Flowdock

This module provides methods for generating HTML that links views to assets such as images, javascripts, stylesheets, and feeds. These methods do not verify the assets exist before linking to them:

image_tag("rails.png")
# => <img alt="Rails" src="/images/rails.png?1230601161" />
stylesheet_link_tag("application")
# => <link href="/stylesheets/application.css?1232285206" media="screen" rel="stylesheet" type="text/css" />

Using asset hosts

By default, Rails links to these assets on the current host in the public folder, but you can direct Rails to link to assets from a dedicated asset server by setting ActionController::Base.asset_host in the application configuration, typically in config/environments/production.rb. For example, you’d define assets.example.com to be your asset host this way:

ActionController::Base.asset_host = "assets.example.com"

Helpers take that into account:

image_tag("rails.png")
# => <img alt="Rails" src="http://assets.example.com/images/rails.png?1230601161" />
stylesheet_link_tag("application")
# => <link href="http://assets.example.com/stylesheets/application.css?1232285206" media="screen" rel="stylesheet" type="text/css" />

Browsers typically open at most two simultaneous connections to a single host, which means your assets often have to wait for other assets to finish downloading. You can alleviate this by using a %d wildcard in the asset_host. For example, “assets%d.example.com”. If that wildcard is present Rails distributes asset requests among the corresponding four hosts “assets0.example.com”, …, “assets3.example.com”. With this trick browsers will open eight simultaneous connections rather than two.

image_tag("rails.png")
# => <img alt="Rails" src="http://assets0.example.com/images/rails.png?1230601161" />
stylesheet_link_tag("application")
# => <link href="http://assets2.example.com/stylesheets/application.css?1232285206" media="screen" rel="stylesheet" type="text/css" />

To do this, you can either setup four actual hosts, or you can use wildcard DNS to CNAME the wildcard to a single asset host. You can read more about setting up your DNS CNAME records from your ISP.

Note: This is purely a browser performance optimization and is not meant for server load balancing. See http://www.die.net/musings/page_load_time/ for background.

Alternatively, you can exert more control over the asset host by setting asset_host to a proc like this:

ActionController::Base.asset_host = Proc.new { |source|
  "http://assets#{Digest::MD5.hexdigest(source).to_i(16) % 2 + 1}.example.com"
}
image_tag("rails.png")
# => <img alt="Rails" src="http://assets1.example.com/images/rails.png?1230601161" />
stylesheet_link_tag("application")
# => <link href="http://assets2.example.com/stylesheets/application.css?1232285206" media="screen" rel="stylesheet" type="text/css" />

The example above generates “http://assets1.example.com” and “http://assets2.example.com”. This option is useful for example if you need fewer/more than four hosts, custom host names, etc.

As you see the proc takes a source parameter. That’s a string with the absolute path of the asset with any extensions and timestamps in place, for example “/images/rails.png?1230601161”.

 ActionController::Base.asset_host = Proc.new { |source|
   if source.starts_with?('/images')
     "http://images.example.com"
   else
     "http://assets.example.com"
   end
 }
image_tag("rails.png")
# => <img alt="Rails" src="http://images.example.com/images/rails.png?1230601161" />
stylesheet_link_tag("application")
# => <link href="http://assets.example.com/stylesheets/application.css?1232285206" media="screen" rel="stylesheet" type="text/css" />

Alternatively you may ask for a second parameter request. That one is particularly useful for serving assets from an SSL-protected page. The example proc below disables asset hosting for HTTPS connections, while still sending assets for plain HTTP requests from asset hosts. If you don’t have SSL certificates for each of the asset hosts this technique allows you to avoid warnings in the client about mixed media.

ActionController::Base.asset_host = Proc.new { |source, request|
  if request.ssl?
    "#{request.protocol}#{request.host_with_port}"
  else
    "#{request.protocol}assets.example.com"
  end
}

You can also implement a custom asset host object that responds to call and takes either one or two parameters just like the proc.

config.action_controller.asset_host = AssetHostingWithMinimumSsl.new(
  "http://asset%d.example.com", "https://asset1.example.com"
)

Customizing the asset path

By default, Rails appends asset’s timestamps to all asset paths. This allows you to set a cache-expiration date for the asset far into the future, but still be able to instantly invalidate it by simply updating the file (and hence updating the timestamp, which then updates the URL as the timestamp is part of that, which in turn busts the cache).

It’s the responsibility of the web server you use to set the far-future expiration date on cache assets that you need to take advantage of this feature. Here’s an example for Apache:

# Asset Expiration
ExpiresActive On
<FilesMatch "\.(ico|gif|jpe?g|png|js|css)$">
  ExpiresDefault "access plus 1 year"
</FilesMatch>

Also note that in order for this to work, all your application servers must return the same timestamps. This means that they must have their clocks synchronized. If one of them drifts out of sync, you’ll see different timestamps at random and the cache won’t work. In that case the browser will request the same assets over and over again even thought they didn’t change. You can use something like Live HTTP Headers for Firefox to verify that the cache is indeed working.

This strategy works well enough for most server setups and requires the least configuration, but if you deploy several application servers at different times - say to handle a temporary spike in load - then the asset time stamps will be out of sync. In a setup like this you may want to set the way that asset paths are generated yourself.

Altering the asset paths that Rails generates can be done in two ways. The easiest is to define the RAILS_ASSET_ID environment variable. The contents of this variable will always be used in preference to calculated timestamps. A more complex but flexible way is to set ActionController::Base.config.asset_path to a proc that takes the unmodified asset path and returns the path needed for your asset caching to work. Typically you’d do something like this in config/environments/production.rb:

# Normally you'd calculate RELEASE_NUMBER at startup.
RELEASE_NUMBER = 12345
config.action_controller.asset_path = proc { |asset_path|
  "/release-#{RELEASE_NUMBER}#{asset_path}"
}

This example would cause the following behavior on all servers no matter when they were deployed:

image_tag("rails.png")
# => <img alt="Rails" src="/release-12345/images/rails.png" />
stylesheet_link_tag("application")
# => <link href="/release-12345/stylesheets/application.css?1232285206" media="screen" rel="stylesheet" type="text/css" />

Changing the asset_path does require that your web servers have knowledge of the asset template paths that you rewrite to so it’s not suitable for out-of-the-box use. To use the example given above you could use something like this in your Apache VirtualHost configuration:

<LocationMatch "^/release-\d+/(images|javascripts|stylesheets)/.*$">
  # Some browsers still send conditional-GET requests if there's a
  # Last-Modified header or an ETag header even if they haven't
  # reached the expiry date sent in the Expires header.
  Header unset Last-Modified
  Header unset ETag
  FileETag None

  # Assets requested using a cache-busting filename should be served
  # only once and then cached for a really long time. The HTTP/1.1
  # spec frowns on hugely-long expiration times though and suggests
  # that assets which never expire be served with an expiration date
  # 1 year from access.
  ExpiresActive On
  ExpiresDefault "access plus 1 year"
</LocationMatch>

# We use cached-busting location names with the far-future expires
# headers to ensure that if a file does change it can force a new
# request. The actual asset filenames are still the same though so we
# need to rewrite the location from the cache-busting location to the
# real asset location so that we can serve it.
RewriteEngine On
RewriteRule ^/release-\d+/(images|javascripts|stylesheets)/(.*)$ /$1/$2 [L]
Show files where this module is defined (5 files)
Register or log in to add new notes.
January 27, 2010
1 thank

Alternative hostname generation method

Instead of using a random number to generate the hostname for the single asset, I prefer using source.hash.modulo, so that a given file is always served from the same host. This makes the content more cacheable by browsers and proxies.

ActionController::Base.asset_host = Proc.new { |source|
  "http://assets#{ source.hash.modulo(9) }.example.com"
}

I didn’t benchmark how long it takes, but String#hash should be reasonably fast.

December 16, 2010
0 thanks

RE: Alternative hostname generation method

According to compute_asset_host as of rails 2.0.0 the %d expansion works in the same way, it’s not a random number generated on each call but the source hash mod 4.

In my brief testing in rails 2.3.9, %d doesn’t appear to work if you use a Proc, so manual generation is of use in that case.