Force asynchronous code to run synchronously dart

Question

I currently have a class that stores images after fetching images asynchronously over GRPC. There is an issue with the handling of events. The problem can be shown here:

import 'dart:async';

main() {
  IntStore intStore = IntStore();

  print("running getInt(A)");
  intStore.getInt("A");

  print("running getInt(A)");
  intStore.getInt("A");

  print("running getInt(B)");
  intStore.getInt("B");

  print("running getInt(A)");
  intStore.getInt("A");

  print("running getInt(C)");
  intStore.getInt("C");

  print("running getInt(D)");
  intStore.getInt("D");
}

class IntStore {
  final Map _store = <String, int>{};
  Future fetchInt(String intName) async {
    print("Fetching: $intName");
    await doSomeWorkAsynchronously(intName);
  }

  Future<int> getInt(String intName) async {
    if (_store.containsKey(intName)) {
      print("Cached: $intName");
      return _store[intName];
    } else {
      await fetchInt(intName);
      return _store[intName];
    }
  }

  Future doSomeWorkAsynchronously(String intName) async {
    await Future.delayed(const Duration(seconds: 3));
    _store[intName] = 3;
    print("Fetched: $intName");
  }
}

which returns:

running getInt(A)
Fetching: A
running getInt(A)
Fetching: A
running getInt(B)
Fetching: B
running getInt(A)
Fetching: A
running getInt(C)
Fetching: C
running getInt(D)
Fetching: D
Fetched: A
Fetched: A
Fetched: B
Fetched: A
Fetched: C
Fetched: D

The problem here is that the work in fetchInt is completed multiple times. This is quite inefficient.

Answer 1

I found Irn's answer in this question helpful. I implemented this to obtain:

import 'dart:async';

main() {
  IntStore intStore = IntStore();
  AsyncMutex mutex = AsyncMutex();

  print("running getInt(A)");
  mutex.run(() => intStore.getInt("A"));

  print("running getInt(A)");
  mutex.run(() => intStore.getInt("A"));

  print("running getInt(B)");
  mutex.run(() => intStore.getInt("B"));

  print("running getInt(A)");
  mutex.run(() => intStore.getInt("A"));

  print("running getInt(C)");
  mutex.run(() => intStore.getInt("C"));

  print("running getInt(D)");
  mutex.run(() => intStore.getInt("D"));
}

class IntStore {
  final Map _store = <String, int>{};

  Future fetchInt(String intName) async {
    print("Fetching: $intName");
    await doSomeWorkAsynchronously(intName);
  }

  Future<int> getInt(String intName) async {
    if (_store.containsKey(intName)) {
      print("Cached: $intName");
      return _store[intName];
    } else {
      await fetchInt(intName);
      return _store[intName];
    }
  }

  Future doSomeWorkAsynchronously(String intName) async {
    await Future.delayed(const Duration(seconds: 3));
    _store[intName] = 3;
    print("Fetched: $intName");
  }
}

class AsyncMutex {
  Future _next = new Future.value(null);

  /// Request [operation] to be run exclusively.
  ///
  /// Waits for all previously requested operations to complete,
  /// then runs the operation and completes the returned future with the
  /// result.
  Future<T> run<T>(Future<T> operation()) {
    var completer = new Completer<T>();
    _next.whenComplete(() {
      completer.complete(new Future<T>.sync(operation));
    });
    return _next = completer.future;
  }
}

which returns

running getInt(A)
running getInt(A)
running getInt(B)
running getInt(A)
running getInt(C)
running getInt(D)
Fetching: A
Fetched: A
Cached: A
Fetching: B
Fetched: B
Cached: A
Fetching: C
Fetched: C
Fetching: D
Fetched: D

where the problem here is that each fetchInt call, while may be necessary, will block each other. This is more inefficient than before. I modified the implementation to be more efficient:

import 'dart:async';

main() {
  IntStore intStore = IntStore();

  print("running getInt(A)");
  intStore.getInt("A");

  print("running getInt(A)");
  intStore.getInt("A");

  print("running getInt(B)");
  intStore.getInt("B");

  print("running getInt(A)");
  intStore.getInt("A");

  print("running getInt(C)");
  intStore.getInt("C");

  print("running getInt(D)");
  intStore.getInt("D");
}

class IntStore {
  final Map _store = <String, int>{};
  final Map _mutexStore = <String, AsyncMutex>{};

  Future<void> fetchInt(String intName) async {
    if (!_store.containsKey(intName)) {
      print("Fetching: $intName");
      await doSomeWorkAsynchronously(intName);
    }
  }

  Future<int> getInt(String intName) async {
    if (_mutexStore.containsKey(intName)) {
      print("Mutex already here: $intName");
      await _mutexStore[intName].run<void>(() => fetchInt(intName));
    } else {
      print("Creating Mutex: $intName");
      _mutexStore[intName] = AsyncMutex();
      await _mutexStore[intName].run<void>(() => fetchInt(intName));
    }
    print("Passing: $intName");
    return _store[intName];
  }

  Future doSomeWorkAsynchronously(String intName) async {
    await Future.delayed(const Duration(seconds: 3));
    _store[intName] = 3;
    print("Fetched: $intName");
  }
}

class AsyncMutex {
  Future _next = new Future.value(null);

  /// Request [operation] to be run exclusively.
  ///
  /// Waits for all previously requested operations to complete,
  /// then runs the operation and completes the returned future with the
  /// result.
  Future<T> run<T>(Future<T> operation()) {
    var completer = new Completer<T>();
    _next.whenComplete(() {
      completer.complete(new Future<T>.sync(operation));
    });
    return _next = completer.future;
  }
}

which returns

running getInt(A)
Creating Mutex: A
running getInt(A)
Mutex already here: A
running getInt(B)
Creating Mutex: B
running getInt(A)
Mutex already here: A
running getInt(C)
Creating Mutex: C
running getInt(D)
Creating Mutex: D
Fetching: A
Fetching: B
Fetching: C
Fetching: D
Fetched: A

And will therefore return each int as fast as possible, keeping concurrency between different calls, but blocking repeated calls for the same int.