Turbolinks 3

Note: Turbolinks 3 isn't production-ready yet. Expect breaking changes before the new version is released.

Turbolinks makes following links in your web application faster. Instead of letting the browser recompile the JavaScript and CSS between each page change, it keeps the current page instance alive and replaces only the body (or parts of) and the title in the head. Think CGI vs persistent process.

This is similar to pjax, but instead of worrying about what element on the page to replace and tailoring the server-side response to fit, we replace the entire body by default, and let you specify which elements to replace on an opt-in basis. This means that you get the bulk of the speed benefits from pjax (no recompiling of the JavaScript or CSS) without having to tailor the server-side response. It just works.

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?

It depends. The more CSS and JavaScript you have, the bigger the benefit of not throwing away the browser instance and recompiling all of it for every page. Just like a CGI script that says "hello world" will be fast, but a CGI script loading Rails on every request will not.

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 library

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 library to work. But it works great with the jQuery or Prototype framework, or whatever else you have.

Events

With Turbolinks pages will change without a full reload, so you can't rely on DOMContentLoaded or 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:before-change a Turbolinks-enabled link has been clicked (see below for more details)
page:fetch starting to fetch a new target page
page:receive the page has been fetched from the server, but not yet parsed
page:before-unload the page has been parsed and is about to be changed
page:after-remove a node (stored in event.data) has been removed from the DOM and should be cleaned up (jQuery data is cleaned up automatically)
page:change the page has been changed to the new version (and on DOMContentLoaded)
page:update is triggered alongside both page:change and jQuery's ajaxSuccess (if jQuery is available - otherwise you can manually trigger it when calling XMLHttpRequest in your own code)
page:load is fired at the end of the loading process.

Handlers bound to the page:before-change event may return false, which will cancel the Turbolinks process.

By default, Turbolinks caches 10 of these page loads. It listens to the popstate event and attempts to 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:before-unload page has been fetched from the cache and is about to be changed
page:change page has changed to the cached page.
page:restore is fired at the end of restore process.

The number of pages Turbolinks caches can be configured to suit your application's needs:

// View the current cache size
Turbolinks.pagesCached();

// Set the cache size
Turbolinks.pagesCached(20);

If you need to make dynamic HTML updates in the current page and want it to be cached properly you can call:

Turbolinks.cacheCurrentPage();

To implement a client-side spinner, you could listen for page:fetch to start it and page:receive to stop it.

// using jQuery for simplicity

$(document).on("page:fetch", startSpinner);
$(document).on("page:receive", stopSpinner);

DOM transformations that are idempotent are best. If you have transformations that are not, bind them to page:load (in addition to the initial page load) instead of page:change (as that would run them again on the cached pages):

// using jQuery for simplicity

$(document).on("ready page:load", nonIdempotentFunction);

Transition Cache: A Speed Boost

Transition Cache, added in v2.2.0, makes loading cached pages instantaneous. Once a user has visited a page, returning later to the page results in an instant load.

For example, if Page A is already cached by Turbolinks and you are on Page B, clicking a link to Page A will immediately display the cached copy of Page A. Turbolinks will then fetch Page A from the server and replace the cached page once the new copy is returned.

To enable Transition Cache, include the following in your javascript:

Turbolinks.enableTransitionCache();

The one drawback is that dramatic differences in appearance between a cached copy and new copy may lead to a jarring affect for the end-user. This will be especially true for pages that have many moving parts (expandable sections, sortable tables, infinite scrolling, etc.).

If you find that a page is causing problems, you can have Turbolinks skip displaying the cached copy by adding data-no-transition-cache to any DOM element on the offending page.

Progress Bar

Because Turbolinks skips the traditional full page reload, browsers won't display their native progress bar when changing pages. To fill this void, Turbolinks offers an optional JavaScript-and-CSS-based progress bar to display page loading progress.

As of Turbolinks 3.0, the progress bar is turned on by default.

To disable (or re-enable) the progress bar, include one of the following in your JavaScript:

Turbolinks.ProgressBar.disable();
Turbolinks.ProgressBar.enable();

The progress bar is implemented on the <html> element's pseudo :before element and can be customized by including CSS with higher specificity than the included styles. For example:

html.turbolinks-progress-bar::before {
  background-color: red !important;
  height: 5px !important;
}

Control the progress bar manually using these methods:

Turbolinks.ProgressBar.start();
Turbolinks.ProgressBar.advanceTo(value); // where value is between 0-100
Turbolinks.ProgressBar.done();

* Note that the Progress Bar API has changed from what it was prior to version 3.0. Consult the CHANGELOG for details.

Initialization

Turbolinks will be enabled only if the server has rendered a GET request.

Some examples, given a standard RESTful resource:

POST :create => resource successfully created => redirect to GET :show
- Turbolinks ENABLED
POST :create => resource creation failed => render :new
- Turbolinks DISABLED

Why not all request types? Some browsers track the request method of each page load, but triggering pushState methods doesn'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 containing a file extension other than .html will automatically be opted out of Turbolinks. So links to /images/panda.gif will just work as expected. To whitelist additional file extensions to be processed by Turbolinks, use Turbolinks.allowLinkExtensions().

Turbolinks.allowLinkExtensions();                 // => ['html']
Turbolinks.allowLinkExtensions('md');             // => ['html', 'md']
Turbolinks.allowLinkExtensions('coffee', 'scss'); // => ['html', 'md', 'coffee', 'scss']

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-method, data-remote, or data-confirm from Rails.

Note: in Turbolinks 3.0, the default behavior of redirect_to is to redirect via Turbolinks (Turbolinks.visit response) for XHR + non-GET requests. You can opt-out of this behavior by passing turbolinks: false to redirect_to.

By default, Turbolinks includes itself in ActionController::Base. To opt out of the Turbolinks features in certain controllers, set config.turbolinks.auto_include to false in application.rb and include Turbolinks::Controller in the controllers where you use Turbolinks.

jquery.turbolinks

If you have a lot of existing JavaScript that binds elements on jQuery.ready(), you can pull the jquery.turbolinks library into your project that will trigger ready() when Turbolinks triggers the page:load event. It may restore functionality of some libraries.

Add the gem to your project, then add the following line to your JavaScript manifest file, after jquery.js but before turbolinks.js:

//= 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>

If those assets change URLs (embed an md5 stamp to ensure this), the page will do a full reload instead of going through Turbolinks. This ensures that all Turbolinks sessions will always be running off your latest JavaScript and CSS.

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

Turbolinks will evaluate any script tags in pages it visits, if those tags do not have a type or if the type is text/javascript. All other script tags will be ignored.

As a rule of thumb when switching to Turbolinks, move all of your javascript tags inside the head and then work backwards, only moving javascript code back to the body if absolutely necessary. If you have any script tags in the body you do not want to be re-evaluated then you can set the data-turbolinks-eval attribute to false:

<script type="text/javascript" data-turbolinks-eval=false>
  console.log("I'm only run once on the initial page load");
</script>

Turbolinks will not re-evaluate script tags on back/forward navigation, unless their data-turbolinks-eval attribute is set to always:

<script type="text/javascript" data-turbolinks-eval=always>
  console.log("I'm run on every page load, including history back/forward");
</script>

Triggering a Turbolinks visit manually

You can use Turbolinks.visit(path) to go to a URL through Turbolinks.

You can also use redirect_to path, turbolinks: true in Rails to perform a redirect via Turbolinks.

Partial replacements with Turbolinks (3.0+)

You can use either Turbolinks.visit(path, options) or Turbolinks.replace(html, options) to trigger partial replacement of nodes in your DOM instead of replacing the entire body.

Turbolinks' partial replacement strategy relies on id attributes specified on individual nodes or a combination of id and data-turbolinks-permanent or data-turbolinks-temporary attributes.

<div id="comments"></div>
<div id="nav" data-turbolinks-permanent></div>
<div id="footer" data-turbolinks-temporary></div>

Any node with an id attribute can be partially replaced. If the id contains a colon, the key before the colon can also be targeted to replace many nodes with a similar prefix.

<div id="comments"></div>
<div id="comments:123"></div>

Turbolinks.visit should be used when you want to perform an XHR request to fetch the latest content from the server and replace all or some of the nodes.

Turbolinks.replace should be used when you already have a response body and want to replace the contents of the current page with it. This is needed for contextual responses like validation errors after a failed create attempt since fetching the page again would lose the validation errors.

<body>
  <div id="sidebar" data-turbolinks-permanent>
    Sidebar, never changes after initial load.
  </div>

  <div id="flash" data-turbolinks-temporary>
    You have <%= @comments.count %> comments.
  </div>

  <section id="comments">
    <%= @comments.each do |comment| %>
      <article id="comments:<%= comment.id %>">
        <h1><%= comment.author %></h1>
        <p><%= comment.body %></p>
      </article>
    <% end %>
  </section>

  <%= form_for Comment.new, remote: true, id: 'new_comment' do |form| %>
    <%= form.text_area :content %>
    <%= form.submit %>
  <% end %>
</body>

<script>
// Will change #flash, #comments, #comments:123
Turbolinks.visit(url, { change: 'comments' });

// Will change #flash, #comments:123
Turbolinks.visit(url, { change: 'comments:123' });

// Will only keep #sidebar
Turbolinks.visit(url)

// Will only keep #sidebar, #flash
Turbolinks.visit(url, { keep: 'flash' });

// Will keep nothing
Turbolinks.visit(url, { flush: true });

// Same as visit() but takes a string or Document, allowing you to
// do inline responses instead of issuing a new GET with Turbolinks.visit.
// This is useful for things like form validation errors or other
// contextualized responses.
Turbolinks.replace(html, options);
</script>

Partial replacement decisions can also be made server-side by using redirect_to or render with change, keep, or flush options.

class CommentsController < ActionController::Base
  def create
    @comment = Comment.new(comment_params)

    if @comment.save
      # This will change #flash, #comments
      redirect_to comments_url, change: 'comments'
      # => Turbolinks.visit('/comments', change: 'comments')
    else
      # Validation failure
      render :new, change: :new_comment
      # => Turbolinks.replace('<%=j render :new %>', change: 'new_comment')
    end
  end
end

# Redirect via Turbolinks when the request is XHR and not GET.
# Will refresh any `data-turbolinks-temporary` nodes.
redirect_to path

# Force a redirect via Turbolinks.
redirect_to path, turbolinks: true

# Force a normal redirection.
redirect_to path, turbolinks: false

# Partially replace any `data-turbolinks-temporary` nodes and nodes with `id`s matching `comments` or `comments:*`.
redirect_to path, change: 'comments'

# Partially replace any `data-turbolinks-temporary` nodes and nodes with `id` not matching `something` and `something:*`.
redirect_to path, keep: 'something'

# Replace the entire `body` of the document, including `data-turbolinks-permanent` nodes.
redirect_to path, flush: true

# Render with Turbolinks when the request is XHR.
# Refresh any `data-turbolinks-temporary` nodes and nodes with `id` matching `new_comment`.
render view, change: 'new_comment'

# Refresh any `data-turbolinks-temporary` nodes and nodes with `id` not matching `something` and `something:*`.
render view, keep: 'something'

# Replace the entire `body` of the document, including `data-turbolinks-permanent` nodes.
render view, flush: true

# Force a render with Turbolinks.
render view, turbolinks: true

# Force a normal render.
render view, turbolinks: false

XHR Request Caching (3.0+)

To prevent browsers from caching Turbolinks requests:

# Globally
Turbolinks.disableRequestCaching();

# Per Request
Turbolinks.visit(url, { cacheRequest: false });

This works just like jQuery.ajax(url, { cache: false }), appending "_#{timestamp}" to the GET parameters.

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.

Compatibility

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.

Do note that existing JavaScript libraries may not all be compatible with Turbolinks out of the box due to the change in instantiation cycle. You might very well have to modify them to work with Turbolinks' new set of events. For help with this, check out the Turbolinks Compatibility project.

Turbolinks works with Rails 3.2 and newer.

Installation

Add gem 'turbolinks' to your Gemfile.
Run bundle install.
Add //= require turbolinks to your Javascript manifest file (usually found at app/assets/javascripts/application.js). If your manifest requires both turbolinks and jQuery, make sure turbolinks is listed after jQuery.
Restart your server and you're now using turbolinks!

Running the tests

Ruby:

rake test:all

BUNDLE_GEMFILE=Gemfile.rails42 bundle
BUNDLE_GEMFILE=Gemfile.rails42 rake test

JavaScript:

bundle install
npm install

script/test   # requires phantomjs >= 2.0
script/server # http://localhost:9292/javascript/index.html

Language Ports

These projects are not affiliated with or endorsed by the Rails Turbolinks team.

Flask Turbolinks (Python Flask)
Django Turbolinks (Python Django)
ASP.NET MVC Turbolinks
PHP Turbolinks Component (Symfony Component)
PHP Turbolinks Package (Laravel Package)
Grails Turbolinks (Grails Plugin)

Credits

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!

wkrsz/turbolinks