README.md

Massive ECS - sparse set ECS with rollbacks

Made for games with deterministic prediction-rollback netcode. Currently distributed as Unity package.

Prediction-rollback netcode has very stable nature, and is mainly used in fast paced online multiplayer games, such as Overwatch and Rocket League.

Overview

This is a library, not a framework. Thus, it does not try to take control of the user codebase or the main game loop.

ECS features:

• Fast and simple ECS without any code generation
• Support сomponents of any type
• No deferred commands execution
• Garbage-free API. Make in-place queries, no caching required
• Groups for SoA multi-component iteration (inspired by EnTT)
• Data pagination for stable resizing during iteration
• Full state serialization and deserialization
• IL2CPP friendly, tested with high stripping level on PC | Android | WebGL
• Unity integration (WIP)

Rollback features:

• This is an extension for ECS and is completely optional to use
• It's natively fast with Array.Copy and cyclic buffers
• Minimalistic API - SaveFrame() and Rollback(frames)
• Support for components with managed data, such as arrays, strings, etc. (see the wiki)

Consider this list a work in progress as well as the project.

Code Examples

struct Player { }
struct Position { public float X; public float Y; }
class Velocity { public float Magnitude; } // Classes work just fine
delegate void ShootingMethod(); // So are the delegates
interface IDontEvenAsk { }

class Program
{
	static void Main()
	{
		var registry = new Registry();

		// Create empty entity
		var enemy = registry.Create();

		// Or with a component
		var player = registry.Create(new Player());

		// Assign components
		registry.Assign(player, new Velocity() { Magnitude = 10f });
		registry.Assign(enemy, new Velocity());
		registry.Assign<Position>(enemy); // Assigns component without initialization

		// Get full entity identifier from player ID.
		// Handy when uniqueness is required, for example, when storing entities for later
		Entity playerEntity = registry.GetEntity(player);

		var deltaTime = 1f / 60f;

		// Iterate using lightweight views
		var view = registry.View();
		view.ForEach((int entityId, ref Position position, ref Velocity velocity) =>
		{
			position.Y += velocity.Magnitude * deltaTime;
			if (position.Y > 5f)
			{
				// Create and destroy any amount of entities during iteration
				registry.Destroy(entityId);
			}
		});

		// Pass extra arguments to avoid boxing
		view.ForEachExtra((registry, deltaTime),
			(ref Position position, ref Velocity velocity,
				(Registry Registry, float DeltaTime) args) =>
			{
				// ...
			});

		// Make queries right in place where they are used.
		// You don't have to cache anything
		registry.View()
			.Filter<Include<Player>, Exclude<Velocity>>()
			.ForEach((ref Position position) =>
			{
				// ...
			});

		// Iterate using foreach
		foreach (var entityId in registry.View().Include<Player, Position>())
		{
			ref var position = ref registry.Get<Position>(entityId);
			// ...
		}

		// Iterate manually over data sets
		var velocities = registry.DataSet<Velocity>();
		for (int i = 0; i < velocities.Count; ++i)
		{
			ref var velocity = ref velocities.Data[i];
			// ...
		}

		// Chain any amount of components in filters
		var filter = registry.Filter<
			Include<int, string, bool, Include<short, byte, uint, Include<ushort>>>,
			Exclude<long, char, float, Exclude<double>>>();

		// Reuse filter variable to reduce code duplication
		// in case of multiple iterations
		registry.View().Filter(filter).ForEach((ref int n, ref bool b) => { });
		registry.View().Filter(filter).ForEach((ref string str) => { });
	}
}

Note

Some APIs are subject to change, but overall the architecture is stable.

Installation

Make sure you have standalone Git installed first. Reboot after installation.
In Unity, open "Window" -> "Package Manager".
Click the "+" sign at the top left corner -> "Add package from git URL..."
Paste this: https://github.com/nilpunch/massive-ecs.git
See minimum required Unity version in the package.json file.

How it works

Each Massive data structure contains cyclic buffer in linear memory. This allows for very fast saving and rollbacking, copying the entire data arrays at once. MassiveRegistry simply uses these Massive data structures internally, so we get the simplest possible ECS with rollbacks.