Not using a browser is a lot faster, so let’s try to avoid it if possible.
click_ujs_link "Delete"
click_ujs_link "Post"
...
def click_ujs_link(name)
link = find(:link, name, {})
visit_with_method link['data-method'], link['href']
end
def visit_with_method(method, url)
page.driver.send method, url
visit page.driver.response.location
end
Just built this little snippet in a hacknight, it simulates a geolcation via capybara, so you can test if you geo-magic actually works 🙂
View
<p>Finding your location... <span id="status">checking...</span></p>
<script>
 jQuery(function () {
  var timeout = (document.location.href.indexOf('test_location') >= 0 ? 100 : 0);
   setTimeout(function(){
     navigator.geolocation.getCurrentPosition(function(){
jQuery('#status').html("found you!");
});
   }, timeout)
 });
</script>
Capybara test via selenium in rspec
 def simulate_location(lat, lng)
   page.driver.browser.execute_script <<-JS
window.navigator.geolocation.getCurrentPosition = function(success){
var position = {"coords" : { "latitude": "#{lat}", "longitude": "#{lng}" }};
success(position);
}
JS
 end
 it "can use location", :js => true do
   visit '/?test_location=true'
   simulate_location 20, 20
   sleep 0.2
page.should have_content "found you!"
 end
Capybara has current_url, which returns an url that is unmatcheable since it includes a randomized port, and current_path with contains neither query nor fragment(aka hash/anchor) and is therefore rather useless.
Behold the perfect current_path_info, which returns the full path and query + fragment.
Code
def current_path_info
current_url.sub(%r{.*?://},'')[%r{[/\?\#].*}] || '/'
end
