Create A Custom Instance Provider

This code is identical to Jimmy’s original code.

public class StructureMapInstanceProvider : IInstanceProvider

{

    private readonly Type _serviceType;

    public StructureMapInstanceProvider(Type serviceType)

    {

        _serviceType = serviceType;

    }

    #region IInstanceProvider Members

    public object GetInstance(InstanceContext instanceContext)

    {

        return GetInstance(instanceContext, null);

    }

    public object GetInstance(InstanceContext instanceContext, Message message)

    {

        return ObjectFactory.GetInstance(_serviceType);

    }

    public void ReleaseInstance(InstanceContext instanceContext, object instance)

    {

    }

    #endregion

}

Create A Custom Service Behaviour Class

This code is identical to Jimmy’s original code.

Create A Custom Service Host

This code is almost identical to Jimmy’s. The only difference is that your custom class should inherit from the WebServiceHost class rather than the normal ServiceHost class for normal WCF Services.

public class StructureMapWebServiceHost : WebServiceHost

{

    public StructureMapWebServiceHost()

    {

    }

    public StructureMapWebServiceHost(object singletonInstance, params Uri[] baseAddresses) : base(singletonInstance, baseAddresses)

    {

    }

    public StructureMapWebServiceHost(Type serviceType, params Uri[] baseAddresses) : base(serviceType, baseAddresses)

    {

    }

    protected override void OnOpening()

    {

        Description.Behaviors.Add(new StructureMapServiceBehaviour());

        base.OnOpening();

    }

}

Create A Custom Service Host Factory

Once again this code is almost identical to Jimmy’s. The only difference is that we inherit from WebServiceHostFactory rather than the usual ServiceHostFactory class.

/// <summary>

/// StructureMap Service Host Factory for WCF REST Services

/// </summary>

public class StructureMapWebServiceHostFactory : WebServiceHostFactory

{

    protected override ServiceHost CreateServiceHost(Type serviceType, Uri[] baseAddresses)

    {

        return new StructureMapWebServiceHost(serviceType, baseAddresses);

    }

}

Wiring Up Our Route

In our Global.asax file create our route using the StructureMapWebServiceHostFactory rather than the normal WebServiceHostFactory.

public class Global : HttpApplication

{

    void Application_Start(object sender, EventArgs e)

    {

        IoC.RegisterDependencies();

        RegisterRoutes();

    }

    private void RegisterRoutes()

    {       

        RouteTable.Routes.Add(new ServiceRoute("ReferenceData", new StructureMapWebServiceHostFactory(), typeof(ReferenceData)));

    }

}

Download it from here..

Ben Hall over at CodeBetter just posted a nice article on using the Castle Dictionary Adapter to isolate your code from the ConfigurationManager class to make your unit tests easier to write.

I thought I might take the opportunity to talk about a technique that we use to isolate our code from ConfigurationManager and also other system classes such as the system DateTime, using the language constructs built within the .Net Framework version 3.5 and above. I’m not saying that this technique is better than the Castle Dictionary Adapter method but it just highlights an alternative approach that is possible without using third party assemblies.

Ok so using Ben’s example, because I’m too lazy to think up my own, we have an application configuration value for our EnableNewsLetterSignup.

<add key="EnableNewsletterSignup" value="false"/>

In our code we pull out the value using the static method on the ConfigurationManager class like so.

public bool Signup(string email)

{

    if (!Boolean.Parse(ConfigurationManager.AppSettings["EnableNewsletterSignup"]))

        return false;

    return true; // technically this would go off to an external web service

} 

Ok so nothing new here but when it comes to testing this code it means that we have to have a configuration file for our unit test project (which is bad) and we can only ever test one path through the code without changing the value of the configuration file. Which means we some untestable code without doing some configuration Kung Fu on our config file which is more work than we should have to do.

Wrapping It Up

The technique we use is to wrap our class up in a static wrapper class which we will call Config. For our corresponding application setting we add a public static field of the same name for clarity, of type Func<string>. Now the important thing is we give it it’s default behaviour which in this case is to go off and use our old friend ConfigurationManager to determine whether or not our newsletter is enabled or not.

public static class Config

{

    public static Func<string> EnableNewsletterSignup = () => ConfigurationManager.AppSettings["EnableNewsletterSignup"];

}

Writing Our Code

Now going back to our code where we use the application setting it will now look something like this.

public bool Signup(string email)

{

    if (!Boolean.Parse(Config.EnableNewsletterSignup()))

        return false;

    return true; // technically this would go off to an external web service

} 

Testing Our Code

So now when we want to unit test this piece of code we simply change the default behaviour of our Config class to the behaviour we want to test. See line 5 below, where we force it to return false for the purpose of our unit test. We now have total control over what our supposedly configuration value is without ever having to touch the file system.

   1: [Test]

   2: public void If_our_newsletterSignup_is_not_enabled_we_should_not_be_signed_up()

   3: {

   4:     // Arrange

   5:     Congig.EnableNewsletterSignup = () => "false";

   6:  

   7:     // Act

   8:     var isSignedUp = _classUnderTest.Signup("emailaddress@foo.com");

   9:  

  10:     // Assert

  11:     Assert.IsFalse(isSignedUp);

  12: }

Other Uses

As I said earlier this technique can also be quite useful for isolating other system dependant classes such as code that uses the DateTime.Now property. Sometimes if you want to logic concerning dates and times it can be pretty hard to control the flow of the code because the DateTime is generated by the system. How for example do you test something if you need to be some date in the future or some date in the past. The code below can be used in the same way to isolate our tests from the underlying call to the DateTime.Now which can be overriden as required to give the behaviour we want.

public static class SystemDateTime

{

   public static Func<DateTime> Now = () => DateTime.Now;

}

Conclusion

This approach offers an alternative solution to using the Castle Dictionary Adapter,  and although it requires writing slightly more code it does mean that it is one less assembly that you have to reference in your project which can be a good thing sometimes.

“The Razor parser and view engine can be instantiated and used outside of the ASP.NET application domain.  This means you can directly instantiate and use it within a unit test project without any dependencies on running ASP.NET.

From a unit testing perspective you can indicate the view template you want to run, supply it with any dependencies, and then pass your own test models/viewmodels to it and have it run and render back a string to you. You could then verify that the correct content came back.  This would isolate the views from your controllers and any data access, and allow you to also isolate them from the runtime environment.  View engines in ASP.NET MVC VNext (both Razor and the .ASPX one) will also support and integrate with dependency injection as well.”

Let’s start with talking about the current testing story with the current MVC framework. Now clearly you want to put as much logic into your controller as possible which is currently testable with the current MVC framework. But on the  modern MVC websites that I’ve been working on there ends up being a considerable amount of JQuery in most pages.

This is where I hear you shout “but what about QUnit you can test your JQuery with that”  Yes you can but do I do it..No I don’t.  I never really got QUnit it always seemed great in theory but in reality it seemed like too much work to implement.

Let me elaborate. All the examples I could find when I was starting out with QUnit involved testing some JavaScript discreet function that did some work and returned a value.  For example a function that added two numbers together and gave you the total back. QUnit works really well for that kind of thing. However I don’t find myself writing much of that kind of  logic in JQuery when I am using MVC. I know its possible to do it in JQuery but most of the time I try to push that kind of logic back into the .NET world and onto the server  where it belongs rather than in the browser.

So typically I find myself writing JQuery that does fairly trivial stuff. Hiding and showing stuff and manipulating items in the Dom. Now when I say trivial it is still important. If you hide your submit button when it should be displayed you have broken your page.

Now it is possible to test this logic with QUnit, but for each test or  group of tests  you had recreate the HTML for your page or part of your page to test your JavaScript against. It is this that always seemed a bit lame to me. You were never really testing against your view proper but a mock up of your view.  Not only is it a lot more work to write your tests initially but there is a lot more maintenance work to do going forward. Any change to your HTML mark up in your view needs to be replicated across your test HTML.  Now I know you have to do maintenance with your .NET tests but most of the time we have compile-time checking. If you change the name a property on a class and you don’t update your unit test it won’t compile. Change the name of a div in your mark up and forget to implement that in your test HTML and there is nothing to tell you that there is a mismatch.

To counter this John Teague over at Los Techies outlines a technique that you can use to generate your markup automatically. To be honest I haven’t used this approach as it still seems like a lot of extra hard work, but it’s worth considering.

Therefore going back to Scott Guthrie’s  comment that we will now be able to  isolate the view from the controller pass in some data and render back a screen  as HTML. This is just fantastic, and for me the last piece in the puzzle with regards to testing. With this functionality it will should be relatively trivial to produce the markup that we can run our QUnit tests against and it means that we are actually testing our view rather than a mock up of it.

Well done Microsoft I think this is a positive step, and I can’t wait to get my hands on the framework. Keep up the good work.

So you want to add a scrollable view to your iPhone/iPad application and you want to use the power of Interface Builder to make it easier. This is how you do it.

In this example we are going add a vertically scrolling window to an iPhone screen in portrait mode.

Start by adding your UIScrollView to your view using Interface Builder. It is not possible to actually scroll the view itself within Interface Builder, so select the UIScrollView and move it up so half the view is off the screen. Drag the edge of the screen to make it longer and then add your extra controls to the bottom. Once you are finished drag the view back to its original position so that you can see the top of the screen again.

Now in Interface Builder select your UIScrollView and look at its size in the Attribute Inspector window. Make a note of the views height, in this example assume it is 756 pixels. Because we don’t want horizontal scrolling in this example the width should be the same as the standard iPhone screen (320 pixels) .

Finally in Interface Builder make sure you add an outlet between your ViewController and your UIScrollView.

Go back to your .NET code and open up your view controller. In your ViewDidLoad method override add the following code to set the initial frame and the content size of you scroll view. Notice that the content size has the full dimension of the entire scrollable view eg. 320 pixels wide and 756 pixels high.

Now fire up your application in the iPhone Simulator and you should now have a working scrollable view. 

The code is shown below.

if (WScript.Arguments.length < 2)

{

   WScript.echo("Client HTTP Request Utility\n");

   WScript.echo("usage: httprequest method uri [options]");

   WScript.echo("  -f filename");

   WScript.echo("  -h headerName headerValue");

   WScript.Quit(1);

}

var method = WScript.Arguments.Item(0);

var uri = WScript.Arguments.Item(1);

var req = new ActiveXObject("MSXML2.XMLHTTP");

var filename = null;

req.Open(method, uri, false);

WScript.echo("********* Request ********* ");

WScript.echo(method + " " + uri + " HTTP/1.1");

for (i=2; i < WScript.Arguments.length; i++)

{

    var option = WScript.Arguments.Item(i);

    if (option == "-f")

        filename = WScript.Arguments.Item(i+1);

    if (option == "-h")

    {

        WScript.echo(WScript.Arguments.Item(i+1) + ": " +

            WScript.Arguments.Item(i+2));

        req.setRequestHeader(WScript.Arguments.Item(i+1),

            WScript.Arguments.Item(i+2));

    }

}

try

{

       if (filename != null)

       {

          var fso = new ActiveXObject("Scripting.FileSystemObject");

          var file = fso.OpenTextFile(filename, 1, false);

          var fileContents = file.ReadAll();

          file.Close();

          WScript.echo(fileContents);

          req.Send(fileContents);

          printResponse(req);

       }

       else

       {

          req.send();

          printResponse(req);

       }

}

catch(e)

{

   WScript.echo("******* Response ********* ");

   WScript.echo(e.message);

}

function printResponse(req)

{

   WScript.echo("******* Response ********* ");

   WScript.echo("HTTP/1.1" + " " + req.status + " " + req.statusText);

   var headers = req.getAllResponseHeaders();

   WScript.echo(headers);

   WScript.echo(req.responseText);

}

To perform a GET simply type from the command line:

HttpUtility.js get http://localhost:1189/Service.svc/

where HttpUtility.js is the file name containing the JavaScript code shown above and the URL of our REST service.

To perform a POST:

HttpUtility.js post http://localhost:1189/Service.svc/ –f message.xml -h Content-Type application/xml

where the –f argument specifies the filename with the message contents and the –h argument indicates the header name and header value.

The source code can be downloaded from here.

Go ahead and take a look you can find it here.

At first glance it looks really nice and simple to learn. Essentially rather than wrapping your code elements in your view with <% %> tags you can simply use an @ symbol. The view engine has syntactic awareness built into it so there is  no need to close off your tags. If you want to declare multiple lines of code you use the following you wrap your code in the following symbols @{  code goes here } This will go a long way to reduce the tag soup that you get in your views with the current out the box MVC view engine.

In Scott Gu’s blog he also states that it will be possible to fully Unit Test your views without the need of a web controller or web server. This sounds really interesting to me and I wonder how far Microsoft will take it as I would love to see this extend into being able to test your JQuery in your views as well, as these days I find this is where the majority of the logic sits. I know it is possible test JQuery with frameworks such as QUnit but I haven’t been altogether comfortable with the amount of duplicated HTML that you have create in order to write your tests…I digress maybe that’s a subject for another post.

Anyway I look forward to it’s release.

This means that if you point your CodeSnippets folder at a folder in your DropBox they will all be magically kept up to date across multiple machines.

I’m going to do the same with my Resharper Template as well!!