constellation

Crates.ioconstellation
lib.rsconstellation
version0.2.0
sourcesrc
created_at2016-12-24 05:26:41.638714
updated_at2017-01-09 16:30:09.329454
descriptionA data-oriented entity component system optimized for cache coherent resource access and parallel system execution.
homepage
repositoryhttps://github.com/TomGillen/constellation/
max_upload_size
id7753
size111,302
Thomas Gillen (TomGillen)

documentation

https://docs.rs/constellation

README

Constellation ECS

A data-oriented entity component system optimized for cache coherent resource access and parallel system execution.

Constellation does not have any native understanding of a "component". Instead, the library is concerned about ensuring safe concurrent access to shared resources, while providing direct access to each resource's own APIs. These resources may store per-entity data such as positions, or may represent application wide services such as asset loaders or input devices.

Systems request read or write access to a set of resources, which can then be scheduled to be potentially executed in parallel by recording them into a SystemCommandBuffer and executing the command buffer within a World.

Crates.io
Documentation

Similar Projects

Examples

Defining Resources:

// Per-entity position data.
struct Position {
    x: f32,
    y: f32,
    z: f32
}

// Store position data into a vector resource
type Positions = VecResource<Position>;

// Per-entity debug names.
struct DebugName {
    name: String
}

// Store debug names in a map resource
type DebugNames = MapResource<DebugName>;

let mut world = World::new();
world.register_resource(Positions::new());
world.register_resource(DebugNames::new());

Update the world with Systems:

let mut update = SystemCommandBuffer::default();
update.queue_systems(|scope| {
    scope.run_r1w1(|ctx, velocities: &Velocities, positions: &mut Positions| {
        println!("Updating positions");
        // iterate through all components for entities with data in both
        // position and velocity resources
        ctx.iter_r1w1(velocities, positions).components(|entity, v, p| {
                p.x += v.x;
                p.y += v.y;
                p.z += v.z;
        });
    });

    scope.run_r2w0(|ctx, names: &DebugNames, positions: &Positions| {
        println!("Printing positions");
        // iterate through all entity IDs for entities with data in both
        // `names` and `positions`
        ctx.iter_r2w0(names, positions).entities(|entity_iter, n, p| {
            // `n` and `p` allow (potentially mutable) access to entity data inside
            // the resource without the ability to add or remove entities from the resource
            // - which would otherwise invalidate the iterator
            for e in entity_iter {
                println!("Entity {} is at {:?}",
                         n.get(e).unwrap().name,
                         p.get(e).unwrap());
            }
        });
    });
});
world.run(&mut update);

Parallel System Execution

Systems queued into a command buffer within a single call to queue_systems may be executed in parallel by the world.

The order in which systems are queued is significant in one way: the scheduler guarantees that any changes to resources will always be observed in the same order in which systems were queued.

For example, given two systems - ReadPositions and WritePositions - if WritePositions was queued before ReadPositions, then it is guarenteed that ReadPositions will see any changes made by WritePositions. Conversely, if the order were to be swapped, then ReadPositions is guaranteed to not observe the changes made by WritePositions.

There is one exception to this. Entity deletions are committed when all concurrently executing systems have completed. This behavior is deterministic, but not always obvious. If you wish to ensure that entity deletions from one system are always seen by a later system, then queue the two systems in separate queue_systems calls.

Commit count: 34

cargo fmt