Optimize query caching, system operations, and component indexing

[Copilot]

Four targeted optimizations for hot paths in the ECS runtime:

Changes

• Query key/mask caching: Memoize query keys (WeakMap) and masks (Map) to eliminate O(n log n) sorting on repeated queries. Queries now check cache first instead of regenerating.

• System enable/disable: Replace O(n) indexOf with O(1) Map-based index lookup. Added _enabledSystemsIndex: Map<System, number> maintained during enable/disable operations.

• Component gathering deduplication: Extract duplicated component collection logic (4 sites) into single _gatherComponents() helper function.

• Component-to-entity index: Maintain Map<ComponentType, Set<Entity>> to skip irrelevant entities in single-component queries. pickOne() now iterates O(matching entities) instead of O(all entities).

Example

// Before: Query key regenerated with sort every call
function _getQueryKey(types: ComponentType[]): string {
    const ids = new Array<number>(length);
    for (let i = 0; i < length; i += 1) { ids[i] = types[i].Id; }
    ids.sort((a, b) => (a - b));
    return ids.join(",");
}

// After: Check cache first, regenerate only on miss
const cached = _queryKeyCache.get(types);
if (cached !== undefined) { return cached; }

All existing tests pass. Added performance benchmarks validating improvements (1000 queries < 100ms, 1000 system toggles < 50ms).

---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.