Opt in read through ActiveRecord caching used in production and extracted from Shopify. IdentityCache lets you specify how you want to cache your model objects, at the model level, and adds a number of convenience methods for accessing those objects through the cache. Memcached is used as the backend cache store, and the database is only hit when a copy of the object cannot be found in Memcached.
IdentityCache keeps track of the objects that have cached indexes and uses an after_commit hook to expire those objects, and any up the tree, when they are changed.
Add this line to your application's Gemfile:
gem 'identity_cache'
gem 'cityhash' # optional, for faster hashing (C-Ruby only)And then execute:
$ bundle
Add the following to your environment/production.rb:
config.identity_cache_store = :mem_cache_store, Memcached::Rails.new(:servers => ["mem1.server.com"])Add an initializer with this code:
IdentityCache.cache_backend = ActiveSupport::Cache.lookup_store(*Rails.configuration.identity_cache_store)class Product < ActiveRecord::Base
include IdentityCache
has_many :images
cache_has_many :images, :embed => true
end
# Fetch the product by its id, the primary index.
@product = Product.fetch(id)
# Fetch the images for the Product. Images are embedded so the product fetch would have already loaded them.
@images = @product.fetch_imagesNote: You must include the IdentityCache module into the classes where you want to use it.
IdentifyCache lets you lookup records by fields other than id. You can have multiple of these indexes with any other combination of fields:
class Product < ActiveRecord::Base
include IdentityCache
cache_index :handle, :unique => true
cache_index :vendor, :product_type
end
# Fetch the product from the cache by the index.
# If the object isn't in the cache it is pulled from the db and stored in the cache.
product = Product.fetch_by_handle(handle)
products = Product.fetch_by_vendor_and_product_type(vendor, product_type)This gives you a lot of freedom to use your objects the way you want to, and doesn't get in your way. This does keep an independent cache copy in Memcached so you might want to watch the number of different caches that are being added.
IdentityCache adds fetch_* methods to the classes that you mark with cache indexes, based on those indexes. The example below will add a fetch_by_domain method to the class.
class Shop < ActiveRecord::Base
include IdentityCache
cache_index :domain
endAssociation caches follow suit and add fetch_* methods based on the indexes added for those associations.
class Product < ActiveRecord::Base
include IdentityCache
has_many :images
has_one :featured_image
cache_has_many :images
cache_has_one :featured_image
end
@product.fetch_featured_image
@product.fetch_imagesIdentityCache can easily embed objects into the parents' cache entry. This means loading the parent object will also load the association and add it to the cache along with the parent. Subsequent cache requests will load the parent along with the association in one fetch. This can again mean some duplication in the cache if you want to be able to cache objects on their own as well, so it should be done with care. This works with both cache_has_many and cache_has_one methods.
class Product < ActiveRecord::Base
include IdentityCache
has_many :images
cache_has_many :images, :embed => true
end
@product = Product.fetch(id)
@product.fetch_imagesWith this code, on cache miss, the product and its associated images will be loaded from the db. All this data will be stored into the single cache key for the product. Later requests will load the entire blob of data; @product.fetch_images will not need to hit the db since the images are loaded with the product from the cache.
IdentityCache tries to figure out both sides of an association whenever it can so it can set those up when rebuilding the object from the cache. In some cases this is hard to determine so you can tell IdentityCache what the association should be. This is most often the case when embedding polymorphic associations. The inverse_name option on cache_has_many and cache_has_one lets you specify the inverse name of the association.
class Metafield < ActiveRecord::Base
belongs_to :owner, :polymorphic => true
end
class Product < ActiveRecord::Base
include IdentityCache
has_many :metafields, :as => 'owner'
cache_has_many :metafields, :inverse_name => :owner
endThe :inverse_name => :owner option tells IdentityCache what the association on the other side is named so that it can correctly set the assocation when loading the metafields from the cache.
For cases where you may not need the entire object to be cached, just an attribute from record, cache_attribute can be used. This will cache the single attribute by the key specified.
class Redirect < ActiveRecord::Base
cache_attribute :target, :by => [:shop_id, :path]
end
Redirect.fetch_target_by_shop_id_and_path(shop_id, path)This will read the attribute from the cache or query the database for the attribute and store it in the cache.
Options:
[:unique] Allows you to say that an index is unique (only one object stored at the index) or not unique, which allows there to be multiple objects matching the index key. The default value is false.
Example:
cache_index :handle
Options:
[:embed] Specifies that the association should be included with the parent when caching. This means the associated objects will be loaded already when the parent is loaded from the cache and will not need to be fetched on their own.
[:inverse_name] Specifies the name of parent object used by the association. This is useful for polymorphic associations when the association is often named something different between the parent and child objects.
Example:
cache_has_many :metafields, :inverse_name => :owner, :embed => true
Options:
[:embed] Specifies that the association should be included with the parent when caching. This means the associated objects will be loaded already when the parent is loaded from the cache and will not need to be fetched on their own.
[:inverse_name] Specifies the name of parent object used by the association. This is useful for polymorphic associations when the association is often named something different between the parent and child objects.
Example:
cache_has_one :configuration, :embed => true
Options:
[:by] Specifies what key(s) you want the attribute cached by. Defaults to :id.
Example:
cache_attribute :target, :by => [:shop_id, :path]
Cache reads and writes can be memoized for a block of code to serve duplicate identity cache requests from memory. This can be done for an http request by adding this around filter in your ApplicationController.
class ApplicationController < ActionController::Base
around_filter :identity_cache_memoization
def identity_cache_memoization
IdentityCache.cache.with_memoization{ yield }
end
endA word of warning. Some versions of rails will silently rescue all exceptions in after_commit hooks. If an after_commit fails before the cache expiry after_commit the cache will not be expired and you will be left with stale data.
Since everything is being marshalled and unmarshalled from Memcached changing Ruby or Rails versions could mean your objects cannot be unmarshalled from Memcached. There are a number of ways to get around this such as namespacing keys when you upgrade or rescuing marshal load errors and treating it as a cache miss. Just something to be aware of if you are using IdentityCache and upgrade Ruby or Rails.
IdentityCache is also very much opt-in by deliberate design. This means IdentityCache does not mess with the way normal Rails associations work, and including it in a model won't change any clients of that model until you switch them to use fetch instead of find. This is because there is no way IdentityCache is ever going to be 100% consistent. Processes die, execeptions happen, and network blips occur, which means there is a chance that some database transaction might commit but the corresponding memcached DEL operation does not make it. This means that you need to think carefully about when you use fetch and when you use find. For example, at Shopify, we never use any fetchers on the path which moves money around, because IdentityCache could simply be wrong, and we want to charge people the right amount of money. We do however use the fetchers on performance critical paths where absolute correctness isn't the most important thing, and this is what IdentityCache is intended for.
Caching is hard. Chances are that if some feature was left out, it was left out on purpose because it didn't make sense to cache in that way. This is used in production at Shopify so we are very opinionated about the types of features we're going to add. Please start the discussion early, before even adding code, so that we can talk about the feature you are proposing and decide if it makes sense in IdentityCache.
Types of contributions we are looking for:
- Bug fixes
- Performance improvements
- Documentation and/or clearer interfaces
- Fork it
- Create your feature branch (
git checkout -b my-new-feature) - Commit your changes (
git commit -am 'Added some feature') - Push to the branch (
git push origin my-new-feature) - Create new Pull Request
Camilo Lopez (@camilo)
Tom Burns (@boourns)
Harry Brundage (@hornairs)
Dylan Smith (@dylanahsmith)
Tobias Lütke (@tobi)
John Duff (@jduff)