HTTP Caching

All major content, product, browse, and search pages are cached for a short and configurable duration by an HTTP cache. This allows the system to serve a large number of requests in a short span of time. This is implemented by the :cache_page method run in a before_action filter:

# storefront/app/controllers/workarea/storefront/pages_controller.rb
module Workarea
  class Storefront::PagesController < Storefront::ApplicationController
    before_action :cache_page
    # ...

Controller/actions which enable this include:

  • Workarea::Storefront::CategoriesController#show
  • Workarea::Storefront::PagesController#show
  • Workarea::Storefront::PagesController#home_page
  • Workarea::Storefront::ProductsController#show
  • Workarea::Storefront::SearchesController#show

This creates a major complication for any data on the page with user or session specific. Examples include login status, cart count, and personalized product recommendations. Workarea solves this problem by loading this content asynchronously via Javascript. Make sure you review any customized functionality to ensure you won't be caching pages with session-specific data.

Using Rack::Cache

Workarea applications use Rack::Cache to store their HTTP cache. Rack::Cache is a HTTP caching Rack middleware. It's backed by Redis, in which Rack::Cache stores all of its metadata and actual cached content.

Out of the box, the effects of Rack::Cache are disabled in test and development environments.

Ignore User-Specific Tracking Parameters in the Query String

Many stores on the Workarea platform use marketing tools such as Listrak or Emarsys to handle abandoned cart emails, marketing campaigns, and so on. These services sometimes use query parameters within the URL to identify each individual user, and this can cause an issue with caching. Rack::Cache will, by default, generate cache keys based on the given URL. So, if you have multiple requests to your /categories/shirts page, like this:

GET /categories/shirts?tracking_id=96a30500-6add-47e1-9ee8-cf3b5052ecf3
GET /categories/shirts?tracking_id=9d9948e3-df0b-4b17-9396-4bbcbac7f4c9
GET /categories/shirts?tracking_id=0ffc0e09-4041-4876-8f8e-05ae87bc3bf3

That will result in multiple entries written to the cache for the same exact request. This is inefficient at best, and dangerous to your performance at worst. To make sure you won't run into this situation, Rack::Cache allows you to configure ignored query parameters, and since Workarea uses its own subclass of Rack::Cache::Key, you'll need to customize the tracking params on that class like so:

Workarea::Cache::RackCacheKey.query_string_ignore = proc { |key, value| key == 'tracking_id' }

If you're using the Workarea::Listrak plugin, you're in luck! 🍀 Recent versions of the plugin include configuration for the cache key out of the box, so you won't need to do anything to get efficient caches in production.

Customizing the Cache Key

There are scenarios where it becomes necessary to vary Rack::Cache entries, like geolocated content blocks or segmentation-based navigation. In these cases, Workarea provides a relatively straight-forward way to add to the cache key. Refer to the documentation in the code for Workarea::Cache::Varies for more information.

Note: Using Workarea::Cache::Varies can have a very adverse affect on application performance, and is not encouraged. Make sure this is what you need, and feel free to reach out on Slack to see if there may be a better way to achieve your goals.

Resources on HTTP Caching

Now on GitHub