Question

I wrote a feature to test the default configuration of my web app using Cucumber and Capybara. Part of the feature looked like this:

And the page has a photo labeled "Device"
And the page has a checkbox labeled "Device"
And I check "Device"
And I submit the form
Then the resulting page has no photo labeled "Device"

It worked great. I want users who have installed and configured the web app on their own servers to be able to run the test to help confirm that their configuration is correct. And “Device” is a string in the default config file that the user can change. It’s an element in an array and they can add to or remove from the array when configuring their instance of the app.

Is this a sensible use of the tools or am I abusing Cucumber and/or Capybara? If it’s sensible, how would I do it? If it’s a bad idea, what tools might I use for this instead, if anything?

Here’s how I got it to work. I’m just not sure this is the best way to do it.

For this to work, the feature would have to look more like this:

And the page has at least 3 photos, let us call the last one "third_photo"

In the corresponding step definition, I use an XPath to pull out the corresponding label string for the first photo and assign it to a Hash object stored in a class variable.

And /^I the page has at least (\d*) photos, let us call the last one "([^\"]*)"$/ do |n, name|
  if ! defined?(@note)
    @note = Hash.new;
  end
   @note[name] = find(:xpath, '//ol[@id="menu"]/li[' +n+ ']/a').text;
end

Subsequent step definitions can now access the value, whatever it was.

So, another feature might be:

Then I uncheck "third_item"

And the corresponding step definition might be:

Then /I uncheck "([^\"]*)"/ do |item|
  uncheck(@note[item])
end

But I don’t feel good about it. If nothing else, I imagine there might be a name collision with another instance variable defined outside the step definitions.

It feels like I’m either Doing It Wrong or else I’m Using The Wrong Tool. What is the right way to do this?

Answer 1

There’s a blog post about using instance variables in step definitions at http://www.cloudspace.com/blog/2010/06/23/using-instance-variables-with-cucumber-steps/.

Commenters talk about the coupling this entails and at least one possible way around it.

In particular, davetron5000 says:

What we do is to not use instance variables at all, but instead
provide a has that shared state can go in. This hash is cleared after
each test run. Not ideal, but it’s a bit cleaner than random instance
variables all over the place (and also ensures a reasonably clean
state before each test)