Using cucumber, capybara, and net/http to check dynamically injected content

bananas.rb

Then /^I should see the API documentation$/ do
  uri = URI.parse('http://api.spire.io')
  request = Net::HTTP::Get.new(uri.request_uri)
  request['Accept'] = 'text/html'

  response = Net::HTTP.start(uri.host, uri.port) {|http|
    http.request(request)
  }

  page.should have_css('.api-reference')

  # ugh.
  script = 'document.getElementsByClassName("api-reference")[0].innerHTML'
  actual = page.evaluate_script(script)

  script = <<-SCRIPT
var div = document.createElement('div')

div.innerHTML = #{ response.body.inspect };

return div.innerHTML;
  SCRIPT

  expected = page.execute_script(script)

  actual.should_not be_nil
  actual.should == expected
end

actual is the content that was injected by the application's javascript

The reason for the crazy expected = page.execute_script(script) on line 24 for ensuring that any transformations on the actual injected content caused by sticking it into the DOM are applied the same to the expected html content. Without this step problems with whitespace and escaped html will cause some serious head scratching...