Testing multistep form with rspec - ruby-on-rails

Question

I'm not very proficient with Capybara and testing tools alike. I'm building a multistep form or a form wizard.
This is how I had a scenario in my head :
User visit signup path
fills in bunch of invalid data and some validation errors should be displayed
fills in valid data, user should be taken to the next step
repeat this step above for n steps with valid data
check if number of users has increased by 1
This is what I've got so far (I managed to get something working) :
describe "signup" do
before { visit signup_path }
let(:submit) { 'Start creating my account' }
let(:next_btn) { 'Next' }
describe "with invalid information" do
it "should not create a user" do
expect { click_button submit }.not_to change(User, :count)
end
describe "should not move to the next step" do
before { click_button submit }
it { should have_content('error') }
it { should have_title('Start here') }
end
end
describe "with valid information wizard step 1" do
before do
fill_in 'First name', with: "Example"
fill_in 'Last name', with: "User"
fill_in 'email', with: "user#example.com"
find("#user_password").set "foobar"
end
it "should move to the next wizard step 2" do
click_button submit
should have_content('We need some more info')
end
it "should have error on the wizard step 2" do
fill_in 'Date of birth', with: "Example"
click_button next_btn
should have_content('error')
end
end
end
This assertion fails should have error on the wizard step 2, it seems to be stuck at step 1 still, I know this from looking at page content and by errors where Dob element can't be found and next button cant be found as well.
Is this even possible with capybara to test sequentially, keeping information step by step?

Answer 1

You should be able to make this work by nesting the describe blocks when moving on to a subsequent step, eg.
describe "signup" do
...
describe "with valid information wizard step 1" do
before do
fill_in 'First name', with: "Example"
fill_in 'Last name', with: "User"
fill_in 'email', with: "user#example.com"
find("#user_password").set "foobar"
click_button submit
end
it "should move to the next wizard step 2" do
should have_content('We need some more info')
end
describe "and filling out the date of birth incorrectly" do
before do
fill_in 'Date of birth', with: "Example"
click_button next_btn
end
it "should have error on the wizard step 2" do
should have_content('error')
end
end
end
end
Now the button clicks are repeated in the nested blocks, so the first example clicks 'submit', and the second example clicks 'submit' and then clicks 'next_btn', etc. Keep in mind that the sequence is repeated for each example so this may slow down your tests. OTOH it more accurately reflects the user interaction, so the tradeoff is probably worth it.

Answer 2

I think this is not a capybara question so much as it is an rspec question. I've used cucumber more than rspec, but so far as I know in my limited experience, each "describe" is an independent test case. I don't think that you can either a. expect these to execute in any particular order or b. share context / state between tests.
Again, my experience has mainly been with cucumber, but I have written tests for this EXACT scenario before- a two page user registration wizard. In cucumber, one writes "scenarios," which are basically short stories that describe a user's interaction with the the system in terms of "given," "when," and "then" (or in other worse, assumptions, actions, and assertions) and these are written in plain English. One then writes Ruby code that maps these English sentences to Ruby code (which might use capybara) to do / test what each scenario describes... What I did was constructed several distinct scenarios.
User visits registration page; fills it out with a whole bunch of errors; check that the system does display the errors
User visits registration page; fills it out correctly; check that the user makes it to the second page
User has already completed the first page; fills out the second page with a whole bunch of errors; check that the system does display the errors
User has already completed the first page; fills out the second page correctly; check that the user makes it to the confirmation page
You can obviously translate those stories into rspec.
The thing that recurs in those scenarios is the filling out (whether with good or bad data) of the two forms. To keep the code DRY, create some helper methods that let you easily populate the form with values that will help you satisfy each of those scenarios.
Does that help any?