Managing Dart Concurrency and Isolates

Contents

• Core Concepts
• Decision Matrix: Async vs. Isolates
• Workflows
  • Implementing Standard Asynchronous UI
  • Offloading Short-Lived Heavy Computation
  • Establishing Long-Lived Worker Isolates
• Examples

Core Concepts

Dart utilizes a single-threaded execution model driven by an Event Loop (comparable to the iOS main loop). By default, all Flutter application code runs on the Main Isolate.

• Asynchronous Operations (async/await): Use for non-blocking I/O tasks (network requests, file access). The Event Loop continues processing other events while waiting for the Future to complete.
• Isolates: Dart's implementation of lightweight threads. Isolates possess their own isolated memory and do not share state. They communicate exclusively via message passing.
• Main Isolate: The default thread where UI rendering and event handling occur. Blocking this isolate causes UI freezing (jank).
• Worker Isolate: A spawned isolate used to offload CPU-bound tasks (e.g., decoding large JSON blobs) to prevent Main Isolate blockage.

Decision Matrix: Async vs. Isolates

Apply the following conditional logic to determine the correct concurrency approach:

• If the task is I/O bound (e.g., HTTP request, database read) -> Use async/await on the Main Isolate.
• If the task is CPU-bound but executes quickly (< 16ms) -> Use async/await on the Main Isolate.
• If the task is CPU-bound, takes significant time, and runs once (e.g., parsing a massive JSON payload) -> Use Isolate.run().
• If the task requires continuous or repeated background processing with multiple messages passed over time -> Use Isolate.spawn() with ReceivePort and SendPort.

Workflows

Implementing Standard Asynchronous UI

Use this workflow to fetch and display non-blocking asynchronous data.

Task Progress:

• Mark the data-fetching function with the async keyword.
• Return a Future<T> from the function.
• Use the await keyword to yield execution until the operation completes.
• Wrap the UI component in a FutureBuilder<T> (or StreamBuilder for streams).
• Handle ConnectionState.waiting, hasError, and hasData states within the builder.
• Run validator -> review UI for loading indicators -> fix missing states.

Offloading Short-Lived Heavy Computation

Use this workflow for one-off, CPU-intensive tasks using Dart 2.19+.

Task Progress:

• Identify the CPU-bound operation blocking the Main Isolate.
• Extract the computation into a standalone callback function.
• Ensure the callback function signature accepts exactly one required, unnamed argument (as per specific architectural constraints).
• Invoke Isolate.run() passing the callback.
• await the result of Isolate.run() in the Main Isolate.
• Assign the returned value to the application state.

Establishing Long-Lived Worker Isolates

Use this workflow for persistent background processes requiring continuous bidirectional communication.

Task Progress:

• Instantiate a ReceivePort on the Main Isolate to listen for messages.
• Spawn the worker isolate using Isolate.spawn(), passing the ReceivePort.sendPort as the initial message.
• In the worker isolate, instantiate its own ReceivePort.
• Send the worker's SendPort back to the Main Isolate via the initial port.
• Store the worker's SendPort in the Main Isolate for future message dispatching.
• Implement listeners on both ReceivePort instances to handle incoming messages.
• Run validator -> review memory leaks -> ensure ports are closed when the isolate is no longer needed.

Examples

Example 1: Asynchronous UI with FutureBuilder

// 1. Define the async operation
Future<String> fetchUserData() async {
  await Future.delayed(const Duration(seconds: 2)); // Simulate network I/O
  return "User Data Loaded";
}

// 2. Consume in the UI
Widget build(BuildContext context) {
  return FutureBuilder<String>(
    future: fetchUserData(),
    builder: (context, snapshot) {
      if (snapshot.connectionState == ConnectionState.waiting) {
        return const CircularProgressIndicator();
      } else if (snapshot.hasError) {
        return Text('Error: ${snapshot.error}');
      } else {
        return Text('Result: ${snapshot.data}');
      }
    },
  );
}

Example 2: Short-Lived Isolate (Isolate.run)

import 'dart:isolate';
import 'dart:convert';

// 1. Define the heavy computation callback
// Note: Adhering to the strict single-argument signature requirement.
List<dynamic> decodeHeavyJson(String jsonString) {
  return jsonDecode(jsonString) as List<dynamic>;
}

// 2. Offload to a worker isolate
Future<List<dynamic>> processDataInBackground(String rawJson) async {
  // Isolate.run spawns the isolate, runs the computation, returns the value, and exits.
  final result = await Isolate.run(() => decodeHeavyJson(rawJson));
  return result;
}

Example 3: Long-Lived Isolate (ReceivePort / SendPort)

import 'dart:isolate';

class WorkerManager {
  late SendPort _workerSendPort;
  final ReceivePort _mainReceivePort = ReceivePort();
  Isolate? _isolate;

  Future<void> initialize() async {
    // 1. Spawn isolate and pass the Main Isolate's SendPort
    _isolate = await Isolate.spawn(_workerEntry, _mainReceivePort.sendPort);

    // 2. Listen for messages from the Worker Isolate
    _mainReceivePort.listen((message) {
      if (message is SendPort) {
        // First message is the Worker's SendPort
        _workerSendPort = message;
        _startCommunication();
      } else {
        // Subsequent messages are data payloads
        print('Main Isolate received: $message');
      }
    });
  }

  void _startCommunication() {
    // Send data to the worker
    _workerSendPort.send("Process this data");
  }

  // 3. Worker Isolate Entry Point
  static void _workerEntry(SendPort mainSendPort) {
    final workerReceivePort = ReceivePort();
    
    // Send the Worker's SendPort back to the Main Isolate
    mainSendPort.send(workerReceivePort.sendPort);

    // Listen for incoming tasks
    workerReceivePort.listen((message) {
      print('Worker Isolate received: $message');
      
      // Perform work and send result back
      final result = "Processed: $message";
      mainSendPort.send(result);
    });
  }

  void dispose() {
    _mainReceivePort.close();
    _isolate?.kill();
  }
}