ploeh blog danish software design

Occasionally I get a question about whether it is reasonable or advisable to let domain objects implement IDataErrorInfo. In summary, my answer is that it's not so much a question about whether it's a leaky abstraction or not, but rather whether it makes sense at all. To me, it doesn't.

Let us first consider the essence of the concept underlying IDataErrorInfo: It provides information about the validity of an object. More specifically, it provides error information when an object is in an invalid state.

This is really the crux of the matter. Domain Objects should be designed so that they cannot be put into invalid states. They should guarantee their invariants.

Let us return to the good old DanishPhoneNumber example. Instead of accepting or representing a Danish phone number as a string or integer, we model it as a Value Object that encapsulates the appropriate domain logic.

More specifically, the class' constructor guarantees that you can't create an invalid instance:

private readonly int number;
 
public DanishPhoneNumber(int number)
{
    if ((number < 112) ||
        (number > 99999999))
    {
        throw new ArgumentOutOfRangeException("number");
    }
    this.number = number;
}

Notice that the Guard Clause guarantees that you can't create an instance with an invalid number, and the readonly keyword guarantees that you can't change the value afterwards. Immutable types make it easier to protect a type's invariants, but it is also possible with mutable types - you just need to place proper Guards in public setters and other mutators, as well as in the constructor.

In any case, whenever a Domain Object guarantees its invariants according to the correct domain logic it makes no sense for it to implement IDataErrorInfo; if it did, the implementation would be trivial, because there would never be an error to report.

Does this mean that IDataErrorInfo is a redundant interface? Not at all, but it is important to realize that it's an Application Boundary concern instead of a Domain concern. At Application Boundaries, data entry errors will happen, and we must be able to cope with them appropriately; we don't want the application to crash by passing unvalidated data to DanishPhoneNumber's constructor.

Does this mean that we should duplicate domain logic at the Application Boundary? That should not be necessary. At first, we can apply a simple refactoring to the DanishPhoneNumber constructor:

public DanishPhoneNumber(int number)
{
    if (!DanishPhoneNumber.IsValid(number))
    {
        throw new ArgumentOutOfRangeException("number");
    }
    this.number = number;
}
 
public static bool IsValid(int number)
{
    return (112 <= number)
        && (number <= 99999999);
}

We now have a public IsValid method we can use to implement an IDataErrorInfo at the Application Boundary. Next steps might be to add a TryParse method.

IDataErrorInfo implementations are often related to input forms in user interfaces. Instead of crashing the application or closing the form, we want to provide appropriate error messages to the user. We can use the Domain Object to provide validation logic, but the concern is completely different: we want the form to stay open until valid data has been entered. Not until all data is valid do we allow the creation of a Domain Object from that data.

In short, if you feel tempted to add IDataErrorInfo to a Domain Class, consider whether you aren't about to violate the Single Responsibility Principle. In my opinion, this is the case, and you would be better off reconsidering the design.

Comments

The last time I presented a sample of an AutoFixture-based unit test, I purposely glossed over the state-based verification that asserted that the resulting state of the basket variable was that the appropriate Pizza was added:

Assert.IsTrue(basket.Pizze.Any(p =>
    p.Name == pizza.Name), "Basket has added pizza.");

The main issue with this assertion is that the implied equality expression is rather weak: we consider a PizzaPresenter instance to be equal to a Pizza instance if their Name properties match.

What if they have other properties (like Size) that don't match? If this is the case, the test would be a false negative. A match would be found in the Pizze collection, but the instances would not truly represent the same pizza.

How do we resolve this conundrum without introducing equality pollution? AutoFixture offers one option in the form of the generic Likeness<TSource, TDestination> class. This class offers convention-based test-specific equality mapping from TSource to TDestination and overriding the Equals method.

One of the ways we can use it is by a convenience extension method. This unit test is a refactoring of the test from the previous post, but now using Likeness:

[TestMethod]
public void AddWillAddToBasket_Likeness()
{
    // Fixture setup
    var fixture = new Fixture();
    fixture.Register<IPizzaMap, PizzaMap>();
 
    var basket = fixture.Freeze<Basket>();
 
    var pizza = fixture.CreateAnonymous<PizzaPresenter>();
    var expectedPizza = 
        pizza.AsSource().OfLikeness<Pizza>();
 
    var sut = fixture.CreateAnonymous<BasketPresenter>();
    // Exercise system
    sut.Add(pizza);
    // Verify outcome
    Assert.IsTrue(basket.Pizze.Any(expectedPizza.Equals));
    // Teardown
}

Notice how the Likeness instance is created with the AsSource() extension method. The pizza instance (of type PizzaPresenter) is the source of the Likeness, whereas the Pizza domain model type is the destination. The expectedPizza instance is of type Likeness<PizzaPresenter, Pizza>.

The Likeness class overrides Equals with a convention-based comparison: if two properties have the same name and type, they are equal if their values are equal. All public properties on the destination must have equal properties on the source.

This allows me to specify the Equals method as the predicate for the Any method in the assertion:

Assert.IsTrue(basket.Pizze.Any(expectedPizza.Equals));

When the Any method evalues the Pizze collection, it executes the Equals method on Likeness, resulting in a convention-based comparison of all public properties and fields on the two instances.

It's possible to customize the comparison to override the behavior for certain properties, but I will leave that to later posts. This post only scratches the surface of what Likeness can do.

To use Likeness, you must add a reference to the Ploeh.SemanticComparison assembly. You can create a new instance using the public constructor, but to use the AsSource extension method, you will need to add a using directive:

using Ploeh.SemanticComparison.Fluent;

Comments

In the next couple of weeks I will be giving a couple of talks in Copenhagen.

At Community Day 2010 I will be giving two talks on respectively Dependency Injection and TDD.

In early June I will be giving a repeat of my previous CNUG TDD talk.

One of Castle Windsor's facilities addresses wiring up of WCF services. So far, the sparse documentation for the WCF Facility seems to indicate that you have to configure your container in a global.asax. That's not much to my liking. First of all, it reeks of ASP.NET, and secondly, it's not going to work if you expose WCF over protocols other than HTTP.

However, now that we know that a custom ServiceHostFactory is effectively a Singleton, a much better alternative is to derive from the WCF Facility's DefaultServiceHost class:

public class FooServiceHostFactory : 
    DefaultServiceHostFactory
{
    public FooServiceHostFactory()
        : base(FooServiceHostFactory.CreateKernel())
    {
    }
 
    private static IKernel CreateKernel()
    {
        var container = new WindsorContainer();
 
        container.AddFacility<WcfFacility>();
 
        container.Register(Component
            .For<FooService>()
            .LifeStyle.Transient);
        container.Register(Component
            .For<IBar>()
            .ImplementedBy<Bar>());
 
        return container.Kernel;
    }
}

Although it feels a little odd to create a container and then not really use it, but only its Kernel property, this works like a charm. It correctly wires up this FooService:

public class FooService : IFooService
{
    private readonly IBar bar;
 
    public FooService(IBar bar)
    {
        if (bar == null)
        {
            throw new ArgumentNullException("bar");
        }
 
        this.bar = bar;
    }
 
    #region IFooService Members
 
    public string Foo()
    {
        return this.bar.Baz;
    }
 
    #endregion
}

However, instead of the static CreateKernel method that creates the IKernel instance, I suggest that the WCF Facility utilizes the Factory Method pattern. As the WCF Facility has not yet been released, perhaps there's still time for that change.

In any case, the WCF Facility saves you from writing a lot of infrastructure code if you would like to wire your WCF services with Castle Windsor.

Comments

For a while I've been wondering about the lifetime behavior of custom ServiceHostFactory classes hosted in IIS. Does IIS create an instance per request? Or a single instance to handle all requests?

I decided to find out, so I wrote a little test service. The conclusion seems to be that there is only a single instance that servers as a factory for all requests. This is very fortunate, since it gives us an excellent place to host a DI Container. The container can then manage the lifetime of all components, including Singletons that will live for the duration of the process.

If you are curious how I arrived at this conclusion, here's the code I wrote. I started out with this custom ServiceHostFactory:

public class PocServiceHostFactory : ServiceHostFactory
{
    private static int number = 1;
 
    public PocServiceHostFactory()
    {
        Interlocked.Increment(
            ref PocServiceHostFactory.number);
    }
 
    protected override ServiceHost CreateServiceHost(
        Type serviceType, Uri[] baseAddresses)
    {
        return new PocServiceHost(
            PocServiceHostFactory.number, serviceType,
            baseAddresses);
    }
}

The idea is that every time a new instance of ServiceHostFactory is created, the static number is incremented.

The PocServiceHostFactory just forwards the number to the PocServiceHost:

public class PocServiceHost : ServiceHost
{
    public PocServiceHost(int number, Type serviceType,
        Uri[] baseAddresses)
        : base(serviceType, baseAddresses)
    {
        foreach (var cd in 
            this.ImplementedContracts.Values)
        {
            cd.Behaviors.Add(
                new NumberServiceInstanceProvider(
                    number));
        }
    }
}

The PocServiceHost just forwards the number to the NumberServiceInstanceProvider:

public class NumberServiceInstanceProvider : 
    IInstanceProvider, IContractBehavior
{
    private readonly int number;
 
    public NumberServiceInstanceProvider(int number)
    {
        this.number = number;
    }
 
    #region IInstanceProvider Members
 
    public object GetInstance(
        InstanceContext instanceContext,
        Message message)
    {
        return this.GetInstance(instanceContext);
    }
 
    public object GetInstance(
        InstanceContext instanceContext)
    {
        return new NumberService(this.number);
    }
 
    public void ReleaseInstance(
        InstanceContext instanceContext,
        object instance)
    {
    }
 
    #endregion
 
    #region IContractBehavior Members
 
    public void AddBindingParameters(
        ContractDescription contractDescription,
        ServiceEndpoint endpoint,
        BindingParameterCollection bindingParameters)
    {
    }
 
    public void ApplyClientBehavior(
        ContractDescription contractDescription,
        ServiceEndpoint endpoint,
        ClientRuntime clientRuntime)
    {
    }
 
    public void ApplyDispatchBehavior(
        ContractDescription contractDescription,
        ServiceEndpoint endpoint,
        DispatchRuntime dispatchRuntime)
    {
        dispatchRuntime.InstanceProvider = this;
    }
 
    public void Validate(
        ContractDescription contractDescription,
        ServiceEndpoint endpoint)
    {
    }
 
    #endregion
}

The relevant part of NumberServiceInstanceProvider is the GetInstanceMethod that simply forwards the number to the NumberService:

public class NumberService : INumberService
{
    private readonly int number;
 
    public NumberService(int number)
    {
        this.number = number;
    }
 
    #region INumberService Members
 
    public int GetNumber()
    {
        return this.number;
    }
 
    #endregion
}

As you can see, NumberService simply returns the injected number.

The experiment is now to host NumberService in IIS using PocServiceHostFactory. If there is only one ServiceHostFactory per application process, we would expect that the same number (2) is returned every time we invoke the GetNumber operation. If, on the other hand, a new instance of ServiceHostFactory is created per request, we would expect the number to increase for every request.

To test this I spun up a few instances of WcfTestClient.exe and invoked the operation. It consistently returns 2 across multiple clients and multiple requests. This supports the hypothesis that there is only one ServiceHostFactory per service process.

Comments

My book contains a section on the Ambient Context pattern that uses a TimeProvider as an example. It's used like this:

this.closedAt = TimeProvider.Current.UtcNow;

Yesterday I was TDDing a state machine that consumes TimeProvider and needed to freeze and advance time at different places in the test. Always on the lookout for making unit tests more readable, I decided to have a little fun with literal extensions and TimeProvider. I ended up with this test:

// Fixture setup
var fixture = new WcfFixture();
 
DateTime.Now.Freeze();
 
fixture.Register(1.Minutes());
var sut = fixture.CreateAnonymous<CircuitBreaker>();
sut.PutInOpenState();
 
2.Minutes().Pass();
// Exercise system
sut.Guard();
// Verify outcome
Assert.IsInstanceOfType(sut.State,
    typeof(HalfOpenCircuitState));
// Teardown

There are several items of note. Imagine that we can freeze time!

DateTime.Now.Freeze();

With the TimeProvider and an extension method, we can:

internal static void Freeze(this DateTime dt)
{
    var timeProviderStub = new Mock<TimeProvider>();
    timeProviderStub.SetupGet(tp => tp.UtcNow).Returns(dt);
    TimeProvider.Current = timeProviderStub.Object;
}

This effectively sets up the TimeProvider to always return the same time.

Later in the test I state that 2 minutes pass:

2.Minutes().Pass();

I particularly like the grammatically correct English. This is accomplished with a combination of a literal extension and changing the state of TimeProvider.

First, the literal extension:

internal static TimeSpan Minutes(this int m)
{
    return TimeSpan.FromMinutes(m);
}

Given the TimeSpan returned from the Minutes method, I can now invoke the Pass extension method:

internal static void Pass(this TimeSpan ts)
{
    var previousTime = TimeProvider.Current.UtcNow;
    (previousTime + ts).Freeze();
}

Note that I just add the TimeSpan to the current time and invoke the Freeze extension method with the new value.

Last, but not least, I should point out that the PutInOpenState method isn't some smelly test-specific method on the SUT, but rather yet another extension method.

I recently had the need to change the lifestyles of all components in a WindsorContainer (read on to the end if you want to know why). This turned out to be amazingly simple to do.

The problem was this: I had numerous components registered in a WindsorContainer, some of them as Singletons, some as Transients and yet again some as PerWebRequest. Configuration was even defined in numerous IWindsorInstallers, including some distributed XML files. I now needed to spin up a second container with the same configuration as the first one, except that the lifestyles should be all Singletons across the board.

This can be easily accomplished by implementing a custom IContributeComponentModelConstruction type. Here's a simple example:

Consider this IWindsorInstaller:

public class FooInstaller : IWindsorInstaller
{
    #region IWindsorInstaller Members
 
    public void Install(IWindsorContainer container,
        IConfigurationStore store)
    {
        container.Register(Component
            .For<IFoo>()
            .ImplementedBy<Foo>()
            .LifeStyle.Transient);
    }
 
    #endregion
}

The important point to notice is that it registers the lifestyle as Transient. In other words, this container will always return new Foo instances:

var container = new WindsorContainer();
container.Install(new FooInstaller());

We can override this behavior by adding this custom IContributeComponentModelConstruction:

public class SingletonEqualizer :
    IContributeComponentModelConstruction
{
    public void ProcessModel(IKernel kernel, 
        ComponentModel model)
    {
        model.LifestyleType = LifestyleType.Singleton;
    }
}

In this very simple example, I always set the lifestyle type to the same value, but obviously we can write as complex code in the ProcessModel method as we would like. We can now configure the container like this:

var container = new WindsorContainer();
container.Kernel.ComponentModelBuilder
    .AddContributor(new SingletonEqualizer());
container.Install(new FooInstaller());

With this configuration we will now get the same instance of Foo every time we Resolve IFoo.

We did I need this? Because my application is a web application and I'm using the PerWebRequest lifestyle in a number of places. However, I needed to spin up a second container that would compose object hierarchies for a background process. This background process needs the same component configuration as the rest of the application, but can't use the PerWebRequest lifestyle as there will be no web request available to the background process. Hence the need to change lifestyles across the board.

About a month ago I decided to migrate from MSTest to xUnit.net, and while I am still in the process, I haven't regretted it yet, and I don't expect to. AutoFixture has already moved over, and I'm slowly migrating all the sample code for my book.

Recently I was asked why, which prompted me to write this post.

I'm not moving away from MSTest for one single reason. It's rather like lots of small reasons.

When I originally started out with TDD, I used nUnit - it was more or less the only unit testing framework available for .NET at the time. When MSTest came, the change was natural, since I worked for Microsoft at the time. This is not the case anymore, but it still took me most of a year to finally abandon MSTest.

There was one thing that really made me cling to MSTest, and that was the IDE integration, but over time, I started to realize that this was the only reason, and even that was getting flaky.

When I started to think about all the things that left me dissatisfied, making the decision was easy:

• First of all, MSTest isn't extensible, but xUnit.net is. In xUnit.net, I can extend the Fact or Theory attributes (and I intent to), while in MSTest, I will have to play with the cards I've been dealt. I think I could live with all the other issues if I could just have this one, but no.
• MSTest has no support for parameterized test. xUnit.net does (via the Theory attribute).
• MSTest has no Assert.Throws, although I requested this feature a long time ago. Now Visual Studio 2010 is out, but Assert.Throws is still nowhere in sight.
• MSTest has no x64 support. Tests always run as x86. Usually it's no big deal, but sometimes it's a really big deal.
• In MSTest, to write unit tests, you must create a special Unit Test Project, and those are only available for C# and VB.net. Good luck trying to write unit tests in a more exotic .NET language (like F# on Visual Studio 2008). xUnit.net doesn't have this problem.
• MSTest uses Test Lists and .vsmdi files to maintain test lists. Why? I don't care, I just want to execute my tests, and the .vsmdi files are in the way. This is particularly bad when you use TFS, but I'm also moving away from TFS, so that wouldn't have continued to be that much of an issue. Still: try having more than one .sln file with unit tests in the same folder, and watch funny things happen because they need to share the same .vsmdi file.
• I suppose it's because of the .vsmdi files, but sometimes I get a Test run error if I delete a test and run the tests immediately after. That's a false positive, if anyone cares.
• MSTest gives special treatment to its own AssertionException, which gets nice formatting in the Test Results window. All other exceptions (like verification exceptions thrown by Moq or Rhino Mocks are rendered near-intelligible because MSTest thinks it's very important to report the fully qualified name of the exception before its message. Most of the time, you have to open the Test Details window to see the exception message.
• Last, but not least, I often get cryptic exception messages like this one: Column 'id_column, runid_column' is constrained to be unique.  Value '8c84fa94-04c1-424b-9868-57a2d4851a1d, d7471c5e-522f-43d3-b2c5-8f5cab55af0e' is already present. This appears in a very annoying modal MessageBox, but clicking OK and retrying usually works, although sometimes it even takes two or three attempts before I can get past this error.

It not one big thing, it's just a lot of small, but very annoying things. After three iterations (VS2005, VS2008 and now VS2010) these issue have still to be addressed, and I got tired of waiting.

So far, I can only say that I have none of these problems with xUnit.net and the IDE integration provided by TestDriven.NET. It's just a much smoother experience with much less friction.

Comments

AutoFixture 1.1 is now available on the CodePlex site! Compared to the Release Candidate, there are no changes.

The 1.1 release page has more details about this particular release, but essentially this is the RC promoted to release status.

Release 1.1 is an interim release that addresses a few issues that appeared since the release of version 1.0. Work continues on AutoFixture 2.0 in parallel.

It seems to me that I've lately encountered a particular mindset towards Dependency Injection (DI). People seem to think that it's only really good for replacing one data access implementation with another. Once you get to that point, you know that the following argument isn't far behind:

“That's all well and good, but we know for certain that we will never exchange [insert name of RDBMS here] with anything else in this application.”

Apart from the hubris of making such a bold statement about the future of any software endeavor, such a statement reveals the narrow view on DI that its only purpose is for replacing data access components - and perhaps for unit testing.

Those are relevant reasons for using DI, but they are only some of the reasons. Let's briefly revisit why we employ DI.

We use DI to enable loose coupling.

DI is only a means to an end. Even if you never intend to replace your database and even if you never want to write a single unit test, DI still offers benefits in form of a more maintainable code base. The loose coupling gives you better separation of concerns because it allows you to apply the Open/Closed Principle.

Example coming right up:

Imagine that we need to implement a PrécisViewModel class with a TopSellers property that returns an IEnumerable<string>. To implement this class, we have a data access component. Let's use the ubiquitous Repository pattern and define IProductRepository to see where that leads us:

public interface IProductRepository
{
    IEnumerable<Product> SelectTopSellers();
}

We can now implement PrécisViewModel like this:

public class PrécisViewModel
{
    private readonly IProductRepository repository;
 
    public PrécisViewModel(IProductRepository repository)
    {
        if (repository == null)
        {
            throw new ArgumentNullException("repository");
        }
 
        this.repository = repository;
    }
 
    public IEnumerable<string> TopSellers
    {
        get
        {
            var topSellers = 
                this.repository.SelectTopSellers();
            return from p in topSellers
                   select p.Name;
        }
    }
}

Nothing fancy is going on here. It's just straight Constructor Injection at work.

Obviously, we can now implement and use a SQL Server-based repository:

var repository = new SqlProductRepository();
var vm = new PrécisViewModel(repository);

So what does all this loose coupling buy us? It doesn't seem to help us a lot.

The real benefit is not yet apparent, but it should become more obvious when we start adding requirements. Let's start with some caching. It turns out that the SelectTopSellers implementation is slow, so we would like to add some caching somewhere.

Where should we add this caching functionality? Without loose coupling, we would more or less be constrained to adding it to either PrécisViewModel or SqlProductRepository, but both have issues:

• First of all we would be violating the Single Responsibility Principle (SRP) in both cases.
• If we implement caching in PrécisViewModel, other consumers of the SelectTopSellers would not benefit from it.
• If we implement caching in SqlProductRepository, it wouldn't be available for any other IProductRepository implementations.

Since the premise for this post is that we will never use any other database than SQL Server, implementing caching directly in SqlProductRepository sounds like the correct choice, but we would still be violating the SRP, and thus making our code more difficult to maintain.

A better solution is to introduce a caching Decorator like this one:

public class CachingProductRepository : IProductRepository
{
    private readonly ICache cache;
    private readonly IProductRepository repository;
 
    public CachingProductRepository(
        IProductRepository repository, ICache cache)
    {
        if (repository == null)
        {
            throw new ArgumentNullException("repository");
        }
        if (cache == null)
        {
            throw new ArgumentNullException("cache");
        }
 
        this.cache = cache;
        this.repository = repository;
    }
 
    #region IProductRepository Members
 
    public IEnumerable<Product> SelectTopSellers()
    {
        return this.cache
            .Retrieve<IEnumerable<Product>>("topSellers",
                this.repository.SelectTopSellers);
    }
 
    #endregion
}

For completeness sake is here the definition of ICache:

public interface ICache
{
    T Retrieve<T>(string key, Func<T> readThrough);
}

The point is that CachingProductRepository extends any IProductRepository we provide to it (including SqlProductRepository) without modifying it. Thus, we have satisfied both the OCP and the SRP.

Just to drive home the point, let us assume that we also wish to record execution times for various methods for purposes of SLA compliance. We can do this by introducing yet another Decorator:

public class PerformanceMeasuringProductRepository : 
    IProductRepository
{
    private readonly IProductRepository repository;
    private readonly IStopwatch stopwatch;
 
    public PerformanceMeasuringProductRepository(
        IProductRepository repository, 
        IStopwatch stopwatch)
    {
        if (repository == null)
        {
            throw new ArgumentNullException("repository");
        }
        if (stopwatch == null)
        {
            throw new ArgumentNullException("stopwatch");
        }
 
        this.repository = repository;
        this.stopwatch = stopwatch;
    }
 
    #region IProductRepository Members
 
    public IEnumerable<Product> SelectTopSellers()
    {
        var timer = this.stopwatch
            .StartMeasuring("SelectTopSellers");
        var topSellers = 
            this.repository.SelectTopSellers();
        timer.StopMeasuring();
        return topSellers;
    }
 
    #endregion
}

Once again, we modified neither SqlProductRepository nor CachingProductRepository to introduce this new feature. We can implement security and auditing features by following the same principle.

To me, this is what loose coupling (and DI) is all about. That we can also replace data access components and unit test using dynamic mocks are very fortunate side effects, but the loose coupling is valuable in itself because it enables us to write more maintainable code.

We don't even need a DI Container to wire up all these repositories (although it sure would could be helpful). Here's how we can do it with Pure DI:

IProductRepository repository =
    new PerformanceMeasuringProductRepository(
        new CachingProductRepository(
            new SqlProductRepository(), new Cache()
            ),
        new RealStopwatch()
    );
var vm = new PrécisViewModel(repository);

The next time someone on your team claims that you don't need DI because the choice of RDBMS is fixed, you can tell them that it's irrelevant. The choice is between DI and Spaghetti Code.

Comments