The goal of this project is to unify Cache APIs and make Strategies easy to implement and sharable across all storage types/adapters
The second goal is to make sure testing of all cache related funciions is easy, meaning caches should be isolated per test and not leak their state to outside tests
The package can be installed by adding elixir_cache to your list of dependencies in mix.exs:
def deps do
[
{:elixir_cache, "~> 0.1.0"}
]
endThe docs can be found at https://hexdocs.pm/elixir_cache.
defmodule MyModule do
use Cache,
adapter: Cache.Redis,
name: :my_name,
sandbox?: Mix.env() === :test,
opts: [...opts]
endIn our application.ex
children = [
{Cache, [MyModule]}
]Now we can use MyModule to call various Cache apis
MyModule.get("key") #> {:ok, nil}
MyModule.put("key", "value") #> :ok
MyModule.get("key") #> {:ok, "value"}Cache.Agent- Simple agent based cachingCache.DETS- Disk persisted caching with detsCache.ETS- Super quick in-memory cache withetsCache.Redis- Caching adapter using Redix & Poolboy, supports Redis JSON and Redis HashesCache.ConCache- Wrapper around ConCache libraryCache.PersistentTerm- Zero-latency reads for rarely-written data via Erlang's:persistent_termCache.Counter- Lock-free atomic integer counters via Erlang's:countersmodule
Some adapters have specific functions such as redis which has hash functions and pipeline functions to make calls easier.
These adapters when used will add extra commands to your cache module.
Cache.Counter adds increment/1,2 and decrement/1,2 to your module:
defmodule MyApp.CounterCache do
use Cache,
adapter: Cache.Counter,
name: :my_app_counter_cache,
opts: []
end
MyApp.CounterCache.increment(:page_views)
MyApp.CounterCache.decrement(:active_users)
{:ok, count} = MyApp.CounterCache.get(:page_views)Strategy adapters compose over existing adapters to provide higher-level caching patterns.
Use a two-element tuple as the adapter option:
use Cache,
adapter: {StrategyModule, UnderlyingAdapterOrConfig},
name: :my_cache,
opts: [strategy_opt: value, underlying_adapter_opt: value]Cache.HashRing- Distribute keys across Erlang cluster nodes via consistent hashing (libring)Cache.MultiLayer- Cascade reads/writes through multiple cache layers (e.g. ETS → Redis) with automatic backfillCache.RefreshAhead- Proactively refresh hot keys in the background before they expire
See the Using Strategy Adapters guide for full details.
Our cache config accepts a sandbox?: boolean. In sandbox mode, the Cache.Sandbox adapter will be used, which is just a simple Agent cache unique to the root process. The Cache.SandboxRegistry is responsible for registering test processes to a
unique instance of the Sandbox adapter cache. This makes it safe in test mode to run all your tests asynchronously!
For test isolation via the Cache.SandboxRegistry to work, you must start the registry in your test/test_helper.exs:
Cache.SandboxRegistry.start_link()
ExUnit.start()Then inside a setup block:
Cache.SandboxRegistry.start([MyCache, CacheItem])For applications with many test files, use Cache.CaseTemplate to define a single CacheCase module that automatically starts sandboxed caches in every test that uses it.
Create a CacheCase module in your test support directory:
# test/support/cache_case.ex
defmodule MyApp.CacheCase do
use Cache.CaseTemplate, default_caches: [MyApp.UserCache, MyApp.SessionCache]
endOr discover caches automatically from a running supervisor:
defmodule MyApp.CacheCase do
use Cache.CaseTemplate, supervisors: [MyApp.Supervisor]
endThen use it in any test file:
defmodule MyApp.SomeTest do
use ExUnit.Case, async: true
use MyApp.CacheCase
# optionally add extra caches just for this file:
# use MyApp.CacheCase, caches: [MyApp.ExtraCache]
endAdapters are very easy to create in this model and are basically just a module that implement the @behaviour Cache
This behaviour adds the following callbacks
put(cache_name, key, ttl, value, opts \\ [])
get(cache_name, key, opts \\ [])
delete(cache_name, key, opts \\ [])
opts_definition() # NimbleOptions definition map
child_spec({cache_name, cache_opts})
Cache.ETS is probably the easiest adapter to follow as a guide as it's a simple Task
Adapter configuration can also be specified at runtime. These options are first passed to the adapter child_spec when starting the adapter and then passed to all runtime function calls.
For example:
# Configure with Module Function
defmodule Cache.Example do
use Cache,
adapter: Cache.Redis,
name: :test_cache_redis,
opts: {Cache.Example, :opts, []}
def opts, do: [host: "localhost", port: 6379]
end
# Configure with callback function
defmodule Cache.Example do
use Cache,
adapter: Cache.Redis,
name: :test_cache_redis,
opts: &Cache.Example.opts/0
def opts, do: [host: "localhost", port: 6379]
end
# Fetch from application config
# config :elixir_cache, Cache.Example, []
defmodule Cache.Example do
use Cache,
adapter: Cache.Redis,
name: :test_cache_redis,
opts: :elixir_cache
end
# Fetch from application config
# config :elixir_cache, :cache_opts, []
defmodule Cache.Example do
use Cache,
adapter: Cache.Redis,
name: :test_cache_redis,
opts: {:elixir_cache, :cache_opts}
endRuntime options can be configured in one of the following formats:
{module, function, args}- Module, function, args{application_name, key}- Application name. This is called asApplication.fetch_env!(application_name, key).application_name- Application name as an atom. This is called asApplication.fetch_env!(application_name, cache_module}).function- Zero arity callback function. For eg.&YourModule.options/0[key: value_type]- Keyword list of options.