Question

I have the following async queue processing routing.

      var commandQueue = new BlockingCollection<MyCommand>();
      commandQueue
            .GetConsumingEnumerable()
            .ToObservable(new LimitedConcurrencyLevelTaskPoolScheduler(5))
            .Subscribe(c =>
                           {
                               try
                               {
                                   ProcessCommand(c);
                               }
                               catch (Exception ex)
                               {
                                   Trace.TraceError(ex.ToString());
                               }
                           }
            );

In one specific scenario (when I’m about to get some data), I need to make sure that my commandQueue is empty before going out and getting the data. This operation is expected to happen synchronously. Basically, I want to do something like

  public void GetData()
  {
     commandQueue.WaitForEmpty(); 

     // could potentially be expressed: 
     // while (commandQueue.Count > 0) Thread.Sleep(10);

     return GoGetTheData()
  }

I realize that in an ideal scenario, all callers will “GetData” async…but sometimes it’s necessary that it happen in a synchronous manner…and so I need to wait for the command queue to be empty to ensure the consistency and up-to-date-ness of my data.

I know how I can do this pretty easily with a ManualResetEvent…but I’d like to know if there’s an easy way with System.Reactive/TPL.

Thanks.

Answer 1

This is a more difficult question than it seems at first. You want BlockingCollection (and the underlying ConcurrentQueue) for producer-consumer job semantics. But you also want to be able to observe what’s happening with these collections, including waiting for the ’empty’ signal.

Best bet is to take a look at JobQueue and ParallelJobQueue from here:

http://social.msdn.microsoft.com/Forums/en-US/rx/thread/2817c6e5-e5a4-4aac-91c1-97ba7de88ff7

Which includes an observable for WhenQueueEmpty and can control the number of simultaneously running jobs and queued jobs (jobs being synonymous in this case with your command concept).