[rspec-users] Question about structuring specs

Ijonas Kisselbach ijonas.kisselbach at gmail.com
Mon Jan 4 08:06:15 EST 2010


Hi David,

Thanks for the suggestions... Yes I am mocking to avoid DB access, but if
the cost is DB access vs. readability: I'll choose readability.

Would you use something like Factory Girl or Machinist to setup the DB? To
avoid too much mocking and to avoid old-skool fixtures.

Thanks for the help.

On Mon, Jan 4, 2010 at 12:56 PM, David Chelimsky <dchelimsky at gmail.com>wrote:

> On Mon, Jan 4, 2010 at 5:26 AM, Ijonas Kisselbach <
> ijonas.kisselbach at gmail.com> wrote:
>
>> Hi David,
>>
>> I see your point about concentrating on outcomes rather than
>> implementation. I suppose who cares about implementation, its the effect
>> code on the "state of the world" that is important.
>>
>> Here's an unmodified sample:
>>   context ", when a new policy is activated on unchanged content" do
>>     before(:each) do
>>       setup_common_mocks
>>     end
>>
>>     it "should create a new policy, the new violation and location,
>> resolve previous violations, recalculate location MD5, and refresh caches"
>> do
>>       # content hasn't changed
>>       @content = mock(:content, :content_md5 => "84290324908230948",
>> :most_recent_violations => [mock(:violation_mr, :update_folder_count =>
>> nil)])
>>       @content_descriptor = mock(:content_descriptor, :contents =>
>> [@content], :most_recent_content => @content, :folder => mock(:folder))
>>
>> ContentDescriptor.should_receive(:by_path_md5_and_site).once.and_return([@content_descriptor])
>>
>>
>>       # Policy is new, gets created once
>>       Policy.should_receive(:find_by_account_category_and_name).once
>>
>> PolicyCategory.should_receive(:find_by_account_and_name).once.and_return(mock(:policy_category))
>>       Policy.should_receive(:create!).once.and_return(mock(:policy))
>>
>>       # below are the changes affected
>>       Violation.should_receive(:resolve_violations).once
>>
>> Violation.should_receive(:recalculate_violation_md5).once.and_return(10)
>>
>> Violation.should_receive(:find_by_content_id_and_policy_id).once.and_return(nil)
>>       violation = mock(:new_violation1, :location_md5= => nil, :save =>
>> true)
>>       Violation.should_receive(:new).once.and_return(violation)
>>       violation.should_receive(:save).once
>>
>>       Location.should_receive(:create!).once.and_return(mock(:location1))
>>
>>       @content.should_receive(:most_recent_violations=).once
>>       @content.should_receive(:unresolved_violation_count=).once
>>       @content.should_receive(:save).once
>>       CacheMaintenance.should_receive(:remove_folder_cache_keys).twice
>>       record
>>     end
>>
>>   end
>>
>> I'm going to try and refactor the specs so that anything that doesn't
>> directly modify state of the app is removed.
>>
>
> Why are you mocking so much here? Why not set up the db in a known state,
> invoke the action you want to invoke (record???) and set expectations about
> the outcomes? If you're concerned about database access and speed, this is a
> case where I think the benefits of just invoking the code outweighs the cost
> of database access. I'm imagining something more like this:
>
> context ", when a new policy is activated on unchanged content" do
>   it "creates a new policy" do
>     record
>     # expect to find the policy by known attributes
>     # something like this:
>     #   Policy.find_by_account_category_and_name(....).should_not be_nil
>   end
>
>   it "creates a new violation" do
>     record
>     # same as the policy - find the Violation by known attributes
>   end
>
>   it "updates most_recent_violations" do
>     record
>     # query for most recent violations and expect the violation
>     # to be found
>   end
>
>   it "updates the violation count" do
>     expect { record }.to change {content.unresolved_violation_count}.by(1)
>   end
>
>   # etc, etc
> end
>
> Yes, this means that the process needs to happen more than once, but each
> example becomes much easier to grok.
>
> WDYT?
>
> Thanks,
>> Ijonas.
>>
>> On Mon, Jan 4, 2010 at 11:16 AM, David Chelimsky <dchelimsky at gmail.com>wrote:
>>
>>>  On Mon, Jan 4, 2010 at 4:33 AM, Ijonas Kisselbach <
>>> ijonas.kisselbach at gmail.com> wrote:
>>>
>>>> Hi,
>>>>
>>>> I'm struggling with structuring my specs describing a large process in
>>>> my app. There are multiple paths of execution through that process each of
>>>> which I'm trying to describe using a different rspec context, eg.
>>>>
>>>> describe Violation do
>>>>    context ", when nothing has changed since the last run"
>>>>    context ", when new content has been created, but policies remain the
>>>> same"
>>>>    context ", when new policies are activated, but content remains the
>>>> same"
>>>> end
>>>>
>>>> Each of the three scenarios/context above have got a bunch of "it
>>>> should..." blocks in it which in turn contain a whole bunch of
>>>> should_receives and should_not_receives on various mocked objects, thereby
>>>> exercising the functionality of the large process.
>>>>
>>>> I would like the context to read as follows:
>>>>
>>>> context ", when new policies are activated, but content remains the
>>>> same" do
>>>>    it "should create the new policy" do
>>>>       # a whole bunch of expectations testing the policy creation part
>>>> of the process
>>>>    end
>>>>    it "should create a new violation and location" do
>>>>      # a whole bunch of expectations testing the violation creation part
>>>> of the process
>>>>    end
>>>>    it "should resolve previous violations" do
>>>>      # a whole bunch of expections testing retrieval of previous
>>>> violations and performing updates on them
>>>>    end
>>>>   ....
>>>> end
>>>>
>>>> The problem is: if I compartmentalize my expectations into the
>>>> individual it-should-blocks then something will fail in the execution of the
>>>> large process, typically caused by a mock not being setup. If I lump all my
>>>> expectations in the before(:each)-block then the whole thing springs to
>>>> life, but I lose my compartmentalization of the specs and the whole thing
>>>> becomes unreadable.
>>>>
>>>> I guess I'm looking for help and advice on how best combat the lumping
>>>> of expectations into the before-block. Should I separate my stubbing from my
>>>> expectations ?
>>>>
>>>> Many thanks for your advice.
>>>>
>>>
>>> I'd need to see the actual code to respond in any precise way here, but
>>> generally, it sounds like you're specifying too much about the
>>> implementation rather than the outcomes. What happens if you eliminate all
>>> of the mocks in these examples and just have expectations like
>>> "Policy.find(@policy_id).should_not be_nil"?
>>>
>>> David
>>>
>>>
>>>>
>>>> (I'm using rspec 1.2.9 and Rails 2.2.2 on OSX)
>>>>
>>>> Regards,
>>>> Ijonas.
>>>
>>>
>>> _______________________________________________
>>> rspec-users mailing list
>>> rspec-users at rubyforge.org
>>> http://rubyforge.org/mailman/listinfo/rspec-users
>>>
>>
>>
>> _______________________________________________
>> rspec-users mailing list
>> rspec-users at rubyforge.org
>> http://rubyforge.org/mailman/listinfo/rspec-users
>>
>
>
> _______________________________________________
> rspec-users mailing list
> rspec-users at rubyforge.org
> http://rubyforge.org/mailman/listinfo/rspec-users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://rubyforge.org/pipermail/rspec-users/attachments/20100104/e3095d81/attachment.html>


More information about the rspec-users mailing list