On 13 Aug 2013, at 13:16, Andy Lindeman <a...@andylindeman.com> wrote:
> The problem with not_to raise_error is that it's a tad confusing. It
> could be interpreted either that you meant no error at all to be raised,
> or an error to be raised but not of the given type.
Hmmm, I always interpreted it as "I don't care what this block does, as long as
it doesn't raise a FooError", which I suppose gives a third possible
interpretation. I'd never considered the other interpretations, so I hadn't
seen the potential for confusion myself until you pointed this out.
> I'd probably write it it as a positive:
>
> expect { actor.method_that_raises_an_error }.to
> raise_error(SpecificTypeOfErrorImExpecting)
>
> This will fail if the error ended up being a DeadActorError or if the
> method wasn't called but didn't raise an error.
>
> Does this help at all? I hope I understood you fully.
So this only solve a problem, but only one of them. The general pattern for
this is that I need to specify two things in my code:
* A certain interesting piece of behaviour happens
* A certain type of error which a naive implementation would raise, isn't raised
In this case, the interesting behaviour is that an ArgumentError is raised, so
I have this example (to save time replying, I'm no longer simplifying it, this
is my actual example):
it "is treated as the multiple arument case" do
result.foo("one", "two", "three")
expect { value }.to raise_error(ArgumentError, /3 for 1/)
end
Now I know that a naive implementation of the code behind `value` will crash
the actor. So I need a separate example to check this is the case. But I don't
want any knowledge of the behaviour of the method leaking into this example, so
I'm currently resorting to this:
it "doesn't crash the actor" do
result.foo("one", "two", "three")
expect {
pass_handlers_to_result rescue nil
begin
pass_handlers_to_result
rescue Exception => e
raise if e.is_a?(Celluloid::DeadActorError)
end
}.to_not raise_error
end
Which I'm sure you will agree, while arguably less ambiguous than `to_not
raise_error(Celluloid::DeadActorError)`, lacks some of its terseness.
Hopefully this details the problem a bit more. Can you think of any
improvements? I'm still at a loss unfortunately.
Cheers
Ash
--
http://www.patchspace.co.uk/
http://www.linkedin.com/in/ashmoran
smime.p7s
Description: S/MIME cryptographic signature
_______________________________________________ rspec-users mailing list rspec-users@rubyforge.org http://rubyforge.org/mailman/listinfo/rspec-users
