MSDN Magazine - August 2008 - (Page 123)

JEFFrEy richtEr ConCurrent AffAirs More AsyncEnumerator Features In my November 2007 Concurrent Affairs column, I discussed the idea of using C# language features to simplify asynchronous pro gramming (see msdn.microsoft.com/magazine/cc163323). In particular, I focused on anonymous methods, lambda expressions, and iterators. Then in my June 2008 column, I introduced my AsyncEnumerator class and explained how it can be used to drive a C# iterator (see msdn.microsoft.com/magazine/cc546608). I also walked through Async Enumerator’s architecture and explained how it works internally. In this column, I want to show you some additional features that AsyncEnumerator provides, such as joining across multiple concurrent asynchronous operations, Asynchronous Program ming Model (APM) support, return values, callback thread con trol, synchronizing access to shared data, automatic discarding of uncompleted operations, and cancellation/timeout support. Along the way, I’ll also walk you through some common programming patterns made possible with AsyncEnumerator. The code at the beginning of the iterator issues several asynchro nous operations and then executes a yield return numOps state ment. This statement tells AsyncEnumerator not to call back into your code until the number of operations indicated by the value in the numOps variable have completed. The code just below the yield return statement then executes a loop to process all of the completed operations. Note that the operations can complete in a different order than they were issued. In order to correlate each result with its request, I passed wr, which identifies the WebRequest object I’m using to Joining Concurrent Asynchronous Operations With AsyncEnumerator, it is easy to initiate concurrent asynchronous applications. initiate the request, in BeginGetResponse’s last argument. Then in the ProcessCompletedWebRequest method, I extract the Web Request object used to initiate the request from the IAsyncResult’s AsyncState property. The code at the bottom of the iterator also issues several asyn chronous operations and then enters a loop to process each op eration as it completes. However, the iterator must first wait until each operation completes. This is accomplished by the yield return 1 statement which tells AsyncEnumerator to call back into the it erator code as soon as a single operation completes. One of the great things about performing asynchronous op erations is that you can perform several of them concurrently, greatly improving your application’s performance. For example, if you initiate three asynchronous Web service requests concur rently, and if each request takes 5 seconds to complete, then the total time your program spends waiting is just 5 seconds. On the other hand, if you perform synchronous Web service requests, then your application waits for each one to complete before initiating the next. So performing three synchronous Web service requests, each of which takes 5 seconds, means that your application waits at least 15 seconds. With AsyncEnumerator, it is extremely easy to initiate several concurrent asynchronous operations. Your code can then process the completed operations after all of them complete or as each one completes. The iterator in Figure 1 shows both techniques for pro cessing completed operations. When I ran it, I got the following output (note that you must wait for the Web request to timeout): All the operations completed: Uri=http://wintellect.com/ ContentLength=41207 Uri=http://www.devscovery.com/ ContentLength=13258 Uri=http://1.1.1.1/ WebException=Unable to connect to remote server An operation completed: Uri=http://wintellect.com/ ContentLength=41208 An operation completed: Uri=http://www.devscovery.com/ ContentLength=13258 An operation completed: Uri=http://1.1.1.1/ WebException=Unable to connect to remote server APM Support In my previous column, I explained how calling Async Enumerator’s Execute method starts the execution of an iterator’s code. However, I also explained that the thread that calls Execute will block until the iterator exits or executes a yield break statement. Having the thread block may hurt an application’s scalability and is something you’d really want to avoid, especially in server applications. It also hurts responsiveness if called by a GUI thread because the iterator takes an indefinite amount of time to execute, Send your questions and comments for Jeffrey to mmsync@microsoft.com. August 2008 123 http://msdn.microsoft.com/magazine/cc163323 http://msdn.microsoft.com/magazine/cc546608

Table of Contents Feed for the Digital Edition of MSDN Magazine - August 2008

MSDN Magazine - August 2008 Toolbox CLR Inside Out Basic Instincts Cutting Edge Patterns in Practice Data 2.0 - Expose And Consume Data In A Web Services World Biztalk EDI - Build A Robust EDI Solution With BizTalk Server Silverlight - Create Animations With XAML And Expression Blend Write On! - Create Web Apps You Can Draw On With Silverlight 2 Wicked Code - Craft Custom Controls For Silverlight 2 Team System Foundations Windows With C++ Concurrent Affairs Going Places { End Bracket }

MSDN Magazine - August 2008

For optimal viewing of this digital publication, please enable JavaScript and then refresh the page. If you would like to try to load the digital publication without using Flash Player detection, please click here.