Do note that this of course means that you'll have a long-running, persistent session with maintained state. That's what's making it so fast. But it also means that you may have to pay additional care not to leak memory or otherwise bloat that long-running state. That should rarely be a problem unless you're doing something really funky, but you do have to be aware of it. Your memory leaking sins will not be swept away automatically by the cleansing page change any more.
How much faster is it really?
In any case, the benefit can be up to twice as fast in apps with lots of JS and CSS. Of course, your mileage may vary, be dependent on your browser version, the moon cycle, and all other factors affecting performance testing. But at least it's a yardstick.
The best way to find out just how fast it is? Try it on your own application. It hardly takes any effort at all.
No jQuery or any other framework
Turbolinks is designed to be as light-weight as possible (so you won't think twice about using it even for mobile stuff). It does not require jQuery or any other framework to work. But it works great with jQuery or Prototype or whatever else have you.
With Turbolinks pages will change without a full reload, so you can't rely on
jQuery.ready() to trigger your code. Instead Turbolinks fires events on
document to provide hooks into the lifecycle of the page:
Load a fresh version of a page from the server:
page:fetch starting to fetch a new target page
page:receive the page has been fetched from the server, but not yet parsed
page:change the page has been parsed and changed to the new version
page:load is fired at the end of the loading process.
Turbolinks caches 10 of these page loads. It listens to the popstate event and attempts restore page state from the cache when it's triggered. When
popstate is fired the following process happens:
Restore a cached page from the client-side cache:
page:change page has changed to the cached page.
page:restore is fired at the end of restore process.
So if you wanted to have a client-side spinner, you could listen for
page:fetch to start it and
page:receive to stop it.
document.addEventListener("page:fetch", startSpinner); document.addEventListener("page:receive", stopSpinner);
DOM transformations that are idempotent are best. If you have transformations that are not, hook them to happen only on
page:load instead of
page:change (as that would run them again on the cached pages).
Turbolinks will be enabled only if the server has rendered a
Some examples, given a standard RESTful resource:
POST :create=> resource successfully created => redirect to
- Turbolinks ENABLED
POST :create=> resource creation failed => render
- Turbolinks DISABLED
Why not all request types? Some browsers track the request method of each page load, but triggering pushState methods don't change this value. This could lead to the situation where pressing the browser's reload button on a page that was fetched with Turbolinks would attempt a
POST (or something other than
GET) because the last full page load used that method.
Opting out of Turbolinks
By default, all internal HTML links will be funneled through Turbolinks, but you can opt out by marking links or their parent container with
data-no-turbolink. For example, if you mark a div with
data-no-turbolink, then all links inside of that div will be treated as regular links. If you mark the body, every link on that entire page will be treated as regular links.
<a href="/">Home (via Turbolinks)</a> <div id="some-div" data-no-turbolink> <a href="/">Home (without Turbolinks)</a> </div>
Note that internal links to files not ending in .html, or having no extension, will automatically be opted out of Turbolinks. So links to /images/panda.gif will just work as expected.
Also, Turbolinks is installed as the last click handler for links. So if you install another handler that calls event.preventDefault(), Turbolinks will not run. This ensures that you can safely use Turbolinks with stuff like
data-confirm from Rails.
page:load event. It may restore functionality of some libraries.
jquery.js but before
//= require jquery.turbolinks
Additional details and configuration options can be found in the jquery.turbolinks README.
Asset change detection
You can track certain assets, like application.js and application.css, that you want to ensure are always of the latest version inside a Turbolinks session. This is done by marking those asset links with data-turbolinks-track, like so:
<link href="/assets/application-9bd64a86adb3cd9ab3b16e9dca67a33a.css" rel="stylesheet" type="text/css" data-turbolinks-track>
When this happens, you'll technically be requesting the same page twice. Once through Turbolinks to detect that the assets changed, and then again when we do a full redirect to that page.
Evaluating script tags
data-turbolinks-eval attribute to
Triggering a Turbolinks visit manually
You can use
Turbolinks.visit(path) to go to a URL through Turbolinks.
Full speed for pushState browsers, graceful fallback for everything else
Like pjax, this naturally only works with browsers capable of pushState. But of course we fall back gracefully to full page reloads for browsers that do not support it.
Turbolinks is designed to work with any browser that fully supports pushState and all the related APIs. This includes Safari 6.0+ (but not Safari 5.1.x!), IE10, and latest Chromes and Firefoxes.
gem 'turbolinks'to your Gemfile.
- Restart your server and you're now using turbolinks!
Thanks to Chris Wanstrath for his original work on Pjax. Thanks to Sam Stephenson and Josh Peek for their additional work on Pjax and Stacker and their help with getting Turbolinks released. Thanks to David Estes and Nick Reed for handling the lion's share of post-release issues and feature requests. And thanks to everyone else who's fixed or reported an issue!