Benchmarks Overview

Objective

Compare ngStato against NgRx v21+ on realistic workloads with transparent methodology.

Known metrics

Bundle size

Package	Gzipped	Raw
@ngstato/core	~3 KB	~12 KB
@ngstato/angular	~1 KB	~4 KB
@ngstato/testing	< 1 KB	~3 KB
ngStato total	~4 KB	~19 KB
@ngrx/store	~8 KB	~30 KB
@ngrx/signals	~12 KB	~45 KB
@ngrx/effects	~5 KB	~18 KB
@ngrx/entity	~4 KB	~15 KB
rxjs (required)	~25 KB	~90 KB
NgRx typical total	~50 KB	~200 KB

ngStato is ~12x lighter when comparing equivalent feature sets.

Lines of code (CRUD store)

A complete CRUD feature store with 5 actions (load, create, update, delete, select), loading/error states, and entity management:

	NgRx SignalStore v21	ngStato
State definition	8 lines	5 lines
Actions (5 CRUD)	~55 lines	~30 lines
Selectors	12 lines	8 lines
Effects/hooks	15 lines	5 lines
Total	~90 lines	~48 lines
Concepts required	9+ (rxMethod, pipe, tap, switchMap, from, tapResponse, patchState, withState, withMethods)	1 (async/await)

Feature comparison

Feature	NgRx v21	ngStato v0.4
State management	✅	✅
Actions (methods)	✅ rxMethod	✅ async/await
Computed/Selectors	✅	✅ (memoized)
Entity adapter	✅	✅ createEntityAdapter + withEntities
Concurrency helpers	✅ via RxJS	✅ native (exclusive, queued, abortable)
Stream integration	✅ RxJS required	✅ RxJS optional (fromStream)
DevTools	✅ Chrome extension	✅ Built-in panel, all browsers
Time-travel	✅ via extension	✅ Built-in with fork-on-dispatch
Action replay	❌	✅ Re-execute any logged action
State export/import	Via extension	✅ JSON file for bug reports
Mobile DevTools	❌	✅
Production safety	⚠️ Manual logOnly	✅ Automatic isDevMode()
Testing utilities	✅ provideMockStore	✅ createMockStore()
Feature composition	✅ signalStoreFeature	✅ mergeFeatures()
Service injection	✅ withProps	✅ withProps() + closures
Inter-store reactions	✅ via Effects	✅ on() multi-action
Persistence	❌ (meta-reducers)	✅ withPersist() built-in
Schematics CLI	✅ ng generate	✅ @ngstato/schematics
ESLint plugin	✅ @ngrx/eslint-plugin	✅ @ngstato/eslint-plugin

Developer experience

Metric	NgRx	ngStato
Time to first store	~15 min	~3 min
Files per feature	3–5 (reducer, actions, effects, selectors, facade)	1
Angular boilerplate	provideMockStore, provideStoreDevtools, etc.	provideStato() — one call
Scaffolding	ng generate @ngrx/schematics:store	ng generate @ngstato/schematics:store
Linting	@ngrx/eslint-plugin	@ngstato/eslint-plugin

Benchmark scenarios (planned)

Scenario 1: CRUD Feature Store

• Create entity adapter with 1000 items
• Perform: load all → create 50 → update 20 → delete 10
• Measure: action latency, memory usage, selector recomputation count

Scenario 2: Search with debounce + cancellation

• Type-ahead search with 300ms debounce
• 50 rapid keystrokes
• Measure: cancelled requests, final result latency, bundle cost

Scenario 3: Concurrent async requests

• 10 concurrent API calls with different strategies
• Compare: exclusive vs NgRx exhaustMap, queued vs concatMap
• Measure: correct behavior, timing, memory

Scenario 4: Large entity updates

• 10,000 entities in normalized state
• Batch update 500 entities
• Measure: update latency, selector recomputation, memory delta

Methodology

• Same backend — identical API contracts
• Same UI behavior — identical acceptance criteria
• Same tooling — Vitest for benchmarks, Chrome DevTools for profiling
• Reproducible — scripts and fixtures in benchmarks/ directory
• Transparent — raw data published alongside results

Status

Initial benchmark suite is in preparation. Results will be published with full methodology and raw data.

For now, the known metrics above (bundle size, lines of code, feature comparison) provide a solid basis for evaluation.