allow stubbing of helper methods

[unorthodoxgeek]

moved the helpers to a helper object held in the API class, which allows it to get stubbed and mocked for testing purposes.

[unorthodoxgeek]

seems like some versions of Ruby have different ways of stubbing, tests fail on 1.8.7 and ree. Rubinius just takes forever to run.

[unorthodoxgeek]

closing until I can get specs to pass.

[unorthodoxgeek]

should now pass

[dblock]

Minor:

• Text should start with a capital H.
• Simplify the example. I would keep only one thing and remove the describe parts and the require, just to get exactly to the point of how to stub something.

[dblock]

I think this is really good, cc: @mbleigh - I definitely think this is a better implementation, and I don't see any immediate holes. Would love you to check this out in detail.

@unorthodoxgeek: please also update CHANGELOG.

[dblock]

This fixes #396 and is an alternative for #397.

[dblock]

I think this should be helpers, because it represents several helpers. When you stub, you stub one of the helpers and it's strange to do .helper.stub(:method), semantically.

[unorthodoxgeek]

applied changes, the helper method on api can't be named helpers as that method name is already taken. I've changed it to a more descriptive name "helpers_container", feel free to suggest a different name.

[mbleigh]

The problem comes in when you have multiple levels of helper methods. Failing spec example:

    it "allows stubbing nested helpers" do
      subject.helpers do
        def foo
          'foo'
        end
      end

      subject.get "/" do
        foo
      end

      subject.namespace :random do
        helpers do
          def foo
            'bar'
          end
        end

        get :what do
          foo
        end
      end

      get '/random/what'
      last_response.body.should == 'bar'

      get '/'
      last_response.body.should == 'foo'

      subject.helper.stub!(:foo).and_return("baz")
      get '/'
      last_response.body.should == 'baz'
      get '/random/what'
      last_response.body.should == 'baz'
    end

[unorthodoxgeek]

any suggestions as to how we should fix this? I'm thinking about a hash... would appreciate your feedback

[dblock]

Maybe we can instead combine the two techniques proposed here and in #397, so leaving helpers alone and injecting a class-level (global) helper block the same way setup did which is accessible externally and therefore can have stubbed methods?

[dblock]

I combined the test above and what you wrote into https://github.com/dblock/grape/blob/spec-helpers/spec/grape/api_spec.rb#L913 (just copy paste). If we can get both to pass, we win. I think @mbleigh's setup approach might work better, we want to inject whatever is being stubbed at runtime. In a way we want to delay the stubs. Maybe this is something that can be done with a bit more finagling with how RSpec mocks work? Like defining stub and stub! ourselves on the helpers module being generated and executing stubs at runtime?

[guilherme]

i'm having this issue. how about the conclusion of the pull request?

[dblock]

@guilherme see above, there's no implementation that works correctly. If you make one, please make a PR. You have to make that test pass that's mentioned above.

[zuk]

Someone put up a $100 bounty to get this resolved: https://www.bountysource.com/issues/655313-can-not-stub-mock-helper-methods/bounties

See #396

[dblock]

I am going to close this PR as it doesn't address the issue (see above).

[coveralls]

Changes Unknown when pulling a13c729 on unorthodoxgeek:master into * on intridea:master*.