Action Controller Streaming

Allows views to be streamed back to the client as they are rendered.

By default, Rails renders views by first rendering the template and then the layout. The response is sent to the client after the whole template is rendered, all queries are made, and the layout is processed.

Streaming inverts the rendering flow by rendering the layout first and subsequently each part of the layout as they are processed. This allows the header of the HTML (which is usually in the layout) to be streamed back to client very quickly, enabling JavaScripts and stylesheets to be loaded earlier than usual.

Several Rack middlewares may not work and you need to be careful when streaming. This is covered in more detail below, see the Middlewares at Streaming section.

Streaming can be added to a given template easily, all you need to do is to pass the :stream option to render.

class PostsController
  def index
    @posts = Post.all
    render stream: true
  end
end

When to use streaming

Streaming may be considered to be overkill for lightweight actions like new or edit. The real benefit of streaming is on expensive actions that, for example, do a lot of queries on the database.

In such actions, you want to delay queries execution as much as you can. For example, imagine the following dashboard action:

def dashboard
  @posts = Post.all
  @pages = Page.all
  @articles = Article.all
end

Most of the queries here are happening in the controller. In order to benefit from streaming you would want to rewrite it as:

def dashboard
  # Allow lazy execution of the queries
  @posts = Post.all
  @pages = Page.all
  @articles = Article.all
  render stream: true
end

Notice that :stream only works with templates. Rendering :json or :xml with :stream won’t work.

Communication between layout and template

When streaming, rendering happens top-down instead of inside-out. Rails starts with the layout, and the template is rendered later, when its yield is reached.

This means that, if your application currently relies on instance variables set in the template to be used in the layout, they won’t work once you move to streaming. The proper way to communicate between layout and template, regardless of whether you use streaming or not, is by using content_for, provide, and yield.

Take a simple example where the layout expects the template to tell which title to use:

<html>
  <head><title><%= yield :title %></title></head>
  <body><%= yield %></body>
</html>

You would use content_for in your template to specify the title:

<%= content_for :title, "Main" %>
Hello

And the final result would be:

<html>
  <head><title>Main</title></head>
  <body>Hello</body>
</html>

However, if content_for is called several times, the final result would have all calls concatenated. For instance, if we have the following template:

<%= content_for :title, "Main" %>
Hello
<%= content_for :title, " page" %>

The final result would be:

<html>
  <head><title>Main page</title></head>
  <body>Hello</body>
</html>

This means that, if you have yield :title in your layout and you want to use streaming, you would have to render the whole template (and eventually trigger all queries) before streaming the title and all assets, which defeats the purpose of streaming. Alternatively, you can use a helper called provide that does the same as content_for but tells the layout to stop searching for other entries and continue rendering.

For instance, the template above using provide would be:

<%= provide :title, "Main" %>
Hello
<%= content_for :title, " page" %>

Resulting in:

<html>
  <head><title>Main</title></head>
  <body>Hello</body>
</html>

That said, when streaming, you need to properly check your templates and choose when to use provide and content_for.

See also ActionView::Helpers::CaptureHelper for more information.

Headers, cookies, session, and flash

When streaming, the HTTP headers are sent to the client right before it renders the first line. This means that, modifying headers, cookies, session or flash after the template starts rendering will not propagate to the client.

Middlewares

Middlewares that need to manipulate the body won’t work with streaming. You should disable those middlewares whenever streaming in development or production. For instance, Rack::Bug won’t work when streaming as it needs to inject contents in the HTML body.

Also Rack::Cache won’t work with streaming as it does not support streaming bodies yet. Whenever streaming Cache-Control is automatically set to “no-cache”.

Errors

When it comes to streaming, exceptions get a bit more complicated. This happens because part of the template was already rendered and streamed to the client, making it impossible to render a whole exception page.

Currently, when an exception happens in development or production, Rails will automatically stream to the client:

"><script>window.location = "/500.html"</script></html>

The first two characters (">) are required in case the exception happens while rendering attributes for a given tag. You can check the real cause for the exception in your logger.

Web server support

Not all web servers support streaming out-of-the-box. You need to check the instructions for each of them.

Unicorn

Unicorn supports streaming but it needs to be configured. For this, you need to create a config file as follow:

# unicorn.config.rb
listen 3000, tcp_nopush: false

And use it on initialization:

unicorn_rails --config-file unicorn.config.rb

You may also want to configure other parameters like :tcp_nodelay.

For more information, please check the documentation.

If you are using Unicorn with NGINX, you may need to tweak NGINX. Streaming should work out of the box on Rainbows.

Passenger

Phusion Passenger with NGINX, offers two streaming mechanisms out of the box.

  1. NGINX response buffering mechanism which is dependent on the value of passenger_buffer_response option (default is “off”).

  2. Passenger buffering system which is always ‘on’ irrespective of the value of passenger_buffer_response.

When passenger_buffer_response is turned “on”, then streaming would be done at the NGINX level which waits until the application is done sending the response back to the client.

For more information, please check the documentation.