[rspec-users] Stub that returns hash values

[Ashley Moran]

On 21 Oct 2009, at 19:18, Stephen Eley wrote:
> On Wed, Oct 21, 2009 at 1:12 PM, Carl Graff <cagraff at cox.net> wrote:
>>
>> In truth, due to my inexperience and confusion, mocks seem to slow my
>> development more than just creating real objects. But since there  
>> has been
>> so much effort to put these into testing frameworks, I think it  
>> must be
>> important to try and learn when it is appropriate to use them.
>
> That is a fallacious line of reasoning.  A lot of effort has also gone
> into American football, every Michael Bay movie, and Windows Vista.
> QED.

To be fair, Carl did say he wanted to "learn when it is appropriate to  
use them" as opposed to just using them indiscriminately.  And the  
answer may be "never".  I suppose it's quite possible that mocks do  
suck, and we just haven't realised yet.  But that's ok, they suck less  
than inspecting instance variables, so I'm happy, at least.

What I would suggest is to not try to learn how to use mocks on  
production code*, unless you have pretty thorough means of integration  
testing.  Incorrect mocking leads to brittle specs* and that leads to  
hidden bugs* and thrashing during refactoring*.

In fact, if you don't use a tool such as Cucumber, I'd recommend  
learning that over mocking first.  (You don't need to use Cucumber,  
but it's more suited for high-level descriptions than RSpec.)  As  
RSpec lets you refactor without risk of breaking units of code,  
Cucumber lets you refactor your code and specs across units.  That  
gives you a metric on how well you're using mocks - unexpected  
Cucumber failures indicate a faulty assumption somewhere.  Of course,  
that assumes your Cucumber features are solid...

Just my coin of small denomination.  Other learning strategies are  
available.

Ashley


* which was my situation

-- 
http://www.patchspace.co.uk/
http://www.linkedin.com/in/ashleymoran
http://aviewfromafar.net/