Tuesday, May 12, 2009

Silverlight 2, WCF, and Lambda Expressions - Part 2

In Part 1, we set up a Silverlight 2 application that consumes a WCF Service. In Part 2, we'll look at a problem with the implementation and different ways of solving it -- ultimately ending with Lambda Expressions. Hopefully by the time we get done, you'll be comfortable with how lambda expressions work and how to use them to make your code cleaner and easy to understand.

Where We Left Off
Here's a quick reminder of what we did last time:
  1. We created a WCF Service that is Silverlight 2 compatible.
  2. We added a reference to the service in a Silverlight application.
  3. We created a UI (Button and ListBox) to call the service and show the results.
  4. We called the WCF Service and hooked up the Completed event to an event handler.

The Page.xaml.cs file we finished with is as follows:

(Click image to enlarge)

And the UI:

(Click image to enlarge)

A Problem
There is a small problem with the UI. After calling the service (by clicking the button), the list gets populated with the names. You can select a name in the list by clicking on it. But if you click the button again, the list refreshes and the selection goes away. We want to update this so that the selected item is maintained when the service is called again.

Solution #1 - Using the Event Handler
The SelectedItem of the ListBox gets cleared when the service is called because we are re-binding the data. From the ListBox's point of view, it is getting a new object to bind to. When the old object is un-bound, the selected item goes away and doesn't come back.

We'll solve this by simply saving off the SelectedIndex of the ListBox before we make the service call. Then after the service call returns, we will reassign the SelectedIndex. (This is assuming that the order of our items is not changing, and we're not getting any new items -- you can extend this basic functionality for more complex situations).

Since we need to save off the SelectedIndex in one method (the Button Click event handler) and then use it in a different method (the Service Completed event handler), we'll need to create a class-level variable to hold this information. See the code below:

(Click image to enlarge)

You'll see 3 new lines of code to get this to work:
  1. private int ehListIndex;
    The class-level variable
  2. ehListIndex = EHList.SelectedIndex;
    Saving the index
  3. EHList.SelectedIndex = ehListIndex;
    Re-assigning the index to the list box

If you re-run the application, you will see that if you make a selection in the list box, that selection is maintained when you call the service multiple times.

Solution #2 - Using an Anonymous Delegate
Now, we'll try this again using an anonymous delegate instead of an explicit event handler. You'll see why this is an advantage in just a bit.

We'll update our XAML by adding another StackPanel to our Grid.

(Click image to enlarge)

You'll notice that this section looks almost identical to the EventHandler UI section. Here's a summary of the differences:

  • The 2nd column of the Grid is specified in the StackPanel.
  • The Names of the ListBox and Button start with AD (for anonymous delegate) instead of EH.
  • The ListBox border is blue (just to make it different).
  • The button Content is "Anonymous Delegate".
  • There is a new Click handler for the button (you can look at the hint in Part 1 about letting IntelliSense create the handler for you).

So, what is an anonymous delegate? In our case, it is an in-line definition of an event handler (for the service callback). This is defined by using the "delegate" keyword, then a set of parentheses with the appropriate parameters, then a set of curly braces with the method implementation. The implementation will look just like in our callback when using the EventHandler.

Here's the initial implementation:

(Click image to enlarge)

What you will notice is that that the first and last lines are the same as our previous button click event. You'll also notice that the parameters for the anonymous delegate ("object" and "GetPeopleCompletedEventArgs") matches the parameters in the explicit event handler we implemented earlier. Finally, the body of the delegate matches the body of the event handler callback.

I know what you're thinking: big deal. Other than "in-lining" some code, exactly what did we gain here? We're about to find out.

We still have the same issue we had earlier with the ListBox SelectedIndex. The advantage to using an anonymous delegate is that any variables in the outer method can be accessed in the delegate implementation. This means that instead of using a class-level variable to store the index, we can use a local variable. Check out the completed code below:

(Click image to enlarge)

You'll see that we now have a local variable named "adListIndex" that is assigned outside of the delegate and then used inside of the delegate. We've just eliminated the need to have a class-level variable.

Solution #3 - Using a Lambda Expression
Let's take this one step further. Instead of using an anonymous delegate, we'll use a lambda expression. First, update the UI. This XAML is very similar to the anonymous delegate section:

(Click image to enlarge)

The same items are updated as above: grid column, element names, border color, button content, and click event.

So, now lets take a look at lambda expressions. There are 2 types of lambda expressions: statement lambdas and expression lambdas. Both use the same syntax (consisting of 3 parts):
  1. Parameters
    These are normally enclosed by parentheses (although, parentheses can be excluded if there is only a single parameter).
    The parameter types may be excluded if the compiler can determine the correct types.
  2. =>
    This is known as the "goes to" operator and denotes that this is a lambda.
  3. Statement(s) or Expression(s)
    A statement lambda has one or more operations wrapped in curly braces. The statement(s) denotes some type of work to be done.
    An expression lambda has one or more expressions (that return true or false) wrapped in curly braces. These are used extensively in LINQ.
    If there is only a single statement or expression, then the curly braces can be excluded.
We'll be working with an Statement Lambda. The first steps to convert the anonymous delegate to a lambda expression are to remove the "delegate" keyword and place the "=>" operator between the parameters and the method implementation. Take a look at the "LEButton_Click" method below and compare it to the first "ADButton_Click" method above:

(Click image to enlarge)

But, as we noted above, if the compiler can determine the parameter types, then we don't need to include them in our code:

(Click image to enlarge)

If you hover over the "s" and "ea" parameters, you will see that Visual Studio recognizes these as being of type "object" and "GetPeopleCompletedEventArgs" respectively. This means that we still have strongly-typed objects. The same is true of "ea.Result". This is the strongly-typed collection that is returned from our WCF Service.

As a final step, we can add the SelectedIndex handling code into the mix:

(Click image to enlarge)

If you run the application, you will see that all three list boxes behave the same way. All three list boxes will keep the selected index between service calls.
What We Learned
We took a look at 3 different ways to solve a particular UI issue. We see that both the Anonymous Delegate and Lambda Expression solutions have the advantage of being able to use a local-scoped variable (thus, keeping our class-level clutter-free). In addition, we see that the Lambda Expression solution has the advantage of not needing to include the parameter types. This is especially helpful if you have a delegate that has the standard 2 parameter signature ("object" and "SomeKindOfEventArgs"); you don't need to know the particular EventArgs class -- just let Visual Studio figure that out for you.
My advice: become familar with Lambda Expressions. You will find that when used judiciously, they can help keep your code understandable and easier to maintain.

Friday, May 1, 2009

Silverlight 2, WCF, and Lambda Expressions - Part 1

This is a 2-part series in using WCF with Silverlight 2. Part 1 is about creating a Silverlight-compatible WCF service and consuming it in a Silverlight application. Part 2 is about lambda expressions, but uses the foundations in Part 1.

Step 1 - Create a New Silverlight Application
First we'll create a new Silverlight application. In Visual Studio, create a New Project, and select the "Silverlight Application" template. I've called mine "SilverlightWithWCF", so that's what you'll see in a couple places throughout the post.

Visual Studio will prompt you with the following:
(Click image to enlarge)

We want to include a new ASP.NET Web Application Project, so we'll leave the defaults as-is. Other options include "ASP.NET Web Site" and possibly "ASP.NET MVC" (if you have the MVC bits installed). In addition to using the ASP.NET application as a test site for the Silverlight application, we will use it to host our WCF service.

Step 2 - Build the WCF Service
We'll add a new WCF Service at this point. Right-click on the Web Application project ("SilverlightWithService.Web" in my case), and choose "Add", then "New Item". Select the WCF Service. We'll name ours "PersonService.svc". See below:

(Click image to enlarge)

You'll notice that there are 2 options we could choose from: "WCF Service" or "Silverlight-enabled WCF Service". You might wonder why we didn't pick the Silverlight one. The answer is that I wanted to look at the standard WCF Service, so that you will know what to do if you already have an existing WCF Service that you want to hook into. I'll point out the difference below when we look at the Web.config file.

Step 2a - The WCF Service Contract
WCF Service consist of 3 main pieces: the Service Contract, the Service Implementation, and the Service Configuration. First, we'll look at the Service Contract.

The contract is simply an interface decorated with a few specific attributes. Visual Studio will create an I[ServiceName].cs file that contains the placeholder contract. If you open the IPersonService.cs file, you'll see an interface that is decorated with the [ServiceContract] attribute. Visual Studio also gives you a sample method called "DoWork". You'll notice that this is decorated with the [OperationContract] attribute. You can think of this as being analogous to the [WebMethod] attribute that we used in ASMX Services. The [OperationContract] attribute denotes which methods will be exposed to the outside world.

We're going to replace the "DoWork" method with a "GetPeople" method. Take a look at the completed code below, and we'll step through it:

(Click image to enlarge)

You'll notice that our method returns a List of Person. Person is a custom class, and so we need to include the definition of the class as well. You'll notice that we have a separate public class called "Person" that is decorated with the [DataContract] attribute. The class consists of 2 public properties (FirstName and LastName) that are each decorated with the [DataMember] attribute. The DataContract and DataMember attributes let the WCF sub-system know that it needs to include this type in the definition of the Service.

On a side note, the [DataContract] and [DataMember] attributes can be excluded in this particular case. .NET 3.5 SP1 will automatically include the type definition for public classes that have automatic properties and a public default constructor. This was added to facilitate Entity Framework. I'll leave it as an exercise for the reader to do further research into this.

The contract only tells what the service will do, not how it does it. For that, we'll need to look at the service implementation.

Step 2b - The WCF Service Implementation
The WCF Service implementation is a class that implements the WCF interface. Open up the "PersonService.svc.cs" file (note: you can simply double-click on the "PersonService.svc", and it will open the .cs file).

You'll see that the class "PersonService" implements "IPersonService" and still has the placeholder "DoWork" method from the stub that Visual Studio created. Delete this method. We'll let Visual Studio add the stubs for the updated interface. Simply right-click on "IPersonService" and select "Implement Interface", then "Implement Interface" again (we won't cover "Implement Interface Explicitly"). This will create a stub for the "GetPeople" method. Here's the implementation:

(Click to enlarge image)

This simply creates a List object and populates it with some Person instances. In real life, you'll probably be doing something more complex (such as accessing a database).

Next comes the configuration.

Step 2c - WCF Service Configuration
Open up the Web.config file, and then navigate to the "system.serviceModel" section (at the bottom). The only thing that you'll need to do is update the binding to "basicHttpBinding" (from "wsHttpBinding"). See below:

(Click image to enlarge)

Remember when I promised to show the difference between "WCF Service" and "Silverlight-enabled WCF Service"? This is it. Silverlight 2 is limited in that it can only use WCF Services that use the "basicHttpBinding". This is the same binding that ASMX services use. (Note: Silverlight 3 will not have this limitation). The "Silverlight-enabled WCF Service" defaults to basicHttpBinding, while "WCF Service" defaults to wsHttpBinding. If you have an existing service, you'll just need to change this value (or add an endpoint that uses this binding).

Note: This is a very important step. If you try to add a Service Reference in a Silverlight application to a WCF Service that is *not* using basicHttpBinding, then the proxy will not work. Even better, you don't get any helpful error messages, just a rather obscure one. So, make sure you don't skip this step.

From here, we'll build everything just to make sure that things look good. This completes our updates to the Web Application. Next, we'll move on to the Silverlight application.

Step 3 - Add the WCF Service Reference to the Silverlight Application
We'll start in the Silverlight application by adding a Service Reference to our newly created service. Right-click on the Silverlight Application project ("SilverlightWithService" in our case), and choose "Add Service Reference". If you click the "Discover" button, it will search out the Services that are included in the Solution. We'll use the namespace "PersonService". See below:

(Click image to enlarge)

Before you click "OK", be sure to check out the "Advanced" button. This opens a dialog that lets you choose how collections are returned from WCF Service calls:

(Click image to enlarge)

You'll see that there are 4 options to choose from: Array, Generic List, Collection, and ObservableCollection. The ObservableCollection is preferred in Silverlight for data binding purposes, but you may want to experiment with the other options as well. It doesn't matter the actual collection type that is returned from the service; it will be converted to the type you select here.

Now that we've added our Service Reference, we'll take a look at the Silverlight UI.

Step 4 - Update the XAML
Next, we'll create the UI. Open up the Page.xaml file. We're going to be making a few changes to the default values and adding some controls of our own. The final layout may look a little odd. This is because we will be adding some more elements to this in Part 2.

Here's the final layout:
(Click image to enlarge)

Let's look at the XAML; then we'll go through it step by step:
(Click image to enlarge)

First, update the Width and Height of the UserControl to 600 and 200 respectively. Then we'll swap out the "Grid" root element for a "Border". Then add a Grid with 3 columns (the other 2 columns will be used in Part 2). Inside the first column is a StackPanel with a ListBox and a Button. The ListBox sets the ItemsSource to "{Binding}". This says that each ListBox item will be bound to each item in the collection. The data binding will use a DataContext that we set in code at runtime.

Next is the Button named "EHButton" (for Event Handler Button -- Part 2 will have 2 additional Lists and Buttons and these names will make more sense). When adding an event handler in XAML, Visual Studio will help you out a bit. If you type "Click=" and hit Ctrl-Space (or let Intellisense pop up on its own), you will have the option "New Handler" to pick from. This will automatically name the handler based on the control name, and then add the stub to the code. In this case, it will create "EHButton_Click" in the code and hook up event.

Now that our UI is laid out, let's put the last of the pieces together.

Step 5 - Call the WCF Service
Here's another short-cut. If you right-click on "EHButton_Click" (the event handler) in the XAML, one of the options is "Navigate to Event Handler". This will open up the Page.xaml.cs file and put the cursor on the "EHButton_Click" event handler.

First, add the WCF Service namespace to the "using" section. This will make the code more readable. The namespace is [projectName].[serviceNamespace]. In our case, it is "SilverlightWithService.PersonService".

Now, we'll go back to the event handler. Start by creating a new proxy object. The proxy is the service name with "Client" appended to the end. In our case it is "PersonServiceClient". Next we'll hook up the callback. Since all Silverlight calls are asynchronous, we have to call the service asynchronously (it's the only choice). Start by hooking up an event handler to the "Completed" event. Visual Studio helps you out with this, too. Type "proxy.GetPeopleCompleted +=", and you'll see Intellisense for creating a new event handler. If you press "Tab" twice, it will create the new handler, and add the implementation. This is a great short-cut. The last step is to call the service using the "Async" method. In this case, just call "proxy.GetPeopleAsync()". See completed button click event handler below:

(Click image to enlarge)

The final step is to implement the callback. This is as simple as getting the result of the call (in this case a List of Person) with "e.Result" in the code. Then we assign this to the DataContext of the ListBox in the UI. One thing to note, if you hover over "e.Result", you will see that it is actually of type ObservableCollection of Person. This is due to the selection we made when we added the Service Reference earlier. The completed Page.xaml.cs file is as follows:

(Click image to enlarge)

Step 6 - Build and Test
Let's build and run. What you should see is that when you click the "Event Handler" button, you will get a list of 5 names in the ListBox. This may take a few seconds for the Service to fire up and run -- remember we're running asynchronously here. Also note that the list only includes the First Name because we did not do a full data template; we just set the DisplayMemeberPath. It's easy enough to enhance yourself, though.

(Click image to enlarge)

You should have a pretty good idea of how to create a WCF Service and consume it with Silverlight 2. We've created a new WCF Service, ensured that it was Silverlight 2 compatible, created a Silverlight 2 application, added the Service Reference, updated the UI, and finally called the Service and bound the results to the UI. In Part 2, we'll be looking at other ways of handling the asynchronous calls.