Thursday, December 31, 2009

Introduction to the Background Worker Component in WPF

(Editor's Note: An updated and expanded version of this article is available for download: Keep Your UI Responsive with the Background Worker Component)

The Problem

We've all experienced it: the application UI that hangs. You get the dreaded "Not Responding" message, and you have to decide if you should wait it out or simply kill the process. If you have long-running processes in your application, you should consider putting them on a separate thread so that your UI remains responsive. However, threading is a daunting subject. We've heard horror stories about race conditions and deadlocks, and needing to use the thread dispatcher to communicate between background threads and UI threads. At this point, it sounds like a subject best left to the experts.

The Solution

Fortunately, the .NET framework provides a simple way to get started in threading with the BackgroundWorker component. This wraps much of the complexity between your background thread and your UI thread without doing any special coding. You can use this component with WinForms and WPF applications. We'll be using it with WPF here.

The BackgroundWorker offers several features which include spawning a background thread, the ability to cancel the background process before it has completed, and the chance to report the progress back to your UI. We'll be looking at all of these features.

The Set Up

We'll start with a fairly simple WPF application that has a long-running process that blocks the application until it has completed. You can download the source code for the application here: http://www.jeremybytes.com/Downloads.aspx. The download includes the starter application and the completed code. The starter application includes the following.

1. A Simple WPF form:


You can find the XAML for this in the download. It consists of 2 Text Boxes (Iterations and Output), a Progress Bar, and 2 Buttons (Start and Cancel).

2. A long-running process (in the code-behind the form):

(click image to enlarge)


You can see that we're using the famously slow process Sleep(100) that loops based on the parameter value.

3. Event-handlers for the buttons (in the code-behind):

(click image to enlarge)


When you run the application and click the Start button, you'll see that the application hangs until the process is finished. If you try to move or resize the window while the process is running, nothing will happen for several seconds. And you'll see the "Not Responding" message if you look in Task Manager:



When the process is completed, you will see the value from the Iterations box mirrored in the Output box. This is simply a confirmation that the process completed. You can try this with different values. Since we are using a Sleep(100), a value of 50 iterations will translate into a 5 second process; a value of 100 is 10 seconds, and so on. Just as a side note, I chose a value of 1/10th of a second for the Sleep rather than the usual 1 second so that we will have a smoother progress update later on.


Adding the BackgroundWorker

The BackgroundWorker is a non-visual component. In the WinForms world, this would mean that we could just drag the BackgroundWorker from the Tool Box onto the Form, and it would show up as a non-visual component. In WPF, things are a little bit different. We need to add the BackgroundWorker as a window resource that we can reference throughout our code. Here's the steps:

1. Add the System.ComponentModel namespace to the XAML. We do this in the markup for the Window. The good news is that Visual Studio IntelliSense helps you out quite a bit with this. We'll give the namespace a "cm" alias so we can reference it easily. Here's the Window markup with the namespace included:

(click image to enlarge)


2. Add a BackgroundWorker as a Window Resource.

(click image to enlarge)



Hooking Things Up

To use the basic functionality of the BackgroundWorker, we need to do a couple of things. First, we need to hook up two event handlers: DoWork and RunWorkerCompleted. These events are much like they sound. To kick off the background process we call the RunWorkerAsync method of the BackgroundWorker and pass any parameters needed. This fires the DoWork event (which is where we'll put our long-running process). The RunWorkerCompleted event fires after that process is done. At that point, we can update our UI and do clean up (if required).

So, let's put our process into the background. We'll start by creating the handlers for the events mentioned above. As a reminder, Visual Studio IntelliSense helps us out quite a bit with this. In our BackgroundWorker markup that we created above, just type "DoWork=" and you'll get the option for "New Event Handler". This will create the stub and give the handler a name based on our component. We'll do the same for "RunWorkerCompleted" and end up with the following XAML:

(click image to enlarge)


Now we'll flip over to the code-behind and implement these handlers. Let's look at the code, then we'll talk through it. Note, in addition to the code below, I have also added a "using System.ComponentModel;" to make things a little less verbose:

(click image to enlarge)


First, you'll see that our DoWork event calls the DoSlowProcess method (our long-running process). You'll note that we are getting an integer argument from the DoWorkEventArgs (we'll see how this gets passed in just a minute). The e.Argument is of type Object, so we have to cast it to the integer type that our method is expecting. Next, you'll note that we're passing the result back to the DoWorkEventArgs in e.Result. This will be used in the next handler.

The RunWorkerCompleted event fires after the long-running process is complete. You can see that the first thing we do is check to see if an error occurred. If not, then we'll go ahead and populate the output box with the result of our method. The e.Result here is actually the same e.Result from the DoWork event. In our case, the DoSlowProcess returns an integer that we populate in the output. In addition, you can see that we are enabling and disabling the buttons as appropriate.

The important thing is what you don't see. Notice that our completed handler is manipulating our UI elements without any use of the Dispatcher or Invoke methods that you need to do if you are handling the threading on your own. Instead, we just reference the elements on the UI thread directly. The BackgroundWorker takes care of all of the complexity on the back end.

Finally, we need to kick off the DoWork event in our Start Button handler. In order to do this, we need a reference to the BackgroundWorker component in our window. The problem is that it is simply a resource right now. The first task is to get a reference to it. We'll put this code at the top of our Window class and modify the constructor:

(click image to enlarge)


In this code, you see that we create a private variable to reference the BackgroundWorker component. Then in the constructor, we pull the component out of the resources by using the FindResource method. Now we can use the component in our code.

Here's our updated code in the Start Button Click event handler:

(click image to enlarge)


You'll notice that instead of calling the DoSlowProcess directly, we are now calling the RunWorkerAsync method of the BackgroundWorker. This method takes an optional object parameter. In our case, we will use this to pass the number of iterations through. This is the value that shows up in the e.Argument of the DoWork handler that we saw above.

The next thing we do is update the button states appropriately. Since the Cancel button is not yet implemented, it won't have too much effect. But we'll get to that in a bit.

Finally, note that we are clearing the output text. Remember that since we are running the process in the background, our UI still remains responsive. We want to clear out the output while the process is running, and then populate it agian after the process is complete. This is handled by the RunWorkerCompleted event that we saw above.

Now we have a fully-functional application with a background process. If you run the application now, you'll notice different behavior from what we saw before. After you click the Start button, the UI remains responsive: you can move and resize the window, type in the boxes, or whatever. When the process if finished, the output box is updated.

But we're not done yet. We still need to look at the Cancel and Progress functions.


Updating Progress

Next we'll look at reporting progress back from our long-running process. One thing to keep in mind if you want to have a progress bar in your UI: you will need to come up with some type of metric for the percent complete. For example, I have used the BackgroundWorker for long-running SQL queries. In this case, I was unable to report percentage because I had no idea exactly how long the process would take. In our sample here, we can use some fairly simple math to report the percentage completed.

The BackgroundWorker has a property we need to set (WorkerReportsProgress) and an event handler (ProgressChanged). These are fairly straight forward to implement. But here's where things get a little complicated. We need to update the progress in our DoSlowProcess method. This means that we need a reference to the BackgroundWorker.

Let's start with the easy parts. First the updated XAML:

(click image to enlarge)


Here we just set the WorkerReportsProgress to True (the default is False) and add the stub for the ProgressChanged event handler. As we did above, we'll just let Visual Studio create a New Event Handler for us.
To implement the event handler, we'll just set the value of the progress in our UI:

(click image to enlarge)


Now we'll make updates to some of our existing code. First, the DoSlowProcess method:

(click image to enlarge)


We've added both BackgroundWorker and DoWorkEventArgs parameters to the DoSlowProcess method. In order to update the progress, we'll add some code to each iteration of the loop. First, we check to make sure that the BackgroundWorker parameter was populated; then we check the WorkerReportsProgress property to see if the BackgroundWorker reports progress. If false, then we'll skip the code. If true, then we calculate the percentage and call the BackgroundWorker.ReportProgress method. This will fire the ProgressChanged event that we implemented above.

Now, since we've added additional parameters to DoSlowProcess, we'll need to update the method call. As a reminder, this was in the DoWork event. Here's the updated code:

(click image to enlarge)


We'll just cast the sender to a BackgroundWorker and pass it on through. We'll just pass the DoWorkEventArgs parameter through as well.

Finally, we'll add a line of code to the RunWorkerCompleted event to zero out the progress bar after it has completed:

(click image to enlarge)


The reason for this is if you are using Windows Vista or Windows 7, the progress bar continues to animate even after it is as 100%. This makes it difficult to tell that the progress is complete. So, we'll just set it back to 0 after it's done.


Now, if we run the application again, we'll see that we have a functional progress bar. The last step is to add cancellation.


Canceling the Background Process

Before implementing cancellation in your application, you will need to take a few things into consideration. First, when you cancel a BackgroundWorker process, there is no event that fires, and the process does not stop immediately. Instead, a cancellation flag gets set on the BackgroundWorker. It is up to your long-running process to check for this flag and to stop running if necessary. In my example above with the long-running SQL query, I could not implement cancellation because there was no "iteration" in my process -- it was simply waiting for the query to return.

In our example here, since we are using a loop, we have a perfect place to check for cancellation and stop our process. Here's an overview of the steps we'll take, then we'll look at each in more detail.

First, we need to set a property on the BackgroundWorker ("WorkerSupportsCancellation"). Then we'll tell the component we want to cancel in the Cancel Button event handler. Next we'll add the cancellation code to our DoSlowProcess method. And finally, we'll make a few changes to the RunWorkerCompleted event handler to behave differently if the process was canceled.

First, the XAML:

(click image to enlarge)

Next, the Cancel Button event handler:

(click image to enlarge)

You can see that we're simply calling the CancelAsync method of the BackgroundWorker.

Next, add the cancellation logic to the DoSlowProcess:

(click image to enlarge)

You can see that we added another conditional to check CancellationPending. If so, then we'll set the e.Cancel property of the DoWorkEventArgs and return from our method.

And finally, the RunWorkerCompleted:

(click image to enlarge)

Here, you can see that we are checking the e.Cancelled property of the EventArgs. If it's true, then we'll put "Canceled" in the output box. One thing you'll note: we are not resetting the progress bar in event of cancellation. This is so that if you stop the process, you can still see how far it got before the cancellation.



Now, when we run the application, we'll see that we have a long-running process that runs in the background (keeping the UI responsive), an updating progress bar, and a working Cancel button.


Wrap Up

The BackgroundWorker component allows us to put long-running processes onto a background thread without the usual complexities of threading. We have seen how we can get progress updates that we can show in our UI as well as how to cancel a process before it has completed. In addition, we've seen that even when updating our UI, we don't have to worry about communicating across threads. It is all handled for us in the component.

Probably the best thing about the BackgroundWorker is that it allows us to get our feet wet in the world of threading in an easy and relatively safe way. Think about this the next time you come across an application that is "Not Responding". And do what you can to keep your UIs responsive for your users.

Happy coding!

Sunday, December 27, 2009

Where's Jeremy?

I'm Still Here

So, it's been a while since I've posted anything. But I've still been working. Coming up at the end of January, I will be doing some presentations as the SoCal Code Camp. The good news is that it's forcing me to put together another 3-4 demos. So, if you check back over the next month, you should see them pop up.

In addition, I've got a website where you can download the sample code for all of the demos that I've presented so far. It's at http://www.jeremybytes.com/. Just check out the Downloads section. (I'll be adding links to my various blog posts as well).

Lots of exciting things coming up.

Happy New Year!

Friday, September 4, 2009

Target Practice - Silverlight 3 / XAML Sample

XAML has intrigued me ever since I took a close look at it. In a previous post, I created an entire WPF application using only XAML (it's trivial, yet functional). When I got started with Silverlight 2, I wanted to replicate that application. Unfortunately, there were limitations to Silverlight 2 (such as a lack of triggers) that kept me from implementing it. I could have built the same application, but it would have required some code-behind to handle the animation.

The good new is that Silverlight 3 has additional support that allows us to replicate the XAML-only version of the application. It does vary a little bit from the WPF version, so even if you've looked through the previous post, you're likely to learn something new. I'll point out the differences as we come across them.

Scenario
We'll be creating a simple shooting gallery-type application. I call it "Target Practice" since it will require some more work to turn it into an actual game. In the final version, we will have a 3 x 3 grid of targets. Each of these is clickable adn will result in a "hit" animation. After 3 seconds, the target will reset to its previous state -- all inside the XAML. You can scroll down to the bottom to see the finished product.

Step 1 - Application Set-Up
We'll start by creating a new Silverlight 3 application. You'll need Visual Studio 2008, plus the Silverlight 3 bits (start here if you need them: http://silverlight.net/GetStarted/). Create a new project by selecting File -> New -> Project. From there, choose the "Silverlight" category under "Visual C#" and then select "Silverlight Application". I named my application "SilverlightTargetPractice".

You'll be asked if you want to create a web application to host the project. Go ahead and leave the defaults (create a new ASP.NET Web Site). This will give you 2 projects: "SilverlightTargetPractice" is your Silverlight application, and "SilverlightTargetPractice.Web" is the hosting web site. Whenever you run the project, you will want to run the web application.

In the Silverlight project, you will find the "MainPage.xaml" file. That's where we are going to start. Open up the xaml file, and you'll see the xaml editor. Note: There is no Silverlight 3 designer support in Visual Studio 2008; for that, you'll need Expression Blend 3. You can see my notes on this in a previous post. Here's the default page the project template creates:

(Click image to enlarge)


We're going to make a couple changes in the "UserControl" tag. Update the following properties:

d:DesignWidth="335"
d:DesignHeight="335"

Next, we'll replace the Grid with a set of nested StackPanels. The nested StackPanels will set up our 3 x 3 grid. As mentioned in the previous post, you can use a Grid here if you'd like, but I chose the StackPanels to keep the Button tags a little "cleaner". For the outer StackPanel, I set the Background to Black. Inside each of the nested StackPanels, I have placed 3 buttons. The buttons have fairly arbitrary names; this is just so we can tell them apart when we're doing the design work. The updated XAML looks like this:

(Click image to enlarge)


If you build and run the application at this point, you'll see this result. (Remember to run the Web application).

(Click image to enlarge)


Step 2 - Templates
Now we're going to take advantage of the "lookless" controls in XAML and replace the default Button template. This will allow us full control over the way our buttons look and behave. This varies a little bit from the WPF sample: in the WPF sample, I used a combination of Templates and Styles; in this sample, we'll use only a Template.

We'll start our Template by adding a local resource to the file. Add a tag "UserControl.Resources" just above the outer StackPanel. As we've seen before, Visual Studio IntelliSense is very helpful when hand-editing XAML; so, let the code-completion do a lot of the work for you. You can also do this entire application in Expression Blend, but that's a lesson for another time. Inside the Resources section, we'll add a ControlTemplate for our buttons.

(Click image to enlarge)


A few key things to point out. The TargetType is set to Button. This means that we can apply this template to Button controls (or their descendents) and only Button controls. Next is the x:Key. This gives a name to the Template that we can use to assign it elsewhere. Next we have a Canvas (our drawing surface with a Margin to space things out a little) and an Ellipse (which is just a DarkRed circle at this point). Note: the Ellipse is named "Outer Red Stripe" in the comments because we will be adding more to this template in just a bit.

Now we'll need to apply our new Template to our Buttons. We do this by assigning the Template property of each button to the StaticResource that we just created. The highlighted code below is repeated for each button.

(Click image to enlarge)


Here's our result (ignore the jpeg artifacts in the picture; it actually looks much better in person):

(Click image to enlarge)


Something to note: the red circles are still buttons. They are clickable (although they don't visibly change when you click on them), and they have standard button events (such as Click). Notice that the Content does not appear in the buttons. This is because our Template does not have a ContentManager, nor do we need one for this application. Because we aren't displaying the Content, we can remove the Content properties from each of our buttons to clean up the code.

Now, let's spice up our visuals.

Step 3 - Eye Candy
Now we'll take the boring red circles and make them into glassy 3-D buttons. As I mentioned in the WPF XAML sample, the best way to learn how to do this is by looking at some Photoshop tutorials. The same techniques (gradient opacity ellipses) can be used in Expression Blend. I take no credit for the "glow" parts of the design template; I copied them from a demo by Walt Ritscher. He's also a good resource for WPF styles and templates.

First, the top glow. This is simply added to our Control Template below the "Outer Red Stripe" code. It is just an ellipse with a radial gradient that moves from white to transparent.

(Click image to enlarge)


The bottom glow is much the same: just an ellipse with a radial gradient. We'll put this directly below the "Top Glow" code.

(Click image to enlarge)


If you build and run the application now, you'll see a distinct difference compared to what we had before:

(Click image to enlarge)


Cool, huh?

Step 4 - Completing the Targets
Next, we just need to complete the visual design of the Targets. This is just a set of ellipses, each smaller than the last, alternating between Red and White.

(Click image to enlarge)


A few things to note here. First you'll see that we are moving the Canvas.Left and Canvas.Top properties because the Top Left corner is considered to be the starting point for drawing on a Canvas. We are moving down and a bit to the right to keep the smaller circle centered on the previous one.

Order is important here. The XAML parser processes items in the order that they appear in the file. This means that the "Outer Red Stripe" is drawn first, and then each smaller circle is drawn on top of it. The same is true of the "glows". We want these on top of the rest, so we specify them last in the file. The Canvas in Silverlight 3 does have a Z-Order property, so you can adjust this if you have a more complex drawing.

We're done with the visuals for our Targets. Now they look like this:

(Click image to enlarge)


Step 5 - The Action
The last step is to add the behavior. This entire section is quite a bit different from the WPF sample, and all brand new to Silverlight 3.

As a reminder, when we click on any of the target buttons, we want it to animate by "falling over" and then reset after 3 seconds. The animation will work by using a Transform. The first step is to define an empty Transform to our Canvas. I have this inside the closing Canvas tag, but below all of the visuals we added above.

(Click image to enlarge)


This is where things really start to get interesting. For the Animation to be triggered by a Click, we need to add some References to our project: Microsoft.Expression.Interactions and System.Windows.Interactivity (noted below).

(Click image to enlarge)


Now we need to add a namespace reference to these assemblies. We add these at the top of the file with the other "xmlns" references. The good news is that you don't have to type all of this in (see the image below). Instead, IntelliSense will help you out. After you type the "=", you will get a pop-up of the assmeblies that are available from the project references. The completed namespace references should look like the image below:

(Click image to enlarge)


Next, we'll add an Interaction section to the Template that includes an EventTrigger. This needs to be placed before the closing Canvas tag. (I have it between the Transform and the end of the Canvas.) The EventTrigger specifies that it acts on the "MouseLeftButtonDown" event. And inside of that, we have an Animation Storyboard. Take a look at the EventTrigger and Animation specified below. Then we'll walk through it step by step.

(Click image to enlarge)


First, note the "i:Interaction.Triggers" tag. The "i:" is specifying that we are using an object in the System.Windows.Interactivity namespace that we referenced above. Next, we define our Event Trigger (also in the "i" namespace) and specify that it will act on the MouseLeftButtonDown event.

Now, we move on to the "im:ControlStoryboardAction" tag. The "im:" is referencing the Microsoft.Blend.Interactions namespace. This is denoting that we want to "Play" the Storyboard when the EventTrigger fires. Finally, we'll define the actual Storyboard in the "ControlStoryboardAction.Storyboard" section.

This contains a standard-looking Animation storyboard that we have named "ClickAnimaton". Let's look at each of the 2 animations in a little more detail.

The "DoubleAnimationUsingKeyFrames" says that we want to animate a double property (in this case the Y-Scale of the Canvas) and that we want to use KeyFrames to control the flow of the animation. The "Storyboard.Target" tells what object the storyboard is acting on (the Canvas in our Template), and the "Storyboard.TargetProperty" tells what property we are acting on. As noted before, we are operating on the Y-Scale of the Canvas by using the RenderTransform we defined earlier. Duration is 1.5 seconds and Auto-Reverse is true. This means that our total animation will be 3 seconds long.

Here's how we want our animation to work: first, we want it to "fall over" quickly (in 0.5 seconds), then stay "laying down" for a while, then pop back up quickly (also 0.5 seconds). They KeyFrames will let us do this. In the first "LinearKeyFrameAnimation", we are giving the starting position of 1, meaning set y-scaling to 100%. In the next, we want the y-scale to be set to 1% (almost invisible) at the half-second mark. The third shows that we want the y-scale to sill be at 1% at the 1.5 second mark. The auto-reverse runs the animation backwards (starting with the third KeyFrame and working up). This gives us a total animated time of 3 seconds.

The second animation is required because the origin of the Canvas is at the Top Left. This means that if we only had the first animation, the target would shrink up (toward the origin). The second animation says that we want to move the center point of the Canvas down so that it will shrink down and "fall over".

You'll have to run it yourself to see the final result. Here's a screen shot of things in action:

(Click image to enlarge)


Round Up
So, Silverlight 3 allows us to do much more in the XAML than Silverlight 2 allowed. We were able to reproduce the WPF XAML sample with a few changes. I expect that by the time .NET 4.0 and the next version of WPF rolls around, that WPF and Silverlight will look pretty much the same. But we'll still have to wait and see with that.

This is a good example of the power of declarative programming with XAML. The more I use it, the more I like it. When I think about trying to do an application like this in WinForms, I just shudder. It would be a mess of custom drawing code, timers, and crazy calculations. XAML gives me the freedom to easily take control over the layout and appearance of my application. And as we've seen here, it's not just limited to "look and feel", we can also program quite a bit of functionality with it.

You can add some code-behind and create a full fledged shooting gallery game that keeps track of scores and gives you a time limit. But I'll leave that as an exercise for the reader.

Happy coding!

Sunday, August 2, 2009

Silverlight 3 -- D'oh! and Woo Hoo!

In July, Microsoft released Silverlight 3. On the one hand, it's good to see the features moving forward at a good pace. On the other hand, we only had 9 months with Silverlight 2. I just thought I was starting to understand how to work with Silverlight 2 effectively, and wham! new version.

The Bad
First, the bad. I was surprised by some of the items that are included in the Silverlight 3 release notes. Fortunately, I read these before installing the bits.

1. Silverlight 2 & Silverlight 3 compatability
Silverlight 2 projects cannot be created with the Silverlight 3 Tools for Visual Studio 2008. To create Silveright 2 projects, uninstall the Silverlight 3 runtime and the Silverlight 3 Tools from Add or Remove Programs and re-install the Silverlight 2 Tools for Visual Studio 2008.
This was a showstopper for me at work. As I've mentioned, I've been developing a Silverlight 2 application at my day job. Due to certain circumstances, I can't upgrade the project to Silverlight 3 (yet). Since I still need to support it in the meantime, I need to keep the Silverlight 2 bits installed on my machine.

There are some blog posts out there with tips on how to switch back and forth between the 2 versions, but it's making things a bit more complicated than I would like. So, the work machine stays on Silverlight 2 (but Silverlight 3 immediately gets installed at home).

2. Design Preview in Visual Studio 2008
Due to performance and rendering issues, the preview window has been disabled in the Silverlight 3 Tools for Visual Studio 2008. The functionality of the XAML editor remains intact, including IntelliSense, error messages, and the ability to drag controls from the Toolbox into the editor. WYSIWYG XAML design can be done by using Expression Blend or Visual Studio 2010.
This limitation is somewhat disappointing to me. I'm a XAML guy. I'm comfortable typing in the design code directly (and IntelliSense works great in Visual Studio). But I did use the preview window to see how things look. Now, I have to have Expression Blend open to check the preview. I wouldn't mind so much if I used Blend more extensively (which I probably should anyway), but right now it seems like a bit of a kludge to have to toggle back and forth.

There are other notes as well, and I would encourage you to read them in full. These are the ones that jumped out at me as being disappointments.

The Good
So, now the good parts. These are just the items that struck me as useful. There are plenty of other new features (such as out-of-browser support and 3-D support that are interesting but not very applicable in my world).

1. WCF Service Support
Silverlight 3 has support for the full WCF stack. This means that we are no longer limited to basicHttpBinding for our WCF services that are consumed by Silverlight. We can use WS* bindings or any of the other bindings.

2. Controls
In addition, there are plenty of new contols available (including a TreeView) and loads of new layout controls. The Silverlight 3 toolbox is starting to look a lot closer to the WPF toolbox (and excels it in a few cases).

3. Improved XAML
New features such as Behaviors and Triggers greatly enhance the ability to define things declaratively in XAML as opposed to having to implement them in code.

Here's my example: One of my first demos on this blog was a WPF application that was written entirely in XAML. It included a visual re-design of buttons and animation that triggered by clicking the button. It was fully functional with absolutely no code-behind. You can see that entry here: WPF Target Practice.

When I first started getting into Silverlight 2, my goal was to reproduce the same application in Silverlight. However, due to limitations (specifically a lack of Triggers), the only way to implement the animation would be in code. I moved forward with learning how to do the animation that way (and implemented some pretty cool visuals in my application at work), but I never posted the results since they missed the goal of using XAML only.

Silverlight 3 has some additional assemblies available that allow you to implement the triggers and fire off animations. So, in an entry in the near future, I'm planning on posting the Silverlight 3 version of Target Practice. So be sure to stay tuned!

Installation
To use Silverlight 3, you'll need Visual Studio 2008, plus the Silverlight 3 bits. Start at the Silverlight site here: http://silverlight.net/GetStarted/. As mentioned above, you want Expression Blend 3 (there are trials available if you don't want to purchase the full suite). Be sure to take a thorough look through the release notes as well.

Wrap Up
So, like most things in life, there is the bad along with the good. I'm hoping that by the time Visual Studio 2010 comes out, we'll have some better options available. In the meantime, the Silverlight 3 does have a lot to offer (I just scratched the surface with a few items that interested me).

Keep learning!

Saturday, June 13, 2009

More Silverlight Part 2 - User Controls & Events

The story so far: In More Silverlight Part 1, we created a Silverlight 2 application and a WCF service. We then used a DataTemplate and various ValueConverters to format and display the data in the ListBox.

Here's what the app looked like when we left off:

(Click image to enlarge)


Now, we're going to move on, first by creating a Detail panel, then abstracting out the ListBox to a User Control, and finally creating an Event that will let us know when a new Person record is selected.

Creating a Detail Panel

We'll start by adding a Detail panel that shows the First Name, Last Name, Birth Date, and Age for a single Person (the Person selected in the ListBox). This will be a StackPanel that is in the 2nd column of the Grid in our Page.xaml.

(Click image to enlarge)


The code is pretty self-explanatory. We set up TextBlocks to serve as labels for the data, and TextBlocks with data binding to hold the data. You'll notice that we are using the same DateConverter and AgeConverter that we used in the ListBox.

Getting Data into the Detail Panel

We want the Detail Panel to update whenever the selection changes in the ListBox. To do this, we'll hook up a "SelectionChanged" event to the ListBox. As a reminder, you can just type "SelectionChanged=" and Visual Studio will give you the option to add a New Event Handler.

(Click image to enlarge)


Now, we'll navigate to the Event Handler in the Page.cs file. What we need to do here is set the DataContext of the DetailPanel. But the question is, what should we bind to? Do we need to make a service call?

The answer comes through the fact that Silverlight data binding is extremely flexible. We can bind to any object, including an object that exists somewhere else on the page. In this case, we want to bind to the Person that is selected in the ListBox. So, we do just that.

(Click image to enlarge)


A couple of things to note here. First, you see that we are getting the SelectedItem for the ListBox and then casting it as a Person. What you should be aware of is that this may result in a null value. The SelectionChanged event will fire whenever the currently selected item changes; this includes when an item is un-selected. When a ListBox is first loaded, the default is to have no selection. This event will fire, and the SelectedItem will be null.

In many circumstances, you will want to check the "var p" to see if it is null after the assignment. In our case, we will want to let p be null (we'll see why in just a bit).

After we get the SelectedItem cast as a Person, we just assign this as the DataContext of our DetailPanel. Pretty straightforward, huh? Here's the result:

(Click image to enlarge)


To see how the nulls are handled in the Event, simply click the Button again. This will clear the selected item in the ListBox, which will in turn clear out the DetailPanel data.

Creating a User Control

But what if we want to re-use the Person ListBox multiple places in our application? Rather than duplicating code, it would be better to abstract out this piece as a separate User Control. That's what we'll do right now.

Step 1: Add a new User Control to the Silverlight project
Right-click on the Silverlight project, select "Add", then "New Item". From the dialog, select "Silverlight User Control" and name it "PersonListControl". This will create a default PersonListControl.xaml and PersonListControl.cs that look very similar to what we started with for Page.xaml/Page.cs. This is because Page is also a Silverlight User Control.

Step 2: Cut the StackPanel xaml from Page.xaml to PersonListControl.xaml
We'll be moving everything from the first Grid column of Page.xaml to the new user control. Just cut the entire StackPanel (the one that has Grid.Column="0"), and paste it over the Grid (LayoutRoot) in PersonListControl.xaml. This will leave us with a StackPanel as the root element of the user control. Note: after pasting it in, you will want to remove the "Grid.Column" attribute. The app will still run with it there, but it's better to keep the xaml as clean as possible.

Step 3: Copy the UserControl.Resources Section
The xaml that we moved in the step above makes use of the Converters we created earlier. This means that our new user control needs to have references to these as well. The easiest way is just to copy these from the Page.xaml. Note that we are doing a copy (not a cut) because these are used in the DetailPanel that is staying on Page.xaml.

Copy the "xmlns:local=" namespace and the entire "UserControl.Resources" sections with all three converters to PersonListControl.xaml.

Step 4: Change the Size
Our new UserControl will need to fit in the first column of our main Grid, so we'll want to change the sizes a bit. Change Width to 240 (to allow for a little bit of padding) and the Height to 300.

Step 5: Move the Event Handlers
The xaml that we moved contained 2 Event Handlers: one for the ListBox selection changed event, and one for the Button click event. Now we'll move the Event Handlers as well.

From the Page.cs file, cut the "PersonList_SelectionChanged" and "GetPeopleButton_Click" methods and paste them into PersonListControl.cs. One thing to note: since the PersonList_SelectionChanged event directly references the DetailPanel, this won't compile. For now, just comment out the 2 lines in this Event Handler. We'll be re-implementing this a little later.

You should be able to build now, but if you run the application, you will find that we still need to place the new user control onto our main Page.

Step 6: Using the User Control
Actually using the new user control on our main Page couldn't be easier. In the spot in our Page.xaml where we used to have a StackPanel (in the first Grid column), we will now have our control:

(Click image to enlarge)


Notice that this uses the "local" prefix (the same as our Converters) since everything is in the same namespace. If we were to split out the User Control into a separate namespace, then we would need to add a new xmlns and give it a different alias to use.

We can now run the application, and the ListBox will look just like it did before. If we click the Button, it will fetch the Person list and display it. The application doesn't work completely, though. We can't get the data into the DetailPanel. For that, we will need to expose an Event in our new User Control.

High Level Event Overview

Before we create the Event, let's take a look at how we'd like it to work. What we want to do is get the Person that is selected in the ListBox and display it in the DetailPanel. The problem is that the ListBox and DetailPanel are now in different scopes. So, we need to get the Person from the ListBox to the DetailPanel somehow. We'll do this by exposing an Event in the ListBox that has the selected Person as part of the EventArgs.

We'll need a custom EventArgs class that will hold the selected Person. Then we'll set up an Event on our new User Control that we can hook in to from our Page.

Custom EventArgs

We'll create the custom EventArgs class in the PersonListControl.cs file. You may want to abstract this out into its own file for other projects, but we won't do that here. We'll descend from RoutedEventArgs (which is a little more specific than EventArgs and is commonly used in Silverlight and WPF). Our custom EventArgs only needs to have a single Property (SelectedPerson) and a contructor that takes a Person as a parameter. Here's the completed code:

(Click image to enlarge)


Declaring the Event Handler

Next, we need to declare the Event in our PersonListControl class. We can do this with a single line of code:

(Click image to enlarge)


Note that there are several ways to make this declaration, including using "delegate" syntax. The syntax used here is the most compact. You'll see that we are using our custom EventArgs class as a generic type. Our Event is called "SelectedPersonChanged".

Invoking the Event Handler

To invoke the Event Handler, the convention is to create a method called OnEventName -- in our case "OnSelectedPersonChanged". Take a look at the code below:

(Click image to enlarge)


If the comments look familiar, it's because they came straight from the Microsoft Help file. They aren't really applicable for this particular application because we're not making descendant classes or doing complex invokation, but it's good to use the "safe" version so that we can get used to it.

You'll see that we are using our custom EventArgs as a parameter for this method. And as noted in the comments, we are making a copy of the handler first. This is because we can only invoke handlers that are actually hooked up to something (i.e. not null). If the handler is null, then we do not want to do anything (and thus we have the null check). The reason that we make the copy is that it is possible that the handler is deleted between the time we check for null and the time we invoke the method. If we make a copy, then we are acting on a snapshot in time, and so eliminate this possibility.

Raising the Event

Now we need to actually raise the event. We will do this in the "PersonList_SelectionChanged" method (the one that we commented out earlier). We will get the SelectedItem from the ListBox and cast it to a Person -- just like the old ListBox SelectionChanged handler. Then we'll use that Person to create the custom EventArgs and pass it to the OnSelectedPersonChanged method.

(Click image to enlarge)


Notice that it is possible for the Person to be null (as we discussed above). This is okay. If there is no SelectedItem, then we will pass the null through the Event and ultimately to the DetailPanel (as we'll see next). Be sure to build everything to make sure that there are now errors. If you run, you still won't see any difference. Yet.

Using the Newly Exposed Event

Now we'll go back to our Page.xaml to hook up the new Event. You'll notice that if you start typing "Select" in the PersonListControl element, IntelliSense will show the Event. Like other events, if you type "SelectedPersonChanged=", Visual Studio will let pop up "New Event Handler" (as we saw above). If you do this, you'll end up with something like this:

(Click image to enlarge)


Finally, we'll navigate to the event handler in the Page.cs file. Since the custom EventArgs exposes a SelectedPerson object, we just need to assign this to the DataContext of the DetailPanel.

(Click image to enlarge)


If you run the application now, you'll see a fully functioning Person ListBox and DetailPanel.

(Click image to enlarge)


This will behave the same as above. If you click the Button again, the DetailPanel is cleared. This is because the SelectedItem is cleared when the ListBox is loaded. Because the SelectedItem changed, the Event fires. And because there is no SelectedItem, the EventArgs contains a null SelectedPerson. This results in a null for the DataContext for the DetailPanel (which is what we want here). You may need to rework this if you need different behavior.

Summary
We've taken our application and abstracted out a significant part of it into a separate User Control. Then we implemented an Event on that User Control that we can use when we add the element to our Page. If we were to have several different forms that needed this same Person ListBox, we can simply add the custom User Control and get all the same functionality. We even have an Event to hook into so we know when the selected Person changes. These are just the basics, of course. For a production app, we would want to add Error handling and other such things.

As you can see, Silverlight has several powerful features that allow us to create flexible and well-designed applications. Since pretty much everything in Silverlight is a User Control, it's possible to nest these elements within each other and come up with some interesting abstractions in our UI.

Happy Coding!