I have a form that I'm testing using Capybara. This form's URL goes to my Braintree sandbox, although I suspect the problem would happen for any remote URL. When Capybara clicks the submit button for the form, the request is routed to the dummy application rather than the remote service.
Here's an example app that reproduces this issue: https://github.com/radar/capybara_remote. Run bundle exec ruby test/form_test.rb and the test will pass, which is not what I'd typically expect.
Why does this happen and is this behaviour that I can rely on always happening?
Mario Visic points out this description in the Capybara documentation:
Furthermore, you cannot use the RackTest driver to test a remote application, or to access remote URLs (e.g., redirects to external sites, external APIs, or OAuth services) that your application might interact with.
But I wanted to know why, so I source dived. Here's my findings:
lib/capybara/node/actions.rb
def click_button(locator)
find(:button, locator).click
end
I don't care about the find
here because that's working. It's the click
that's more interesting. That method is defined like this:
lib/capybara/node/element.rb
def click
wait_until { base.click }
end
I don't know what base
is, but I see the method is defined twice more in lib/capybara/rack_test/node.rb
and lib/capybara/selenium/node.rb
. The tests are using Rack::Test
and not Selenium, so it's probably the former:
lib/capybara/rack_test/node.rb
def click
if tag_name == 'a'
method = self["data-method"] if driver.options[:respect_data_method]
method ||= :get
driver.follow(method, self[:href].to_s)
elsif (tag_name == 'input' and %w(submit image).include?(type)) or
((tag_name == 'button') and type.nil? or type == "submit")
Capybara::RackTest::Form.new(driver, form).submit(self)
end
end
The tag_name
is probably not a link -- because it's a button we're clicking -- so it falls to the elsif
. It's definitely an input
tag with type == "submit"
, so then let's see what Capybara::RackTest::Form
does:
lib/capybara/rack_test/form.rb
def submit(button)
driver.submit(method, native['action'].to_s, params(button))
end
Ok then. driver
is probably the Rack::Test
driver for Capybara. What's that doing?
lib/capybara/rack_test/driver.rb
def submit(method, path, attributes)
browser.submit(method, path, attributes)
end
What is this mysterious browser? It's defined in the same file thankfully:
def browser
@browser ||= Capybara::RackTest::Browser.new(self)
end
Let's look at what this class's submit
method does.
lib/capybara/rack_test/browser.rb
def submit(method, path, attributes)
path = request_path if not path or path.empty?
process_and_follow_redirects(method, path, attributes, {'HTTP_REFERER' => current_url})
end
process_and_follow_redirects
does what it says on the box:
def process_and_follow_redirects(method, path, attributes = {}, env = {})
process(method, path, attributes, env)
5.times do
process(:get, last_response["Location"], {}, env) if last_response.redirect?
end
raise Capybara::InfiniteRedirectError, "redirected more than 5 times, check for infinite redirects." if last_response.redirect?
end
So does process
:
def process(method, path, attributes = {}, env = {})
new_uri = URI.parse(path)
method.downcase! unless method.is_a? Symbol
if new_uri.host
@current_host = "#{new_uri.scheme}://#{new_uri.host}"
@current_host << ":#{new_uri.port}" if new_uri.port != new_uri.default_port
end
if new_uri.relative?
if path.start_with?('?')
path = request_path + path
elsif not path.start_with?('/')
path = request_path.sub(%r(/[^/]*$), '/') + path
end
path = current_host + path
end
reset_cache!
send(method, path, attributes, env.merge(options[:headers] || {}))
end
Time to break out the debugger and see what method
is here. Sticking a binding.pry
before the final line in that method, and a require 'pry'
in the test. It turns out method
is :post
and, for interest's sake, new_uri
is a URI
object with our remote form's URL.
Where's this post
method coming from? method(:post).source_location
tells me:
["/Users/ryan/.rbenv/versions/1.9.3-p374/lib/ruby/1.9.1/forwardable.rb", 199]
That doesn't seem right... Does Capybara have a def post
somewhere?
capybara (master)★ack "def post"
lib/capybara/rack_test/driver.rb
76: def post(*args, &block); browser.post(*args, &block); end
Cool. We know that browser is a
Capybara::RackTest::Browser` object. The class beginning gives the next hint:
class Capybara::RackTest::Browser
include ::Rack::Test::Methods
I know that Rack::Test::Methods
comes with a post
method. Time to dive into that gem.
lib/rack/test.rb
def post(uri, params = {}, env = {}, &block)
env = env_for(uri, env.merge(:method => "POST", :params => params))
process_request(uri, env, &block)
end
Ignoring env_for
for the time being, what does process_request
do?
lib/rack/test.rb
def process_request(uri, env)
uri = URI.parse(uri)
uri.host ||= @default_host
@rack_mock_session.request(uri, env)
if retry_with_digest_auth?(env)
auth_env = env.merge({
"HTTP_AUTHORIZATION" => digest_auth_header,
"rack-test.digest_auth_retry" => true
})
auth_env.delete('rack.request')
process_request(uri.path, auth_env)
else
yield last_response if block_given?
last_response
end
end
Hey, @rack_mock_session
looks interesting. Where's that defined?
rack-test (master)★ack "@rack_mock_session ="
lib/rack/test.rb
40: @rack_mock_session = mock_session
42: @rack_mock_session = MockSession.new(mock_session)
In two places, very close to each other. What's on and around these lines?
def initialize(mock_session)
@headers = {}
if mock_session.is_a?(MockSession)
@rack_mock_session = mock_session
else
@rack_mock_session = MockSession.new(mock_session)
end
@default_host = @rack_mock_session.default_host
end
Ok then, so it ensures it is a MockSession
object. What's MockSession
and how is its request
method defined?
def request(uri, env)
env["HTTP_COOKIE"] ||= cookie_jar.for(uri)
@last_request = Rack::Request.new(env)
status, headers, body = @app.call(@last_request.env)
headers["Referer"] = env["HTTP_REFERER"] || ""
@last_response = MockResponse.new(status, headers, body, env["rack.errors"].flush)
body.close if body.respond_to?(:close)
cookie_jar.merge(last_response.headers["Set-Cookie"], uri)
@after_request.each { |hook| hook.call }
if @last_response.respond_to?(:finish)
@last_response.finish
else
@last_response
end
end
I'm going to go right ahead here and assume @app
is the Rack application stack. By calling the call
method, the request is routed directly to this stack, rather going out to the world.
I conclude that this behaviour looks like its intentional and that I can indeed rely on it being that way.
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With