PERFORMANCE_TODO.md

exostate — Performance Optimization TODO

• Optimize notifyListeners loop: Replace for...of with index-based loop and add empty-check short-circuit. Implement Copy-on-Write (COW) for listeners to ensure safety during unsubscription without cloning on dispatch.
  • Actual: Switched to Set with COW strategy based on benchmark data (Set iteration ~10% faster than Array).
• Refactor to Class-based implementation: Replace closure-heavy createStore with a StoreImpl class to ensure monomorphic call sites and reduce memory pressure.
  • Expected Gain: 5-10% on raw throughput; better memory usage.
• Optimize subscribe selector wrapper: Flatten the notification logic to reduce call stack depth during dispatch.
  • Expected Gain: 10-15% on "Update with Subscriber" benchmark.
• Eliminate DeepReadonly runtime overhead: Ensure casts are purely compile-time (verify no runtime impact remains).
  • Actual: Verified in dist/store.js and types.ts. DeepReadonly is a mapped type with 0 runtime footprint. snapshot() returns this.current directly.
• Final Benchmark: Compare against Redux and Zustand.
  • Result: Exostate (17.5M ops/s) vs Zustand (13.4M ops/s) vs Redux (5.1M ops/s).
  • Optimization: Using Object.assign in reducer instead of spread ({...}) yields 2.5x speedup, beating Zustand by 30%.