# 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
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
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
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
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
At least partially addresses #6282.
Resources are currently stored as a dedicated Resource archetype (ID 1). This allows for easy code reusability, but unnecessarily adds 72 bytes (on 64-bit systems) to the struct that is only used for that one archetype. It also requires several fields to be `pub(crate)` which isn't ideal.
This should also remove one sparse-set lookup from fetching, inserting, and removing resources from a `World`.
## Solution
- Add `Resources` parallel to `Tables` and `SparseSets` and extract the functionality used by `Archetype` in it.
- Remove `unique_components` from `Archetype`
- Remove the `pub(crate)` on `Archetype::components`.
- Remove `ArchetypeId::RESOURCE`
- Remove `Archetypes::resource` and `Archetypes::resource_mut`
---
## Changelog
Added: `Resources` type to store resources.
Added: `Storages::resource`
Removed: `ArchetypeId::RESOURCE`
Removed: `Archetypes::resource` and `Archetypes::resources`
Removed: `Archetype::unique_components` and `Archetypes::unique_components_mut`
## Migration Guide
Resources have been moved to `Resources` under `Storages` in `World`. All code dependent on `Archetype::unique_components(_mut)` should access it via `world.storages().resources()` instead.
All APIs accessing the raw data of individual resources (mutable *and* read-only) have been removed as these APIs allowed for unsound unsafe code. All usages of these APIs should be changed to use `World::{get, insert, remove}_resource`.
# Objective
There is currently no good way of getting the width (# of components) of a table outside of `bevy_ecs`.
# Solution
Added the methods `Table::{component_count, component_capacity}`
For consistency and clarity, renamed `Table::{len, capacity}` to `entity_count` and `entity_capacity`.
## Changelog
- Added the methods `Table::component_count` and `Table::component_capacity`
- Renamed `Table::len` and `Table::capacity` to `entity_count` and `entity_capacity`
## Migration Guide
Any use of `Table::len` should now be `Table::entity_count`. Any use of `Table::capacity` should now be `Table::entity_capacity`.
# Objective
remove `insert_resource_with_id` because `insert_resource_by_id` exists and does almost exactly the same thing
blocked on #5587 because otherwise we will leak a resource when it's inserted
## Solution
remove the function and also add a safety invariant of to `insert_resource_by_id` that the id be valid for the world.
I didn't see any discussion in #4447 about this safety invariant being left off in favor of a panic so I'm curious if there was one or if it just seemed nicer to have less safety invariants for callers to uphold 😅
---
## Changelog
- safety invariant added to `insert_resource_by_id` requiring the id to be valid for world
## Migration Guide
- audit any calls to `insert_resource_by_id` making sure that the id is valid for the world
Co-authored-by: Carter Anderson <mcanders1@gmail.com>
Remove unnecessary calls to `iter()`/`iter_mut()`.
Mainly updates the use of queries in our code, docs, and examples.
```rust
// From
for _ in list.iter() {
for _ in list.iter_mut() {
// To
for _ in &list {
for _ in &mut list {
```
We already enable the pedantic lint [clippy::explicit_iter_loop](https://rust-lang.github.io/rust-clippy/stable/) inside of Bevy. However, this only warns for a few known types from the standard library.
## Note for reviewers
As you can see the additions and deletions are exactly equal.
Maybe give it a quick skim to check I didn't sneak in a crypto miner, but you don't have to torture yourself by reading every line.
I already experienced enough pain making this PR :)
Co-authored-by: devil-ira <justthecooldude@gmail.com>
# Objective
`SAFETY` comments are meant to be placed before `unsafe` blocks and should contain the reasoning of why in this case the usage of unsafe is okay. This is useful when reading the code because it makes it clear which assumptions are required for safety, and makes it easier to spot possible unsoundness holes. It also forces the code writer to think of something to write and maybe look at the safety contracts of any called unsafe methods again to double-check their correct usage.
There's a clippy lint called `undocumented_unsafe_blocks` which warns when using a block without such a comment.
## Solution
- since clippy expects `SAFETY` instead of `SAFE`, rename those
- add `SAFETY` comments in more places
- for the last remaining 3 places, add an `#[allow()]` and `// TODO` since I wasn't comfortable enough with the code to justify their safety
- add ` #![warn(clippy::undocumented_unsafe_blocks)]` to `bevy_ecs`
### Note for reviewers
The first commit only renames `SAFETY` to `SAFE` so it doesn't need a thorough review.
cb042a416e..55cef2d6fa is the diff for all other changes.
### Safety comments where I'm not too familiar with the code
774012ece5/crates/bevy_ecs/src/entity/mod.rs (L540-L546)774012ece5/crates/bevy_ecs/src/world/entity_ref.rs (L249-L252)
### Locations left undocumented with a `TODO` comment
5dde944a30/crates/bevy_ecs/src/schedule/executor_parallel.rs (L196-L199)5dde944a30/crates/bevy_ecs/src/world/entity_ref.rs (L287-L289)5dde944a30/crates/bevy_ecs/src/world/entity_ref.rs (L413-L415)
Co-authored-by: Jakob Hellermann <hellermann@sipgate.de>
The first leak:
```rust
#[test]
fn blob_vec_drop_empty_capacity() {
let item_layout = Layout:🆕:<Foo>();
let drop = drop_ptr::<Foo>;
let _ = unsafe { BlobVec::new(item_layout, Some(drop), 0) };
}
```
this is because we allocate the swap scratch in blobvec regardless of what the capacity is, but we only deallocate if capacity is > 0
The second leak:
```rust
#[test]
fn panic_while_overwriting_component() {
let helper = DropTestHelper::new();
let res = panic::catch_unwind(|| {
let mut world = World::new();
world
.spawn()
.insert(helper.make_component(true, 0))
.insert(helper.make_component(false, 1));
println!("Done inserting! Dropping world...");
});
let drop_log = helper.finish(res);
assert_eq!(
&*drop_log,
[
DropLogItem::Create(0),
DropLogItem::Create(1),
DropLogItem::Drop(0),
]
);
}
```
this is caused by us not running the drop impl on the to-be-inserted component if the drop impl of the overwritten component panics
---
managed to figure out where the leaks were by using this 10/10 command
```
cargo --quiet test --lib -- --list | sed 's/: test$//' | MIRIFLAGS="-Zmiri-disable-isolation" xargs -n1 cargo miri test --lib -- --exact
```
which runs every test one by one rather than all at once which let miri actually tell me which test had the leak 🙃
# Objective
Speed up entity moves between tables by reducing the number of copies conducted. Currently three separate copies are conducted: `src[index] -> swap scratch`, `src[last] -> src[index]`, and `swap scratch -> dst[target]`. The first and last copies can be merged by directly using the copy `src[index] -> dst[target]`, which can save quite some time if the component(s) in question are large.
## Solution
This PR does the following:
- Adds `BlobVec::swap_remove_unchecked(usize, PtrMut<'_>)`, which is identical to `swap_remove_and_forget_unchecked`, but skips the `swap_scratch` and directly copies the component into the provided `PtrMut<'_>`.
- Build `Column::initialize_from_unchecked(&mut Column, usize, usize)` on top of it, which uses the above to directly initialize a row from another column.
- Update most of the table move APIs to use `initialize_from_unchecked` instead of a combination of `swap_remove_and_forget_unchecked` and `initialize`.
This is an alternative, though orthogonal, approach to achieve the same performance gains as seen in #4853. This (hopefully) shouldn't run into the same Miri limitations that said PR currently does. After this PR, `swap_remove_and_forget_unchecked` is still in use for Resources and swap_scratch likely still should be removed, so #4853 still has use, even if this PR is merged.
## Performance
TODO: Microbenchmark
This PR shows similar improvements to commands that add or remove table components that result in a table move. When tested on `many_cubes sphere`, some of the more command heavy systems saw notable improvements. In particular, `prepare_uniform_components<T>`, this saw a reduction in time from 1.35ms to 1.13ms (a 16.3% improvement) on my local machine, a similar if not slightly better gain than what #4853 showed [here](https://github.com/bevyengine/bevy/pull/4853#issuecomment-1159346106).
The command heavy `Extract` stage also saw a smaller overall improvement:
---
## Changelog
Added: `BlobVec::swap_remove_unchecked`.
Added: `Column::initialize_from_unchecked`.
# Objective
Closes#1557. Partially addresses #3362.
Cleanup the public facing API for storage types. Most of these APIs are difficult to use safely when directly interfacing with these types, and is also currently impossible to interact with in normal ECS use as there is no `World::storages_mut`. The majority of these types should be easy enough to read, and perhaps mutate the contents, but never structurally altered without the same checks in the rest of bevy_ecs code. This both cleans up the public facing types and helps use unused code detection to remove a few of the APIs we're not using internally.
## Solution
- Mark all APIs that take `&mut T` under `bevy_ecs::storage` as `pub(crate)` or `pub(super)`
- Cleanup after it all.
Entire type visibility changes:
- `BlobVec` is `pub(super)`, only storage code should be directly interacting with it.
- `SparseArray` is now `pub(crate)` for the entire type. It's an implementation detail for `Table` and `(Component)SparseSet`.
- `TableMoveResult` is now `pub(crate)
---
## Changelog
TODO
## Migration Guide
Dear God, I hope not.
# Objective
Following #4855, `Column` is just a parallel `BlobVec`/`Vec<UnsafeCell<ComponentTicks>>` pair, which is identical to the dense and ticks vecs in `ComponentSparseSet`, which has some code duplication with `Column`.
## Solution
Replace dense and ticks in `ComponentSparseSet` with a `Column`.
# Objective
`debug_assert!` macros must still compile properly in release mode due to how they're implemented. This is causing release builds to fail.
## Solution
Change them to `assert!` macros inside `#[cfg(debug_assertions)]` blocks.
# Objective
Use less memory to store SparseSet components.
## Solution
Change `ComponentSparseSet` to only use `Entity::id` in it's key internally, and change the usize value in it's SparseArray to use u32 instead, as it cannot have more than u32::MAX live entities stored at once.
This should reduce the overhead of storing components in sparse set storage by 50%.
# Objective
The `ComponentId` in `Column` is redundant as it's stored in parallel in the surrounding `SparseSet` all the time.
## Solution
Remove it. Add `SparseSet::iter(_mut)` to parallel `HashMap::iter(_mut)` to allow iterating pairs of columns and their IDs.
---
## Changelog
Added: `SparseSet::iter` and `SparseSet::iter_mut`.
# Objective
Even if bevy itself does not provide any builtin scripting or modding APIs, it should have the foundations for building them yourself.
For that it should be enough to have APIs that are not tied to the actual rust types with generics, but rather accept `ComponentId`s and `bevy_ptr` ptrs.
## Solution
Add the following APIs to bevy
```rust
fn EntityRef::get_by_id(ComponentId) -> Option<Ptr<'w>>;
fn EntityMut::get_by_id(ComponentId) -> Option<Ptr<'_>>;
fn EntityMut::get_mut_by_id(ComponentId) -> Option<MutUntyped<'_>>;
fn World::get_resource_by_id(ComponentId) -> Option<Ptr<'_>>;
fn World::get_resource_mut_by_id(ComponentId) -> Option<MutUntyped<'_>>;
// Safety: `value` must point to a valid value of the component
unsafe fn World::insert_resource_by_id(ComponentId, value: OwningPtr);
fn ComponentDescriptor::new_with_layout(..) -> Self;
fn World::init_component_with_descriptor(ComponentDescriptor) -> ComponentId;
```
~~This PR would definitely benefit from #3001 (lifetime'd pointers) to make sure that the lifetimes of the pointers are valid and the my-move pointer in `insert_resource_by_id` could be an `OwningPtr`, but that can be adapter later if/when #3001 is merged.~~
### Not in this PR
- inserting components on entities (this is very tied to types with bundles and the `BundleInserter`)
- an untyped version of a query (needs good API design, has a large implementation complexity, can be done in a third-party crate)
Co-authored-by: Jakob Hellermann <hellermann@sipgate.de>
# Objective
- We do a lot of function pointer calls in a hot loop (clearing entities in render). This is slow, since calling function pointers cannot be optimised out. We can avoid that in the cases where the function call is a no-op.
- Alternative to https://github.com/bevyengine/bevy/pull/2897
- On my machine, in `many_cubes`, this reduces dropping time from ~150μs to ~80μs.
## Solution
- Make `drop` in `BlobVec` an `Option`, recording whether the given drop impl is required or not.
- Note that this does add branching in some cases - we could consider splitting this into two fields, i.e. unconditionally call the `drop` fn pointer.
- My intuition of how often types stored in `World` should have non-trivial drops makes me think that would be slower, however.
N.B. Even once this lands, we should still test having a 'drop_multiple' variant - for types with a real `Drop` impl, the current implementation is definitely optimal.
# Objective
- (Eventually) reduce noise in reporting access conflicts between unordered systems.
- `SystemStage` only looks at unfiltered `ComponentId` access, any conflicts reported are potentially `false`.
- the systems could still be accessing disjoint archetypes
- Comparing systems' filtered access sets can maybe avoid that (for statically known component types).
- #4204
## Solution
- Modify `SparseSetIndex` trait to require `PartialEq`, `Eq`, and `Hash` (all internal types except `BundleId` already did).
- Add `is_compatible` and `get_conflicts` methods to `FilteredAccessSet<T>`
- (existing method renamed to `get_conflicts_single`)
- Add docs for those and all the other methods while I'm at it.
1. change `PtrMut::as_ptr(self)` and `OwnedPtr::as_ptr(self)` to take `&self`, otherwise printing the pointer will prevent doing anything else afterwards
2. make all `as_ptr` methods safe. There's nothing unsafe about obtaining a pointer, these kinds of methods are safe in std as well [str::as_ptr](https://doc.rust-lang.org/stable/std/primitive.str.html#method.as_ptr), [Rc::as_ptr](https://doc.rust-lang.org/stable/std/rc/struct.Rc.html#method.as_ptr)
3. rename `offset`/`add` to `byte_offset`/`byte_add`. The unprefixed methods in std add in increments of `std::mem::size_of::<T>`, not in bytes. There's a PR for rust to add these byte_ methods https://github.com/rust-lang/rust/pull/95643 and at the call site it makes it much more clear that you need to do `.byte_add(i * layout_size)` instead of `.add(i)`
# Objective
The pointer types introduced in #3001 are useful not just in `bevy_ecs`, but also in crates like `bevy_reflect` (#4475) or even outside of bevy.
## Solution
Extract `Ptr<'a>`, `PtrMut<'a>`, `OwnedPtr<'a>`, `ThinSlicePtr<'a, T>` and `UnsafeCellDeref` from `bevy_ecs::ptr` into `bevy_ptr`.
**Note:** `bevy_ecs` still reexports the `bevy_ptr` as `bevy_ecs::ptr` so that crates like `bevy_transform` can use the `Bundle` derive without needing to depend on `bevy_ptr` themselves.
# Objective
The `Ptr` types gives free access to the underlying `NonNull<u8>`, which adds more publicly visible pointer wrangling than there needs to be. There are also a few edge cases where Ptr types could be more readily utilized for properly validating the soundness of ECS operations.
## Solution
- Replace `*Ptr(Mut)::inner` with `cast` which requires a concrete type to give the pointer. This function could also have a `debug_assert` with an alignment check to ensure that the pointer is aligned properly, but is currently not included.
- Use `OwningPtr::read` in ECS macros over casting the inner pointer around.
# Objective
`bevy_ecs` has large amounts of unsafe code which is hard to get right and makes it difficult to audit for soundness.
## Solution
Introduce lifetimed, type-erased pointers: `Ptr<'a>` `PtrMut<'a>` `OwningPtr<'a>'` and `ThinSlicePtr<'a, T>` which are newtypes around a raw pointer with a lifetime and conceptually representing strong invariants about the pointee and validity of the pointer.
The process of converting bevy_ecs to use these has already caught multiple cases of unsound behavior.
## Changelog
TL;DR for release notes: `bevy_ecs` now uses lifetimed, type-erased pointers internally, significantly improving safety and legibility without sacrificing performance. This should have approximately no end user impact, unless you were meddling with the (unfortunately public) internals of `bevy_ecs`.
- `Fetch`, `FilterFetch` and `ReadOnlyFetch` trait no longer have a `'state` lifetime
- this was unneeded
- `ReadOnly/Fetch` associated types on `WorldQuery` are now on a new `WorldQueryGats<'world>` trait
- was required to work around lack of Generic Associated Types (we wish to express `type Fetch<'a>: Fetch<'a>`)
- `derive(WorldQuery)` no longer requires `'w` lifetime on struct
- this was unneeded, and improves the end user experience
- `EntityMut::get_unchecked_mut` returns `&'_ mut T` not `&'w mut T`
- allows easier use of unsafe API with less footguns, and can be worked around via lifetime transmutery as a user
- `Bundle::from_components` now takes a `ctx` parameter to pass to the `FnMut` closure
- required because closure return types can't borrow from captures
- `Fetch::init` takes `&'world World`, `Fetch::set_archetype` takes `&'world Archetype` and `&'world Tables`, `Fetch::set_table` takes `&'world Table`
- allows types implementing `Fetch` to store borrows into world
- `WorldQuery` trait now has a `shrink` fn to shorten the lifetime in `Fetch::<'a>::Item`
- this works around lack of subtyping of assoc types, rust doesnt allow you to turn `<T as Fetch<'static>>::Item'` into `<T as Fetch<'a>>::Item'`
- `QueryCombinationsIter` requires this
- Most types implementing `Fetch` now have a lifetime `'w`
- allows the fetches to store borrows of world data instead of using raw pointers
## Migration guide
- `EntityMut::get_unchecked_mut` returns a more restricted lifetime, there is no general way to migrate this as it depends on your code
- `Bundle::from_components` implementations must pass the `ctx` arg to `func`
- `Bundle::from_components` callers have to use a fn arg instead of closure captures for borrowing from world
- Remove lifetime args on `derive(WorldQuery)` structs as it is nonsensical
- `<Q as WorldQuery>::ReadOnly/Fetch` should be changed to either `RO/QueryFetch<'world>` or `<Q as WorldQueryGats<'world>>::ReadOnly/Fetch`
- `<F as Fetch<'w, 's>>` should be changed to `<F as Fetch<'w>>`
- Change the fn sigs of `Fetch::init/set_archetype/set_table` to match respective trait fn sigs
- Implement the required `fn shrink` on any `WorldQuery` implementations
- Move assoc types `Fetch` and `ReadOnlyFetch` on `WorldQuery` impls to `WorldQueryGats` impls
- Pass an appropriate `'world` lifetime to whatever fetch struct you are for some reason using
### Type inference regression
in some cases rustc may give spurrious errors when attempting to infer the `F` parameter on a query/querystate this can be fixed by manually specifying the type, i.e. `QueryState:🆕:<_, ()>(world)`. The error is rather confusing:
```rust=
error[E0271]: type mismatch resolving `<() as Fetch<'_>>::Item == bool`
--> crates/bevy_pbr/src/render/light.rs:1413:30
|
1413 | main_view_query: QueryState::new(world),
| ^^^^^^^^^^^^^^^ expected `bool`, found `()`
|
= note: required because of the requirements on the impl of `for<'x> FilterFetch<'x>` for `<() as WorldQueryGats<'x>>::Fetch`
note: required by a bound in `bevy_ecs::query::QueryState::<Q, F>::new`
--> crates/bevy_ecs/src/query/state.rs:49:32
|
49 | for<'x> QueryFetch<'x, F>: FilterFetch<'x>,
| ^^^^^^^^^^^^^^^ required by this bound in `bevy_ecs::query::QueryState::<Q, F>::new`
```
---
Made with help from @BoxyUwU and @alice-i-cecile
Co-authored-by: Boxy <supbscripter@gmail.com>
For some keys, it is too expensive to hash them on every lookup. Historically in Bevy, we have regrettably done the "wrong" thing in these cases (pre-computing hashes, then re-hashing them) because Rust's built in hashed collections don't give us the tools we need to do otherwise. Doing this is "wrong" because two different values can result in the same hash. Hashed collections generally get around this by falling back to equality checks on hash collisions. You can't do that if the key _is_ the hash. Additionally, re-hashing a hash increase the odds of collision!
#3959 needs pre-hashing to be viable, so I decided to finally properly solve the problem. The solution involves two different changes:
1. A new generalized "pre-hashing" solution in bevy_utils: `Hashed<T>` types, which store a value alongside a pre-computed hash. And `PreHashMap<K, V>` (which uses `Hashed<T>` internally) . `PreHashMap` is just an alias for a normal HashMap that uses `Hashed<T>` as the key and a new `PassHash` implementation as the Hasher.
2. Replacing the `std::collections` re-exports in `bevy_utils` with equivalent `hashbrown` impls. Avoiding re-hashes requires the `raw_entry_mut` api, which isn't stabilized yet (and may never be ... `entry_ref` has favor now, but also isn't available yet). If std's HashMap ever provides the tools we need, we can move back to that. The latest version of `hashbrown` adds support for the `entity_ref` api, so we can move to that in preparation for an std migration, if thats the direction they seem to be going in. Note that adding hashbrown doesn't increase our dependency count because it was already in our tree.
In addition to providing these core tools, I also ported the "table identity hashing" in `bevy_ecs` to `raw_entry_mut`, which was a particularly egregious case.
The biggest outstanding case is `AssetPathId`, which stores a pre-hash. We need AssetPathId to be cheaply clone-able (and ideally Copy), but `Hashed<AssetPath>` requires ownership of the AssetPath, which makes cloning ids way more expensive. We could consider doing `Hashed<Arc<AssetPath>>`, but cloning an arc is still a non-trivial expensive that needs to be considered. I would like to handle this in a separate PR. And given that we will be re-evaluating the Bevy Assets implementation in the very near future, I'd prefer to hold off until after that conversation is concluded.
What is says on the tin.
This has got more to do with making `clippy` slightly more *quiet* than it does with changing anything that might greatly impact readability or performance.
that said, deriving `Default` for a couple of structs is a nice easy win
#3457 adds the `doc_markdown` clippy lint, which checks doc comments to make sure code identifiers are escaped with backticks. This causes a lot of lint errors, so this is one of a number of PR's that will fix those lint errors one crate at a time.
This PR fixes lints in the `bevy_ecs` crate.
# Objective
- Storages are used to store the ECS data.
- They're undocumented.
## Solution
- Add some very basic docs.
## Notes
- Some of this was hard to immediately understand when reading the code, so suggestions on improvements / things to add are particularly welcome.
# Objective
I thought I'd have a go a trying to fix#2597.
Hopefully fixes#2597.
## Solution
I reused the memory pointed to by the value parameter, that is already required by `insert` to not be dropped, to contain the extracted value while dropping it.
This implements the most minimal variant of #1843 - a derive for marker trait. This is a prerequisite to more complicated features like statically defined storage type or opt-out component reflection.
In order to make component struct's purpose explicit and avoid misuse, it must be annotated with `#[derive(Component)]` (manual impl is discouraged for compatibility). Right now this is just a marker trait, but in the future it might be expanded. Making this change early allows us to make further changes later without breaking backward compatibility for derive macro users.
This already prevents a lot of issues, like using bundles in `insert` calls. Primitive types are no longer valid components as well. This can be easily worked around by adding newtype wrappers and deriving `Component` for them.
One funny example of prevented bad code (from our own tests) is when an newtype struct or enum variant is used. Previously, it was possible to write `insert(Newtype)` instead of `insert(Newtype(value))`. That code compiled, because function pointers (in this case newtype struct constructor) implement `Send + Sync + 'static`, so we allowed them to be used as components. This is no longer the case and such invalid code will trigger a compile error.
Co-authored-by: = <=>
Co-authored-by: TheRawMeatball <therawmeatball@gmail.com>
Co-authored-by: Carter Anderson <mcanders1@gmail.com>
## Objective
The upcoming Bevy Book makes many references to the API documentation of bevy.
Most references belong to the first two chapters of the Bevy Book:
- bevyengine/bevy-website#176
- bevyengine/bevy-website#182
This PR attempts to improve the documentation of `bevy_ecs` and `bevy_app` in order to help readers of the Book who want to delve deeper into technical details.
## Solution
- Add crate and level module documentation
- Document the most important items (basically those included in the preludes), with the following style, where applicable:
- **Summary.** Short description of the item.
- **Second paragraph.** Detailed description of the item, without going too much in the implementation.
- **Code example(s).**
- **Safety or panic notes.**
## Collaboration
Any kind of collaboration is welcome, especially corrections, wording, new ideas and guidelines on where the focus should be put in.
---
### Related issues
- Fixes#2246
This upstreams the code changes used by the new renderer to enable cross-app Entity reuse:
* Spawning at specific entities
* get_or_spawn: spawns an entity if it doesn't already exist and returns an EntityMut
* insert_or_spawn_batch: the batched equivalent to `world.get_or_spawn(entity).insert_bundle(bundle)`
* Clearing entities and storages
* Allocating Entities with "invalid" archetypes. These entities cannot be queried / are treated as "non existent". They serve as "reserved" entities that won't show up when calling `spawn()`. They must be "specifically spawned at" using apis like `get_or_spawn(entity)`.
In combination, these changes enable the "render world" to clear entities / storages each frame and reserve all "app world entities". These can then be spawned during the "render extract step".
This refactors "spawn" and "insert" code in a way that I think is a massive improvement to legibility and re-usability. It also yields marginal performance wins by reducing some duplicate lookups (less than a percentage point improvement on insertion benchmarks). There is also some potential for future unsafe reduction (by making BatchSpawner and BatchInserter generic). But for now I want to cut down generic usage to a minimum to encourage smaller binaries and faster compiles.
This is currently a draft because it needs more tests (although this code has already had some real-world testing on my custom-shaders branch).
I also fixed the benchmarks (which currently don't compile!) / added new ones to illustrate batching wins.
After these changes, Bevy ECS is basically ready to accommodate the new renderer. I think the biggest missing piece at this point is "sub apps".
# Objective
There is currently a 1-to-1 mapping between components and real rust types. This means that it is impossible for multiple components to be represented by the same rust type or for a component to not have a rust type at all. This means that component types can't be defined in languages other than rust like necessary for scripting or sandboxed (wasm?) plugins.
## Solution
Refactor `ComponentDescriptor` and `Bundle` to remove `TypeInfo`. `Bundle` now uses `ComponentId` instead. `ComponentDescriptor` is now always created from a rust type instead of through the `TypeInfo` indirection. A future PR may make it possible to construct a `ComponentDescriptor` from it's fields without a rust type being involved.
When dropping the data, we originally only checked the size of an individual item instead of the size of the allocation. However with a capacity of 0, we attempt to deallocate a pointer which was not the result of allocation. That is, an item of `Layout { size_: 8, align_: 8 }` produces an array of `Layout { size_: 0, align_: 8 }` when `capacity = 0`.
Fixes#2294
Continuing the work on reducing the safety footguns in the code, I've removed one extra `UnsafeCell` in favour of safe `Cell` usage inisde `ComponentTicks`. That change led to discovery of misbehaving component insert logic, where data wasn't properly dropped when overwritten. Apart from that being fixed, some method names were changed to better convey the "initialize new allocation" and "replace existing allocation" semantic.
Depends on #2221, I will rebase this PR after the dependency is merged. For now, review just the last commit.
Co-authored-by: Carter Anderson <mcanders1@gmail.com>
I've noticed that we are overusing interior mutability of the Table data, where in many cases we already own a unique reference to it. That prompted a slight refactor aiming to reduce number of safety constraints that must be manually upheld. Now the majority of those are just about avoiding bound checking, which is relatively easy to prove right.
Another aspect is reducing the complexity of Table struct. Notably, we don't ever use archetypes stored there, so this whole thing goes away. Capacity and grow amount were mostly superficial, as we are already using Vecs inside anyway, so I've got rid of those too. Now the overall table capacity is being driven by the internal entity Vec capacity. This has a side effect of automatically implementing exponential growth pattern for BitVecs reallocations inside Table, which to my measurements slightly improves performance in tests that are heavy on inserts. YMMV, but I hope that those tests were at least remotely correct.
- simplified code around archetype generations a little bit, as the special case value is not actually needed
- removed unnecessary UnsafeCell around pointer value that is never updated through shared references
- fixed and added a test for correct drop behaviour when removing sparse components through remove_bundle command
Removing the checks on this line https://github.com/bevyengine/bevy/blob/main/crates/bevy_sprite/src/frustum_culling.rs#L64 and running the "many_sprites" example revealed two corner case bugs in bevy_ecs. The first, a simple and honest missed line introduced in #1471. The other, an insidious monster that has been there since the ECS v2 rewrite, just waiting for the time to strike:
1. #1471 accidentally removed the "insert" line for sparse set components with the "mutated" bundle state. Re-adding it fixes the problem. I did a slight refactor here to make the implementation simpler and remove a branch.
2. The other issue is nastier. ECS v2 added an "archetype graph". When determining what components were added/mutated during an archetype change, we read the FromBundle edge (which encodes this state) on the "new" archetype. The problem is that unlike "add edges" which are guaranteed to be unique for a given ("graph node", "bundle id") pair, FromBundle edges are not necessarily unique:
```rust
// OLD_ARCHETYPE -> NEW_ARCHETYPE
// [] -> [usize]
e.insert(2usize);
// [usize] -> [usize, i32]
e.insert(1i32);
// [usize, i32] -> [usize, i32]
e.insert(1i32);
// [usize, i32] -> [usize]
e.remove::<i32>();
// [usize] -> [usize, i32]
e.insert(1i32);
```
Note that the second `e.insert(1i32)` command has a different "archetype graph edge" than the first, but they both lead to the same "new archetype".
The fix here is simple: just remove FromBundle edges because they are broken and store the information in the "add edges", which are guaranteed to be unique.
FromBundle edges were added to cut down on the number of archetype accesses / make the archetype access patterns nicer. But benching this change resulted in no significant perf changes and the addition of get_2_mut() for archetypes resolves the access pattern issue.
Fixes#1692
Alternative to #1696
This ensures that the capacity actually grows in increments of grow_amount, and also ensures that Table capacity is always <= column and entity vec capacity.
Debug logs that describe the new logic (running the example in #1692)
[out.txt](https://github.com/bevyengine/bevy/files/6173808/out.txt)
# Problem Definition
The current change tracking (via flags for both components and resources) fails to detect changes made by systems that are scheduled to run earlier in the frame than they are.
This issue is discussed at length in [#68](https://github.com/bevyengine/bevy/issues/68) and [#54](https://github.com/bevyengine/bevy/issues/54).
This is very much a draft PR, and contributions are welcome and needed.
# Criteria
1. Each change is detected at least once, no matter the ordering.
2. Each change is detected at most once, no matter the ordering.
3. Changes should be detected the same frame that they are made.
4. Competitive ergonomics. Ideally does not require opting-in.
5. Low CPU overhead of computation.
6. Memory efficient. This must not increase over time, except where the number of entities / resources does.
7. Changes should not be lost for systems that don't run.
8. A frame needs to act as a pure function. Given the same set of entities / components it needs to produce the same end state without side-effects.
**Exact** change-tracking proposals satisfy criteria 1 and 2.
**Conservative** change-tracking proposals satisfy criteria 1 but not 2.
**Flaky** change tracking proposals satisfy criteria 2 but not 1.
# Code Base Navigation
There are three types of flags:
- `Added`: A piece of data was added to an entity / `Resources`.
- `Mutated`: A piece of data was able to be modified, because its `DerefMut` was accessed
- `Changed`: The bitwise OR of `Added` and `Changed`
The special behavior of `ChangedRes`, with respect to the scheduler is being removed in [#1313](https://github.com/bevyengine/bevy/pull/1313) and does not need to be reproduced.
`ChangedRes` and friends can be found in "bevy_ecs/core/resources/resource_query.rs".
The `Flags` trait for Components can be found in "bevy_ecs/core/query.rs".
`ComponentFlags` are stored in "bevy_ecs/core/archetypes.rs", defined on line 446.
# Proposals
**Proposal 5 was selected for implementation.**
## Proposal 0: No Change Detection
The baseline, where computations are performed on everything regardless of whether it changed.
**Type:** Conservative
**Pros:**
- already implemented
- will never miss events
- no overhead
**Cons:**
- tons of repeated work
- doesn't allow users to avoid repeating work (or monitoring for other changes)
## Proposal 1: Earlier-This-Tick Change Detection
The current approach as of Bevy 0.4. Flags are set, and then flushed at the end of each frame.
**Type:** Flaky
**Pros:**
- already implemented
- simple to understand
- low memory overhead (2 bits per component)
- low time overhead (clear every flag once per frame)
**Cons:**
- misses systems based on ordering
- systems that don't run every frame miss changes
- duplicates detection when looping
- can lead to unresolvable circular dependencies
## Proposal 2: Two-Tick Change Detection
Flags persist for two frames, using a double-buffer system identical to that used in events.
A change is observed if it is found in either the current frame's list of changes or the previous frame's.
**Type:** Conservative
**Pros:**
- easy to understand
- easy to implement
- low memory overhead (4 bits per component)
- low time overhead (bit mask and shift every flag once per frame)
**Cons:**
- can result in a great deal of duplicated work
- systems that don't run every frame miss changes
- duplicates detection when looping
## Proposal 3: Last-Tick Change Detection
Flags persist for two frames, using a double-buffer system identical to that used in events.
A change is observed if it is found in the previous frame's list of changes.
**Type:** Exact
**Pros:**
- exact
- easy to understand
- easy to implement
- low memory overhead (4 bits per component)
- low time overhead (bit mask and shift every flag once per frame)
**Cons:**
- change detection is always delayed, possibly causing painful chained delays
- systems that don't run every frame miss changes
- duplicates detection when looping
## Proposal 4: Flag-Doubling Change Detection
Combine Proposal 2 and Proposal 3. Differentiate between `JustChanged` (current behavior) and `Changed` (Proposal 3).
Pack this data into the flags according to [this implementation proposal](https://github.com/bevyengine/bevy/issues/68#issuecomment-769174804).
**Type:** Flaky + Exact
**Pros:**
- allows users to acc
- easy to implement
- low memory overhead (4 bits per component)
- low time overhead (bit mask and shift every flag once per frame)
**Cons:**
- users must specify the type of change detection required
- still quite fragile to system ordering effects when using the flaky `JustChanged` form
- cannot get immediate + exact results
- systems that don't run every frame miss changes
- duplicates detection when looping
## [SELECTED] Proposal 5: Generation-Counter Change Detection
A global counter is increased after each system is run. Each component saves the time of last mutation, and each system saves the time of last execution. Mutation is detected when the component's counter is greater than the system's counter. Discussed [here](https://github.com/bevyengine/bevy/issues/68#issuecomment-769174804). How to handle addition detection is unsolved; the current proposal is to use the highest bit of the counter as in proposal 1.
**Type:** Exact (for mutations), flaky (for additions)
**Pros:**
- low time overhead (set component counter on access, set system counter after execution)
- robust to systems that don't run every frame
- robust to systems that loop
**Cons:**
- moderately complex implementation
- must be modified as systems are inserted dynamically
- medium memory overhead (4 bytes per component + system)
- unsolved addition detection
## Proposal 6: System-Data Change Detection
For each system, track which system's changes it has seen. This approach is only worth fully designing and implementing if Proposal 5 fails in some way.
**Type:** Exact
**Pros:**
- exact
- conceptually simple
**Cons:**
- requires storing data on each system
- implementation is complex
- must be modified as systems are inserted dynamically
## Proposal 7: Total-Order Change Detection
Discussed [here](https://github.com/bevyengine/bevy/issues/68#issuecomment-754326523). This proposal is somewhat complicated by the new scheduler, but I believe it should still be conceptually feasible. This approach is only worth fully designing and implementing if Proposal 5 fails in some way.
**Type:** Exact
**Pros:**
- exact
- efficient data storage relative to other exact proposals
**Cons:**
- requires access to the scheduler
- complex implementation and difficulty grokking
- must be modified as systems are inserted dynamically
# Tests
- We will need to verify properties 1, 2, 3, 7 and 8. Priority: 1 > 2 = 3 > 8 > 7
- Ideally we can use identical user-facing syntax for all proposals, allowing us to re-use the same syntax for each.
- When writing tests, we need to carefully specify order using explicit dependencies.
- These tests will need to be duplicated for both components and resources.
- We need to be sure to handle cases where ambiguous system orders exist.
`changing_system` is always the system that makes the changes, and `detecting_system` always detects the changes.
The component / resource changed will be simple boolean wrapper structs.
## Basic Added / Mutated / Changed
2 x 3 design:
- Resources vs. Components
- Added vs. Changed vs. Mutated
- `changing_system` runs before `detecting_system`
- verify at the end of tick 2
## At Least Once
2 x 3 design:
- Resources vs. Components
- Added vs. Changed vs. Mutated
- `changing_system` runs after `detecting_system`
- verify at the end of tick 2
## At Most Once
2 x 3 design:
- Resources vs. Components
- Added vs. Changed vs. Mutated
- `changing_system` runs once before `detecting_system`
- increment a counter based on the number of changes detected
- verify at the end of tick 2
## Fast Detection
2 x 3 design:
- Resources vs. Components
- Added vs. Changed vs. Mutated
- `changing_system` runs before `detecting_system`
- verify at the end of tick 1
## Ambiguous System Ordering Robustness
2 x 3 x 2 design:
- Resources vs. Components
- Added vs. Changed vs. Mutated
- `changing_system` runs [before/after] `detecting_system` in tick 1
- `changing_system` runs [after/before] `detecting_system` in tick 2
## System Pausing
2 x 3 design:
- Resources vs. Components
- Added vs. Changed vs. Mutated
- `changing_system` runs in tick 1, then is disabled by run criteria
- `detecting_system` is disabled by run criteria until it is run once during tick 3
- verify at the end of tick 3
## Addition Causes Mutation
2 design:
- Resources vs. Components
- `adding_system_1` adds a component / resource
- `adding system_2` adds the same component / resource
- verify the `Mutated` flag at the end of the tick
- verify the `Added` flag at the end of the tick
First check tests for: https://github.com/bevyengine/bevy/issues/333
Second check tests for: https://github.com/bevyengine/bevy/issues/1443
## Changes Made By Commands
- `adding_system` runs in Update in tick 1, and sends a command to add a component
- `detecting_system` runs in Update in tick 1 and 2, after `adding_system`
- We can't detect the changes in tick 1, since they haven't been processed yet
- If we were to track these changes as being emitted by `adding_system`, we can't detect the changes in tick 2 either, since `detecting_system` has already run once after `adding_system` :(
# Benchmarks
See: [general advice](https://github.com/bevyengine/bevy/blob/master/docs/profiling.md), [Criterion crate](https://github.com/bheisler/criterion.rs)
There are several critical parameters to vary:
1. entity count (1 to 10^9)
2. fraction of entities that are changed (0% to 100%)
3. cost to perform work on changed entities, i.e. workload (1 ns to 1s)
1 and 2 should be varied between benchmark runs. 3 can be added on computationally.
We want to measure:
- memory cost
- run time
We should collect these measurements across several frames (100?) to reduce bootup effects and accurately measure the mean, variance and drift.
Entity-component change detection is much more important to benchmark than resource change detection, due to the orders of magnitude higher number of pieces of data.
No change detection at all should be included in benchmarks as a second control for cases where missing changes is unacceptable.
## Graphs
1. y: performance, x: log_10(entity count), color: proposal, facet: performance metric. Set cost to perform work to 0.
2. y: run time, x: cost to perform work, color: proposal, facet: fraction changed. Set number of entities to 10^6
3. y: memory, x: frames, color: proposal
# Conclusions
1. Is the theoretical categorization of the proposals correct according to our tests?
2. How does the performance of the proposals compare without any load?
3. How does the performance of the proposals compare with realistic loads?
4. At what workload does more exact change tracking become worth the (presumably) higher overhead?
5. When does adding change-detection to save on work become worthwhile?
6. Is there enough divergence in performance between the best solutions in each class to ship more than one change-tracking solution?
# Implementation Plan
1. Write a test suite.
2. Verify that tests fail for existing approach.
3. Write a benchmark suite.
4. Get performance numbers for existing approach.
5. Implement, test and benchmark various solutions using a Git branch per proposal.
6. Create a draft PR with all solutions and present results to team.
7. Select a solution and replace existing change detection.
Co-authored-by: Brice DAVIER <bricedavier@gmail.com>
Co-authored-by: Carter Anderson <mcanders1@gmail.com>
Removes `get_unchecked` and `get_unchecked_mut` from `Tables` and `Archetypes` collections in favor of safe Index implementations. This fixes a safety error in `Archetypes::get_id_or_insert()` (which previously relied on TableId being valid to be safe ... the alternative was to make that method unsafe too). It also cuts down on a lot of unsafe and makes the code easier to look at. I'm not sure what changed since the last benchmark, but these numbers are more favorable than my last tests of similar changes. I didn't include the Components collection as those severely killed perf last time I tried. But this does inspire me to try again (just in a separate pr)!
Note that the `simple_insert/bevy_unbatched` benchmark fluctuates a lot on both branches (this was also true for prior versions of bevy). It seems like the allocator has more variance for many small allocations. And `sparse_frag_iter/bevy` operates on such a small scale that 10% fluctuations are common.
Some benches do take a small hit here, but I personally think its worth it.
This also fixes a safety error in Query::for_each_mut, which needed to mutably borrow Query (aaahh!).
