classic producer consumer pattern using blockingco

问题:

Please see below pseudo code

//Single or multiple Producers produce using below method
    void Produce(object itemToQueue)
    {
        concurrentQueue.enqueue(itemToQueue);
        consumerSignal.set;
    }

    //somewhere else we have started a consumer like this
    //we have only one consumer
    void StartConsumer()
    {
        while (!concurrentQueue.IsEmpty())
        {
            if (concurrentQueue.TrydeQueue(out item))
            {
                //long running processing of item
            }
        }
        consumerSignal.WaitOne();
    }

How do I port this pattern I have used since time immemorial to use taskfactory created tasks and the new signalling features of net 4. In other words if someone were to write this pattern using net 4 how would it look like ? Pseudo code is fine. Iam already using .net 4 concurrentQueue as you can see. How do I use a task and possibly use some newer signalling mechanism if possible. thanks

Solution to my problem below thanks to Jon/Dan. Sweet. No manual signalling or while(true) or while(itemstoProcess) type loops like the old days

//Single or multiple Producers produce using below method
 void Produce(object itemToQueue)
 {
     blockingCollection.add(item);
 }

 //somewhere else we have started a consumer like this
 //this supports multiple consumers !
 task(StartConsuming()).Start; 

 void StartConsuming()
 {
     foreach (object item in blockingCollection.GetConsumingEnumerable())
     {
                //long running processing of item
     }
 }

cancellations are handled using cancel tokens

回答1:

You would use BlockingCollection<T>. There's an example in the documentation.

That class is specifically designed to make this trivial.

回答2:

Your second block of code looks better. But, starting a Task and then immediately waiting on it is pointless. Just call Take and then process the item that is returned directly on the consuming thread. That is how the producer-consumer pattern is meant to be done. If you think the processing of work items is intensive enough to warrant more consumers then by all means start more consumers. BlockingCollection is safe multiple producers and multiple consumers.

public class YourCode
{
  private BlockingCollection<object> queue = new BlockingCollection<object>();

  public YourCode()
  {
    var thread = new Thread(StartConsuming);
    thread.IsBackground = true;
    thread.Start();
  }

  public void Produce(object item)
  {
    queue.Add(item);
  }

  private void StartConsuming()
  {
    while (true)
    {
      object item = queue.Take();
      // Add your code to process the item here.
      // Do not start another task or thread. 
    }
  }
}

回答3:

I've used a pattern before that creates a sort of 'on-demand' queue consumer (based on consuming from a ConcurrentQueue):

        private void FireAndForget(Action fire)
        {
            _firedEvents.Enqueue(fire);
            lock (_taskLock)
            {
                if (_launcherTask == null)
                {
                    _launcherTask = new Task(LaunchEvents);
                    _launcherTask.ContinueWith(EventsComplete);
                    _launcherTask.Start();
                }
            }
        }

        private void LaunchEvents()
        {
            Action nextEvent;

            while (_firedEvents.TryDequeue(out nextEvent))
            {
                if (_synchronized)
                {
                    var syncEvent = nextEvent;
                    _mediator._syncContext.Send(state => syncEvent(), null);
                }
                else
                {
                    nextEvent();                        
                }

                lock (_taskLock)
                {
                    if (_firedEvents.Count == 0)
                    {
                        _launcherTask = null;
                        break;
                    }
                }
            }
        }

        private void EventsComplete(Task task)
        {
            if (task.IsFaulted && task.Exception != null)
            {
                 // Do something with task Exception here
            }
        }