app's schedule to allow for them to be run on separate threads in the future
- Fixes#6990
## Solution
- add a run method to `SubApp` that runs the schedule
- change the name of `sub_app_runner` to extract to make it clear that this function is only for extracting data between the main app and the sub app
- remove the extract stage from the sub app schedule so it can be run separately. This is done by adding a `setup` method to the `Plugin` trait that runs after all plugin build methods run. This is required to allow the extract stage to be removed from the schedule after all the plugins have added their systems to the stage. We will also need the setup method for pipelined rendering to setup the render thread. See e3267965e1/crates/bevy_render/src/pipelined_rendering.rs (L57-L98)
## Changelog
- Separate SubApp Extract stage from running the sub app schedule.
## Migration Guide
### SubApp `runner` has conceptually been changed to an `extract` function.
The `runner` no longer is in charge of running the sub app schedule. It's only concern is now moving data between the main world and the sub app. The `sub_app.app.schedule` is now run for you after the provided function is called.
```rust
// before
fn main() {
let sub_app = App::empty();
sub_app.add_stage(MyStage, SystemStage::parallel());
App::new().add_sub_app(MySubApp, sub_app, move |main_world, sub_app| {
extract(app_world, render_app);
render_app.app.schedule.run();
});
}
// after
fn main() {
let sub_app = App::empty();
sub_app.add_stage(MyStage, SystemStage::parallel());
App::new().add_sub_app(MySubApp, sub_app, move |main_world, sub_app| {
extract(app_world, render_app);
// schedule is automatically called for you after extract is run
});
}
```
# Objective
- Remove redundant gamepad events
- Simplify consuming gamepad events.
- Refactor: Separate handling of gamepad events into multiple systems.
## Solution
- Removed `GamepadEventRaw`, and `GamepadEventType`.
- Added bespoke `GamepadConnectionEvent`, `GamepadAxisChangedEvent`, and `GamepadButtonChangedEvent`.
- Refactored `gamepad_event_system`.
- Added `gamepad_button_event_system`, `gamepad_axis_event_system`, and `gamepad_connection_system`, which update the `Input` and `Axis` resources using their corresponding event type.
Gamepad events are now handled in their own systems and have their own types.
This allows for querying for gamepad events without having to match on `GamepadEventType` and makes creating handlers for specific gamepad event types, like a `GamepadConnectionEvent` or `GamepadButtonChangedEvent` possible.
We remove `GamepadEventRaw` by filtering the gamepad events, using `GamepadSettings`, _at the source_, in `bevy_gilrs`. This way we can create `GamepadEvent`s directly and avoid creating `GamepadEventRaw` which do not pass the user defined filters.
We expose ordered `GamepadEvent`s and we can respond to individual gamepad event types.
## Migration Guide
- Replace `GamepadEvent` and `GamepadEventRaw` types with their specific gamepad event type.
# Objective
- Fixes https://github.com/bevyengine/bevy/discussions/6338
This PR allows for smooth transitions between different animations.
## Solution
- This PR uses very simple linear blending of animations.
- When starting a new animation, you can give it a duration, and throughout that duration, the previous and the new animation are being linearly blended, until only the new animation is running.
- I'm aware of https://github.com/bevyengine/rfcs/pull/49 and https://github.com/bevyengine/rfcs/pull/51, which are more complete solutions to this problem, but they seem still far from being implemented. Until they're ready, this PR allows for the most basic use case of blending, i.e. smoothly transitioning between different animations.
## Migration Guide
- no bc breaking changes
# Objective
- Storage buffers are useful and not currently supported by the `AsBindGroup` derive which means you need to expand the macro if you need a storage buffer
## Solution
- Add a new `#[storage]` attribute to the derive `AsBindGroup` macro.
- Support and optional `read_only` parameter that defaults to false when not present.
- Support visibility parameters like the texture and sampler attributes.
---
## Changelog
- Add a new `#[storage(index)]` attribute to the derive `AsBindGroup` macro.
Co-authored-by: IceSentry <IceSentry@users.noreply.github.com>
# Objective
- Avoid slower than necessary first frame after spawning many entities due to them not having `Aabb`s and so being marked visible
- Avoids unnecessarily large system and VRAM allocations as a consequence
## Solution
- I noticed when debugging the `many_cubes` stress test in Xcode that the `MeshUniform` binding was much larger than it needed to be. I realised that this was because initially, all mesh entities are marked as being visible because they don't have `Aabb`s because `calculate_bounds` is being run in `PostUpdate` and there are no system commands applications before executing the visibility check systems that need the `Aabb`s. The solution then is to run the `calculate_bounds` system just before the previous system commands are applied which is at the end of the `Update` stage.
Spiritual successor to #5205.
Actual successor to #6865.
# Objective
Currently, system params are defined using three traits: `SystemParam`, `ReadOnlySystemParam`, `SystemParamState`. The behavior for each param is specified by the `SystemParamState` trait, while `SystemParam` simply defers to the state.
Splitting the traits in this way makes it easier to implement within macros, but it increases the cognitive load. Worst of all, this approach requires each `MySystemParam` to have a public `MySystemParamState` type associated with it.
## Solution
* Merge the trait `SystemParamState` into `SystemParam`.
* Remove all trivial `SystemParam` state types.
* `OptionNonSendMutState<T>`: you will not be missed.
---
- [x] Fix/resolve the remaining test failure.
## Changelog
* Removed the trait `SystemParamState`, merging its functionality into `SystemParam`.
## Migration Guide
**Note**: this should replace the migration guide for #6865.
This is relative to Bevy 0.9, not main.
The traits `SystemParamState` and `SystemParamFetch` have been removed, and their functionality has been transferred to `SystemParam`.
```rust
// Before (0.9)
impl SystemParam for MyParam<'_, '_> {
type State = MyParamState;
}
unsafe impl SystemParamState for MyParamState {
fn init(world: &mut World, system_meta: &mut SystemMeta) -> Self { ... }
}
unsafe impl<'w, 's> SystemParamFetch<'w, 's> for MyParamState {
type Item = MyParam<'w, 's>;
fn get_param(&mut self, ...) -> Self::Item;
}
unsafe impl ReadOnlySystemParamFetch for MyParamState { }
// After (0.10)
unsafe impl SystemParam for MyParam<'_, '_> {
type State = MyParamState;
type Item<'w, 's> = MyParam<'w, 's>;
fn init_state(world: &mut World, system_meta: &mut SystemMeta) -> Self::State { ... }
fn get_param<'w, 's>(state: &mut Self::State, ...) -> Self::Item<'w, 's>;
}
unsafe impl ReadOnlySystemParam for MyParam<'_, '_> { }
```
The trait `ReadOnlySystemParamFetch` has been replaced with `ReadOnlySystemParam`.
```rust
// Before
unsafe impl ReadOnlySystemParamFetch for MyParamState {}
// After
unsafe impl ReadOnlySystemParam for MyParam<'_, '_> {}
```
# Objective
- The doctest for `Mut::map_unchanged` uses a fake function `set_if_not_equal` to demonstrate usage.
- Now that #6853 has been merged, we can use `Mut::set_if_neq` directly instead of mocking it.
(github made me type out a message for the commit which looked like it was for the pr, sorry)
# Objective
- Add a way to get all of the input devices of an `Axis`, primarily useful for looping through them
## Solution
- Adds `Axis<T>::devices()` which returns a `FixedSizeIterator<Item = &T>`
- Adds a (probably unneeded) `test_axis_devices` test because tests are cool.
---
## Changelog
- Added `Axis<T>::devices()` method
## Migration Guide
Not a breaking change.
# Objective
`SystemParam` `Local`s documentation currently leaves out information that should be documented.
- What happens when multiple `SystemParam`s within the same system have the same `Local` type.
- What lifetime parameter is expected by `Local`.
## Solution
- Added sentences to documentation to communicate this information.
- Renamed `Local` lifetimes in code to `'s` where they previously were not. Users can get complicated incorrect suggested fixes if they pass the wrong lifetime. Some instance of the code had `'w` indicating the expected lifetime might not have been known to those that wrote the code either.
Co-authored-by: iiYese <83026177+iiYese@users.noreply.github.com>
# Objective
- Fixes#7081.
## Solution
- Moved functionality from kitchen sink plugin `CorePlugin` to separate plugins, `TaskPoolPlugin`, `TypeRegistrationPlugin`, `FrameCountPlugin`. `TaskPoolOptions` resource should now be used with `TaskPoolPlugin`.
## Changelog
Minimal changes made (code kept in `bevy_core/lib.rs`).
## Migration Guide
- `CorePlugin` broken into separate plugins. If not using `DefaultPlugins` or `MinimalPlugins` `PluginGroup`s, the replacement for `CorePlugin` is now to add `TaskPoolPlugin`, `TypeRegistrationPlugin`, and `FrameCountPlugin` to the app.
## Notes
- Consistent with Bevy goal "modularity over deep integration" but the functionality of `TypeRegistrationPlugin` and `FrameCountPlugin` is weak (the code has to go somewhere, though!).
- No additional tests written.
# Objective
It is currently possible to break reference counting for assets by creating a strong `HandleUntyped` and then modifying the `id` field before dropping the handle. This should not be allowed.
## Solution
Change the `id` field visibility to private and add a getter instead. The same change was previously done for `Handle<T>` in #6176, but `HandleUntyped` was forgotten.
---
## Migration Guide
- Instead of directly accessing the ID of a `HandleUntyped` as `handle.id`, use the new getter `handle.id()`.
# Objective
- When using `Color::hex` for the first time, I was confused by the fact that I can't specify colors using #, which is much more familiar.
- In the code editor (if there is support) there is a preview of the color, which is very convenient.
## Solution
- Allow you to enter colors like `#ff33f2` and use the `.strip_prefix` method to delete the `#` character.
# Objective
- Fix#4200
Currently, `#[derive(SystemParam)]` publicly exposes each field type, which makes it impossible to encapsulate private fields.
## Solution
Previously, the fields were leaked because they were used as an input generic type to the macro-generated `SystemParam::State` struct. That type has been changed to store its state in a field with a specific type, instead of a generic type.
---
## Changelog
- Fixed a bug that caused `#[derive(SystemParam)]` to leak the types of private fields.
# Objective
- Set the cursor grab mode after the window is built, fix#7007, clean some conversion code.
## Solution
- Set the cursor grab mode after the window is built.
# Objective
Fixes#6891
## Solution
Replaces deserializing map keys as `&str` with deserializing them as `String`.
This bug seems to occur when using something like `File` or `BufReader` rather than bytes or a string directly (I only tested `File` and `BufReader` for `rmp-serde` and `serde_json`). This might be an issue with other `Read` impls as well (except `&[u8]` it seems).
We already had passing tests for Message Pack but none that use a `File` or `BufReader`. This PR also adds or modifies tests to check for this in the future.
This change was also based on [feedback](https://github.com/bevyengine/bevy/pull/4561#discussion_r957385136) I received in a previous PR.
---
## Changelog
- Fix bug where scene deserialization using certain readers could fail (e.g. `BufReader`, `File`, etc.)
# Objective
Speed up animation by leveraging all threads in `ComputeTaskPool`.
## Solution
This PR parallelizes animation sampling across all threads.
To ensure that this is safely done, all animation is predicated with an ancestor query to ensure that there is no conflicting `AnimationPlayer` above each animated hierarchy that may cause this to alias.
Unlike the RFC, this does not add support for reflect based "animate anything", but only extends the existing `AnimationPlayer` to support high numbers of animated characters on screen at once.
## Performance
This cuts `many_foxes`'s frame time on my machine by a full millisecond, from 7.49ms to 6.5ms. (yellow is this PR, red is main).
---
## Changelog
Changed: Animation sampling now runs fully multi-threaded using threads from `ComputeTaskPool`.
Changed: `AnimationPlayer` that are on a child or descendant of another entity with another player will no longer be run.
# Objective
- Fixes#5529
## Solution
- Add assosciated constants named DEFAULT to as many types as possible
- Add const to as many methods in bevy_ui as possible
I have not applied the same treatment to the bundles in bevy_ui as it would require going into other bevy crates to implement const defaults for structs in bevy_text or relies on UiImage which calls HandleUntyped.typed() which isn't const safe.
Alternatively the defaults could relatively easily be turned into a macro to regain some of the readability and conciseness at the cost of explicitness.
Such a macro that partially implements this exists as a crate here: [const-default](https://docs.rs/const-default/latest/const_default/derive.ConstDefault.html) but does not support enums.
Let me know if there's anything I've missed or if I should push further into other crates.
Co-authored-by: Carter Anderson <mcanders1@gmail.com>
# Objective
Speed up the render phase of rendering. Simplify the trait structure for render commands.
## Solution
- Merge `EntityPhaseItem` into `PhaseItem` (`EntityPhaseItem::entity` -> `PhaseItem::entity`)
- Merge `EntityRenderCommand` into `RenderCommand`.
- Add two associated types to `RenderCommand`: `RenderCommand::ViewWorldQuery` and `RenderCommand::WorldQuery`.
- Use the new associated types to construct two `QueryStates`s for `RenderCommandState`.
- Hoist any `SQuery<T>` fetches in `EntityRenderCommand`s into the aformentioned two queries. Batch fetch them all at once.
## Performance
`main_opaque_pass_3d` is slightly faster on `many_foxes` (427.52us -> 401.15us)
The shadow pass node is also slightly faster (344.52 -> 338.24us)
## Future Work
- Can we hoist the view level queries out of the core loop?
---
## Changelog
Added: `PhaseItem::entity`
Added: `RenderCommand::ViewWorldQuery` associated type.
Added: `RenderCommand::ItemorldQuery` associated type.
Added: `Draw<T>::prepare` optional trait function.
Removed: `EntityPhaseItem` trait
## Migration Guide
TODO
# Objective
- The #7064 PR had poor performance on an M1 Max in MacOS due to significant overuse of registers resulting in 'register spilling' where data that would normally be stored in registers on the GPU is instead stored in VRAM. The latency to read from/write to VRAM instead of registers incurs a significant performance penalty.
- Use of registers is a limiting factor in shader performance. Assignment of a struct from memory to a local variable can incur copies. Passing a variable that has struct type as an argument to a function can also incur copies. As such, these two cases can incur increased register usage and decreased performance.
## Solution
- Remove/avoid a number of assignments of light struct type data to local variables.
- Remove/avoid a number of passing light struct type variables/data as value arguments to shader functions.
# Objective
- The recently merged PR #7013 does not allow multiple `RenderPhase`s to share the same `RenderPass`.
- Due to the introduced overhead we want to minimize the number of `RenderPass`es recorded during each frame.
## Solution
- Take a constructed `TrackedRenderPass` instead of a `RenderPassDiscriptor` as a parameter to the `RenderPhase::render` method.
---
## Changelog
To enable multiple `RenderPhases` to share the same `TrackedRenderPass`,
the `RenderPhase::render` signature has changed.
```rust
pub fn render<'w>(
&self,
render_pass: &mut TrackedRenderPass<'w>,
world: &'w World,
view: Entity)
```
Co-authored-by: Kurt Kühnert <51823519+kurtkuehnert@users.noreply.github.com>
# Objective
`Query::get` and other random access methods require looking up `EntityLocation` for every provided entity, then always looking up the `Archetype` to get the table ID and table row. This requires 4 total random fetches from memory: the `Entities` lookup, the `Archetype` lookup, the table row lookup, and the final fetch from table/sparse sets. If `EntityLocation` contains the table ID and table row, only the `Entities` lookup and the final storage fetch are required.
## Solution
Add `TableId` and table row to `EntityLocation`. Ensure it's updated whenever entities are moved around. To ensure `EntityMeta` does not grow bigger, both `TableId` and `ArchetypeId` have been shrunk to u32, and the archetype index and table row are stored as u32s instead of as usizes. This should shrink `EntityMeta` by 4 bytes, from 24 to 20 bytes, as there is no padding anymore due to the change in alignment.
This idea was partially concocted by @BoxyUwU.
## Performance
This should restore the `Query::get` "gains" lost to #6625 that were introduced in #4800 without being unsound, and also incorporates some of the memory usage reductions seen in #3678.
This also removes the same lookups during add/remove/spawn commands, so there may be a bit of a speedup in commands and `Entity{Ref,Mut}`.
---
## Changelog
Added: `EntityLocation::table_id`
Added: `EntityLocation::table_row`.
Changed: `World`s can now only hold a maximum of 2<sup>32</sup>- 1 archetypes.
Changed: `World`s can now only hold a maximum of 2<sup>32</sup> - 1 tables.
## Migration Guide
A `World` can only hold a maximum of 2<sup>32</sup> - 1 archetypes and tables now. If your use case requires more than this, please file an issue explaining your use case.
# Objective
There isn't really a way to test that code using bevy_reflect compiles or doesn't compile for certain scenarios. This would be especially useful for macro-centric PRs like #6511 and #6042.
## Solution
Using `bevy_ecs_compile_fail_tests` as reference, added the `bevy_reflect_compile_fail_tests` crate.
Currently, this crate contains a very simple test case. This is so that we can get the basic foundation of this crate agreed upon and merged so that more tests can be added by other PRs.
### Open Questions
- [x] Should this be added to CI? (Answer: Yes)
---
## Changelog
- Added the `bevy_reflect_compile_fail_tests` crate for testing compilation errors
# Objective
Solve #5464
## Solution
Adds a `SystemInformationDiagnosticsPlugin` to add diagnostics.
Adds `Cargo.toml` flags to fix building on different platforms.
---
## Changelog
Adds `sysinfo` crate to `bevy-diagnostics`.
Changes in import order are due to clippy.
Co-authored-by: l1npengtul <35755164+l1npengtul@users.noreply.github.com>
Co-authored-by: IceSentry <c.giguere42@gmail.com>
# Objective
`TEXTURE_ADAPTER_SPECIFIC_FORMAT_FEATURES` was already included in `adapter.features()` on non-wasm target, and since it is the default value for `WgpuSettings.features`, the subsequent code will also combine into this feature:
b6066c30b6/crates/bevy_render/src/renderer/mod.rs (L155-L156)
# Objective
Bevy uses custom `Ptr` types so the rust borrow checker can help ensure lifetimes are correct, even when types aren't known. However, these types don't benefit from the automatic lifetime coercion regular rust references enjoy
## Solution
Add a couple methods to Ptr, PtrMut, and MutUntyped to allow for easy usage of these types in more complex scenarios.
## Changelog
- Added `as_mut` and `as_ref` methods to `MutUntyped`.
- Added `shrink` and `as_ref` methods to `PtrMut`.
## Migration Guide
- `MutUntyped::into_inner` now marks things as changed.
# Objective
All `RenderPhases` follow the same render procedure.
The same code is duplicated multiple times across the codebase.
## Solution
I simply extracted this code into a method on the `RenderPhase`.
This avoids code duplication and makes setting up new `RenderPhases` easier.
---
## Changelog
### Changed
You can now set up the rendering code of a `RenderPhase` directly using the `RenderPhase::render` method, instead of implementing it manually in your render graph node.
# Objective
- Loading a gltf files prints many errors
```
ERROR bevy_ecs::world: Unable to send event `bevy_hierarchy::events::HierarchyEvent`
Event must be added to the app with `add_event()`
https://docs.rs/bevy/*/bevy/app/struct.App.html#method.add_event
```
- Loading a gltf file create a world for a scene where events are not registered. Executing hierarchy commands on that world should not print error
## Solution
- Revert part of #6921
- don't use `world.send_event` / `world.send_event_batch` from commands
# Objective
- We already log the adapter info on startup when bevy_render is present. It would be nice to have more info about the system to be able to ask users to submit it in bug reports
## Solution
- Use the `sysinfo` crate to get all the information
- I made sure it _only_ gets the required informations to avoid unnecessary system request
- Add a system that logs this on startup
- This system is currently in `bevy_diagnostics` because I didn't really know where to put it.
Here's an example log from my system:
```log
INFO bevy_diagnostic: SystemInformation { os: "Windows 10 Pro", kernel: "19044", cpu: "AMD Ryzen 7 5800X 8-Core Processor", core_count: "8", memory: "34282242 KB" }
```
---
## Changelog
- Added a new default log when starting a bevy app that logs the system information
# Objective
It is often necessary to update an entity's parent while keeping its GlobalTransform static. Currently it is cumbersome and error-prone (two questions in the discord `#help` channel in the past week)
- Part 1 of #5475
- Part 2: #7024.
## Solution
- Add a `reparented_to` method to `GlobalTransform`
---
## Changelog
- Add a `reparented_to` method to `GlobalTransform`
# Objective
Resolve#6156.
The most common type of command is one that runs for a single entity. Built-in commands like this can be ergonomically added to the command queue using the `EntityCommands` struct. However, adding custom entity commands to the queue is quite cumbersome. You must first spawn an entity, store its ID in a local, then construct a command using that ID and add it to the queue. This prevents method chaining, which is the main benefit of using `EntityCommands`.
### Example (before)
```rust
struct MyCustomCommand(Entity);
impl Command for MyCustomCommand { ... }
let id = commands.spawn((...)).id();
commmands.add(MyCustomCommand(id));
```
## Solution
Add the `EntityCommand` trait, which allows directly adding per-entity commands to the `EntityCommands` struct.
### Example (after)
```rust
struct MyCustomCommand;
impl EntityCommand for MyCustomCommand { ... }
commands.spawn((...)).add(MyCustomCommand);
```
---
## Changelog
- Added the trait `EntityCommand`. This is a counterpart of `Command` for types that execute code for a single entity.
## Future Work
If we feel its necessary, we can simplify built-in commands (such as `Despawn`) to use this trait.
# Objective
Any closure with the signature `FnOnce(&mut World)` implicitly implements the trait `Command` due to a blanket implementation. However, this implementation unnecessarily has the `Sync` bound, which limits the types that can be used.
## Solution
Remove the bound.
---
## Changelog
- `Command` closures no longer need to implement the marker trait `std::marker::Sync`.
# Objective
The documentation for camera priority is very confusing at the moment, it requires a bit of "double negative" kind of thinking.
# Solution
Flipping the wording on the documentation to reflect more common usecases like having an overlay camera and also renaming it to "order", since priority implies that it will override the other camera rather than have both run.
Consolidation of all the feedback about #6271 as well as the addition of an "unconditionally visible" mode.
# Objective
The current implementation of the `Visibility` struct simply wraps a boolean.. which seems like an odd pattern when rust has such nice enums that allow for more expression using pattern-matching.
Additionally as it stands Bevy only has two settings for visibility of an entity:
- "unconditionally hidden" `Visibility { is_visible: false }`,
- "inherit visibility from parent" `Visibility { is_visible: true }`
where a root level entity set to "inherit" is visible.
Note that given the behaviour, the current naming of the inner field is a little deceptive or unclear.
Using an enum for `Visibility` opens the door for adding an extra behaviour mode. This PR adds a new "unconditionally visible" mode, which causes an entity to be visible even if its Parent entity is hidden. There should not really be any performance cost to the addition of this new mode.
--
The recently added `toggle` method is removed in this PR, as its semantics could be confusing with 3 variants.
## Solution
Change the Visibility component into
```rust
enum Visibility {
Hidden, // unconditionally hidden
Visible, // unconditionally visible
Inherited, // inherit visibility from parent
}
```
---
## Changelog
### Changed
`Visibility` is now an enum
## Migration Guide
- evaluation of the `visibility.is_visible` field should now check for `visibility == Visibility::Inherited`.
- setting the `visibility.is_visible` field should now directly set the value: `*visibility = Visibility::Inherited`.
- usage of `Visibility::VISIBLE` or `Visibility::INVISIBLE` should now use `Visibility::Inherited` or `Visibility::Hidden` respectively.
- `ComputedVisibility::INVISIBLE` and `SpatialBundle::VISIBLE_IDENTITY` have been renamed to `ComputedVisibility::HIDDEN` and `SpatialBundle::INHERITED_IDENTITY` respectively.
Co-authored-by: Carter Anderson <mcanders1@gmail.com>
# Objective
- Be able to name the type that `ManualEventReader::iter/iter_with_id` returns and `EventReader::iter/iter_with_id` by proxy.
Currently for the purpose of https://github.com/bevyengine/bevy/pull/5719
## Solution
- Create a custom `Iterator` type.
# Objective
- alternative to #2895
- as mentioned in #2535 the uuid based ids in the render module should be replaced with atomic-counted ones
## Solution
- instead of generating a random UUID for each render resource, this implementation increases an atomic counter
- this might be replaced by the ids of wgpu if they expose them directly in the future
- I have not benchmarked this solution yet, but this should be slightly faster in theory.
- Bevymark does not seem to be affected much by this change, which is to be expected.
- Nothing of our API has changed, other than that the IDs have lost their IMO rather insignificant documentation.
- Maybe the documentation could be added back into the macro, but this would complicate the code.
# Objective
- Describe the objective or issue this PR addresses.
SpritePipelineKey could use more constification.
## Solution
Constify SpritePipelineKey implementation.
## Changelog
Co-authored-by: AxiomaticSemantics <117950168+AxiomaticSemantics@users.noreply.github.com>
# Objective
Align the hierarchy API between `EntityCommands` and `EntityMut`.
Added missing methods to `EntityMut`.
Replaced the duplicate `Command` implementations with the ones on `EntityMut` (e.g. The `AddChild` command is now just `world.entity_mut(..).add_child(..)`)
Fixed `update_old_parents` not sending `ChildAdded` events.
This PR does not add `add_children` to `EntityMut` as I would like to remove it from `EntityCommands` instead in #6942.
## Changelog
* Added `add_child`, `set_parent` and `remove_parent` to `EntityMut`
* Fixed missing `ChildAdded` events
Co-authored-by: devil-ira <justthecooldude@gmail.com>
# Objective
* Currently, the `SystemParam` derive does not support types with const generic parameters.
* If you try to use const generics, the error message is cryptic and unhelpful.
* Continuation of the work started in #6867 and #6957.
## Solution
Allow const generic parameters to be used with `#[derive(SystemParam)]`.
# Objective
Fixes#4729.
Continuation of #4854.
## Solution
Add documentation to `ParamSet` and its methods. Includes examples suggested by community members in the original PR.
Co-authored-by: Nanox19435 <50684926+Nanox19435@users.noreply.github.com>
Co-authored-by: JoJoJet <21144246+JoJoJet@users.noreply.github.com>
# Objective
Fixes#4231.
## Solution
This PR implements the solution suggested by @bjorn3 : Use an internal property within `App` to detect `App::run()` calls from `Plugin::build()`.
---
## Changelog
- panic when App::run() is called from Plugin::build()
# Objective
Upgrade to Taffy 0.2
## Solution
Do it
## Changelog
Upgraded to Taffy 0.2, improving UI layout performance significantly and adding the flexbox `gap` property and `AlignContent::SpaceEvenly`.
## Notes
`many_buttons` is 8% faster! speed improvements for more highly nested UIs will be much more dramatic. Great work, Team Taffy.
# Objective
* The `SystemParam` derive internally uses tuples, which means it is constrained by the 16-field limit on `all_tuples`.
* The error message if you exceed this limit is abysmal.
* Supercedes #5965 -- this does the same thing, but is simpler.
## Solution
If any tuples have more than 16 fields, they are folded into tuples of tuples until they are under the 16-field limit.
# Objective
Currently, only named structs can be used with the `SystemParam` derive macro.
## Solution
Remove the restriction. Tuple structs and unit structs are now supported.
---
## Changelog
+ Added support for tuple structs and unit structs to the `SystemParam` derive macro.
# Objective
Fixes#6862 (oh hey good catch @alice-i-cecile)
Bevy was failing to print events from `info!()` and friends to the console if the `trace_tracy` feature was enabled. It shouldn't be doing that.
## Solution
The problem was this per-layer filter that was added in #4320 to suppress a noisy per-frame event (which Tracy requires in order to properly close out a frame):
- The problem event's target was `"bevy_render::renderer"`, not `"tracy"`. - So, the filter wasn't specifically targeting the noisy event.
- Without a default, `tracing_subscriber::filter::Targets` will remove _everything_ that doesn't match an explicit target rule. - So, the filter _was_ silencing the noisy event, along with everything else.
This PR changes that filter to do what was probably intended in #4320: suppress ~any events more verbose than `ERROR` from `bevy_render::renderer`~ the one problematically noisy event, but allow anything else that already made it through the top-level filter_layer.
Also, adds a comment to clarify the intent of that filter, since it's otherwise a bit opaque and required some research.
---
## Changelog
Fixed a bug that hid console log messages when the `trace_tracy` feature was enabled.
The Camera link in the UiCameraConfig was not rendered properly by the documentation.
# Objective
- In the UiCameraConfig page (https://docs.rs/bevy/latest/bevy/prelude/struct.UiCameraConfig.html), a link to the Camera page (https://docs.rs/bevy/latest/bevy/render/camera/struct.Camera.html) is broken.
## Solution
- It seems that when using URL fragment specifiers, backtick should not be used. It might be an issue of rust itself. Replacing the URL fragment specifier `[`Camera`]: bevy_render:📷:Camera` with `[Camera]: bevy_render:📷:Camera` solves this.
# Objective
- `bevy_ptr::{Ptr, PtrMut, OwnedPtr}` wrap raw pointers and should be printable using pointer formatting.
## Solution
- Add a `core::fmt::Pointer` impl for `Ptr`, `PtrMut` and `OwnedPtr` based on the wrapped `NonNull` pointer.
---
## Changelog
- Added a `core::fmt::Pointer` impl to `Ptr`, `PtrMut` and `OwnedPtr`.
Co-authored-by: MrGunflame <mrgunflame@protonmail.com>
## Objective
Bevy UI uses a `MeasureFunc` that preserves the aspect ratio of text, not just images. This means that the extent of flex-items containing text may be calculated incorrectly depending on the ratio of the text size compared to the size of its containing node.
Fixes#6748
Related to #6724
with Bevy 0.9:
with this PR (accurately matching the behavior of Flexbox):
## Solution
Only perform the aspect ratio calculations if the uinode contains an image.
## Changelog
* Added a field `preserve_aspect_ratio` to `CalculatedSize`
* The `MeasureFunc` only preserves the aspect ratio when `preserve_aspect_ratio` is true.
* `update_image_calculated_size_system` sets `preserve_aspect_ratio` to true for nodes with images.
# Objective
The `WgpuSettings` resource is only used during plugin build. Move it into the `RenderPlugin` struct.
Changing these settings requires re-initializing the render context, which is currently not supported.
If it is supported in the future it should probably be more explicit than changing a field on a resource, maybe something similar to the `CreateWindow` event.
## Migration Guide
```rust
// Before (0.9)
App::new()
.insert_resource(WgpuSettings { .. })
.add_plugins(DefaultPlugins)
// After (0.10)
App::new()
.add_plugins(DefaultPlugins.set(RenderPlugin {
wgpu_settings: WgpuSettings { .. },
}))
```
Co-authored-by: devil-ira <justthecooldude@gmail.com>
# Objective
Following #4402, extract systems run on the render world instead of the main world, and allow retained state operations on it's resources. We're currently extracting to `ExtractedJoints` and then copying it twice during Prepare. Once into `SkinnedMeshJoints` and again into the actual GPU buffer.
This makes #4902 obsolete.
## Solution
Cut out the middle copy and directly extract joints into `SkinnedMeshJoints` and remove `ExtractedJoints` entirely.
This also removes the per-frame allocation that is being made to send `ExtractedJoints` into the render world.
## Performance
On my local machine, this halves the time for `prepare_skinned _meshes` on `many_foxes` (195.75us -> 93.93us on average).
---
## Changelog
Added: `BufferVec::truncate`
Added: `BufferVec::extend`
Changed: `SkinnedMeshJoints::build` now takes a `&mut BufferVec` instead of a `&mut Vec` as a parameter.
Removed: `ExtractedJoints`.
## Migration Guide
`ExtractedJoints` has been removed. Read the bound bones from `SkinnedMeshJoints` instead.
# Objective
Fix#1991. Allow users to have a bit more control over the creation and finalization of the threads in `TaskPool`.
## Solution
Add new methods to `TaskPoolBuilder` that expose callbacks that are called to initialize and finalize each thread in the `TaskPool`.
Unlike the proposed solution in #1991, the callback is argument-less. If an an identifier is needed, `std:🧵:current` should provide that information easily.
Added a unit test to ensure that they're being called correctly.
# Objective
- Dynamic scene builder can build scenes without components, if they didn't have any matching the type registry
- Those entities are not really useful in the final `DynamicScene`
## Solution
- Add a method `remove_empty_entities` that will remove empty entities. It's not called by default when calling `build`, I'm not sure if that's a good idea or not.
# Objective
There is currently no way to iterate over key/value pairs inside an `EntityMap`, which makes the usage of this struct very awkward. I couldn't think of a good reason why the `iter()` function should not be exposed, considering the interface already exposes `keys()` and `values()`, so I made this PR.
## Solution
Implement `iter()` for `EntityMap` in terms of its inner map type.
# Objective
Some settings were only applied in windowed mode.
Fix the issue in #6933
# Solution
Always apply the settings.
Co-authored-by: devil-ira <justthecooldude@gmail.com>
# Objective
Remove a method with an unfortunate name and questionable usefulness.
Added in #4708
It doesn't make sense to me for us to provide a method to work around a limitation of closures when we can simply, *not* use a closure.
The limitation in this case is not being able to initialize a variable from inside a closure:
```rust
let child_id;
commands.spawn_empty().with_children(|parent| {
// Error: passing uninitalized variable to a closure.
child_id = parent.spawn_empty().id();
});
// Do something with child_id
```
The docs for `add_children` suggest the following:
```rust
let child_id = commands
.spawn_empty()
.add_children(|parent| parent.spawn_empty().id());
```
I would instead suggest using the following snippet.
```rust
let parent_id = commands.spawn_empty().id();
let child_id = commands.spawn_empty().set_parent(parent_id).id();
// To be fair, at the time of #4708 this would have been a bit more cumbersome since `set_parent` did not exist.
```
Using `add_children` gets more unwieldy when you also want the `parent_id`.
```rust
let parent_commands = commands.spawn_empty();
let parent_id = parent_commands.id();
let child_id = parent_commands.add_children(|parent| parent.spawn_empty().id());
```
### The name
I see why `add_children` is named that way, it's the non-builder variant of `with_children` so it kinda makes sense,
but now the method name situation for `add_child`, `add_children` and `push_children` is *rather* unfortunate.
Removing `add_children` and renaming `push_children` to `add_children` in one go is kinda bleh, but that way we end up with the matching methods `add_child` and `add_children`.
Another reason to rename `push_children` is that it's trying to mimick the `Vec` api naming but fails because `push` is for single elements. I guess it should have been `extend_children_from_slice`, but lets not name it that :)
### Questions
~~Should `push_children` be renamed in this PR? This would make the migration guide easier to deal with.~~
Let's do that later.
Does anyone know of a way to do a simple text/regex search through all the github repos for usage of `add_children`?
That way we can have a better idea of how this will affect users. My guess is that usage of `add_children` is quite rare.
## Migration Guide
The method `add_children` on `EntityCommands` was removed.
If you were using `add_children` over `with_children` to return data out of the closure you can use `set_parent` or `add_child` to avoid the closure instead.
Co-authored-by: devil-ira <justthecooldude@gmail.com>
# Objective
`AsBindGroup` can't be used as a trait object because of the constraint `Sized` and because of the associated function.
This is a problem for [`bevy_atmosphere`](https://github.com/JonahPlusPlus/bevy_atmosphere) because it needs to use a trait that depends on `AsBindGroup` as a trait object, for switching out different shaders at runtime. The current solution it employs is reimplementing the trait and derive macro into that trait, instead of constraining to `AsBindGroup`.
## Solution
Remove the `Sized` constraint from `AsBindGroup` and add the constraint `where Self: Sized` to the associated function `bind_group_layout`. Also change `PreparedBindGroup<T: AsBindGroup>` to `PreparedBindGroup<T>` and use it as `PreparedBindGroup<Self::Data>` instead of `PreparedBindGroup<Self>`.
This weakens the constraints, but increases the flexibility of `AsBindGroup`.
I'm not entirely sure why the `Sized` constraint was there, because it worked fine without it (maybe @cart wasn't aware of use cases for `AsBindGroup` as a trait object or this was just leftover from legacy code?).
---
## Changelog
- `AsBindGroup` can be used as a trait object.
# Objective
[Rust 1.66](https://blog.rust-lang.org/inside-rust/2022/12/12/1.66.0-prerelease.html) is coming in a few days, and bevy doesn't build with it.
Fix that.
## Solution
Replace output from a trybuild test, and fix a few new instances of `needless_borrow` and `unnecessary_cast` that are now caught.
## Note
Due to the trybuild test, this can't be merged until 1.66 is released.
# Objective
A separate `tracing` span for running a system's commands is created, even if the system doesn't have commands. This is adding extra measuring overhead (see #4892) where it's not needed.
## Solution
Move the span into `ParallelCommandState` and `CommandQueue`'s `SystemParamState::apply`. To get the right metadata for the span, a additional `&SystemMeta` parameter was added to `SystemParamState::apply`.
---
## Changelog
Added: `SystemMeta::name`
Changed: Systems without `Commands` and `ParallelCommands` will no longer show a "system_commands" span when profiling.
Changed: `SystemParamState::apply` now takes a `&SystemMeta` parameter in addition to the provided `&mut World`.
# Objective
Change detection can be spuriously triggered by setting a field to the same value as before. As a result, a common pattern is to write:
```rust
if *foo != value {
*foo = value;
}
```
This is confusing to read, and heavy on boilerplate.
Adopted from #5373, but untangled and rebased to current `bevy/main`.
## Solution
1. Add a method to the `DetectChanges` trait that implements this boilerplate when the appropriate trait bounds are met.
2. Document this minor footgun, and point users to it.
## Changelog
* added the `set_if_neq` method to avoid triggering change detection when the new and previous values are equal. This will work on both components and resources.
## Migration Guide
If you are manually checking if a component or resource's value is equal to its new value before setting it to avoid triggering change detection, migrate to the clearer and more convenient `set_if_neq` method.
## Context
Related to #2363 as it avoids triggering change detection, but not a complete solution (as it still requires triggering it when real changes are made).
Co-authored-by: Zoey <Dessix@Dessix.net>
# Objective
The following code:
```rs
use bevy::prelude::Image;
use image::{ DynamicImage, GenericImage, Rgba };
fn main() {
let mut dynamic_image = DynamicImage::new_rgb32f(1, 1);
dynamic_image.put_pixel(0, 0, Rgba([1, 1, 1, 1]));
let image = Image::from_dynamic(dynamic_image, false); // Panic!
println!("{image:?}");
}
```
Can cause an assertion failed:
```
thread 'main' panicked at 'assertion failed: `(left == right)`
left: `16`,
right: `14`: Pixel data, size and format have to match', .../bevy_render-0.9.1/src/texture/image.rs:209:9
stack backtrace:
...
4: core::panicking::assert_failed<usize,usize>
at /rustc/897e37553bba8b42751c67658967889d11ecd120/library/core/src/panicking.rs:181
5: bevy_render::texture::image::Image::new
at .../bevy_render-0.9.1/src/texture/image.rs:209
6: bevy_render::texture::image::Image::from_dynamic
at .../bevy_render-0.9.1/src/texture/image_texture_conversion.rs:159
7: bevy_test::main
at ./src/main.rs:8
...
```
It seems to be cause by a copypasta in `crates/bevy_render/src/texture/image_texture_conversion.rs`. Let's fix it.
## Solution
```diff
// DynamicImage::ImageRgb32F(image) => {
- let a = u16::max_value();
+ let a = 1f32;
```
This will fix the conversion.
---
## Changelog
- Fixed the alpha channel of the `image::DynamicImage::ImageRgb32F` to `bevy_render::texture::Image` conversion in `bevy_render::texture::Image::from_dynamic()`.
# Objective
- https://github.com/bevyengine/bevy/pull/5364 Added a few features to the AsBindGroup derive, but if you don't know they exist they aren't documented anywhere.
## Solution
- Document the new arguments in the doc block for the derive.
# Objective
Speed up bundle insertion and spawning from a bundle.
## Solution
Use the same technique used in #6800 to remove the branch on storage type when writing components from a `Bundle` into storage.
- Add a `StorageType` argument to the closure on `Bundle::get_components`.
- Pass `C::Storage::STORAGE_TYPE` into that argument.
- Match on that argument instead of reading from a `Vec<StorageType>` in `BundleInfo`.
- Marked all implementations of `Bundle::get_components` as inline to encourage dead code elimination.
The `Vec<StorageType>` in `BundleInfo` was also removed as it's no longer needed. If users were reliant on this, they can either use the compile time constants or fetch the information from `Components`. Should save a rather negligible amount of memory.
## Performance
Microbenchmarks show a slight improvement to inserting components into existing entities, as well as spawning from a bundle. Ranging about 8-16% faster depending on the benchmark.
```
group main soft-constant-write-components
----- ---- ------------------------------
add_remove/sparse_set 1.08 1019.0±80.10µs ? ?/sec 1.00 944.6±66.86µs ? ?/sec
add_remove/table 1.07 1343.3±20.37µs ? ?/sec 1.00 1257.3±18.13µs ? ?/sec
add_remove_big/sparse_set 1.08 1132.4±263.10µs ? ?/sec 1.00 1050.8±240.74µs ? ?/sec
add_remove_big/table 1.02 2.6±0.05ms ? ?/sec 1.00 2.5±0.08ms ? ?/sec
get_or_spawn/batched 1.15 401.4±17.76µs ? ?/sec 1.00 349.3±11.26µs ? ?/sec
get_or_spawn/individual 1.13 732.1±43.35µs ? ?/sec 1.00 645.6±41.44µs ? ?/sec
insert_commands/insert 1.12 623.9±37.48µs ? ?/sec 1.00 557.4±34.99µs ? ?/sec
insert_commands/insert_batch 1.16 401.4±17.00µs ? ?/sec 1.00 347.4±12.87µs ? ?/sec
insert_simple/base 1.08 416.9±5.60µs ? ?/sec 1.00 385.2±4.14µs ? ?/sec
insert_simple/unbatched 1.06 934.5±44.58µs ? ?/sec 1.00 881.3±47.86µs ? ?/sec
spawn_commands/2000_entities 1.09 190.7±11.41µs ? ?/sec 1.00 174.7±9.15µs ? ?/sec
spawn_commands/4000_entities 1.10 386.5±25.33µs ? ?/sec 1.00 352.3±18.81µs ? ?/sec
spawn_commands/6000_entities 1.10 586.2±34.42µs ? ?/sec 1.00 535.3±27.25µs ? ?/sec
spawn_commands/8000_entities 1.08 778.5±45.15µs ? ?/sec 1.00 718.0±33.66µs ? ?/sec
spawn_world/10000_entities 1.04 1026.4±195.46µs ? ?/sec 1.00 985.8±253.37µs ? ?/sec
spawn_world/1000_entities 1.06 103.8±20.23µs ? ?/sec 1.00 97.6±18.22µs ? ?/sec
spawn_world/100_entities 1.15 11.4±4.25µs ? ?/sec 1.00 9.9±1.87µs ? ?/sec
spawn_world/10_entities 1.05 1030.8±229.78ns ? ?/sec 1.00 986.2±231.12ns ? ?/sec
spawn_world/1_entities 1.01 105.1±23.33ns ? ?/sec 1.00 104.6±31.84ns ? ?/sec
```
---
## Changelog
Changed: `Bundle::get_components` now takes a `FnMut(StorageType, OwningPtr)`. The provided storage type must be correct for the component being fetched.
# Objective
Partially address #3492.
## Solution
Document the remaining undocumented members of `bevy_utils` and set `warn(missing_docs)` on the crate level. Also enabled `clippy::undocumented_unsafe_blocks` as a warning on the crate to keep it in sync with `bevy_ecs`'s warnings.
# Objective
```rust
// makes clippy complain about 'taking a mutable reference to a `const` item'
let color = *Color::RED.set_a(0.5);
// Now you can do
let color = Color::RED.with_a(0.5);
```
## Changelog
Added `with_r`, `with_g`, `with_b`, and `with_a` to `Color`.
Co-authored-by: devil-ira <justthecooldude@gmail.com>
# Objective
- Fixes https://github.com/bevyengine/bevy/issues/6417
## Solution
- clear_trackers was not being called on the render world. This causes the removed components vecs to continuously grow. This PR adds clear trackers to the end of RenderStage::Cleanup
## Migration Guide
The call to `clear_trackers` in `App` has been moved from the schedule to App::update for the main world and calls to `clear_trackers` have been added for sub_apps in the same function. This was due to needing stronger guarantees. If clear_trackers isn't called on a world it can lead to memory leaks in `RemovedComponents`.
# Objective
* Implementing a custom `SystemParam` by hand requires implementing three traits -- four if it is read-only.
* The trait `SystemParamFetch<'w, 's>` is a workaround from before we had generic associated types, and is no longer necessary.
## Solution
* Combine the trait `SystemParamFetch` with `SystemParamState`.
* I decided to remove the `Fetch` name and keep the `State` name, since the former was consistently conflated with the latter.
* Replace the trait `ReadOnlySystemParamFetch` with `ReadOnlySystemParam`, which simplifies trait bounds in generic code.
---
## Changelog
- Removed the trait `SystemParamFetch`, moving its functionality to `SystemParamState`.
- Replaced the trait `ReadOnlySystemParamFetch` with `ReadOnlySystemParam`.
## Migration Guide
The trait `SystemParamFetch` has been removed, and its functionality has been transferred to `SystemParamState`.
```rust
// Before
impl SystemParamState for MyParamState {
fn init(world: &mut World, system_meta: &mut SystemMeta) -> Self { ... }
}
impl<'w, 's> SystemParamFetch<'w, 's> for MyParamState {
type Item = MyParam<'w, 's>;
fn get_param(...) -> Self::Item;
}
// After
impl SystemParamState for MyParamState {
type Item<'w, 's> = MyParam<'w, 's>; // Generic associated types!
fn init(world: &mut World, system_meta: &mut SystemMeta) -> Self { ... }
fn get_param<'w, 's>(...) -> Self::Item<'w, 's>;
}
```
The trait `ReadOnlySystemParamFetch` has been replaced with `ReadOnlySystemParam`.
```rust
// Before
unsafe impl ReadOnlySystemParamFetch for MyParamState {}
// After
unsafe impl<'w, 's> ReadOnlySystemParam for MyParam<'w, 's> {}
```
# Objective
It's not clear to users how to handle `!Sync` types as components and resources in the absence of engine level support for them.
## Solution
Added a section to `Component`'s and `Resource`'s type level docs on available options for making a type `Sync` when it holds `!Sync` fields, linking `bevy_utils::synccell::SyncCell` and the currently unstable `std::sync::Exclusive`.
Also added a compile_fail doctest that illustrates how to apply `SyncCell`. These will break when/if #6572 gets merged, at which point these docs should be updated.
# Objective
This is an adoption of #5792. Fixes#5791.
## Solution
Implemented all the required reflection traits for `VecDeque`, taking from `Vec`'s impls.
---
## Changelog
Added: `std::collections::VecDeque` now implements `Reflect` and all relevant traits.
Co-authored-by: james7132 <contact@jamessliu.com>
# Objective
- Get rid of giant match statement to get PixelInfo.
- This will allow for supporting any texture that is uncompressed, instead of people needing to PR in any textures that are supported in wgpu, but not bevy.
## Solution
- More conservative alternative to https://github.com/bevyengine/bevy/pull/6788, where we don't try to make some of the calculations correct for compressed types.
- Delete `PixelInfo` and get the pixel_size directly from wgpu. Data from wgpu is here: https://docs.rs/wgpu-types/0.14.0/src/wgpu_types/lib.rs.html#2359
- Panic if the texture is a compressed type. An integer byte size of a pixel is no longer a valid concept when talking about compressed textures.
- All internal usages use `pixel_size` and not `pixel_info` and are on uncompressed formats. Most of these usages are on either explicit texture formats or slightly indirectly through `TextureFormat::bevy_default()`. The other uses are in `TextureAtlas` and have other calculations that assumes the texture is uncompressed.
## Changelog
- remove `PixelInfo` and get `pixel_size` from wgpu
## Migration Guide
`PixelInfo` has been removed. `PixelInfo::components` is equivalent to `texture_format.describe().components`. `PixelInfo::type_size` can be gotten from `texture_format.describe().block_size/ texture_format.describe().components`. But note this can yield incorrect results for some texture types like Rg11b10Float.
# Objective
Adds a cylinder shape. Fixes#2282.
## Solution
- I added a custom cylinder shape, taken from [here](https://github.com/rparrett/typey_birb/blob/main/src/cylinder.rs) with permission from @rparrett.
- I also added the cylinder shape to the `3d_shapes` example scene.
---
## Changelog
- Added cylinder shape
Co-Authored-By: Rob Parrett <robparrett@gmail.com>
Co-Authored-By: davidhof <7483215+davidhof@users.noreply.github.com>
# Objective
Fixes#6224, add ``dbg``, ``info``, ``warn`` and ``error`` system piping adapter variants to expand #5776, which call the corresponding re-exported [bevy_log macros](https://docs.rs/bevy/latest/bevy/log/macro.info.html) when the result is an error.
## Solution
* Added ``dbg``, ``info``, ``warn`` and ``error`` system piping adapter variants to ``system_piping.rs``.
* Modified and added tests for these under examples in ``system_piping.rs``.
# Objective
- Make `AudioOutput` a `Resource`.
## Solution
- Do not store `OutputStream` in the struct.
- `mem::forget` `OutputStream`.
---
## Changelog
### Added
- `AudioOutput` is now a `Resource`.
## Migration Guide
- Use `Res<AudioOutput<Source>>` instead of `NonSend<AudioOutput<Source>>`. Same for `Mut` variants.
# Objective
Resolves#4597 (based on the work from #6056 and a refresh of #4147)
When using reflection, we may often end up in a scenario where we have a Dynamic representing a certain type. Unfortunately, we can't just call `MyType::from_reflect` as we do not have knowledge of the concrete type (`MyType`) at runtime.
Such scenarios happen when we call `Reflect::clone_value`, use the reflection deserializers, or create the Dynamic type ourselves.
## Solution
Add a `ReflectFromReflect` type data struct.
This struct allows us to easily convert Dynamic representations of our types into their respective concrete instances.
```rust
#[derive(Reflect, FromReflect)]
#[reflect(FromReflect)] // <- Register `ReflectFromReflect`
struct MyStruct(String);
let type_id = TypeId::of::<MyStruct>();
// Register our type
let mut registry = TypeRegistry::default();
registry.register::<MyStruct>();
// Create a concrete instance
let my_struct = MyStruct("Hello world".to_string());
// `Reflect::clone_value` will generate a `DynamicTupleStruct` for tuple struct types
let dynamic_value: Box<dyn Reflect> = my_struct.clone_value();
assert!(!dynamic_value.is::<MyStruct>());
// Get the `ReflectFromReflect` type data from the registry
let rfr: &ReflectFromReflect = registry
.get_type_data::<ReflectFromReflect>(type_id)
.unwrap();
// Call `FromReflect::from_reflect` on our Dynamic value
let concrete_value: Box<dyn Reflect> = rfr.from_reflect(&dynamic_value);
assert!(concrete_value.is::<MyStruct>());
```
### Why this PR?
###### Why now?
The three main reasons I closed#4147 were that:
1. Registering `ReflectFromReflect` is clunky (deriving `FromReflect` *and* registering `ReflectFromReflect`)
2. The ecosystem and Bevy itself didn't seem to pay much attention to deriving `FromReflect`
3. I didn't see a lot of desire from the community for such a feature
However, as time has passed it seems 2 and 3 are not really true anymore. Bevy is internally adding lots more `FromReflect` derives, which should make this feature all the more useful. Additionally, I have seen a growing number of people look for something like `ReflectFromReflect`.
I think 1 is still an issue, but not a horrible one. Plus it could be made much, much better using #6056. And I think splitting this feature out of #6056 could lead to #6056 being adopted sooner (or at least make the need more clear to users).
###### Why not just re-open #4147?
The main reason is so that this PR can garner more attention than simply re-opening the old one. This helps bring fresh eyes to the PR for potentially more perspectives/reviews.
---
## Changelog
* Added `ReflectFromReflect`
Co-authored-by: Gino Valente <49806985+MrGVSV@users.noreply.github.com>
# Objective
#6547 accidentally broke change detection for SparseSet components by using `Ticks::from_tick_cells` with the wrong argument order.
## Solution
Use the right argument order. Add a regression test.
# Objective
The window event types currently don't support reflection. This PR adds support to them (as requested [here](https://github.com/bevyengine/bevy/issues/6223#issuecomment-1273852329)).
## Solution
Implement `Reflect` + `FromReflect` for window event types. Relevant traits are also being reflected with `#[reflect(...)]` attributes.
Additionally, this PR derives `Reflect` + `FromReflect` for `WindowDescriptor` and the types it depends on so that `CreateWindow` events can be fully manipulated through reflection.
Finally, this PR adds `FromReflect` for `PathBuf` as a value type, which is needed for `FileDragAndDrop`.
This adds the "glam" feature to the `bevy_reflect` dependency for package `bevy_window`. Since `bevy_window` transitively depends on `glam` already, all this brings in are the reflection `impl`s.
## Open questions
Should `app.register_type::<PathBuf>();` be moved to `CorePlugin`? I added it to `WindowPlugin` because that's where it's used and `CorePlugin` doesn't seem to register all the missing std types, but it would also make sense in `CorePlugin` I believe since it's a commonly used type.
---
## Changelog
Added:
- Implemented `Reflect` + `FromReflect` for window events and related types. These types are automatically registered when adding the `WindowPlugin`.
# Objective
- Fixes#6841
- In some case, the number of maximum storage buffers is `u32::MAX` which doesn't fit in a `i32`
## Solution
- Add an option to have a `u32` in a `ShaderDefVal`
having `doc(hidden)` on the read only version of a generated mutable world query leads to docs on the readonly item having a dead link. It also makes it annoying to have nice docs for libraries attempting to expose derived `WorldQuery` structs as re-exporting the read only item does not cause it to appear in docs even though it would be intended for users to know about the read only world query and use it.
# Objective
Fixes#6866.
## Solution
Docs now should describe what the _front_, _first_, _back_, and _last_ elements are for an implementor of the `bevy::reflect::list::List` Trait. Further, the docs should describe how `bevy::reflect::list::List::push` and `bevy::reflect::list::List::pop` should act on these elements.
Co-authored-by: Linus Käll <linus.kall.business@gmail.com>
# Objective
Prevent future unsoundness that was seen in #6623.
## Solution
Newtype both indexes in `Archetype` and `Table` as `ArchetypeRow` and `TableRow`. This avoids weird numerical manipulation on the indices, and can be stored and treated opaquely. Also enforces the source and destination of where these indices at a type level.
---
## Changelog
Changed: `Archetype` indices and `Table` rows have been newtyped as `ArchetypeRow` and `TableRow`.
# Objective
`EntityRef::get` and friends all type erase calls to fetch the target components by using passing in the `TypeId` instead of using generics. This is forcing a lookup to `Components` to fetch the storage type. This adds an extra memory lookup and forces a runtime branch instead of allowing the compiler to optimize out the unused branch.
## Solution
Leverage `Component::Storage::STORAGE_TYPE` as a constant instead of fetching the metadata from `Components`.
## Performance
This has a near 2x speedup for all calls to `World::get`. Microbenchmark results from my local machine. `Query::get_component`, which uses `EntityRef::get` internally also show a slight speed up. This has closed the gap between `World::get` and `Query::get` for the same use case.
```
group entity-ref-generics main
----- ------------------- ----
query_get_component/50000_entities_sparse 1.00 890.6±40.42µs ? ?/sec 1.10 980.6±28.22µs ? ?/sec
query_get_component/50000_entities_table 1.00 968.5±73.73µs ? ?/sec 1.08 1048.8±31.76µs ? ?/sec
query_get_component_simple/system 1.00 703.2±4.37µs ? ?/sec 1.00 702.1±6.13µs ? ?/sec
query_get_component_simple/unchecked 1.02 855.8±8.98µs ? ?/sec 1.00 843.1±8.19µs ? ?/sec
world_get/50000_entities_sparse 1.00 202.3±3.15µs ? ?/sec 1.85 374.0±20.96µs ? ?/sec
world_get/50000_entities_table 1.00 193.0±1.78µs ? ?/sec 2.02 389.2±26.55µs ? ?/sec
world_query_get/50000_entities_sparse 1.01 162.4±2.23µs ? ?/sec 1.00 161.3±0.95µs ? ?/sec
world_query_get/50000_entities_table 1.00 199.9±0.63µs ? ?/sec 1.00 200.2±0.74µs ? ?/sec
```
This should also, by proxy, speed up the `ReflectComponent` APIs as most of those use `World::get` variants internally.
# Objective
- Fixes#3004
## Solution
- Replaced all the types with their fully quallified names
- Replaced all trait methods and inherent methods on dyn traits with their fully qualified names
- Made a new file `fq_std.rs` that contains structs corresponding to commonly used Structs and Traits from `std`. These structs are replaced by their respective fully qualified names when used inside `quote!`
# Objective
- Fixes#6711
## Solution
- Change the `path` function parameter of `dynamically_load_plugin` and `DynamicPluginExt::load_plugin` to a generic with `AsRef<OsStr>` bound
# Objective
`prepare_asset` for Image has an alternate path for texture creation that is used when the image is not compressed and does not contain mipmaps. This additional code path is unnecessary as `render_device.create_texture_with_data()` will handle both cases correctly.
## Solution
Use `render_device.create_texture_with_data()` in all cases.
Tested successfully with the following examples:
- load_gltf
- render_to_texture
- texture
- 3d_shapes
- sprite
- sprite_sheet
- array_texture
- shader_material_screenspace_texture
- skybox (though this already would use the `create_texture_with_data()` branch anyway)
# Objective
The methods `World::change_tick` and `World::read_change_tick` lack documentation and have confusingly similar behavior.
## Solution
Add documentation and clarify the distinction between the two functions.
The PR fixes the interface of `EventReader::clear`. Currently, the method consumes the reader, which makes it unusable.
## Changelog
- `EventReader::clear` now takes a mutable reference instead of consuming the event reader.
## Migration Guide
`EventReader::clear` now takes a mutable reference instead of consuming the event reader. This means that `clear` now needs explicit mutable access to the reader variable, which previously could have been omitted in some cases:
```rust
// Old (0.9)
fn clear_events(reader: EventReader<SomeEvent>) {
reader.clear();
}
// New (0.10)
fn clear_events(mut reader: EventReader<SomeEvent>) {
reader.clear();
}
```
Co-authored-by: Carter Anderson <mcanders1@gmail.com>
# Objective
- Fixes#6812.
## Solution
- Replaced `World::read_change_ticks` with `World::change_ticks` within `bevy_ecs` crate in places where `World` references were mutable.
---
# Objective
Partially addresses #5504. Allow users to get an `Iterator<Item = EntityRef<'a>>` over all entities in the `World`.
## Solution
Change `World::iter_entities` to return an iterator of `EntityRef` instead of `Entity`.
Not sure how to tackle making an `Iterator<Item = EntityMut<'_>>` without being horribly unsound. Might need to wait for `LendingIterator` to stabilize so we can ensure only one of them is valid at a given time.
---
## Changelog
Changed: `World::iter_entities` now returns an iterator of `EntityRef` instead of `Entity`.
# Objective
Currently, the `SystemParam` derive forces you to declare the lifetime parameters `<'w, 's>`, even if you don't use them.
If you don't follow this structure, the error message is quite nasty.
### Example (before):
```rust
#[derive(SystemParam)]
pub struct EventWriter<'w, 's, E: Event> {
events: ResMut<'w, Events<E>>,
// The derive forces us to declare the `'s` lifetime even though we don't use it,
// so we have to add this `PhantomData` to please rustc.
#[system_param(ignore)]
_marker: PhantomData<&'s ()>,
}
```
## Solution
* Allow the user to omit either lifetime.
* Emit a descriptive error if any lifetimes used are invalid.
### Example (after):
```rust
#[derive(SystemParam)]
pub struct EventWriter<'w, E: Event> {
events: ResMut<'w, Events<E>>,
}
```
---
## Changelog
* The `SystemParam` derive is now more flexible, allowing you to omit unused lifetime parameters.
Updates the requirements on [tracing-chrome](https://github.com/thoren-d/tracing-chrome) to permit the latest version.
<details>
<summary>Release notes</summary>
<p><em>Sourced from <a href="https://github.com/thoren-d/tracing-chrome/releases">tracing-chrome's releases</a>.</em></p>
<blockquote>
<h2>Release v0.7.0</h2>
<ul>
<li>Add <code>start_new</code> to <code>FlushGuard</code>. You can now generate multiple traces in one run!</li>
<li>Clean up dependencies</li>
<li>Make events thread-scoped</li>
</ul>
</blockquote>
</details>
<details>
<summary>Commits</summary>
<ul>
<li><a href="8760d81206"><code>8760d81</code></a> Prepare for 0.7 (<a href="https://github-redirect.dependabot.com/thoren-d/tracing-chrome/issues/16">#16</a>)</li>
<li><a href="a30bfb78d2"><code>a30bfb7</code></a> Update documentation (<a href="https://github-redirect.dependabot.com/thoren-d/tracing-chrome/issues/15">#15</a>)</li>
<li><a href="3fe24ff44d"><code>3fe24ff</code></a> Save thread names for start_new (<a href="https://github-redirect.dependabot.com/thoren-d/tracing-chrome/issues/14">#14</a>)</li>
<li><a href="fa8a0ff8ba"><code>fa8a0ff</code></a> Adding "Start New" feature that allows a user to finalize writing to the (<a href="https://github-redirect.dependabot.com/thoren-d/tracing-chrome/issues/11">#11</a>)</li>
<li><a href="d3059d66b3"><code>d3059d6</code></a> Directly depend on crossbeam_channel (<a href="https://github-redirect.dependabot.com/thoren-d/tracing-chrome/issues/12">#12</a>)</li>
<li><a href="441dba5c21"><code>441dba5</code></a> change scope of instant events to thread (<a href="https://github-redirect.dependabot.com/thoren-d/tracing-chrome/issues/13">#13</a>)</li>
<li>See full diff in <a href="https://github.com/thoren-d/tracing-chrome/compare/v0.6.0...v0.7.0">compare view</a></li>
</ul>
</details>
<br />
Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting `@dependabot rebase`.
[//]: # (dependabot-automerge-start)
[//]: # (dependabot-automerge-end)
---
<details>
<summary>Dependabot commands and options</summary>
<br />
You can trigger Dependabot actions by commenting on this PR:
- `@dependabot rebase` will rebase this PR
- `@dependabot recreate` will recreate this PR, overwriting any edits that have been made to it
- `@dependabot merge` will merge this PR after your CI passes on it
- `@dependabot squash and merge` will squash and merge this PR after your CI passes on it
- `@dependabot cancel merge` will cancel a previously requested merge and block automerging
- `@dependabot reopen` will reopen this PR if it is closed
- `@dependabot close` will close this PR and stop Dependabot recreating it. You can achieve the same result by closing it manually
- `@dependabot ignore this major version` will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)
- `@dependabot ignore this minor version` will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)
- `@dependabot ignore this dependency` will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)
</details>
# Objective
Fixes#6827
## Solution
Use the `Color::rgba_linear` function instead of the `Color::rgba` function to correctly interpret colors from glTF files in the linear color space rather than the incorrect sRGB color space
# Objective
> Followup to [this](https://github.com/bevyengine/bevy/pull/6755#discussion_r1032671178) comment
Rearrange the impls in the `impls/std.rs` file.
The issue was that I had accidentally misplaced the impl for `Option<T>` and put it between the `Cow<'static, str>` impls. This is just a slight annoyance and readability issue.
## Solution
Move the `Option<T>` and `&'static Path` impls around to be more readable.
# Objective
The soundness of the ECS `World` partially relies on the correctness of the state of `Entities` stored within it. We're currently allowing users to (unsafely) mutate it, as well as readily construct it without using a `World`. While this is not strictly unsound so long as users (including `bevy_render`) safely use the APIs, it's a fairly easy path to unsoundness without much of a guard rail.
Addresses #3362 for `bevy_ecs::entity`. Incorporates the changes from #3985.
## Solution
Remove `Entities`'s `Default` implementation and force access to the type to only be through a properly constructed `World`.
Additional cleanup for other parts of `bevy_ecs::entity`:
- `Entity::index` and `Entity::generation` are no longer `pub(crate)`, opting to force the rest of bevy_ecs to use the public interface to access these values.
- `EntityMeta` is no longer `pub` and also not `pub(crate)` to attempt to cut down on updating `generation` without going through an `Entities` API. It's currently inaccessible except via the `pub(crate)` Vec on `Entities`, there was no way for an outside user to use it.
- Added `Entities::set`, an unsafe `pub(crate)` API for setting the location of an Entity (parallel to `Entities::get`) that replaces the internal case where we need to set the location of an entity when it's been spawned, moved, or despawned.
- `Entities::alloc_at_without_replacement` is only used in `World::get_or_spawn` within the first party crates, and I cannot find a public use of this API in any ecosystem crate that I've checked (via GitHub search).
- Attempted to document the few remaining undocumented public APIs in the module.
---
## Changelog
Removed: `Entities`'s `Default` implementation.
Removed: `EntityMeta`
Removed: `Entities::alloc_at_without_replacement` and `AllocAtWithoutReplacement`.
Co-authored-by: james7132 <contact@jamessliu.com>
Co-authored-by: James Liu <contact@jamessliu.com>
# Objective
- Since #5900 3d examples fail in wasm
```
ERROR crates/bevy_render/src/render_resource/pipeline_cache.rs:660 failed to process shader: Unknown shader def: 'AVAILABLE_STORAGE_BUFFER_BINDINGS'
```
## Solution
- Fix it by always adding the shaderdef `AVAILABLE_STORAGE_BUFFER_BINDINGS` with the actual value, instead of 3 when 3 or more were available
# Objective
- Support textures in `Rgb9e5Ufloat` format.
## Solution
- Add `TextureFormatPixelInfo` for `Rgb9e5Ufloat`.
Tested this with a `Rgb9e5Ufloat` encoded KTX2 texture.
# Objective
- Every usage of `DrawFunctionsInternals::get_id()` was followed by a `.unwrap()`. which just adds boilerplate.
## Solution
- Introduce a fallible version of `DrawFunctionsInternals::get_id()` and use it where possible.
- I also took the opportunity to improve the error message a little in the case where it fails.
---
## Changelog
- Added `DrawFunctionsInternals::id()`
# Objective
Document `bevy_ecs::archetype` and and declutter the public documentation for the module by making types non-`pub`.
Addresses #3362 for `bevy_ecs::archetype`.
## Solution
- Add module level documentation.
- Add type and API level documentation for all public facing types.
- Make `ArchetypeId`, `ArchetypeGeneration`, and `ArchetypeComponentId` truly opaque IDs that are not publicly constructable.
- Make `AddBundle` non-pub, make `Edges::get_add_bundle` return a `Option<ArchetypeId>` and fork the existing function into `Edges::get_add_bundle_internal`.
- Remove `pub(crate)` on fields that have a corresponding pub accessor function.
- Removed the `Archetypes: Default` impl, opting for a `pub(crate) fn new` alternative instead.
---
## Changelog
Added: `ArchetypeGeneration` now implements `Ord` and `PartialOrd`.
Removed: `Archetypes`'s `Default` implementation.
Removed: `Archetype::new` and `Archetype::is_empty`.
Removed: `ArchetypeId::new` and `ArchetypeId::value`.
Removed: `ArchetypeGeneration::value`
Removed: `ArchetypeIdentity`.
Removed: `ArchetypeComponentId::new` and `ArchetypeComponentId::value`.
Removed: `AddBundle`. `Edges::get_add_bundle` now returns `Option<ArchetypeId>`
# Objective
- The documentation describing different ways to spawn an Entity is missing reference to "method" for "Spawn an entity with components".
## Solution
- Update the documentation to add the reference to `World::spawn`.
# Objective
* Fix#6668
* There is no need to panic when a parenting operation is redundant, as no invalid state is entered.
## Solution
Make `push_children` idempotent.
# Objective
One of the use-cases for the function `Entity::from_raw` is creating placeholder entity ids, which are meant to be overwritten later. If we use a constant for this instead of `from_raw`, it is more ergonomic and self-documenting.
## Solution
Add a constant that returns an entity ID with an index of `u32::MAX` and a generation of zero. Users are instructed to overwrite this value before using it.
# Objective
When a global tracing subscriber has already been set, `LogPlugin` panics with an error message explaining this. However, if a global logger has already been set, it simply panics on an unwrap.
#6426 mentiones the panic and has been fixed by unique plugins, but the panic can still occur if a logger has been set through different means or multiple apps are created, as in #4934. The solution to that specific case isn't clear; this PR only fixes the missing error message.
## Solution
- ~add error message to panic~
- turn into warning
# Objective
- Reduce confusion around uniform bindings in materials. I've seen multiple people on discord get confused by it because it uses a struct that is named the same in the rust code and the wgsl code, but doesn't contain the same data. Also, the only reason this works is mostly by chance because the memory happens to align correctly.
## Solution
- Remove the confusing parts of the doc
## Notes
It's not super clear in the diff why this causes confusion, but essentially, the rust code defines a `CustomMaterial` struct with a color and a texture, but in the wgsl code the struct with the same name only contains the color. People are confused by it because the struct in wgsl doesn't need to be there.
You _can_ have complex structs on each side and the macro will even combine it for you if you reuse a binding index, but as it is now, this example seems to confuse more than help people.
# Objective
Fixes#6739
## Solution
Implement the required traits. They cannot be implemented for `Path` directly, since it is a dynamically-sized type.
# Objective
Fixes#6642
In a way that doesn't create any breaking changes, as a possible way to fix the above in a patch release.
## Solution
Don't actually remove font atlases when `max_font_atlases` is exceeded. Add a warning instead.
Keep `TextError::ExceedMaxTextAtlases` and `TextSettings` as-is so we don't break anything.
This is a bit of a cop-out, but the problems revealed by #6642 seem very challenging to fix properly.
Maybe follow up later with something more like https://github.com/rparrett/bevy/commits/remove-max-font-atlases later, if this is the direction we want to go.
## Note
See previous attempt at a "simple fix" that only solved some of the issues: #6666
# Objective
> Part of #6573
When serializing a `DynamicScene` we end up treating almost all non-value types as though their type data doesn't exist. This is because when creating the `DynamicScene` we call `Reflect::clone_value` on the components, which generates a Dynamic type for all non-value types.
What this means is that the `glam` types are treated as though their `ReflectSerialize` registrations don't exist. However, the deserializer _does_ pick up the registration and attempts to use that instead. This results in the deserializer trying to operate on "malformed" data, causing this error:
```
WARN bevy_asset::asset_server: encountered an error while loading an asset: Expected float
```
## Solution
Ideally, we should better handle the serialization of possibly-Dynamic types. However, this runs into issues where the `ReflectSerialize` expects the concrete type and not a Dynamic representation, resulting in a panic:
0aa4147af6/crates/bevy_reflect/src/type_registry.rs (L402-L413)
Since glam types are so heavily used in Bevy (specifically in `Transform` and `GlobalTransform`), it makes sense to just a quick fix in that enables them to be used properly in scenes while a proper solution is found.
This PR simply removes all `ReflectSerialize` and `ReflectDeserialize` registrations from the glam types that are reflected as structs.
---
## Changelog
- Remove `ReflectSerialize` and `ReflectDeserialize` registrations from most glam types
## Migration Guide
This PR removes `ReflectSerialize` and `ReflectDeserialize` registrations from most glam types. This means any code relying on either of those type data existing for those glam types will need to not do that.
This also means that some serialized glam types will need to be updated. For example, here is `Affine3A`:
```rust
// BEFORE
(
"glam::f32::affine3a::Affine3A": (1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0),
// AFTER
"glam::f32::affine3a::Affine3A": (
matrix3: (
x_axis: (
x: 1.0,
y: 0.0,
z: 0.0,
),
y_axis: (
x: 0.0,
y: 1.0,
z: 0.0,
),
z_axis: (
x: 0.0,
y: 0.0,
z: 1.0,
),
),
translation: (
x: 0.0,
y: 0.0,
z: 0.0,
),
)
)
```
# Objective
Many types in `bevy_render` implemented `Reflect` but were not registered.
## Solution
Register all types in `bevy_render` that impl `Reflect`.
This also registers additional dependent types (i.e. field types).
> Note: Adding these dependent types would not be needed using something like #5781😉
---
## Changelog
- Register missing `bevy_render` types in the `TypeRegistry`:
- `camera::RenderTarget`
- `globals::GlobalsUniform`
- `texture::Image`
- `view::ComputedVisibility`
- `view::Visibility`
- `view::VisibleEntities`
- Register additional dependent types:
- `view::ComputedVisibilityFlags`
- `Vec<Entity>`
# Objective
- Fixes https://github.com/bevyengine/bevy/issues/6603
## Solution
- `Task`s will cancel when dropped, but wait until they return Pending before they actually get canceled. That means that if a task panics, it's possible for that error to get propagated to the scope and the scope gets dropped, while scoped tasks in other threads are still running. This is a big problem since scoped task can hold life-timed values that are dropped as the scope is dropped leading to UB.
---
## Changelog
- changed `Scope` to use `FallibleTask` and await the cancellation of all remaining tasks when it's dropped.
# Objective
- Reverts #5730.
- Fixes#6173, fixes#6596.
## Solution
Remove the warning entirely.
## Changelog
You will no longer be spammed about
> Missed 31 `bevy_input:🐭:MouseMotion` events. Consider
reading from the `EventReader` more often (generally the best
solution) or calling Events::update() less frequently
(normally this is called once per frame). This problem is most
likely due to run criteria/fixed timesteps or consuming events
conditionally. See the Events documentation for
more information.
when you miss events. These warnings were often (but not always) a false positive. You can still check this manually by using `ManualEventReader::missed_events`
# Objective
Remove an obscure and inconsistent bit of API.
Simplify the `WorldChildBuilder` code.
No idea why this even exists.
An example of the removed API:
```rust
world.spawn_empty().with_children(|parent| {
parent.spawn_empty();
parent.push_children(&[some_entity]); // Does *not* add children to the parent.
// It's actually identical to:
parent.spawn_empty().push_children(&[some_entity]);
});
world.spawn_empty().with_children(|parent| {
// This just panics.
parent.push_children(&[some_entity]);
});
```
This exists only on `WorldChildBuilder`; `ChildBuilder` does not have this API.
Yeet.
## Migration Guide
Hierarchy editing methods such as `with_children` and `push_children` have been removed from `WorldChildBuilder`.
You can edit the hierarchy via `EntityMut` instead.
Co-authored-by: devil-ira <justthecooldude@gmail.com>
# Objective
Fixes#6713
Binary deserialization is failing for unit structs as well as structs with all ignored/skipped fields.
## Solution
Add a check for the number of possible fields in a struct before deserializing. If empty, don't attempt to deserialize any fields (as there will be none).
Note: ~~This does not apply to enums as they do not properly handle skipped fields (see #6721).~~ Enums still do not properly handle skipped fields, but I decided to include the logic for it anyways to account for `#[reflect(ignore)]`'d fields in the meantime.
---
## Changelog
- Fix bug where deserializing unit structs would fail for non-self-describing formats
# Objective
Consider the test
```rust
let cell = world.cell();
let _value_a = cell.resource_mut::<A>();
let _value_b = cell.resource_mut::<A>();
```
Currently, this will roughly execute
```rust
// first call
let value = unsafe {
self.world
.get_non_send_unchecked_mut_with_id(component_id)?
};
return Some(WorldBorrowMut::new(value, archetype_component_id, self.access)))
// second call
let value = unsafe {
self.world
.get_non_send_unchecked_mut_with_id(component_id)?
};
return Some(WorldBorrowMut::new(value, archetype_component_id, self.access)))
```
where `WorldBorrowMut::new` will panic if the resource is already borrowed.
This means, that `_value_a` will be created, the access checked (OK), then `value_b` will be created, and the access checked (`panic`).
For a moment, both `_value_a` and `_value_b` existed as `&mut T` to the same location, which is insta-UB as far as I understand it.
## Solution
Flip the order so that `WorldBorrowMut::new` first checks the access, _then_ fetches creates the value. To do that, we pass a `impl FnOnce() -> Mut<T>` instead of the `Mut<T>` directly:
```rust
let get_value = || unsafe {
self.world
.get_non_send_unchecked_mut_with_id(component_id)?
};
return Some(WorldBorrowMut::new(get_value, archetype_component_id, self.access)))
```
# Objective
- shaders defs can now have a `bool` or `int` value
- `#if SHADER_DEF <operator> 3`
- ok if `SHADER_DEF` is defined, has the correct type and pass the comparison
- `==`, `!=`, `>=`, `>`, `<`, `<=` supported
- `#SHADER_DEF` or `#{SHADER_DEF}`
- will be replaced by the value in the shader code
---
## Migration Guide
- replace `shader_defs.push(String::from("NAME"));` by `shader_defs.push("NAME".into());`
- if you used shader def `NO_STORAGE_BUFFERS_SUPPORT`, check how `AVAILABLE_STORAGE_BUFFER_BINDINGS` is now used in Bevy default shaders
# Objective
`add_node_edge` and `add_slot_edge` are fallible methods, but are always used with `.unwrap()`.
`input_node` is often unwrapped as well.
This points to having an infallible behaviour as default, with an alternative fallible variant if needed.
Improves readability and ergonomics.
## Solution
- Change `add_node_edge` and `add_slot_edge` to panic on error.
- Change `input_node` to panic on `None`.
- Add `try_add_node_edge` and `try_add_slot_edge` in case fallible methods are needed.
- Add `get_input_node` to still be able to get an `Option`.
---
## Changelog
### Added
- `try_add_node_edge`
- `try_add_slot_edge`
- `get_input_node`
### Changed
- `add_node_edge` is now infallible (panics on error)
- `add_slot_edge` is now infallible (panics on error)
- `input_node` now panics on `None`
## Migration Guide
Remove `.unwrap()` from `add_node_edge` and `add_slot_edge`.
For cases where the error was handled, use `try_add_node_edge` and `try_add_slot_edge` instead.
Remove `.unwrap()` from `input_node`.
For cases where the option was handled, use `get_input_node` instead.
Co-authored-by: Torstein Grindvik <52322338+torsteingrindvik@users.noreply.github.com>
# Objective
Fixes#4697. Hierarchical propagation of properties, currently only Transform -> GlobalTransform, can be a very expensive operation. Transform propagation is a strict dependency for anything positioned in world-space. In large worlds, this can take quite a bit of time, so limiting it to a single thread can result in poor CPU utilization as it bottlenecks the rest of the frame's systems.
## Solution
- Move transforms without a parent or a child (free-floating (Global)Transform) entities into a separate parallel system.
- Chunk the hierarchy based on the root entities and process it in parallel with `Query::par_for_each_mut`.
- Utilize the hierarchy's specific properties introduced in #4717 to allow for safe use of `Query::get_unchecked` on multiple threads. Assuming each child is unique in the hierarchy, it is impossible to have an aliased `&mut GlobalTransform` so long as we verify that the parent for a child is the same one propagated from.
---
## Changelog
Removed: `transform_propagate_system` is no longer `pub`.
Without this fix, piped systems containing exclusive systems fail to run, giving a runtime panic.
With this PR, running piped systems that contain exclusive systems now works.
## Explanation of the bug
This is because, unless overridden, the default implementation of `run` from the `System` trait simply calls `run_unsafe`. That is not valid for exclusive systems. They must always be called via `run`, as `run_unsafe` takes `&World` instead of `&mut World`.
Trivial reproduction example:
```rust
fn main() {
App::new()
.add_plugins(DefaultPlugins)
.add_system(exclusive.pipe(another))
.run();
}
fn exclusive(_world: &mut World) {}
fn another() {}
```
If you run this, you will get a panic 'Cannot run exclusive systems with a shared World reference' and the backtrace shows how bevy (correctly) tries to call the `run` method (because the system is exclusive), but it is the implementation from the `System` trait (because `PipeSystem` does not have its own), which calls `run_unsafe` (incorrect):
- 3: <bevy_ecs::system::system_piping::PipeSystem<SystemA,SystemB> as bevy_ecs::system::system::System>::run_unsafe
- 4: bevy_ecs::system::system::System::run
# Objective
Allow more use cases where the user may benefit from both `ExtractComponentPlugin` _and_ `UniformComponentPlugin`.
## Solution
Add an associated type to `ExtractComponent` in order to allow specifying the output component (or bundle).
Make `extract_component` return an `Option<_>` such that components can be extracted only when needed.
What problem does this solve?
`ExtractComponentPlugin` allows extracting components, but currently the output type is the same as the input.
This means that use cases such as having a settings struct which turns into a uniform is awkward.
For example we might have:
```rust
struct MyStruct {
enabled: bool,
val: f32
}
struct MyStructUniform {
val: f32
}
```
With the new approach, we can extract `MyStruct` only when it is enabled, and turn it into its related uniform.
This chains well with `UniformComponentPlugin`.
The user may then:
```rust
app.add_plugin(ExtractComponentPlugin::<MyStruct>::default());
app.add_plugin(UniformComponentPlugin::<MyStructUniform>::default());
```
This then saves the user a fair amount of boilerplate.
## Changelog
### Changed
- `ExtractComponent` can specify output type, and outputting is optional.
Co-authored-by: Torstein Grindvik <52322338+torsteingrindvik@users.noreply.github.com>
Add a method to rotate a transform to point towards a direction.
Also updated the docs to link to `forward` and `up` instead of mentioning local negative `Z` and local `Y`.
Unfortunately, links to methods don't work in rust-analyzer :(
Co-authored-by: Devil Ira <justthecooldude@gmail.com>
# Objective
Latest Release, "bevy 0.9" move the FrameCount updater into RenderPlugin, it leads to user who only run app with Core/Minimal Plugin cannot get the right number of FrameCount, it always return 0.
As for use cases like a server app, we don't want to add render dependencies to the app.
More detail in #6656
## Solution
- Move the `update_frame_count` into CorePlugin
# Objective
This add a ctor to `Box` to aid the creation of non-centred boxes. The PR adopts @rezural's work on PR #3322, taking into account the feedback on that PR from @james7132.
## Solution
`Box::from_corners()` creates a `Box` from two opposing corners and automatically determines the min and max extents to ensure that the `Box` is well-formed.
Co-authored-by: rezural <rezural@protonmail.com>
# Objective
- Bevy should be usable to create 'overlay' type apps, where the input is not captured by Bevy, but passed down/into a target app, or to allow passive displays/widgets etc.
## Solution
- the `winit:🪟:Window` already has a `set_cursor_hittest()` which basically does this for mouse input events, so I've exposed it (trying to copy the style laid out in the existing wrappings, and added a simple demo.
---
## Changelog
- Added `hittest` to `WindowAttributes`
- Added the `hittest`'s setters/getters
- Modified the `WindowBuilder`
- Modifed the `WindowDescriptor`'s `Default` impl.
- Added an example `cargo run --example fallthrough`
# Objective
Fixes#4884. `ComponentTicks` stores both added and changed ticks contiguously in the same 8 bytes. This is convenient when passing around both together, but causes half the bytes fetched from memory for the purposes of change detection to effectively go unused. This is inefficient when most queries (no filter, mutating *something*) only write out to the changed ticks.
## Solution
Split the storage for change detection ticks into two separate `Vec`s inside `Column`. Fetch only what is needed during iteration.
This also potentially also removes one blocker from autovectorization of dense queries.
EDIT: This is confirmed to enable autovectorization of dense queries in `for_each` and `par_for_each` where possible. Unfortunately `iter` has other blockers that prevent it.
### TODO
- [x] Microbenchmark
- [x] Check if this allows query iteration to autovectorize simple loops.
- [x] Clean up all of the spurious tuples now littered throughout the API
### Open Questions
- ~~Is `Mut::is_added` absolutely necessary? Can we not just use `Added` or `ChangeTrackers`?~~ It's optimized out if unused.
- ~~Does the fetch of the added ticks get optimized out if not used?~~ Yes it is.
---
## Changelog
Added: `Tick`, a wrapper around a single change detection tick.
Added: `Column::get_added_ticks`
Added: `Column::get_column_ticks`
Added: `SparseSet::get_added_ticks`
Added: `SparseSet::get_column_ticks`
Changed: `Column` now stores added and changed ticks separately internally.
Changed: Most APIs returning `&UnsafeCell<ComponentTicks>` now returns `TickCells` instead, which contains two separate `&UnsafeCell<Tick>` for either component ticks.
Changed: `Query::for_each(_mut)`, `Query::par_for_each(_mut)` will now leverage autovectorization to speed up query iteration where possible.
## Migration Guide
TODO
# Objective
Fix#6453.
## Solution
Use the solution mentioned in the issue by catching the unwind and dropping the error. Wrap the `executor.try_tick` calls with `std::catch::unwind`.
Ideally this would be moved outside of the hot loop, but the mut ref to the `spawned` future is not `UnwindSafe`.
This PR only addresses the bug, we can address the perf issues (should there be any) later.
# Objective
Fix#5149
## Solution
Instead of returning the **total count** of elements in the `QueryIter` in
`size_hint`, we return the **count of remaining elements**. This
Fixes#5149 even when #5148 gets merged.
- https://github.com/bevyengine/bevy/issues/5149
- https://github.com/bevyengine/bevy/pull/5148
---
## Changelog
- Fix partially consumed `QueryIter` and `QueryCombinationIter` having invalid `size_hint`
Co-authored-by: Nicola Papale <nicopap@users.noreply.github.com>
# Objective
Make core pipeline graphic nodes, including `BloomNode`, `FxaaNode`, `TonemappingNode` and `UpscalingNode` public.
This will allow users to construct their own render graphs with these build-in nodes.
## Solution
Make them public.
Also put node names into bevy's core namespace (`core_2d::graph::node`, `core_3d::graph::node`) which makes them consistent.
# Objective
Fix android touch events being flipped. Only removed test for android, don't have ios device to test with. Tested with emulator and physical device.
## Solution
Remove check, no longer needed with coordinate change in 0.9
# Objective
Delete `ImageMode`. It doesn't do anything except mislead people into thinking it controls the aspect ratio of images somehow.
Fixes#3933 and #6637
## Solution
Delete `ImageMode`
## Changelog
Removes the `ImageMode` enum.
Removes the `image_mode` field from `ImageBundle`
Removes the `With<ImageMode>` query filter from `image_node_system`
Renames `image_node_system` to` update_image_calculated_size_system`
# Objective
Fixes#6661
## Solution
Make `SceneSpawner::spawn_dynamic` return `InstanceId` like other functions there.
---
## Changelog
Make `SceneSpawner::spawn_dynamic` return `InstanceId` instead of `()`.
When I was upgrading to 0.9 noticed there were some changes to the timer, mainly the `TimerMode`. When switching from the old `is_repeating()` and `set_repeating()` to the new `mode()` and `set_mode()` noticed the docs still had the old description.
# Objective
BlobVec currently relies on a scratch piece of memory allocated at initialization to make a temporary copy of a component when using `swap_remove_and_{forget/drop}`. This is potentially suboptimal as it writes to a, well-known, but random part of memory instead of using the stack.
## Solution
As the `FIXME` in the file states, replace `swap_scratch` with a call to `swap_nonoverlapping::<u8>`. The swapped last entry is returned as a `OwnedPtr`.
In theory, this should be faster as the temporary swap is allocated on the stack, `swap_nonoverlapping` allows for easier vectorization for bigger types, and the same memory is used between the swap and the returned `OwnedPtr`.
# Objective
* Enable `Res` and `Query` parameter mutual exclusion
* Required for https://github.com/bevyengine/bevy/pull/5080
The `FilteredAccessSet::get_conflicts` methods didn't work properly with
`Res` and `ResMut` parameters. Because those added their access by using
the `combined_access_mut` method and directly modifying the global
access state of the FilteredAccessSet. This caused an inconsistency,
because get_conflicts assumes that ALL added access have a corresponding
`FilteredAccess` added to the `filtered_accesses` field.
In practice, that means that SystemParam that adds their access through
the `Access` returned by `combined_access_mut` and the ones that add
their access using the `add` method lived in two different universes. As
a result, they could never be mutually exclusive.
## Solution
This commit fixes it by removing the `combined_access_mut` method. This
ensures that the `combined_access` field of FilteredAccessSet is always
updated consistently with the addition of a filter. When checking for
filtered access, it is now possible to account for `Res` and `ResMut`
invalid access. This is currently not needed, but might be in the
future.
We add the `add_unfiltered_{read,write}` methods to replace previous
usages of `combined_access_mut`.
We also add improved Debug implementations on FixedBitSet so that their
meaning is much clearer in debug output.
---
## Changelog
* Fix `Res` and `Query` parameter never being mutually exclusive.
## Migration Guide
Note: this mostly changes ECS internals, but since the API is public, it is technically breaking:
* Removed `FilteredAccessSet::combined_access_mut`
* Replace _immutable_ usage of those by `combined_access`
* For _mutable_ usages, use the new `add_unfiltered_{read,write}` methods instead of `combined_access_mut` followed by `add_{read,write}`
# Objective
Make core types in ECS smaller. The column sparse set in Tables is never updated after creation.
## Solution
Create `ImmutableSparseSet` which removes the capacity fields in the backing vec's and the APIs for inserting or removing elements. Drops the size of the sparse set by 3 usizes (24 bytes on 64-bit systems)
## Followup
~~After #4809, Archetype's component SparseSet should be replaced with it.~~ This has been done.
---
## Changelog
Removed: `Table::component_capacity`
## Migration Guide
`Table::component_capacity()` has been removed as Tables do not support adding/removing columns after construction.
Co-authored-by: Carter Anderson <mcanders1@gmail.com>
# Objective
Archetype is a deceptively large type in memory. It stores metadata about which components are in which storage in multiple locations, which is only used when creating new Archetypes while moving entities.
## Solution
Remove the redundant `Box<[ComponentId]>`s and iterate over the sparse set of component metadata instead. Reduces Archetype's size by 4 usizes (32 bytes on 64-bit systems), as well as the additional allocations for holding these slices.
It'd seem like there's a downside that the origin archetype has it's component metadata iterated over twice when creating a new archetype, but this change also removes the extra `Vec<ArchetypeComponentId>` allocations when creating a new archetype which may amortize out to a net gain here. This change likely negatively impacts creating new archetypes with a large number of components, but that's a cost mitigated by the fact that these archetypal relationships are cached in Edges and is incurred only once for each edge created.
## Additional Context
There are several other in-flight PRs that shrink Archetype:
- #4800 merges the entities and rows Vecs together (shaves off 24 bytes per archetype)
- #4809 removes unique_components and moves it to it's own dedicated storage (shaves off 72 bytes per archetype)
---
## Changelog
Changed: `Archetype::table_components` and `Archetype::sparse_set_components` return iterators instead of slices. `Archetype::new` requires iterators instead of parallel slices/vecs.
## Migration Guide
Do I still need to do this? I really hope people were not relying on the public facing APIs changed here.
# Objective
`ComputedVisibility` could afford to be smaller/faster. Optimizing the size and performance of operations on the component will positively benefit almost all extraction systems.
This was listed as one of the potential pieces of future work for #5310.
## Solution
Merge both internal booleans into a single `u8` bitflag field. Rely on bitmasks to evaluate local, hierarchical, and general visibility.
Pros:
- `ComputedVisibility::is_visible` should be a single bitmask test instead of two.
- `ComputedVisibility` is now only 1 byte. Should be able to fit 100% more per cache line when using dense iteration.
Cons:
- Harder to read.
- Setting individual values inside `ComputedVisiblity` require bitmask mutations.
This should be a non-breaking change. No public API was changed. The only publicly visible effect is that `ComputedVisibility` is now 1 byte instead of 2.
# Objective
Fixes#6594
## Solution
- `New` function for `Size` is now a `const` function :)
## Changelog
- `New` function for `Size` is now a `const` function
## Migration Guide
- Nothing has been changed
# Objective
In bevy 0.8 you could list all resources using `world.archetypes().resource().components()`. As far as I can tell the resource archetype has been replaced with the `Resources` storage, and it would be nice if it could be used to iterate over all resource component IDs as well.
## Solution
- add `fn Resources::iter(&self) -> impl Iterator<Item = (ComponentId, &ResourceData)>`
# Objective
- Derive Clone and Debug for `AssetPlugin`
- Make it possible to log asset server settings
- And get an owned instance if wrapping `AssetPlugin` in another plugin. See: 129224ef72/src/web_asset_plugin.rs (L45)
# Objective
> Part of #6573
`Children` was not being properly deserialized in scenes. This was due to a missing registration on `SmallVec<[Entity; 8]>`, which is used by `Children`.
## Solution
Register `SmallVec<[Entity; 8]>`.
---
## Changelog
- Registered `SmallVec<[Entity; 8]>`
# Objective
Fixes#6030, making ``serde`` optional.
## Solution
This was solved by making a ``serialize`` feature that can activate ``serde``, which is now optional.
When ``serialize`` is deactivated, the ``Plugin`` implementation for ``ScenePlugin`` does nothing.
Co-authored-by: Linus Käll <linus.kall.business@gmail.com>
# Objective
Currently, `Ptr` and `PtrMut` can only be constructed via unsafe code. This means that downgrading a reference to an untyped pointer is very cumbersome, despite being a very simple operation.
## Solution
Define conversions for easily and safely constructing untyped pointers. This is the non-owned counterpart to `OwningPtr::make`.
Before:
```rust
let ptr = unsafe { PtrMut::new(NonNull::from(&mut value).cast()) };
```
After:
```rust
let ptr = PtrMut::from(&mut value);
```
Co-authored-by: Carter Anderson <mcanders1@gmail.com>
# Objective
I needed a window which is always on top, to create a overlay app.
## Solution
expose the `always_on_top` property of winit in bevy's `WindowDescriptor` as a boolean flag
---
## Changelog
### Added
- add `WindowDescriptor.always_on_top` which configures a window to stay on top.
# Objective
`ScalingMode::Auto` for cameras only targets min_height and min_width, or as the docs say it `Use minimal possible viewport size while keeping the aspect ratio.`
But there is no ScalingMode that targets max_height and Max_width or `Use maximal possible viewport size while keeping the aspect ratio.`
## Solution
Added `ScalingMode::AutoMax` that does the exact opposite of `ScalingMode::Auto`
---
## Changelog
Renamed `ScalingMode::Auto` to `ScalingMode::AutoMin`.
## Migration Guide
just rename `ScalingMode::Auto` to `ScalingMode::AutoMin` if you are using it.
Co-authored-by: Lixou <82600264+DasLixou@users.noreply.github.com>
Add a method to get the focused window.
Use this instead of `WindowFocused` events in `close_on_esc`.
Seems that the OS/window manager might not always send focused events on application startup.
Sadly, not a fix for #5646.
Co-authored-by: devil-ira <justthecooldude@gmail.com>
# Objective
Fixes #3225, Allow for flippable UI Images
## Solution
Add flip_x and flip_y fields to UiImage, and swap the UV coordinates accordingly in ui_prepare_nodes.
## Changelog
* Changes UiImage to a struct with texture, flip_x, and flip_y fields.
* Adds flip_x and flip_y fields to ExtractedUiNode.
* Changes extract_uinodes to extract the flip_x and flip_y values from UiImage.
* Changes prepare_uinodes to swap the UV coordinates as required.
* Changes UiImage derefs to texture field accesses.
# Objective
Fixes#6615.
`BlobVec` does not respect alignment for zero-sized types, which results in UB whenever a ZST with alignment other than 1 is used in the world.
## Solution
Add the fn `bevy_ptr::dangling_with_align`.
---
## Changelog
+ Added the function `dangling_with_align` to `bevy_ptr`, which creates a well-aligned dangling pointer to a type whose alignment is not known at compile time.
# Objective
- Implements removal of entries from a `dyn Map`
- Fixes#6563
## Solution
- Adds a `remove` method to the `Map` trait which takes in a `&dyn Reflect` key and returns the value removed if it was present.
---
## Changelog
- Added `Map::remove`
## Migration Guide
- Implementors of `Map` will need to implement the `remove` method.
Co-authored-by: radiish <thesethskigamer@gmail.com>
# Objective
The [documentation for `ButtonSettingsError`](https://docs.rs/bevy/0.9.0/bevy/input/gamepad/enum.ButtonSettingsError.html) incorrectly describes the valid range of values as `0.0..=2.0`, probably because it was copied from `AxisSettingsError`. The actual range, as seen in the functions that return it and in its own `thiserror` description, is `0.0..=1.0`.
## Solution
Update the doc comments to reflect the correct range.
Co-authored-by: Sol Toder <ajaxgb@gmail.com>
# Objective
Fix#6548. Most of these methods were already made `const` in #5688. `Entity::to_bits` is the only one that remained.
## Solution
Make it const.
# Objective
- Fix#3606
- Fix#4579
- Fix#3380
## Solution
When running on a Linux machine with some AMD or Intel device, when calling
`surface.get_current_texture()`, ignore `wgpu::SurfaceError::Timeout` errors.
## Alternative
An alternative solution found in the `wgpu` examples is:
```rust
let frame = surface
.get_current_texture()
.or_else(|_| {
render_device.configure_surface(surface, &swap_chain_descriptor);
surface.get_current_texture()
})
.expect("Error reconfiguring surface");
window.swap_chain_texture = Some(TextureView::from(frame));
```
See: <94ce76391b/wgpu/examples/framework.rs (L362-L370)>
Veloren [handles the Timeout error the way this PR proposes to handle it](https://github.com/gfx-rs/wgpu/issues/1218#issuecomment-1092056971).
The reason I went with this PR's solution is that `configure_surface` seems to be quite an expensive operation, and it would run every frame with the wgpu framework solution, despite the fact it works perfectly fine without `configure_surface`.
I know this looks super hacky with the linux-specific line and the AMD check, but my understanding is that the `Timeout` occurrence is specific to a quirk of some AMD drivers on linux, and if otherwise met should be considered a bug.
Co-authored-by: Carter Anderson <mcanders1@gmail.com>
# Objective
- Closes#5262
- Fix color banding caused by quantization.
## Solution
- Adds dithering to the tonemapping node from #3425.
- This is inspired by Godot's default "debanding" shader: https://gist.github.com/belzecue/
- Unlike Godot:
- debanding happens after tonemapping. My understanding is that this is preferred, because we are running the debanding at the last moment before quantization (`[f32, f32, f32, f32]` -> `f32`). This ensures we aren't biasing the dithering strength by applying it in a different (linear) color space.
- This code instead uses and reference the origin source, Valve at GDC 2015
## Additional Notes
Real time rendering to standard dynamic range outputs is limited to 8 bits of depth per color channel. Internally we keep everything in full 32-bit precision (`vec4<f32>`) inside passes and 16-bit between passes until the image is ready to be displayed, at which point the GPU implicitly converts our `vec4<f32>` into a single 32bit value per pixel, with each channel (rgba) getting 8 of those 32 bits.
### The Problem
8 bits of color depth is simply not enough precision to make each step invisible - we only have 256 values per channel! Human vision can perceive steps in luma to about 14 bits of precision. When drawing a very slight gradient, the transition between steps become visible because with a gradient, neighboring pixels will all jump to the next "step" of precision at the same time.
### The Solution
One solution is to simply output in HDR - more bits of color data means the transition between bands will become smaller. However, not everyone has hardware that supports 10+ bit color depth. Additionally, 10 bit color doesn't even fully solve the issue, banding will result in coherent bands on shallow gradients, but the steps will be harder to perceive.
The solution in this PR adds noise to the signal before it is "quantized" or resampled from 32 to 8 bits. Done naively, it's easy to add unneeded noise to the image. To ensure dithering is correct and absolutely minimal, noise is adding *within* one step of the output color depth. When converting from the 32bit to 8bit signal, the value is rounded to the nearest 8 bit value (0 - 255). Banding occurs around the transition from one value to the next, let's say from 50-51. Dithering will never add more than +/-0.5 bits of noise, so the pixels near this transition might round to 50 instead of 51 but will never round more than one step. This means that the output image won't have excess variance:
- in a gradient from 49 to 51, there will be a step between each band at 49, 50, and 51.
- Done correctly, the modified image of this gradient will never have a adjacent pixels more than one step (0-255) from each other.
- I.e. when scanning across the gradient you should expect to see:
```
|-band-| |-band-| |-band-|
Baseline: 49 49 49 50 50 50 51 51 51
Dithered: 49 50 49 50 50 51 50 51 51
Dithered (wrong): 49 50 51 49 50 51 49 51 50
```
You can see from above how correct dithering "fuzzes" the transition between bands to reduce distinct steps in color, without adding excess noise.
### HDR
The previous section (and this PR) assumes the final output is to an 8-bit texture, however this is not always the case. When Bevy adds HDR support, the dithering code will need to take the per-channel depth into account instead of assuming it to be 0-255. Edit: I talked with Rob about this and it seems like the current solution is okay. We may need to revisit once we have actual HDR final image output.
---
## Changelog
### Added
- All pipelines now support deband dithering. This is enabled by default in 3D, and can be toggled in the `Tonemapping` component in camera bundles. Banding is a graphical artifact created when the rendered image is crunched from high precision (f32 per color channel) down to the final output (u8 per channel in SDR). This results in subtle gradients becoming blocky due to the reduced color precision. Deband dithering applies a small amount of noise to the signal before it is "crunched", which breaks up the hard edges of blocks (bands) of color. Note that this does not add excess noise to the image, as the amount of noise is less than a single step of a color channel - just enough to break up the transition between color blocks in a gradient.
Co-authored-by: Carter Anderson <mcanders1@gmail.com>
# Objective
- Make the many foxes not unnecessarily bright. Broken since #5666.
- Fixes#6528
## Solution
- In #5666 normalisation of normals was moved from the fragment stage to the vertex stage. However, it was not added to the vertex stage for skinned normals. The many foxes are skinned and their skinned normals were not unit normals. which made them brighter. Normalising the skinned normals fixes this.
---
## Changelog
- Fixed: Non-unit length skinned normals are now normalized.
This reverts commit 8429b6d6ca as discussed in #6522.
I tested that the game_menu example works as it should.
Co-authored-by: devil-ira <justthecooldude@gmail.com>
# Objective
Attempting to build `bevy_tasks` produces the following error:
```
error[E0599]: no method named `is_finished` found for struct `async_executor::Task` in the current scope
--> /[...]]/bevy/crates/bevy_tasks/src/task.rs:51:16
|
51 | self.0.is_finished()
| ^^^^^^^^^^^ method not found in `async_executor::Task<T>`
```
It looks like this was introduced along with `Task::is_finished`, which delegates to `async_task::Task::is_finished`. However, the latter was only introduced in `async-task` 4.2.0; `bevy_tasks` does not explicitly depend on `async-task` but on `async-executor` ^1.3.0, which in turn depends on `async-task` ^4.0.0.
## Solution
Add an explicit dependency on `async-task` ^4.2.0.
# Objective
Copy `send_event` and friends from `World` to `WorldCell`.
Clean up `bevy_winit` using `WorldCell::send_event`.
## Changelog
Added `send_event`, `send_event_default`, and `send_event_batch` to `WorldCell`.
Co-authored-by: devil-ira <justthecooldude@gmail.com>
# Objective
- it would be useful to inspect these structs using reflection
## Solution
- derive and register reflect
- Note that `#[reflect(Component)]` requires `Default` (or `FromWorld`) until #6060, so I implemented `Default` for `Tonemapping` with `is_enabled: false`
# Objective
The `load_internal_asset` macro is helpful when creating rendering plugins, but it doesn't support load binary assets (like those compiled as spir-v).
## Solution
Add a `load_internal_binary_asset` macro that use `include_bytes!`.
# Objective
Fixes#5393
## Solution
- Add padding to `GlobalsUniform` / `Globals` to make it 16-byte aligned.
Still not super clear on whether this is a `naga` thing or an `encase` thing or what. But now that we're offering `globals` up to users and #5393 is not just breaking an example, maybe we should do this sort of workaround?
* Move the despawn debug log from `World::despawn` to `EntityMut::despawn`.
* Move the despawn non-existent warning log from `Commands::despawn` to `World::despawn`.
This should make logging consistent regardless of which of the three `despawn` methods is used.
Co-authored-by: devil-ira <justthecooldude@gmail.com>
# Objective
Using `Reflect` we can easily switch between a specific reflection trait object, such as a `dyn Struct`, to a `dyn Reflect` object via `Reflect::as_reflect` or `Reflect::as_reflect_mut`.
```rust
fn do_something(value: &dyn Reflect) {/* ... */}
let foo: Box<dyn Struct> = Box::new(Foo::default());
do_something(foo.as_reflect());
```
However, there is no way to convert a _boxed_ reflection trait object to a `Box<dyn Reflect>`.
## Solution
Add a `Reflect::into_reflect` method which allows converting a boxed reflection trait object back into a boxed `Reflect` trait object.
```rust
fn do_something(value: Box<dyn Reflect>) {/* ... */}
let foo: Box<dyn Struct> = Box::new(Foo::default());
do_something(foo.into_reflect());
```
---
## Changelog
- Added `Reflect::into_reflect`
# Objective
Some render plugins, like [bevy-hikari](https://github.com/cryscan/bevy-hikari) require to set `CameraRenderGraph`. In order to switch between render graphs I need to insert a new `CameraRenderGraph` component. It's not very ergonomic.
## Solution
Add `CameraRenderGraph::set` like in [Name](https://docs.rs/bevy/latest/bevy/core/struct.Name.html).
---
## Changelog
### Added
- `CameraRenderGraph::set`.
# Objective
There is no way to gen an owned value of `Reflect`.
## Solution
Add it! This was originally a part of #6421, but @MrGVSV asked me to create a separate for it to implement reflect diffing.
---
## Changelog
### Added
- `Reflect::reflect_owned` to get an owned version of `Reflect`.
# Objective
`NodeBundle` contains an `image` field, which can be misleading, because if you do supply an image there, nothing will be shown to screen. You need to use an `ImageBundle` instead.
## Solution
* `image` (`UiImage`) field is removed from `NodeBundle`,
* extraction stage queries now make an optional query for `UiImage`, if one is not found, use the image handle that is used as a default by `UiImage`: c019a60b39/crates/bevy_ui/src/ui_node.rs (L464)
* touching up docs for `NodeBundle` to help guide what `NodeBundle` should be used for
* renamed `entity.rs` to `node_bundle.rs` as that gives more of a hint regarding the module's purpose
* separating `camera_config` stuff from the pre-made UI node bundles so that `node_bundle.rs` makes more sense as a module name.
# Objective
- adding a new `.register` should not overwrite old type data
- separate crates should both be able to register the same type
I ran into this while debugging why `register::<Handle<T>>` removed the `ReflectHandle` type data from a prior `register_asset_reflect`.
## Solution
- make `register` do nothing if called again for the same type
- I also removed some unnecessary duplicate registrations
# Objective
When an error causes `debug_checked_unreachable` to be called, the panic message unhelpfully points to the function definition instead of the place that caused the error.
## Solution
Add the `#[track_caller]` attribute in debug mode.
# Objective
Fixes: https://github.com/bevyengine/bevy/issues/6466
Summary: The UI Scaling example dynamically scales the UI which will dynamically allocate fonts to the font atlas surpassing the protective limit, throwing a panic.
## Solution
- Set TextSettings.allow_dynamic_font_size = true for the UI Scaling example. This is the ideal solution since the dynamic changes to the UI are not continuous yet still discrete.
- Update the panic text to reflect ui scaling as a potential cause
# Objective
Bevy UI (and third party plugins) currently have no good way to position themselves after all post processing effects. They currently use the tonemapping node, but this is not adequate if there is anything after tonemapping (such as FXAA).
## Solution
Add a logical `END_MAIN_PASS_POST_PROCESSING` RenderGraph node that main pass post processing effects position themselves before, and things like UIs can position themselves after.
# Objective
This PR fixes#5789, by enabling movable (and scalable) directional light shadow volumes.
## Solution
This PR changes `ExtractedDirectionalLight` to hold a copy of the `DirectionalLight` entity's `GlobalTransform`, instead of just a `direction` vector. This allows the shadow map volume (as defined by the light's `shadow_projection` field) to be transformed honoring translation _and_ scale transforms, and not just rotation.
It also augments the texel size calculation (used to determine the `shadow_normal_bias`) so that it now takes into account the upper bound of the x/y/z scale of the `GlobalTransform`.
This change makes the directional light extraction code more consistent with point and spot lights (that already use `transform`), and allows easily moving and scaling the shadow volume along with a player entity based on camera distance/angle, immediately enabling more real world use cases until we have a more sophisticated adaptive implementation, such as the one described in #3629.
**Note:** While it was previously possible to update the projection achieving a similar effect, depending on the light direction and distance to the origin, the fact that the shadow map camera was always positioned at the origin with a hardcoded `Vec3::Y` up value meant you would get sub-optimal or inconsistent/incorrect results.
---
## Changelog
### Changed
- `DirectionalLight` shadow volumes now honor translation and scale transforms
## Migration Guide
- If your directional lights were positioned at the origin and not scaled (the default, most common scenario) no changes are needed on your part; it just works as before;
- If you previously had a system for dynamically updating directional light shadow projections, you might now be able to simplify your code by updating the directional light entity's transform instead;
- In the unlikely scenario that a scene with directional lights that previously rendered shadows correctly has missing shadows, make sure your directional lights are positioned at (0, 0, 0) and are not scaled to a size that's too large or too small.
Co-authored-by: JMS55 <47158642+JMS55@users.noreply.github.com>
Co-authored-by: Carter Anderson <mcanders1@gmail.com>
Co-authored-by: DGriffin91 <github@dgdigital.net>
`EntityMut::remove_children` does not call `self.update_location()` which is unsound.
Verified by adding the following assertion, which fails when running the tests.
```rust
let before = self.location();
self.update_location();
assert_eq!(before, self.location());
```
I also removed incorrect messages like "parent entity is not modified" and the unhelpful "Inserting a bundle in the children entities may change the parent entity's location if they were of the same archetype" which might lead people to think that's the *only* thing that can change the entity's location.
# Changelog
Added `EntityMut::world_scope`.
Co-authored-by: devil-ira <justthecooldude@gmail.com>
Allow passing `Vec`s of glam vector types as vertex attributes.
Alternative to #4548 and #2719
Also used some macros to cut down on all the repetition.
# Migration Guide
Implementations of `From<Vec<[u16; 4]>>` and `From<Vec<[u8; 4]>>` for `VertexAttributeValues` have been removed.
I you're passing either `Vec<[u16; 4]>` or `Vec<[u8; 4]>` into `Mesh::insert_attribute` it will now require wrapping it with right the `VertexAttributeValues` enum variant.
Co-authored-by: devil-ira <justthecooldude@gmail.com>
# Objective
Closes#5934
Currently it is not possible to de/serialize data to non-self-describing formats using reflection.
## Solution
Add support for non-self-describing de/serialization using reflection.
This allows us to use binary formatters, like [`postcard`](https://crates.io/crates/postcard):
```rust
#[derive(Reflect, FromReflect, Debug, PartialEq)]
struct Foo {
data: String
}
let mut registry = TypeRegistry::new();
registry.register::<Foo>();
let input = Foo {
data: "Hello world!".to_string()
};
// === Serialize! === //
let serializer = ReflectSerializer::new(&input, ®istry);
let bytes: Vec<u8> = postcard::to_allocvec(&serializer).unwrap();
println!("{:?}", bytes); // Output: [129, 217, 61, 98, ...]
// === Deserialize! === //
let deserializer = UntypedReflectDeserializer::new(®istry);
let dynamic_output = deserializer
.deserialize(&mut postcard::Deserializer::from_bytes(&bytes))
.unwrap();
let output = <Foo as FromReflect>::from_reflect(dynamic_output.as_ref()).unwrap();
assert_eq!(expected, output); // OK!
```
#### Crates Tested
- ~~[`rmp-serde`](https://crates.io/crates/rmp-serde)~~ Apparently, this _is_ self-describing
- ~~[`bincode` v2.0.0-rc.1](https://crates.io/crates/bincode/2.0.0-rc.1) (using [this PR](https://github.com/bincode-org/bincode/pull/586))~~ This actually works for the latest release (v1.3.3) of [`bincode`](https://crates.io/crates/bincode) as well. You just need to be sure to use fixed-int encoding.
- [`postcard`](https://crates.io/crates/postcard)
## Future Work
Ideally, we would refactor the `serde` module, but I don't think I'll do that in this PR so as to keep the diff relatively small (and to avoid any painful rebases). This should probably be done once this is merged, though.
Some areas we could improve with a refactor:
* Split deserialization logic across multiple files
* Consolidate helper functions/structs
* Make the logic more DRY
---
## Changelog
- Add support for non-self-describing de/serialization using reflection.
Co-authored-by: Gino Valente <49806985+MrGVSV@users.noreply.github.com>
# Objective
- Adds a bloom pass for HDR-enabled Camera3ds.
- Supersedes (and all credit due to!) https://github.com/bevyengine/bevy/pull/3430 and https://github.com/bevyengine/bevy/pull/2876
## Solution
- A threshold is applied to isolate emissive samples, and then a series of downscale and upscaling passes are applied and composited together.
- Bloom is applied to 2d or 3d Cameras with hdr: true and a BloomSettings component.
---
## Changelog
- Added a `core_pipeline::bloom::BloomSettings` component.
- Added `BloomNode` that runs between the main pass and tonemapping.
- Added a `BloomPlugin` that is loaded as part of CorePipelinePlugin.
- Added a bloom example project.
Co-authored-by: JMS55 <47158642+JMS55@users.noreply.github.com>
Co-authored-by: Carter Anderson <mcanders1@gmail.com>
Co-authored-by: DGriffin91 <github@dgdigital.net>
# Objective
Alternative to #6424Fixes#6226
Fixes spawning empty bundles
## Solution
Add `BundleComponentStatus` trait and implement it for `AddBundle` and a new `SpawnBundleStatus` type (which always returns an Added status). `write_components` is now generic on `BundleComponentStatus` instead of taking `AddBundle` directly. This means BundleSpawner can now avoid needing AddBundle from the Empty archetype, which means BundleSpawner no longer needs a reference to the original archetype.
In theory this cuts down on the work done in `write_components` when spawning, but I'm seeing no change in the spawn benchmarks.
# Objective
The UI pass in HDR breaks currently because the color attachment format does not match the HDR ViewTarget.
## Solution
Specialize the UI pipeline on "hdr-ness" and select the appropriate format (like we do in the other built in pipelines).
# Objective
- Fixes#4019
- Fix lighting of double-sided materials when using a negative scale
- The FlightHelmet.gltf model's hose uses a double-sided material. Loading the model with a uniform scale of -1.0, and comparing against Blender, it was identified that negating the world-space tangent, bitangent, and interpolated normal produces incorrect lighting. Discussion with Morten Mikkelsen clarified that this is both incorrect and unnecessary.
## Solution
- Remove the code that negates the T, B, and N vectors (the interpolated world-space tangent, calculated world-space bitangent, and interpolated world-space normal) when seeing the back face of a double-sided material with negative scale.
- Negate the world normal for a double-sided back face only when not using normal mapping
### Before, on `main`, flipping T, B, and N
<img width="932" alt="Screenshot 2022-08-22 at 15 11 53" src="https://user-images.githubusercontent.com/302146/185965366-f776ff2c-cfa1-46d1-9c84-fdcb399c273c.png">
### After, on this PR
<img width="932" alt="Screenshot 2022-08-22 at 15 12 11" src="https://user-images.githubusercontent.com/302146/185965420-8be493e2-3b1a-4188-bd13-fd6b17a76fe7.png">
### Double-sided material without normal maps
https://user-images.githubusercontent.com/302146/185988113-44a384e7-0b55-4946-9b99-20f8c803ab7e.mp4
---
## Changelog
- Fixed: Lighting of normal-mapped, double-sided materials applied to models with negative scale
- Fixed: Lighting and shadowing of back faces with no normal-mapping and a double-sided material
## Migration Guide
`prepare_normal` from the `bevy_pbr::pbr_functions` shader import has been reworked.
Before:
```rust
pbr_input.world_normal = in.world_normal;
pbr_input.N = prepare_normal(
pbr_input.material.flags,
in.world_normal,
#ifdef VERTEX_TANGENTS
#ifdef STANDARDMATERIAL_NORMAL_MAP
in.world_tangent,
#endif
#endif
in.uv,
in.is_front,
);
```
After:
```rust
pbr_input.world_normal = prepare_world_normal(
in.world_normal,
(material.flags & STANDARD_MATERIAL_FLAGS_DOUBLE_SIDED_BIT) != 0u,
in.is_front,
);
pbr_input.N = apply_normal_mapping(
pbr_input.material.flags,
pbr_input.world_normal,
#ifdef VERTEX_TANGENTS
#ifdef STANDARDMATERIAL_NORMAL_MAP
in.world_tangent,
#endif
#endif
in.uv,
);
```
# Objective
Replace `WorldQueryGats` trait with actual gats
## Solution
Replace `WorldQueryGats` trait with actual gats
---
## Changelog
- Replaced `WorldQueryGats` trait with actual gats
## Migration Guide
- Replace usage of `WorldQueryGats` assoc types with the actual gats on `WorldQuery` trait
Respect mipmap_filter when create ImageDescriptor with linear()/nearest()
# Objective
Fixes#6348
## Migration Guide
This PR changes default `ImageSettings` and may lead to unexpected behaviour for existing projects with mipmapped textures. Users should provide custom `ImageSettings` resource with `mipmap_filter=FilterMode::Nearest` if they want to keep old behaviour.
Co-authored-by: Yakov Borevich <j.borevich@gmail.com>
# Objective
Currently we are limiting the amount of direction lights in a scene to one.
## Solution
Increase the amount of direction lights from 1 to 10.
This still is not a perfect solution, but should unblock many use cases.
We could probably just store the directional lights similar to the point lights in an storage buffer, allowing for an variable amount of directional lights.
Co-authored-by: Kurt Kühnert <51823519+Ku95@users.noreply.github.com>
# Objective
Right now, the `TaskPool` implementation allows panics to permanently kill worker threads upon panicking. This is currently non-recoverable without using a `std::panic::catch_unwind` in every scheduled task. This is poor ergonomics and even poorer developer experience. This is exacerbated by #2250 as these threads are global and cannot be replaced after initialization.
Removes the need for temporary fixes like #4998. Fixes#4996. Fixes#6081. Fixes#5285. Fixes#5054. Supersedes #2307.
## Solution
The current solution is to wrap `Executor::run` in `TaskPool` with a `catch_unwind`, and discarding the potential panic. This was taken straight from [smol](404c7bcc0a/src/spawn.rs (L44))'s current implementation. ~~However, this is not entirely ideal as:~~
- ~~the signaled to the awaiting task. We would need to change `Task<T>` to use `async_task::FallibleTask` internally, and even then it doesn't signal *why* it panicked, just that it did.~~ (See below).
- ~~no error is logged of any kind~~ (See below)
- ~~it's unclear if it drops other tasks in the executor~~ (it does not)
- ~~This allows the ECS parallel executor to keep chugging even though a system's task has been dropped. This inevitably leads to deadlock in the executor.~~ Assuming we don't catch the unwind in ParallelExecutor, this will naturally kill the main thread.
### Alternatives
A final solution likely will incorporate elements of any or all of the following.
#### ~~Log and Ignore~~
~~Log the panic, drop the task, keep chugging. This only addresses the discoverability of the panic. The process will continue to run, probably deadlocking the executor. tokio's detatched tasks operate in this fashion.~~
Panics already do this by default, even when caught by `catch_unwind`.
#### ~~`catch_unwind` in `ParallelExecutor`~~
~~Add another layer catching system-level panics into the `ParallelExecutor`. How the executor continues when a core dependency of many systems fails to run is up for debate.~~
`async_task::Task` bubbles up panics already, this will transitively push panics all the way to the main thread.
#### ~~Emulate/Copy `tokio::JoinHandle` with `Task<T>`~~
~~`tokio::JoinHandle<T>` bubbles up the panic from the underlying task when awaited. This can be transitively applied across other APIs that also use `Task<T>` like `Query::par_for_each` and `TaskPool::scope`, bubbling up the panic until it's either caught or it reaches the main thread.~~
`async_task::Task` bubbles up panics already, this will transitively push panics all the way to the main thread.
#### Abort on Panic
The nuclear option. Log the error, abort the entire process on any thread in the task pool panicking. Definitely avoids any additional infrastructure for passing the panic around, and might actually lead to more efficient code as any unwinding is optimized out. However gives the developer zero options for dealing with the issue, a seemingly poor choice for debuggability, and prevents graceful shutdown of the process. Potentially an option for handling very low-level task management (a la #4740). Roughly takes the shape of:
```rust
struct AbortOnPanic;
impl Drop for AbortOnPanic {
fn drop(&mut self) {
abort!();
}
}
let guard = AbortOnPanic;
// Run task
std::mem::forget(AbortOnPanic);
```
---
## Changelog
Changed: `bevy_tasks::TaskPool`'s threads will no longer terminate permanently when a task scheduled onto them panics.
Changed: `bevy_tasks::Task` and`bevy_tasks::Scope` will propagate panics in the spawned tasks/scopes to the parent thread.
# Objective
Add consistent UI rendering and interaction where deep nodes inside two different hierarchies will never render on top of one-another by default and offer an escape hatch (z-index) for nodes to change their depth.
## The problem with current implementation
The current implementation of UI rendering is broken in that regard, mainly because [it sets the Z value of the `Transform` component based on a "global Z" space](https://github.com/bevyengine/bevy/blob/main/crates/bevy_ui/src/update.rs#L43) shared by all nodes in the UI. This doesn't account for the fact that each node's final `GlobalTransform` value will be relative to its parent. This effectively makes the depth unpredictable when two deep trees are rendered on top of one-another.
At the moment, it's also up to each part of the UI code to sort all of the UI nodes. The solution that's offered here does the full sorting of UI node entities once and offers the result through a resource so that all systems can use it.
## Solution
### New ZIndex component
This adds a new optional `ZIndex` enum component for nodes which offers two mechanism:
- `ZIndex::Local(i32)`: Overrides the depth of the node relative to its siblings.
- `ZIndex::Global(i32)`: Overrides the depth of the node relative to the UI root. This basically allows any node in the tree to "escape" the parent and be ordered relative to the entire UI.
Note that in the current implementation, omitting `ZIndex` on a node has the same result as adding `ZIndex::Local(0)`. Additionally, the "global" stacking context is essentially a way to add your node to the root stacking context, so using `ZIndex::Local(n)` on a root node (one without parent) will share that space with all nodes using `Index::Global(n)`.
### New UiStack resource
This adds a new `UiStack` resource which is calculated from both hierarchy and `ZIndex` during UI update and contains a vector of all node entities in the UI, ordered by depth (from farthest from camera to closest). This is exposed publicly by the bevy_ui crate with the hope that it can be used for consistent ordering and to reduce the amount of sorting that needs to be done by UI systems (i.e. instead of sorting everything by `global_transform.z` in every system, this array can be iterated over).
### New z_index example
This also adds a new z_index example that showcases the new `ZIndex` component. It's also a good general demo of the new UI stack system, because making this kind of UI was very broken with the old system (e.g. nodes would render on top of each other, not respecting hierarchy or insert order at all).
---
## Changelog
- Added the `ZIndex` component to bevy_ui.
- Added the `UiStack` resource to bevy_ui, and added implementation in a new `stack.rs` module.
- Removed the previous Z updating system from bevy_ui, because it was replaced with the above.
- Changed bevy_ui rendering to use UiStack instead of z ordering.
- Changed bevy_ui focus/interaction system to use UiStack instead of z ordering.
- Added a new z_index example.
## ZIndex demo
Here's a demo I wrote to test these features
https://user-images.githubusercontent.com/1060971/188329295-d7beebd6-9aee-43ab-821e-d437df5dbe8a.mp4
Co-authored-by: Carter Anderson <mcanders1@gmail.com>
This reverts commit 53d387f340.
# Objective
Reverts #6448. This didn't have the intended effect: we're now getting bevy::prelude shown in the docs again.
Co-authored-by: Alejandro Pascual <alejandro.pascual.pozo@gmail.com>
# Objective
- Right now re-exports are completely hidden in prelude docs.
- Fixes#6433
## Solution
- We could show the re-exports without inlining their documentation.
# Objective
Fixes#6059, changing all incorrect occurrences of ``id`` in the ``entity`` module to ``index``:
* struct level documentation,
* ``id`` struct field,
* ``id`` method and its documentation.
## Solution
Renaming and verifying using CI.
Co-authored-by: Edvin Kjell <43633999+Edwox@users.noreply.github.com>
# Objective
In some scenarios it can be useful to check if a task has been finished without polling it. I added a function called `is_finished` to check if a task has been finished.
## Solution
Since `async_task` supports it out of the box, it is just a simple wrapper function.
---
# Objective
- Add post processing passes for FXAA (Fast Approximate Anti-Aliasing)
- Add example comparing MSAA and FXAA
## Solution
When the FXAA plugin is added, passes for FXAA are inserted between the main pass and the tonemapping pass. Supports using either HDR or LDR output from the main pass.
---
## Changelog
- Add a new FXAANode that runs after the main pass when the FXAA plugin is added.
Co-authored-by: Carter Anderson <mcanders1@gmail.com>
For `derive(WorldQuery)`, there are three structs generated, `Item`, `Fetch` and `State`.
These inherit the visibility of the derived structure, thus `#![warn(missing_docs)]` would
warn about missing documentation for these structures.
- [ ] I'd like some advice on what to write here, as I personally don't really understand `Fetch` nor `State`.
# Objective
Currently, `bevy_dynamic_plugin` simply panics on error. This makes it impossible to handle failures in applications that use this feature.
For example, I'd like to build an optional expansion for my game, that may not be distributed to all users. I want to use `bevy_dynamic_plugin` for loading it. I want my game to try to load it on startup, but continue without it if it cannot be loaded.
## Solution
- Make the `dynamically_load_plugin` function return a `Result`, so it can gracefully return loading errors.
- Create an error enum type, to provide useful information about the kind of error. This adds `thiserror` to the dependencies of `bevy_dynamic_plugin`, but that dependency is already used in other parts of bevy (such as `bevy_asset`), so not a big deal.
I chose not to change the behavior of the builder method in the App extension trait. I kept it as panicking. There is no clean way (that I'm aware of) to make a builder-style API that has fallible methods. So it is either a panic or a warning. I feel the panic is more appropriate.
---
## Changelog
### Changed
- `bevy_dynamic_plugin::dynamically_load_plugin` now returns `Result` instead of panicking, to allow for error handling
# Objective
* Add benchmarks for `Query::get_many`.
* Speed up `Query::get_many`.
## Solution
Previously, `get_many` and `get_many_mut` used the method `array::map`, which tends to optimize very poorly. This PR replaces uses of that method with loops.
## Benchmarks
| Benchmark name | Execution time | Change from this PR |
|--------------------------------------|----------------|---------------------|
| query_get_many_2/50000_calls_table | 1.3732 ms | -24.967% |
| query_get_many_2/50000_calls_sparse | 1.3826 ms | -24.572% |
| query_get_many_5/50000_calls_table | 2.6833 ms | -30.681% |
| query_get_many_5/50000_calls_sparse | 2.9936 ms | -30.672% |
| query_get_many_10/50000_calls_table | 5.7771 ms | -36.950% |
| query_get_many_10/50000_calls_sparse | 7.4345 ms | -36.987% |
# Objective
Add documentation `#[world_query(ignore)]`. Fixes#6283.
---
I've only described it's behavior so far (which appears to be the same as with `system_param`). Is there another use-case for this besides with `PhantomData`? I could only find a single usage of this construct on GitHub, which is [here](ffcb816927/bevy/examples/ecs/custom_query_param.rs (L102)).
I was also wondering if it would make sense to add a usage example to the `custom_query_example`? 🤔 That's why it's currently still in there.
Co-authored-by: Lucas Jenß <243719+x3ro@users.noreply.github.com>
# Objective
`bevy_core` is missing a feature corresponding to the `serialize` feature on the `bevy` crate. Similar to #6378 and https://github.com/bevyengine/bevy/pull/6379 to serialize `Name` easily.
## Solution
Add this feature and hand-written serialization for `Name` (to avoid storing `hash` field).
---
## Changelog
### Added
* `Serialize` and `Deserialize` derives for `Name` under `serialize` feature.
# Objective
Post processing effects cannot read and write to the same texture. Currently they must own their own intermediate texture and redundantly copy from that back to the main texture. This is very inefficient.
Additionally, working with ViewTarget is more complicated than it needs to be, especially when working with HDR textures.
## Solution
`ViewTarget` now stores two copies of the "main texture". It uses an atomic value to track which is currently the "main texture" (this interior mutability is necessary to accommodate read-only RenderGraph execution).
`ViewTarget` now has a `post_process_write` method, which will return a source and destination texture. Each call to this method will flip between the two copies of the "main texture".
```rust
let post_process = render_target.post_process_write();
let source_texture = post_process.source;
let destination_texture = post_process.destination;
```
The caller _must_ read from the source texture and write to the destination texture, as it is assumed that the destination texture will become the new "main texture".
For simplicity / understandability `ViewTarget` is now a flat type. "hdr-ness" is a property of the `TextureFormat`. The internals are fully private in the interest of providing simple / consistent apis. Developers can now easily access the main texture by calling `view_target.main_texture()`.
HDR ViewTargets no longer have an "ldr texture" with `TextureFormat::bevy_default`. They _only_ have their two "hdr" textures. This simplifies the mental model. All we have is the "currently active hdr texture" and the "other hdr texture", which we flip between for post processing effects.
The tonemapping node has been rephrased to use this "post processing pattern". The blit pass has been removed, and it now only runs a pass when HDR is enabled. Notably, both the input and output texture are assumed to be HDR. This means that tonemapping behaves just like any other "post processing effect". It could theoretically be moved anywhere in the "effect chain" and continue to work.
In general, I think these changes will make the lives of people making post processing effects much easier. And they better position us to start building higher level / more structured "post processing effect stacks".
---
## Changelog
- `ViewTarget` now stores two copies of the "main texture". Calling `ViewTarget::post_process_write` will flip between copies of the main texture.
# Objective
- `ReflectDefault` can be used to create default values for reflected types
- `std` primitives that are `Default`-constructable should register `ReflectDefault`
## Solution
- register `ReflectDefault`
# Objective
- Fixes #6311
- Make it clearer what should be done in the example (close the Bevy app window)
## Solution
- Remove the second windowed Bevy App [since winit does not support this](https://github.com/rust-windowing/winit/blob/v0.27.4/src/event_loop.rs#L82-L83)
- Add title to the Bevy window asking the user to close it
This is more of a quick fix to have a working example. It would be nicer if we had a small real usecase for this functionality.
Another alternativ that I tried out: If we want to showcase a second Bevy app as it was before, we could still do this as long as one of them does not have a window. But I don't see how this is helpful in the context of the example, so I stuck with only one Bevy app and a simple print afterwards.
# Objective
Entities are unique, however, this is not reflected in the scene format. Currently, entities are stored in a list where a user could inadvertently create a duplicate of the same entity.
## Solution
Switch from the list representation to a map representation for entities.
---
## Changelog
* The `entities` field in the scene format is now a map of entity ID to entity data
## Migration Guide
The scene format now stores its collection of entities in a map rather than a list:
```rust
// OLD
(
entities: [
(
entity: 12,
components: {
"bevy_transform::components::transform::Transform": (
translation: (
x: 0.0,
y: 0.0,
z: 0.0
),
rotation: (0.0, 0.0, 0.0, 1.0),
scale: (
x: 1.0,
y: 1.0,
z: 1.0
),
),
},
),
],
)
// NEW
(
entities: {
12: (
components: {
"bevy_transform::components::transform::Transform": (
translation: (
x: 0.0,
y: 0.0,
z: 0.0
),
rotation: (0.0, 0.0, 0.0, 1.0),
scale: (
x: 1.0,
y: 1.0,
z: 1.0
),
),
},
),
},
)
```
# Objective
- Make it impossible to add a plugin twice
- This is going to be more a risk for plugins with configurations, to avoid things like `App::new().add_plugins(DefaultPlugins).add_plugin(ImagePlugin::default_nearest())`
## Solution
- Panic when a plugin is added twice
- It's still possible to mark a plugin as not unique by overriding `is_unique`
- ~~Simpler version of~~ #3988 (not simpler anymore because of how `PluginGroupBuilder` implements `PluginGroup`)
# Objective
- Bevy main crashs on Safari mobile
- On Safari mobile, calling winit_window.set_cursor_grab(true) fails as the API is not implemented (as there is no cursor on Safari mobile, the api doesn't make sense there). I don't know about other mobile browsers
## Solution
- Do not call the api to release cursor grab on window creation, as the cursor is not grabbed anyway at this point
- This is #3617 which was lost in #6218
# Objective
Fixes#6378
`bevy_transform` is missing a feature corresponding to the `serialize` feature on the `bevy` crate.
## Solution
Adds a `serialize` feature to `bevy_transform`.
Derives `serde::Serialize` and `Deserialize` when feature is enabled.
# Objective
Currently toggling an `AudioSink` (for example from a game menu) requires writing
```rs
if sink.is_paused() {
sink.play();
} else {
sink.pause();
}
```
It would be nicer if we could reduce this down to a single line
```rs
sink.toggle();
```
## Solution
Add an `AudioSink::toggle` method which does exactly that.
---
## Changelog
- Added `AudioSink::toggle` which can be used to toggle state of a sink.
# Objective
Add methods to `Query<&Children>` and `Query<&Parent>` to iterate over descendants and ancestors, respectively.
## Changelog
* Added extension trait for `Query` in `bevy_hierarchy`, `HierarchyQueryExt`
* Added method `iter_descendants` to `Query<&Children>` via `HierarchyQueryExt` for iterating over the descendants of an entity.
* Added method `iter_ancestors` to `Query<&Parent>` via `HierarchyQueryExt` for iterating over the ancestors of an entity.
Co-authored-by: devil-ira <justthecooldude@gmail.com>
- Freeing unused memory held by visible entities
- Fixed comment style
# Objective
With Rust 1.56 it's possible to shrink vectors to a specified capacity. Visibility system had a comment before asking for that feature to free unused memory by a vector if its capacity is two times larger than the length.
## Solution
Shrinking the vector of visible entities to the nearest power of 2 elements next to `len()`, if capacity exceeds it more than two times.
# Objective
- Time have `Reflect`, but doesn't have `FromReflect`.
## Solution
- Add it for `Timer`, `Stopwatch` and `TimerMode`.
---
## Changelog
### Added
* `FromReflect` derive for `Timer`, `Stopwatch` and `TimerMode`.
# Objective
Bevy still has many instances of using single-tuples `(T,)` to create a bundle. Due to #2975, this is no longer necessary.
## Solution
Search for regex `\(.+\s*,\)`. This should have found every instance.
# Objective
Fix the soundness issue outlined in #5866. In short the problem is that `query.to_readonly().get_component_mut::<T>()` can provide unsound mutable access to the component. This PR is an alternative to just removing the offending api. Given that `to_readonly` is a useful tool, I think this approach is a preferable short term solution. Long term I think theres a better solution out there, but we can find that on its own time.
## Solution
Add what amounts to a "dirty flag" that marks Queries that have been converted to their read-only variant via `to_readonly` as dirty. When this flag is set to true, `get_component_mut` will fail with an error, preventing the unsound access.
# Objective
Following discussion on #3536 and #3522, `Handle::as_weak()` takes a type `U`, reinterpreting the handle as of another asset type while keeping the same ID. This is mainly used today in font atlas code. This PR does two things:
- Rename the method to `cast_weak()` to make its intent more clear
- Actually change the type uuid in the handle if it's not an asset path variant.
## Migration Guide
- Rename `Handle::as_weak` uses to `Handle::cast_weak`
The method now properly sets the associated type uuid if the handle is a direct reference (e.g. not a reference to an `AssetPath`), so adjust you code accordingly if you relied on the previous behavior.
# Objective
Currently for entities we serialize only `id`. But this is not very expected behavior. For example, in networking, when the server sends its state, it contains entities and components. On the client, I create new objects and map them (using `EntityMap`) to those received from the server (to know which one matches which). And if `generation` field is missing, this mapping can be broken. Example:
1. Server sends an entity `Entity{ id: 2, generation: 1}` with components.
2. Client puts the received entity in a map and create a new entity that maps to this received entity. The new entity have different `id` and `generation`. Let's call it `Entity{ id: 12, generation: 4}`.
3. Client sends a command for `Entity{ id: 12, generation: 4}`. To do so, it maps local entity to the one from server. But `generation` field is 0 because it was omitted for serialization on the server. So it maps to `Entity{ id: 2, generation: 0}`.
4. Server receives `Entity{ id: 2, generation: 0}` which is invalid.
In my game I worked around it by [writing custom serialization](https://github.com/dollisgame/dollis/blob/master/src/core/network/entity_serde.rs) and using `serde(with = "...")`. But it feels like a bad default to me.
Using `Entity` over a custom `NetworkId` also have the following advantages:
1. Re-use `MapEntities` trait to map `Entity`s in replicated components.
2. Instead of server `Entity <-> NetworkId ` and `Entity <-> NetworkId`, we map entities only on client.
3. No need to handling uniqueness. It's a rare case, but makes things simpler. For example, I don't need to query for a resource to create an unique ID.
Closes#6143.
## Solution
Use default serde impls. If anyone want to avoid wasting memory on `generation`, they can create a new type that holds `u32`. This is what Bevy do for [DynamicEntity](https://docs.rs/bevy/latest/bevy/scene/struct.DynamicEntity.html) to serialize scenes. And I don't see any use case to serialize an entity id expect this one.
---
## Changelog
### Changed
- Entity now serializes / deserializes `generation` field.
## Migration Guide
- Entity now fully serialized. If you want to serialze only `id`, as it was before, you can create a new type that wraps `u32`.
# Objective
- fix new clippy lints before they get stable and break CI
## Solution
- run `clippy --fix` to auto-fix machine-applicable lints
- silence `clippy::should_implement_trait` for `fn HandleId::default<T: Asset>`
## Changes
- always prefer `format!("{inline}")` over `format!("{}", not_inline)`
- prefer `Box::default` (or `Box::<T>::default` if necessary) over `Box::new(T::default())`
# Objective
^ enable this
Concretely, I need to
- list all handle ids for an asset type
- fetch the asset as `dyn Reflect`, given a `HandleUntyped`
- when encountering a `Handle<T>`, find out what asset type that handle refers to (`T`'s type id) and turn the handle into a `HandleUntyped`
## Solution
- add `ReflectAsset` type containing function pointers for working with assets
```rust
pub struct ReflectAsset {
type_uuid: Uuid,
assets_resource_type_id: TypeId, // TypeId of the `Assets<T>` resource
get: fn(&World, HandleUntyped) -> Option<&dyn Reflect>,
get_mut: fn(&mut World, HandleUntyped) -> Option<&mut dyn Reflect>,
get_unchecked_mut: unsafe fn(&World, HandleUntyped) -> Option<&mut dyn Reflect>,
add: fn(&mut World, &dyn Reflect) -> HandleUntyped,
set: fn(&mut World, HandleUntyped, &dyn Reflect) -> HandleUntyped,
len: fn(&World) -> usize,
ids: for<'w> fn(&'w World) -> Box<dyn Iterator<Item = HandleId> + 'w>,
remove: fn(&mut World, HandleUntyped) -> Option<Box<dyn Reflect>>,
}
```
- add `ReflectHandle` type relating the handle back to the asset type and providing a way to create a `HandleUntyped`
```rust
pub struct ReflectHandle {
type_uuid: Uuid,
asset_type_id: TypeId,
downcast_handle_untyped: fn(&dyn Any) -> Option<HandleUntyped>,
}
```
- add the corresponding `FromType` impls
- add a function `app.register_asset_reflect` which is supposed to be called after `.add_asset` and registers `ReflectAsset` and `ReflectHandle` in the type registry
---
## Changelog
- add `ReflectAsset` and `ReflectHandle` types, which allow code to use reflection to manipulate arbitrary assets without knowing their types at compile time
fixes https://github.com/bevyengine/bevy/issues/5944
Uses the second solution:
> 2. keep track of the old viewport in the computed_state, and if camera.viewport != camera.computed_state.old_viewport, then update the projection. This is more reliable, but needs to store two UVec2s more in the camera (probably not a big deal).
# Objective
Clean up code surrounding fetch by pulling out the common parts into the iteration code.
## Solution
Merge `Fetch::table_fetch` and `Fetch::archetype_fetch` into a single API: `Fetch::fetch(&mut self, entity: &Entity, table_row: &usize)`. This provides everything any fetch requires to internally decide which storage to read from and get the underlying data. All of these functions are marked as `#[inline(always)]` and the arguments are passed as references to attempt to optimize out the argument that isn't being used.
External to `Fetch`, Query iteration has been changed to keep track of the table row and entity outside of fetch, which moves a lot of the expensive bookkeeping `Fetch` structs had previously done internally into the outer loop.
~~TODO: Benchmark, docs~~ Done.
---
## Changelog
Changed: `Fetch::table_fetch` and `Fetch::archetype_fetch` have been merged into a single `Fetch::fetch` function.
## Migration Guide
TODO
Co-authored-by: Brian Merchant <bhmerchang@gmail.com>
Co-authored-by: Saverio Miroddi <saverio.pub2@gmail.com>
# Objective
Currently, `DynamicSceneBuilder` keeps track of entities via a `HashMap`. This has an unintended side-effect in that, when building the full `DynamicScene`, we aren't guaranteed any particular order.
In other words, inserting Entity A then Entity B can result in either `[A, B]` or `[B, A]`. This can be rather annoying when running tests on scenes generated via the builder as it will work sometimes but not other times. There's also the potential that this might unnecessarily clutter up VCS diffs for scene files (assuming they had an intentional order).
## Solution
Store `DynamicSceneBuilder`'s entities in a `Vec` rather than a `HashMap`.
---
## Changelog
* Stablized entity order in `DynamicSceneBuilder` (0.9.0-dev)
# Objective
Bevy's internal plugins have lots of execution-order ambiguities, which makes the ambiguity detection tool very noisy for our users.
## Solution
Silence every last ambiguity that can currently be resolved.
Each time an ambiguity is silenced, it is accompanied by a comment describing why it is correct. This description should be based on the public API of the respective systems. Thus, I have added documentation to some systems describing how they use some resources.
# Future work
Some ambiguities remain, due to issues out of scope for this PR.
* The ambiguity checker does not respect `Without<>` filters, leading to false positives.
* Ambiguities between `bevy_ui` and `bevy_animation` cannot be resolved, since neither crate knows that the other exists. We will need a general solution to this problem.
# Objective
- Fixes#5876 .
## Solution
- added pub use statements to re-export the following traits in bevy_audio: rodio::source::Source, rodio::Sample, rodio::cpal::Sample.
- rodio::cpal::Sample was re-exported as CpalSample to avoid naming conflict with rodio::Sample.
# Objective
Currently scenes define components using a list:
```rust
[
(
entity: 0,
components: [
{
"bevy_transform::components::transform::Transform": (
translation: (
x: 0.0,
y: 0.0,
z: 0.0
),
rotation: (0.0, 0.0, 0.0, 1.0),
scale: (
x: 1.0,
y: 1.0,
z: 1.0
),
),
},
{
"my_crate::Foo": (
text: "Hello World",
),
},
{
"my_crate::Bar": (
baz: 123,
),
},
],
),
]
```
However, this representation has some drawbacks (as pointed out by @Metadorius in [this](https://github.com/bevyengine/bevy/pull/4561#issuecomment-1202215565) comment):
1. Increased nesting and more characters (minor effect on overall size)
2. More importantly, by definition, entities cannot have more than one instance of any given component. Therefore, such data is best stored as a map— where all values are meant to have unique keys.
## Solution
Change `components` to store a map of components rather than a list:
```rust
[
(
entity: 0,
components: {
"bevy_transform::components::transform::Transform": (
translation: (
x: 0.0,
y: 0.0,
z: 0.0
),
rotation: (0.0, 0.0, 0.0, 1.0),
scale: (
x: 1.0,
y: 1.0,
z: 1.0
),
),
"my_crate::Foo": (
text: "Hello World",
),
"my_crate::Bar": (
baz: 123
),
},
),
]
```
#### Code Representation
This change only affects the scene format itself. `DynamicEntity` still stores its components as a list. The reason for this is that storing such data as a map is not really needed since:
1. The "key" of each value is easily found by just calling `Reflect::type_name` on it
2. We should be generating such structs using the `World` itself which upholds the one-component-per-entity rule
One could in theory create manually create a `DynamicEntity` with duplicate components, but this isn't something I think we should focus on in this PR. `DynamicEntity` can be broken in other ways (i.e. storing a non-component in the components list), and resolving its issues can be done in a separate PR.
---
## Changelog
* The scene format now uses a map to represent the collection of components rather than a list
## Migration Guide
The scene format now uses a map to represent the collection of components. Scene files will need to update from the old list format.
<details>
<summary>Example Code</summary>
```rust
// OLD
[
(
entity: 0,
components: [
{
"bevy_transform::components::transform::Transform": (
translation: (
x: 0.0,
y: 0.0,
z: 0.0
),
rotation: (0.0, 0.0, 0.0, 1.0),
scale: (
x: 1.0,
y: 1.0,
z: 1.0
),
),
},
{
"my_crate::Foo": (
text: "Hello World",
),
},
{
"my_crate::Bar": (
baz: 123,
),
},
],
),
]
// NEW
[
(
entity: 0,
components: {
"bevy_transform::components::transform::Transform": (
translation: (
x: 0.0,
y: 0.0,
z: 0.0
),
rotation: (0.0, 0.0, 0.0, 1.0),
scale: (
x: 1.0,
y: 1.0,
z: 1.0
),
),
"my_crate::Foo": (
text: "Hello World",
),
"my_crate::Bar": (
baz: 123
),
},
),
]
```
</details>
# Objective
Currently, Bevy only supports rendering to the current "surface texture format". This means that "render to texture" scenarios must use the exact format the primary window's surface uses, or Bevy will crash. This is even harder than it used to be now that we detect preferred surface formats at runtime instead of using hard coded BevyDefault values.
## Solution
1. Look up and store each window surface's texture format alongside other extracted window information
2. Specialize the upscaling pass on the current `RenderTarget`'s texture format, now that we can cheaply correlate render targets to their current texture format
3. Remove the old `SurfaceTextureFormat` and `AvailableTextureFormats`: these are now redundant with the information stored on each extracted window, and probably should not have been globals in the first place (as in theory each surface could have a different format).
This means you can now use any texture format you want when rendering to a texture! For example, changing the `render_to_texture` example to use `R16Float` now doesn't crash / properly only stores the red component:
