Thanks a lot David, this helps a lot. I have been trying to do just as
you recommended and letting the stories drive every aspect implicitly
without me explicitly defining the model methods in it.
The confusion I was having was that when you say you are going to follow
the failing stories to drive the process the first thing that usually
will fail will be in a model.. Meaning, you usually have a given clause
that uses the model to set up the DB. Pat Maddox, in his Story Runner
screencast, showed how to effectively stub out the model in the story
runner with a "class ModelName; end" type approach. However, this does
not always work depending on the type of methods you will be calling on
the model... Hmm.. actually I guess you could just add something like
this, that would make that class/object act like a mock that receives
anything:
#Top of story file that uses UndefinedModel
class UndefinedModel
def initialize(*args); end
def method_missing
#I eat all methods
end
end
So stuff like this should work just fine:
Given....
u = UndefinedModel.new(:some_attribute => "value"
u.some_other_method
Doing this I suppose would allow the story runner to pass all of the
steps until it hit the view steps which in that case, like you
suggested, you could drop down to the view specs and then work your way
towards defining the real model.
I'll play around this approach and see where it takes me. It should be
easy to leverage the current mocking/stubbing framework and create some
helpers to allow for such a process to be relatively painless. Thanks
again!
-Ben
David Chelimsky wrote:
> On Nov 8, 2007 12:17 AM, Ben Mabey <[EMAIL PROTECTED]> wrote:
>
>> David Chelimsky wrote:
>>
>
> <snip>
>
>
>>> ... the philosohy that Dan is
>>> espousing: expressing stories in the business domain rather than the
>>> UI domain (btw Dan, that was brilliantly put).
>>>
>> What exactly do you and Dan mean when you say that the stories should be
>> expressed in the business domain rather than the UI domain? (BTW, is
>> there a link to a Dan North post abou that?)
>>
>
> Sorry - there are two different threads going on about this topic
> right now - the other one is on the rspec-devel list and that was
> where Dan posted that statement:
>
> http://rubyforge.org/pipermail/rspec-devel/2007-November/004259.html
>
> Keep in mind that User Stories are oft described as "a token for a
> conversation", and that even with acceptance criteria spelled out,
> whether the button says "Create New Account" or "Enter" is rarely of
> sufficient business value to express that in a story. That's the sort
> of detail that comes out when you say to your customer "the entry form
> is done, why don't you come by so I can show it to you," as opposed to
> in the iteration planning meeting.
>
> In the end it really does boil down to what the customer feels is
> necessary in order to accept the software. If the customer really DOES
> care about what the button says, he or she may want that expressed in
> a story. But even then, saying "And I press Create New Account" still
> leaves things flexible enough so that you could be describing a web
> app, a GUI app, a touch-screen app, etc. The only thing this doesn't
> work for would be a command line app. So maybe "And I tell the system
> to Create New Account" would leave room for that as well. Or maybe,
> "And I fold my arms and command the system to Create New Account." I
> kinda like that one!
>
>
>> From my understanding you
>> are saying that the steps that say "user types in such and such" and
>> user "hits the login button" are too UI specific and don't really have
>> much to do with the business domain. Is that correct?
>>
>
> Unless the business IS software, like a text editor, yes.
>
>
>> I think that this is a part of writing stories that I am somewhat
>> struggling with, meaning how specific should these stories be in
>> explaining the view? At RubyConf during your presentation (great job,
>> BTW!) you mentioned how you like to spec out your views first.
>>
>
> Ah - confusion. I DO like to spec my views first - when I get down to
> the Spec Framework.
>
> In our example at RailsConf EU, we talked about a Cup Tracker (as in
> Rugby, which was going on at the time). Take a look at the example
> story:
>
> http://rspec.rubyforge.org/svn/branches/railsconfeu2007/tags/chapter1/stories/plan_cup.rb
>
> This story really rides this whole line very nicely. We're describing
> software that presents a view of something abstract, and we do so with
> some detail about what it presents - but there is nothing in it that
> says "when I click this button" or "when I enter this text."
>
> Now look at this version (same story, w/ the steps filled in):
>
> http://rspec.rubyforge.org/svn/branches/railsconfeu2007/tags/chapter4/stories/plan_cup.rb
>
> Note how the implementations of the steps are starting to get into the
> views. That's where that level of detail starts to play out.
>
> Now look at this:
>
> http://rspec.rubyforge.org/svn/branches/railsconfeu2007/tags/chapter4/spec/views/cups/show.rhtml_spec.rb
>
> That spec was arrived at with the very granular red-green-refactor
> process of TDD with a focus, naturally, on behaviour, design and
> documentation (It just occurs to me that behaviour, design and
> documentation are B, D and D - interesting ...).
>
> Now some people look at that spec and scream "holy crap - look at all
> the duplication - all that mocking - so brittle!." I won't disagree
> that there is lots of duplication with other parts of the system, a
> lot of mocking, and changes to the views would require changes to this
> spec (brittle). And, looking back at this, I might want to break that
> spec up into more granular examples. While it speaks very well looking
> at the code, running it would only tell you that "/cups/show.rhtml
> should display the chart."
>
> But check this out! The implementation of that spec, with all that
> mocking, IS the spec for the controller and model. That single spec
> tells us about the structure that the model should expose (NOT the
> actual structure necessarily - just what it should expose to things
> like views that want the models to be easy to use) and what the
> controller should provide for the view.
>
> So this is all about discovery - starting from the outside and moving
> in. Implementing the steps that describe domain concepts help us to
> discover some details of the views. Implementing the view specs help
> us to discover the services we want from our controller and the APIs
> we want from our model. I find that to be very compelling.
>
> <snip>
>
>
>> Where is the happy medium, in your opinion, between
>> keeping the stories concise and letting them drive every aspect of the
>> development? Maybe I am wrong in saying that the stories should drive
>> EVERY aspect of the development?
>>
>
> Well, they should drive every aspect, but not directly. You're not
> going to describe database table structure in your stories, but in the
> end those structures end up as they do because of the stories.
>
>
>>>> The Selenium integration is also an interesting idea that you might want
>>>> to read about if you haven't seen this post. I'm still trying to
>>>> decided if using Selenium is worth the extra effort (my main goal would
>>>> be to test the JS during the acceptance tests)
>>>>
>
> <snip>
>
>
>>> Generally speaking, in-browser tests tend to be incredibly brittle and
>>> run dog-slow. They simply have no choice but to be tied directly to
>>> low-level implementation details like html structure and element IDs.
>>>
>>> So I'd recommend only testing what you absolutely must in-browser. But
>>> given our ajax-laden world, "what you absolutely must" may well turn
>>> out to be a lot!
>>>
>
> <snip>
>
>
>> I have heard that Selenium suites can become out of hand and end of
>> being more hassle than what there worth. Thanks for the recommendation,
>> I think that sounds like a sensible one.
>>
>> Thanks for taking the time to discuss this,
>>
>
> My pleasure!
>
> Cheers,
> David
>
>
>> Ben
>>
> _______________________________________________
> rspec-users mailing list
> rspec-users@rubyforge.org
> http://rubyforge.org/mailman/listinfo/rspec-users
>
_______________________________________________
rspec-users mailing list
rspec-users@rubyforge.org
http://rubyforge.org/mailman/listinfo/rspec-users
