BennyCache
BennyCache is a model, data and method caching library that uses the ActiveRecord API, but does not try to get in between you and ActiveRecord. The main motivation for creating BennyCache was to make it possible to implicitly and efficiently clear the cache of one record when changes to related records are made.
For example, suppose an Agent has a set of Items in its Inventory. We have an agent's data is populated in the cache, and we want to change the agent's inventory, either add a new item, or update the remaining ammo of a weapon. With BennyCache, we can update an individual Item and the Agent's inventory cache is cleared. We do not have to load the agent into memory, and the agent's basic information remains unchanged in the cache -- only the Inventory data for that Agent is flushed.
BennyCache uses Rails.cache if available, or you can provide your own caching engine. Otherwise, it uses an internal memory cache by default. The internal memory cache is meant for testing and evaluation purposes only.
Contrasting BennyCache with other similar caching tools:
Differences
- BennyCache is marginally aware of ActiveRecord, but doesn't touch the internals, so it should be forward-compatible, or as much as it can be.
- Usage of BennyCache is explicit: It doesn't try to do hide itself from the code.
- Method caching in BennyCache was a bit inspired by CacheMethod. CacheMethod is a more robust solution for method caching, especially if you are passing complex data structures as parameters.
Installation
Add this line to your application's Gemfile:
gem 'benny_cache'
And then execute:
$ bundle
Or install it yourself as:
$ gem install benny_cache
Usage
BennyCache will cache three separate but related types of information.
- Model Cache Model Caches are a cached representation of a model
- Data Cache Data Caches are cached representation of data related to a model, but not the model itself.
- Method Cache Method Caches are cached result of a call to model method.
Model, data and method caches are independent of each other. A model is cached and uncached independently of the data cache or method cache related ot the model. You can cache a model method without caching the model itself.
Another important concept in BennyCache is the Related Index When related indexes are defined, a link is created between one model and another model's data or method caches. For example, when a "skill" (an instance of the Skill class) is upgraded, the related Robot will automatically have its skills method cache cleared, without having to directly reference a Robot model. The relationship is created by defining a method_index in the Robot class, and a related_index in the Skill class.
Defining the cache store
By default, BennyCache uses Rails.cache store if available. You can explicitly set the cache store by calling:
BennyCache::Config.store = Rails.cache
The cache is expected to support the #read
, #write
, #delete
and #fetch
methods of the
ActiveSuport::Cache::Store
interface.
If Rails.cache is not defined, and you don't explicitly initialize a cache, BennyCache uses an internal
BennyCache::Cache
object, which is just an in-memory key-value hash. The internal cache
object is not intended for production use.
Model Indexes
Include BennyCache::Model
into your ActiveRecord model and declare your indexes:
class Robot < ActiveRecord::Base
include BennyCache::Model
benny_model_index :user_id
end
class Location < ActiveRecord::Base
include BennyCache::Model
benny_model_index [:x, :y]
end
Once that is done, loading item from the cache is easy:
robot = Robot.benny_model_cache(user_id: current_user.id)
location = Location.benny_model_cache(123) # cached by primary key
location = Location.benny_model_cache(x: 30, y: 50) # cached by coordinates
Calling these will populate the my_model object in the cache with different keys. If you
change location
and call #save()
or #destroy()
, all instances will be cleared from the cache.
Data Indexes
A data index is a piece of data related to a specific model that is created by a block of code. Data caches are maintained separate from the model cache itself. Data caches are useful for data about a model that is expensive to load and/or calculate, or changes infrequently in relation to the model itself.
Data caches and model caches are independent of each other. A model cache is maintained independently of any data caches related to the model.
To use data indexes, you must first declare a data index key. When loading the key, you also pass a block that is used to populate the cache.
class MyModel < ActiveRecord::Base
include BennyCache::Model
benny_data_index :my_data_index
end
Then, when using the model...
my_model.benny_data_cache(:my_data_index) {
self.expensive_data_to_calculate()
}
The return value of self.expensive_data_to_calculate() is used to populate the cache for the :my_data_index key. Further calls my_model.benny_data_cache(:my_data_index) will return the cached value until the cache is cleared. Usually, some external process will invalidate a data cache.
Clearing data index cache
To manually clear a data index cache for a model, you need the primary key of the model and use a class method. You do not need to instantiate the model.
MyModel.benny_data_cache_delete(123, :my_data_index)
If changes to one model might need to invalidate the data caches of another mother, this can be managed automatically
with the BennyCache::Related
mixin describe below.
Method Indexes
A method index is a cached result of a call to a model method. Like Data caches, Method caches are created and deleted separate from the model cache itself. Also like Data caches, Method caches are useful for caching data about a model that is expensive to load or calculate, or changes at different intervals in relation to the model itself.
To use method indexes, declare a method index. Method indexes use ruby method aliasing, so the source method must be defined before declaring the method index.
class MyModel < ActiveRecord::Base
include BennyCache::Model
def method_name # source method first
[expensive_code]
end
benny_method_index :method_name # index second
end
Arguments passed to a cached method are hashed to create a unique signature per parameter list, and the cached value is based on the method name and the args hash sig. In the following example:
rv1 = agent.method_name :foo
rv2 = agent.method_name :bar
If agent#method_name is declared as a method_index, rv1 and and rv2 will be two different cached values.
Local caching
When a BennyCache method index is called, benny cache keeps an model-specific copy of the cache in local memory, so multiple calls to the same method return the same object with the same object_id. This supports in process updates to the data. For example:
rv1 = agent.method_name #=> [:a, :b, :c]
rv1.push :d
rv2 = agent.method_name #=> [:a, :b, :c, :d]
rv1.size == 4 #=> true
rv1.object_id == rv2.object_id #=> true
This behavior works for my needs, but may not suit all users. I may add the ability to change this behavior.
I have not fully tested the benny_method_index functionality with all of the ActiveRelation's varied functionality. Using the two together and exercising different parts of ActiveRelation may have unexpected results. However, in my simple case, where I use basic :has_many relationships and don't use #where, #include, etc, it works the way I need it to work.
Simple use cases should work without issue, but passing complex data structures to cached methods may confuse BennyCache. For more robust method caching, checkout out CacheMethod.
Clearing model index cache
Method indexes will cache data on per-args_hash basis, but clearing the cache for a model index is more of a shotgun approach: clearing a model index cache will clear all cached data for all args hashes.
To manually clear a method index cache for a model, you need the primary key of the model, the method name, and use a class method. You do not need to instantiate the model.
MyModel.benny_method_cache_delete(123, :method_name)
If changes to one model might need to invalidate the data caches of another mother, this can be managed
with the BennyCache::Related
mixin.
Related Indexes
Related indexes are used when you know that a change to one model will need to invalidate a data or method cache of another model. Defining a related index will make the cache invalidation automatic.
BennyCache::Related installs and after_save/after_destroy callback to clear the related data indexes of other models.
Defined the two classes like so, a class that uses BennyCache::Model with a data_index, and a class that uses BennyCache::Related that defines a benny_related_index that points to the main class's date_index
class MainModel < ActiveRecord::Base
has_many :related_models
include BennyCache::Model
benny_data_index :my_related_items # data cache
benny_model_index :my_related_method # method cache
end
class RelatedModel < ActiveRecord::Base
belongs_to :main_model
include BennyCache::Related
":main_model_id/MainModel/my_related_items"
":main_model_id/MainModel/my_related_method"
end
The benny_related_index call sets up all RelatedModel instances to clear the :my_related_items
data index of the
MainModel instance withe a primary key of related_model.main_model_id
whenever the RelatedModel instance is created,
updated, or deleted.
The benny_related_method call sets up all RelatedModel instances to clear the :my_related_method
method cache data of the
MainModel instance withe a primary key of related_model.main_model_id
whenever the RelatedModel instance is created,
updated, or deleted.
Cache namespacing
Internally, BennyCache uses the class name of the classes using BennyCache::Model as part of the key name for caching. Sometimes that might not be what you want, and will need to explicitly declare the namespace. This happens when you use classes that take advantage of ActiveRecord's single table inheritance:
class Location < ActiveRecord::Base
include BennyCache::Model
benny_cache_ns 'Location'
end
class IndustrialComplex < Location
end
class RecreationArea < Location
end
By declaring the above namespace, these two calls reference the same key:
l = Location.benny_cache_model 123
l = RecreationArea.benny_cache_model 123
Bugs
There are probably bugs. Until there is an issue tracking system, send email to
[email protected]
and put 'BennyCache' in the subject.
Contributing
- 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