Now that i have presented a simple ControlStateMachine i can raise the bar a little. A statemachine that handles commands. Here is how a developer should be able to initialize this machine:

sut.WhenIn(States.Loading)
 .On(Commands.Next)
  .Do(() => Console.WriteLine("got next command while loading..."))
  .Do(() => Console.WriteLine("doing it again..."))
 .On(Commands.Previous)
  .Do(() => Console.WriteLine("got previous command while loading..."));

sut.WhenIn(States.Ready)
 .On(Commands.Previous)
  .Do(() => Console.WriteLine("got previous command while ready..."));

So how should we define our methods to accomplish this initialization style? Let’s begin with identifying the methods we need.

• WhenIn(TSTate state)
• On(TCommand command)
• Do(Action action)

Next thing to do is analyze in which sequence these methods can be called:

Ok, now that we have clarified the requirements a little we can start working on a solution. Let’s start with defining an interface for each of the methods:

interface IChooseState<tstate, TCommand> { Q1 WhenIn(TState state); }
interface IChooseCommand<tstate, TCommand> { Q2 On(TCommand command); }
interface IChooseAction<tstate, TCommand> { Q3 Do(Action action); }

From WhenIn we need to be able to call On. Thus Q1 = IChooseCommand<TState, TCommand>. Q2 is also easily solved because from On we only have to be able to call Do, thus Q2 = IChooseAction<TState, TCommand>.

From Do we should be able to call both On and Do. We can do that by defining another interface which has both methods:

interface IChooseCommandAndAction<tstate, TCommand> : IChooseCommand<tstate, TCommand>, IChooseAction<tstate, TCommand> { }

Now that we have found answers for Q1, Q2 and Q3 we can define the API for initializing our StateMachine as following:

IChooseCommand<tstate, TCommand> WhenIn(TState state);
IChooseAction<tstate, TCommand> On(TCommand command);
IChooseCommandAndAction<tstate, TCommand> Do(Action action);

Now tell me about your strategy for implementing a fluent interface!