System

In one line: a unit of logic with declared data access — you say what it reads and writes, the engine works out what can run at the same time.

A system is a class you derive from one of three bases and give two methods: Configure (declare it — Name, Phase, Input, Reads/Writes) and Execute (run it against a TickContext). The runtime gives each system its own transaction per tick, created on its worker thread and committed by the scheduler — you never call Commit.

Three shapes, picked by what the work looks like: CallbackSystem (non-entity work — input, timers, spawning), QuerySystem (do something to every entity in a view — the workhorse, parallelisable), PipelineSystem (bulk data-parallel sweeps). The declared access is what lets the scheduler parallelise safely.

How it relates

• Runtime — runs your systems every tick, in parallel.
• Scheduler & phases — turns declared access into a safe execution graph.
• View — a QuerySystem's input set.
• Transaction / PointInTimeAccessor — how a system touches data (transactional, or lock-free parallel reads).
• TickContext — the per-tick object Execute receives (transaction, accessor, entities, side-transactions).

In the API

• CallbackSystem · QuerySystem · PipelineSystem — the three bases.
• SystemBuilder — the Configure builder (Reads/Writes/Input/Parallel/…).
• TickContext — what Execute receives (Transaction / Accessor / Entities / DeltaTime).

Learn & use

• Narrative: Guide ch.5 — systems & the tick loop
• Feature detail: declarative system scheduling · system types