DrillSergeant 0.0.3-alpha

DrillSergeant

.net behavior driven testing written by developers, for developers.

Summary

DrillSergeant is a behavior testing library that empowers developers to apply BDD practices with minimal amount of friction. Simply import the package and write your behaviors in familiar C# syntax.

Getting Started

For a complete example of a feature, see this

Creating a Behavior

Creating a behavior is very simple:

[Behavior, Theory, InputData(1,1)]
public Behavior MyBehaviorTest(int value1, int value2)
{
  var behavior = new Behavior<Context,Input>()
    // Configure behavior here...
    
  return behavior;
}

Behaviors are regular test methods that are decorated with the [Behavior] attribute and return an instance of a Behavior class. Because [Behavior] is built on top of [Theory] one or more inputs must be provided to the test. This can be done through any xunit data discovery mechanism (e.g. [InlineData], [MemberData], etc...).

Context and Input

DrillSergeant is built on top of xunit and makes use of [Theory] based tests. As a result behavior tests require both a context to hold state throughout the test and input to drive the test. These are typically defined using the C# record type:

public record Context();
public record Input();

Each step within a behavior returns an updated context, which is then fed into the next step. Input on the other hand is immutable and does not change between steps. While it's not a requirement to use the record type, it is preferred because they're immutable by default and have syntax that makes copying them easy:

return context with { UpdatedField = "new_value" };

Configuring Input and Context

Use the WithInput() and WithContext() methods to configure the behavior:

var behavior = new Behavior<Context,Input>()
  .WithInput(() => new Input())
  .WithContext(() => new Context());

The WithInput() method is used to map the arguments passed to the test method to the Input type used by the test. Likewise, the WithContext() method is used to establish the initial context for the first step.

Configuring Steps

Individual steps can be configured dependeing on the level of granularity required.

Inline Steps

Inline steps are the simplest type of step. An inline step can be added simply by calling Given()/When()/Then() and passing in a lambda:

Given("My step", (c,i) => {
    // Perform some action
    return c with { /* changes */ };
});

All steps pass the context and input as the first two parameters. To pass additional dependencies you can call one of the generic overrides:

Given<MyDependency>("My step", (c,i, dep) => {
    // Perform some action
    return c with { /* changes */ };
});

Inline steps are convenient when you need a one-off step that won't be reused in other behaviors.

Lambda Steps

Lambda steps are ideal for situations where a step needs to be reused for multiple behaviors within a single class:

public LambdaStep<Context,Input> MyStep =>
    new GivenLambdaStep<Context,Input>()
		.Named("My step")
	    .Handle( (c,i) => {
		    // Perform some action.
			return c with { /* changes */ };
		});

As you can see, the syntax is nearly identical to an inline step.

Class Steps

Class steps are the most flexible type of step and best used when a particular step needs to be reused between multiple features. To create a class step, override the desired verb and fill in the step method:

public class MyStep<Context,Input> : GivenStep<Context,Input>
{
  public override Context Given(Context context, Input input)
  {
      // Perform some action.
	  return context with { /* changes */ };
  }
}

Unlike inline and lambda steps, class steps are convention based. By default, The GivenStep, WhenStep, and ThenStep provide virtual methods for convenience, but it is not required to use them. Internally, DrillSergeant will pick a matching verb method with the most parameters. For example:

public class MyStep<Context,Input> : GivenStep<Context,Input>
{
    // DrillSergeant will not excute this.
    public override Context Given(Context context, Input input)
    {
        // Perform some action.
	    return context with { /* changes */ };
    }
  
    // DrillSergeant will execute this.
    public override Context Given(Context context, Input input, MyDependency dependency)
    {
        // Perform some action.
	    return context with { /* changes */ };
    }
}

Configuring the Resolver

DrillSergeant has first-class support for dependency injection. When writing a behavior method, simply prefix any dependency with the [Inject] attribute:

public Behavior MyBehavior([Inject] MyDependency dependency)
{
    // ...
}

The [Inject] parameter is needed so that DrillSergeant can differentiate between input parameters passed by [Theory] and what it needs to inject.

Dependency resolution is handled with the IDependencyResolver interface, which contains a single method: object Resolve(Type type). By default DrillSergeant will satisfy dependencies by instantiating new instances of them via the Activator.CreateInstance() method. To override this: a custom resolver can be configured in the behavior class:

[BehaviorResolverSetup]
public IDependencyResolver SetupResolver()
{
  var resolver = A.Fake<IDependencyResolver>();
  
  A.CallTo(() => resolver.Resove(typeof(MyDependency))).Returns(new MyDependency);
}

In this example, we're using the mocking library FakeItEasy to create a resolver that returns instances of the required dependency, but for more advanced scenarios a real DI container can be substituted in its place.

Note: The resolver is scoped to each test case and does not share data between tests. To share data, use xunit's ClassFixture and CollectionFixture fixtures. **Note: The name of the method here is unimportant. DrillSergeant will look for the first public method returning an IDependencyResolver that is marked with the [BehaviorSetupResolver] attribute.c

Best Practices

No Logic in Behaviors

Internally DrillSergeant will execute the behavior method prior to executing any test cases, therefore it's important not to write any logic within the behavior itself. The behavior should only be a single a single return statement.

public Behavior MyBehavior([Inject] MyDependency dependency)
{
    dependency.Initialize(); // Don't do this!!!
    
    return new Behavior<Context,Input>();
}

If you need to perform initialization on dependencies prior to running the test, they should be performed within the SetupResolver() method.

Don't Mutate Context. Instead Return New Context

Future editions of DrillSergeant will keep a running history of context as it executes behavior. Therefore it's a good idea to make sure that when returning from a step that a new context is created based on the previous step. This can be accomplished easily using the with construct.

Don't Mutate Input. Just Don't.

The input for each behavior scenario is tied to the test results in xunit. Once the input has been mapped with the WithInput() method, do not modify it again.

Don't Map Input or Context More Than Once

The WithInput() and WithContext() methods are only executed once prior to executing any steps. Calling either method multiple times with only result in the previous handler being overwritten.