Bevy ECS Query Execution Order and Determinism

Query Iteration Order in Bevy ECS

Bevy's ECS stores component data in archetypes, which are contiguous memory layouts matching component combinations. A query iterates over every archetype that matches its component filter, yielding entities in the order those components appear in the archetype's internal arrays. This order is not a globally stable entity sequence; it reflects the archetype's current layout, which the ECS rearranges whenever entities are spawned, despawned, or move between archetypes by adding or removing components.

Confirmed facts versus likely explanation

• Confirmed: The query iterator walks archetypes sequentially and yields entities from each archetype's storage order. Bevy does not provide a built-in flag or schedule setting that makes this iteration deterministic across frames or runs.
• Likely explanation: Because archetype layouts shift whenever component membership changes, the position of a given entity within the iterator can change from frame to frame, especially in scenes with frequent spawn/despawn or component addition/removal.

System execution order, controlled by the Bevy schedule, determines when systems run relative to one another, but it does not influence the internal ordering of entities within a query's iterator. Parallel system execution can overlap query processing, yet within any single system the iterator remains sequential and follows the archetype layout at the moment the system executes.

Minimal steps for this case

1. Run the same system across multiple frames with controlled entity churn (spawn, despawn, add/remove components). Log the entity IDs returned by the query each frame.
2. If the sequence changes, the iteration order is unstable for your use case.
3. To obtain deterministic ordering, collect the query's entities into a collection, then sort by a stable key—such as a custom index component, entity generation bits, or a frame‑assigned sequence number.
4. Verify the sort result stays consistent across repeated runs on the same ECS state.

Bevy's query API and archetype internals have evolved between major releases. The behavior described here applies to the version you are using; earlier or later versions may expose slightly different iteration mechanics or additional API flags.

If you can provide the Bevy version and whether you have noticed order changes under typical component churn, I can refine the sorting strategy or check for version‑specific ordering guarantees.