Rspec has two different types of tests that are very similar to integration tests; request specs and feature specs. The difference between them has never been particularly clear to me and seems to be a point of confusion for many others as well.
Feature specs and Request specs are both are part of rspec-rails and are built on a similar foundation.
Request specs are built on top of Rails integration tests, which offer an API with methods that mirror what you find in HTTP: “get”, “put”, “post”, etc.
It gives you a “response” object for you to make your assertions against.
Feature specs are built on top of Capybara. The goal of Capybara is to offer an API that “simulates user behaviour”. To that end their API employs language that is not far from what you would use when directing someone to use a website over the phone:
visit '/sessions/new' fill_in 'Login', :with => 'email@example.com' fill_in 'Password', :with => 'password' click_link 'Sign in' page.should have_content 'Success'
Under the covers “visit” is just calling Rack::Test methods but there is no low-level, HTTP oriented language (request, response, get, put, etc…) used in the API itself. Capybara is consciously omitting them as Jonas Nicklas (Capybara’s author) explains on his blog.
Feature specs exist to provide a place for the type of high level testing that Capybara enables. The Relishapp site describes them as “high-level tests meant to exercise slices of functionality through an application. They should drive the application only via its external interface, usually web pages.” This unfortunately jargon-laden description, really only made sense after some digging.
Which brings us to Request specs. These are for those moments when the details of HTTP are exactly what you care about; like when you are testing an API.
get "/todays-forcast" expect(response.body).to have_content "sunny"
Here the details of HTTP are front and centre, because in an API they matter. This is why request specs are referred to as ” a thin wrapper around Rails’ integration tests”, the focus is on providing just some Rspec sugar to use with the existing HTTP oriented methods.
The confusion around these things seem to stem from the early days of the project. When Capybara was first created it was intended to be used within request specs. That created a bunch of confusion because of the competing syntaxes ( get/visit, response/page) which resulted in the creation of feature specs to give a proper home to this type of testing.
Unfortunately people are still following old instructions and this usage is still common which muddies the water when trying to understand which is which. Hopefully this adds a little clarity.