an_reactive_state

An industrial-grade, call-stack synchronous reactive graph engine (Signals) for pure Dart and Flutter, built for ultimate evaluation performance and absolute memory safety.

an_reactive_state discards the heavy macro-task/micro-task scheduling overhead found in traditional asynchronous stream models (such as Stream or Kotlin Flow). Instead, it runs entirely on an ultra-fast CPU Call-Stack synchronous reactive topology. Deeply integrated with the Cancellable lifecycle tree, it fundamentally eliminates async context interleaving bugs and memory leaks.

🌟 Key Killer Features

  • ⚡ Explicit State Roles: Separates reactive nodes into RState<T> (writable atoms) and ComputedState<T> (read-only derivations). This architecture ensures strict data flow control and prevents accidental writes to calculated nodes at the compiler level.
  • ⛓️ Linked-List Evaluation Context: Replaces expensive dynamic array reallocations with a lightweight, doubly-linked-list node switching architecture. PUSH/POP operations for reactive tracking run at a strict, deterministic O(1) time complexity.
  • 🛡️ Zone-Safe Async Isolation: Leverages Dart's native Zone local storage mechanism. When multiple asynchronous tasks interleave and modify states concurrently, the dependency tracking and evaluation scope remain strictly isolated per execution branch, preventing global state pollution.
  • 🛑 Zero-Leak Lifecycle Self-Destruction: Fully bound to the Cancellable token tree. When a host container is disposed of, all parent-child listener registrations are permanently purged within a single atomic microsecond, leaving zero memory overhead.
  • 🚫 Circular Dependency Watchdog: Before entering the call-stack evaluation domain, it crawls upwards from the current node to check its ancestors. If a dependency loop is detected, it throws a transparent StateError immediately, eliminating any risk of Stack Overflow crashes.
  • 🔍 Lazy Evaluation & Memoization: Implements a highly efficient "Dirty Propagation" mechanism. Derived computation states do not calculate values eagerly upon initialization. They run only when a downstream node or listener explicitly invokes get value or addListener, caching the output until upstreams flag it as dirty again.
  • ⚖️ Equality Guard: Supports custom equality comparers. Even if your calculation closure generates a brand-new instance (such as a fresh List or object mapping), it will block downstream propagation if the content remains identical, completely terminating downstream evaluation ripples.
  • 📦 Synchronous Batching & Untracking: Provides global batch and untracked closures. You can execute multiple rapid updates inside a batch block while ensuring downstream effects are triggered exactly once, executing synchronously right before exiting the block—with no microtask latency.

📦 Installation

Add the following dependencies to your pubspec.yaml:

dependencies:
  an_reactive_state: ^1.0.0
  cancellable: ^2.6.0     # The foundation of industrial lifecycle management

🚀 Core Usage Guide

1. Basic Read/Write, Derived State, and Side Effects (effect)

In an_reactive_state, raw state sources and high-order calculation pipelines share a uniform interface.

import 'package:cancellable/cancellable.dart';
import 'package:an_reactive_state/an_reactive_state.dart';

void main() {
  final scope = Cancellable();

  // 1. Declare raw state atoms (Sources)
  final price = RState<double>(initialValue: 99.0, cancellable: scope);
  final count = RState<int>(initialValue: 1, cancellable: scope);

  // 2. Declare a high-order derived calculation state (Computed)
  final totalPrice = ComputedState<double>(
    cancellable: scope,
    computer: () => price.value * count.value, // Automatically tracks price and count
  );

  // 3. Register a side effect: whatever reactive states it reads inside the block,
  // it will automatically re-run whenever those states change.
  effect(() {
    print('【UI Effect Card】Current total checkout amount: \${totalPrice.value}');
  }, cancellable: scope);

  print('--- Modifying Quantity ---');
  count.value = 3; // Executed synchronously inside the current call stack

  scope.cancel(); // Everything self-destructs; subsequent writes are safely blocked.
}

2. Postponing Notifications (batch) & Reading Without Tracking (untracked)

Wrap multiple independent value adjustments inside a batch to prevent computational jitter. Use untracked to pull snapshots of data safely without contaminating the active evaluation graph.

void main() {
  final scope = Cancellable();

  final width = RState<int>(initialValue: 10, cancellable: scope);
  final height = RState<int>(initialValue: 5, cancellable: scope);
  final isLogging = RState<bool>(initialValue: true, cancellable: scope);

  final areaReport = ComputedState<String>(
    cancellable: scope,
    computer: () {
      // Core control: Read the switch silently inside an untracked closure.
      // This ensures areaReport does NOT establish a dependency on isLogging!
      final logActive = untracked(() => isLogging.value);
      if (logActive) {
        untracked(() => print('【System Internal】Evaluating reactive matrix area...'));
      }
      return 'Current Area: \${width.value * height.value} sq.m';
    },
  );

  areaReport.addListener(() => print('【External Screen】Updated -> \${areaReport.value}'));

  print('\n--- Scenario: Triggering Batch Updates ---');
  batch(() {
    print('(Starting sequential assignments inside batch...)');
    width.value = 20; // Upstream flags change, downstream evaluation holds
    height.value = 10; // Upstream flags change, downstream evaluation holds
    print('(Exiting batch scope shortly)');
  });
  // Downstream triggers EXACTLY ONCE on the call-stack right at the block exit!

  scope.cancel();
}

🏗️ Domain Lifecycle Management

In true production-grade architectures, wrapping your reactive states inside a manageable domain controller gives you a clean state tree with absolute memory safety.

import 'package:cancellable/cancellable.dart';
import 'package:an_reactive_state/an_reactive_state.dart';

class OrderCartManager {
  // Establish a single master cancellable token for this business module
  final Cancellable _bag = Cancellable();

  late final RState<double> productPrice;
  late final RState<int> quantity;
  late final ComputedState<double> finalPay;

  OrderCartManager() {
    productPrice = RState<double>(initialValue: 15.0, cancellable: _bag);
    quantity = RState<int>(initialValue: 1, cancellable: _bag);

    finalPay = ComputedState<double>(
      cancellable: _bag,
      // Inject Equality Guard to ensure content matching blocks computational duplication
      equals: (a, b) => a == b,
      computer: () => productPrice.value * quantity.value,
    );

    // Bind automated domain side-effects
    effect(() {
      print('【Domain Event Report】Current final payment amount flowed to: \${finalPay.value}');
    }, cancellable: _bag);
  }

  void addItems(int countToAdd) {
    quantity.value += countToAdd;
  }

  // Explicitly wipe bindings completely when this manager clears up
  void dispose() {
    print('【Architecture Lifecycle】Manager is clearing; triggering global state graph flush!');
    _bag.cancel(); // All map registries and closures are cleared in 0μs, guaranteeing 0 leaks.
  }
}

🛠️ Functional Advanced Trick: Freeze / Once Memoization

You do not even need to extend the internal methods of ComputedState. Thanks to our highly sensitive Dual-Buffered Graph Diffing algorithm, a developer can write a self-rewriting closure inside the computer to achieve an auto-detaching one-time snapshot that calculates only once, breaks all parent connections, and permanently freezes itself:


final ComputedState<String> lazyAndFrozenConfig = ComputedState<String>(
  cancellable: scope,
  computer: () {
    String? snapshot;
    return () {
      if (snapshot != null) {
        // Ensuing access pathways drop into the memoized cache branch.
        // Because no upstreams (like remoteIp.value) are invoked here,
        // our dynamic graph diff engine automatically realizes those parent dependencies 
        // are no longer needed, permanently severing listeners from the parent list!
        return untracked(() => snapshot!);
      }
      print('【Extreme Optimization】Executing heavy configuration parsing exactly once...');
      snapshot = 'Server Endpoint: IP=\${remoteIp.value}, PORT=\${remotePort.value}';
      return snapshot!;
    };
  }(), // Immediately run the outer closure to bind the self-detaching inner function
);

Issues

If you encounter issues, here are some tips for debug, if nothing helps report to issue tracker on GitHub:

Libraries

an_reactive_state