Entity Component System as a Relational Model: Archetypes, Component-Centric Relations, and Normal Forms

The Entity Component System (ECS) is a data-oriented software architectural pattern for the efficient organization of entities and associated components and is used in particular in performance-critical applications such as video games and real-time simulations.In the literature, common storage models of ECS systems are frequently associated with relational data structures; an explicit normal-form-theoretic classification, however, is rarely found.This paper therefore examines two widespread ECS storage organizations from the perspective of the relational model: the archetype-based approach and a column-oriented organization that is often implemented in practice using Sparse Sets and that is referred to here, in an implementation-agnostic manner, as component-centric.It is shown that, under the model assumptions made, archetype-based relations satisfy the 1st through 5th normal forms as well as Boyce-Codd Normal Form.Since archetype-based relations generally exhibit nontrivial join dependencies, their vertical decomposition into irreducible, component-centric relations that satisfy the 6th normal form is subsequently considered.Based on established concepts of database theory as well as current ECS research, the paper shows that relational notions such as keys, dependencies, and irreducible decomposition are suitable for formally describing central differences between archetype-based and component-centric storage organizations;at the same time, analogies to other application areas of database theory become apparent, such as table partitioning and Anchor Modeling.The relational ECS model presented here thus provides a foundation for classifying design and modeling decisions in ECS systems, formulating structural and runtime-related invariants for ensuring consistency, and describing optimization approaches for query evaluation.