Re: [NMock2-Dev] Refactorable calls

[Nat P. <nat...@gm...>]

On 23/01/2008, Steve Mitcham <Ste...@ty...> wrote:
> I've looked a jMock's implementation and I've looked at Rhino. I don't
> really like the two line setups with the Call/LastCall that is the same as
> the Rhino expectations.  I'm trying to think of an 'NMock-like'
> implementation of the same thing that keeps the fluent interface going.

JMock's API was a desperate attempt to work around Java's butt-ugly
syntax for closures.  C# has a cleaner syntax, so you could do
something like the example below.

   A mockA = mockery.mock(typeof(A));
   B mockB = mockery.mock(typeof(B));

   mockery.Checking(delegate {
       mockA.DoSomething(EqualTo(B), NotNull);
       Once; mockB.GetStuff(); Will(Return("stuff"));
   });


The Checking method would put the mockery into a "record" state, call
the delegate, and then put the mockery back into a "check" state.

That would be much easier to implement in C# than Java because C# has
real generics and closures.

Even better would be something like:

mockery.Checking(delegate {
    ... run the test in here
}.WillCause(delegate {
    ... expectations in here
});

Which makes the expectations read more like assertions. It puts the
tested code and the expectations in causal order, which is easier to
read I think.

--Nat