[rspec-users] Testing a rails controller *outside* of spec/controllers

Scott Taylor scott at railsnewbie.com
Thu Mar 11 01:41:42 EST 2010

On Mar 11, 2010, at 1:26 AM, Nicolás Sanguinetti wrote:

> We have an app that has "extensions" that are built as rails engines.
> Each engine needs some code in the controllers, which we solved as
> "include this module, call a method". In order to test it, though, we
> don't want to "copy" the test for how that works into each engine, so
> the idea is to write the tests once, in 'spec/lib/extensions_spec.rb',
> where we are testing the other stuff that's common for all the
> extensions (which are defined on lib/extensions.rb, thus the spec's
> location)
> In order to do it, I'm declaring a 'sample' controller in the specs
> (just include the module, call the method), and then writing a couple
> specs for what we are doing. Problem is, I can't get rspec to
> understand that that example group is a controller example groups.
> I have
>    class MyExampleController < ApplicationController
>      include Extensions::ControllerHelpers
>      call_awesome_method # this, among other things, defines an
> instance method on the controller that I want to test
>    end
> inside my spec, and after that I tried:
> 1) describe("controller helpers", :type => :controller) {
> controller_name "my_example"; ... }
> It complains about controller_name not being a method of this dynamic
> subclass of ActiveSupport::TestCase.
> After some time reading the source I figured that :type => :controller
> pretty much just works for config.include/config.extend. So on to the
> next thing:
> 2) describe MyExampleController do...end # aka, duh, the obvious one.
> It complained about @controller being nil. Ended up reading most of
> action_controller/test_case, and rspec-rails, with no success.
> 3) move that *outside* the big top-level describe(Extensions). That at
> least got the specs running, but they failed because @controller is
> nil. Again. So "Ok, it's not setting the controller class at any
> point, let's force it", and I added the 'controller_name ...' call
> (also tried going low-level to calling
> ActiveSupport::TestCase.tests(klass), but that didn't work either.)
> 4) move all that into it's own file (maybe, for some reason which
> wouldn't make much sense, two different types of specs can't live
> together in the same file)
> Nope, same errors.
> 5) move the file into spec/controllers
> Bang! It works. But it's ugly that it's there, because I'm just
> testing something "related" to controllers. (Granted, writing a
> controller spec isn't the best idea ever for this, but it's the
> straightest way to get a controller instanced and that responds to all
> the routing helpers -- which is, in the end, what I need to call.)
> Basically, I read a lot of source code for both rspec-rails and rails'
> own testing code, trying to figure out a way around of this, and I'm
> running out of options. An alternative would be to include the module
> and call the method on a mock instead of a real controller -- but
> 'call_awesome_method' actually does some stuff with the rails router,
> and I really don't want to start writing all those huge mocks for just
> testing a simple method (since the 2.3.x router is definitely not
> something that I would characterize as 'overly testable')
> So the question is: is there a way I force rspec into saying *this is
> a controller spec, treat it as such* when the describe block is not
> defined on spec/controllers? And (Or) where does the relevant code to
> decide if it's a controller spec (i.e., it should subclass
> ControllerExampleGroup) or otherwise live? I can manage to work my way
> around from there :)
> Sorry for the rant, it's past 4am here and I'm not sure I can think
> straight at this anymore (or that all of this makes much sense)

What version of rspec are you using?


> Cheers,
> -foca
> _______________________________________________
> rspec-users mailing list
> rspec-users at rubyforge.org
> http://rubyforge.org/mailman/listinfo/rspec-users

More information about the rspec-users mailing list