ConcurrentQueue in TPL

This article will present some simple code to make use of the ConcurrentQueue collection that lives in the System.Collections.Concurrent namespace in TPL. The collections available in this namespace are the following:

• ConcurrentBag
• ConcurrentQueue
• ConcurrentStack
• ConcurrentDictionary
• BlockingCollection

This article will focus on the ConcurrentQueue collection. This class is much like a Queue, but it is also thread safe and therefore supports concurrency in a better manner than the ordinary Queue collection. To enqueue items in the ConcurrentQueue, one uses the Enqueue method. To dequeue items in the ConcurrentQueue, one uses the TryDequeue Method. If this method returns true, the out parameter will contain the object returned from the ConcurrentQueue. Make note also that as the ordinary Queue we add and remove items from the queue in the ordinary First-In First-Out manner (FIFO), obviously because it is called a queue. The next code shows how to use the ConcurrentQueue:

 class Program
    {

        static void Main(string[] args)
        {

            ConcurrentQueue<int> queue = new ConcurrentQueue<int>(); 
            foreach (var i in Enumerable.Range(0, 1000))
            {
                queue.Enqueue(i); 
            }

            Parallel.ForEach(queue, new ParallelOptions { 
                MaxDegreeOfParallelism = System.Environment.ProcessorCount },
                q =>
            {
                int currentItem;
                if (queue.TryDequeue(out currentItem))
                    Console.WriteLine("Got item: {0}", currentItem);
            }); 

            Console.WriteLine("Press any key to continue ...");
            Console.ReadKey(); 

        }

In the code above, the code methods Enqueue and TryDequeue is used of the ConcurrentQueue class. Also note the use of the out parameter and the fact that TryDequeue returns a boolean flag that shows if the dequeueing was successful or not. There is a also a method called TryPeek which will inspect the item returned from the front of the queue, but it will not remove the item from the queue. Often, one will use ConcurrentDictionary also and BlockingCollection is often used to implement Producer-Consumer patterns in a parallel world. ConcurrentStack and ConcurrentBag will also be used in many situations. Of all the concurrent classes, ConcurrentBag looks the simplest to use. The nice part of these classes is that we do not have think too much about locking access to them, since this is already implemented. Actually, do not lock access to these collections at all, one only has to understand each collection and call the correct methods upon them to interact with the collection. Problems around synchronization and parallel access will already be resolved.

Continuing tasks in TPL

This article will present code where a task is continued with another task in TPL. Let's look at some code for this:

  class Program
    {
        static void Main(string[] args)
        {
            var task = Task.Factory.StartNew(() =>
            {
                Console.WriteLine("Running the first task ...");
                Thread.Sleep(2000);
                Console.WriteLine("First task done");
            }).ContinueWith((antecendent) =>
             {
                 Console.WriteLine("Continuing the second task ...");
                 Thread.Sleep(3000);
                 Console.WriteLine("Second task done");
             }, TaskContinuationOptions.NotOnFaulted | TaskContinuationOptions.NotOnCanceled);

            task.Wait(); 

            Console.WriteLine("Press any key to continue ...");
            Console.ReadKey(); 

        }
    }

A task is started with Task.Factory.StartNew and the empty lambda contains a block following that contains the code that will run the work of the first task. Now, the ContinueWith extension method is used next to continue with the next task. I have specified TaskContinuationOptions here with NotOnFaulted and NotOnCanceled, this is an overload and one does not have to specify this always. An often used continuing TaskContinuationOption is OnlyOnFaulted. Let's look last at some code that actually returns some data and gives that data to the next task continuing the work. This makes it possible to partition the work performed in much like a pipeline, one calls this kind of parallelism as dataflow parallelism.

 class Program
    {
        static void Main(string[] args)
        {
            var task = Task.Factory.StartNew(() =>
            {
                Console.WriteLine("Running the first task ...");
                Thread.Sleep(2000);
                Console.WriteLine("First task done, returning some result");
                return 42;
            }).ContinueWith((antecendent) =>
             {
                 Console.WriteLine("Continuing the second task ...");
                 Thread.Sleep(3000);
                 Console.WriteLine("Got result from first task: {0}", antecendent.Result);
                 Console.WriteLine("Second task done");
                 return antecendent.Result * 2; 
             }).ContinueWith((antecendent) => {
                 Console.WriteLine("Continuing the third task ...");
                 Thread.Sleep(3000);
                 Console.WriteLine("Got result from second task: {0}", antecendent.Result);
                 Console.WriteLine("Second task done");
                 return antecendent.Result * 2; 
             });  

            task.Wait();

            Console.WriteLine("Final result: {0}", task.Result);

            Console.WriteLine("Press any key to continue ...");
            Console.ReadKey();

        }

In the code above, one not only continue a task with another task, but then continue with a third task. There can be arbitrary number of tasks being chained in this way. We propagate the result from the previous example by accessing the Result of the antecendent task (previous task). This makes it possible to partition a problem into stages - a pipeline of tasks doing individual processing to create the final result, much like how a factory works. It is possible to propagate values that are not only value types such as int, but also Reference types (arbitrary complex classes). This shows that TPL is able to help you compose the total amount of work to be processed into stages in a pipeline fashion in a simple way. Keep in mind that there are more optimized way of doing this following a more advanced Producer-Consumer approach if one wishes to implement a true pipeline with TPL.

Simple child tasks in TPL

This short article will present some code that displays how to perform the steps necessary to create a child task in the Task Parallel Library (TPL). To create a child task, simple specify the TaskCreationOptions set to TaskCreationOptions.AttachedToParent. The simple console application presents this next:

    var nums = ParallelEnumerable.Range(0, 10000);

    var parent = Task.Factory.StartNew(() =>
            {
                Console.WriteLine("Parent task starting");
                Thread.Sleep(2000);
                Console.WriteLine("Parent task done");

                   Task.Factory.StartNew(() =>
                    {
                        Console.WriteLine("Child task starting");
                        Thread.Sleep(5000);
                        Console.WriteLine("Child task done");
                    
                    }, TaskCreationOptions.AttachedToParent);               

            });

      parent.Wait(); 

      Console.WriteLine("Press any key to continue ...");
      Console.ReadKey(); 
                     

The console app code above waits for the parent task and will not continue to the next line until not only the parent task is performed and executed, but also its child tasks are finished. In other ways, a task can have multiple child tasks. However, only the child tasks that specifies a TaskCreationOptions set to AttachedToParent will actually be waited upon. Also note that creating a next Task is done using the Task.Factory.StartNew calls. This is the preferred option to create new tasks, but one can also instantiate a Task with the new keyword, but then you have to remember to start the task also. Task.Factory.StartNew not only instantiates a new task but also starts it immediately. In the next short article, the ContinueWith construct in TPL will be presented. This is a construct that allows one to pipeline tasks, one after another. This is known as dataflow parallelism.