[rspec-users] [RSpec] Testing a helper with or without model layer?

[Bruno Cardoso]

Hi David

> Even though the code is in a helper, it depends heavily on the model. 
> This exhibits a code smell called Feature Envy, in which one object (the 
> helper) does some computation but another object (the CfgInterface 
> model) has all the data. Based on that, one might argue this method 
> belongs on the model anyhow, in which case the argument of stubbing out 
> the model layer makes little sense.

I see your point and in this case you are probably right but what if my 
helper was more heavy on the processing and still very data dependent ? 
Should I use fixtures on the helper or should I stub/mock whatever I 
need to avoid the model layer within the helper ?

> There are also several expectations that overlap. It doesn't matter that 
> the class is a Hash - what matters is that it behaves like a hash, which 
> is demonstrated in other expectations. The fact that the keys exist is 
> also demonstrated in the expectations that access those keys.

Good point, thanks.

> Based on those comments, I'd probably do something like 
> http://gist.github.com/506366. Then I'd figure out how to swap in 
> factories for the fixtures - right now there's no way to understand what 
> 'CONS_ALL_ACCOUNT' means without looking elsewhere.

'CONS_ALL_ACCOUNT' actually has no impact when using stub/mocks, 
'CONS_ALL_ACCOUNT' was the input parameter the model was using to fetch 
data from the BD, in the second form of the test with stubs/mocks, this 
input parameter has no meaning since I'm controlling all return 
parameters.
-- 
Posted via http://www.ruby-forum.com/.