Re: [NMock2-Dev] Feature Request: "Stub" should have alower prioritythen "Expect"

[<rdr...@ar...>]

Currently my tests are not brittle... I was explaining the the missing ability to override an existing stub would make the tests more brittle. 

I finally gave in and will "reverse the setup", but i still feel its odd...  and it will cause me to add "extra" initialisation code into my test methods... (Which was hidden inside the [setup] before, since it was unimportant to the test)

I feel if JMock had the ability to override, and the sequence in which the expections are evaluated is reversed, then there is a feature that has been removed... Its a feature I expected so naturally, i was quite confused when it was not there...

Dont get me wrong though... I love NMock and enjoy using it... Or else i wouldnt complain ;)

----- Original Nachricht ----
Von:     Nat Pryce <nat...@gm...>
An:      NMock2 Development Discussion <nmo...@li...>
Datum:   12.09.2007 21:44
Betreff: Re: [NMock2-Dev] Feature Request: "Stub" should have alower
	prioritythen "Expect"

> One thing that might be worth investigating is why the tests are so
> brittle.  Why does a change to a class (adding a new event) cause many
> tests that are testing other aspects of that class to fail.  Maybe
> events can be split up somehow and tested separately.
> 
> --Nat
> 
> On 12/09/2007, rdr...@ar... <rdr...@ar...> wrote:
> > The problem still remains...
> >
> > I would have to do a lot more "Copy&Paste" inheritance if i am unable to
> override a speccific setup. For a simple view with 5 Events i would have to
> create 5 different setup functions, which would all differe in one line. And
> there is no way to refactor this out.
> >
> > Thats 25 (n^2) Lines of mock setup vs 10 (2 * n) lines... and the number
> only gets bigger for more events. The code is not complex, but it leads to
> maintenance problems if there is a change...
> >
> > If I want to add another event to the view, this would cause no less then
> 5 failed tests, and force me to update 5 Setup-functions, and create a new
> one from scratch (by Copy and Paste). on the other Version I would only have
> to update the generic setup function, and then add one event in my
> test-method... (or use the hack which I find ugly...)
> >
> > I like my testcases very short and simple... I usualy feel bad when they
> reach 10 lines, since I consider that to complex...
> >
> > Hope this makes sense...
> >
> > Heiko