Benchmarks

Test-drive real rendering behavior in the companion app:
Live bench → https://acoolhq.github.io/rts-bench/

• RTS (store + selector) — baseline makeStore + useStoreSelector.
• RTS + Batching — same store, but updates wrapped in batch to coalesce bursts triggered by controllers.
• RTS + Context (createContextSync) — provider + selectors built from the library.
• Context + useState — naïve context that re-renders every leaf on any change.
• Context + useReducer — similar baseline using a reducer.

All variants end up with the same final selector value—React will bail if nothing changed—but wall time diverges because every store update still emits a notification. The batching row shows how collapsing those notify → read → compare cycles into one pass keeps the UI just as stable while spending far less time doing redundant work.

Scenarios you can trigger in the bench:

• Unrelated updates — change/hammer a field unrelated to the items multiple times
• Single index updates — repeatedly bump one items[i].
• Random churn — touch random indices many times.

Run the bench in production mode for realistic timings.