stockpile

code

github.com/halostatue/stockpile/

bugs

github.com/halostatue/stockpile/issues

continuous integration

<img src=“https://travis-ci.org/halostatue/stockpile.png” />

Description

Stockpile is a simple key-value store connection manager framework. Stockpile itself does not implement a connection manager, but places expectations for implemented connection managers. So far, only Redis has been implemented (stockpile-redis).

Stockpile also provides an adapter so that its functionality can be accessed from within a module.

Features

  • Stockpile manages key-value store connections. There are two variants:

    • wide (Stockpile.new, the default), where additional client connections are new instances of the client library;

    • narrow (Stockpile.new(narrow: true)), where additional client connections use the same client library instance.

  • Stockpile can also be injected into a module (Stockpile.inject!(self, options = {})), which gives the module cache management and adapter methods (.cache and .cache_adapter, by default).

Requirements

The desired key-value store must already be installed and/or specified in your Gemfile.

Synopsis

wide = Stockpile.new # A Stockpile to Redis.
wide.connection.set('hello', 'world') # => 'OK'
wide.connection.get('hello') # => 'world'

# Connections are independent from one another.
wide.connection_for(:other) != wide.connection # => true

# Or set ENV['STOCKPILE_CONNECTION_WDITH'] = 'narrow'
narrow = Stockpile.new(narrow: true) # A 'narrow' Stockpile to Redis.
narrow.connection_for(:other) == narrow.connection # => true

# Special Redis::Namespace handling for Resque. Assumes that redis-namespace
# has been installed, as well.
narrow.connection_for(:resque) != narrow.connection # => true
narrow.connection_for(:resque).redis == narrow.connection # => true

# Show a Stockpile with no adapter capabilities, but name the method
# stockpile, not cache. This will still usefully manage connections.
module Cacher
  Stockpile.inject!(self, method: :stockpile, adaptable: false)
end
Cacher.respond_to?(:stockpile) # => true
Cacher.respond_to?(:stockpile_adapter) # => false
Cacher.stockpile.connection.set('hello', 'world') # => 'OK'
Cacher.stockpile.connection.get('hello') # => 'world'

# Now a Stockpile with adapter capabilities.
module Jobber
  module LastRunTime
    def last_run_time(key, value = nil)
      if value
        connection.hset(__method__, key, value.utc.iso8601)
      else
        value = connection.hget(__method__, key)
        Time.parse(value) if value
      end
    end
  end

  Stockpile.inject!(self)
end
Jobber.respond_to?(:cache) # => true
Jobber.respond_to?(:cache_adapter) # => true

# Four ways:
# 1. Adapt Jobber.cache to recognize #last_run_time.
Jobber.cache_adapter(Jobber::LastRunTime)
Jobber.cache.last_run_time('hello', t = Time.now) # => true
Jobber.cache.last_run_time('hello') # => approximately t

# 2. Adapt Jobber.cache and another module to recognize #last_run_time.
module Foo; end
Jobber.cache_adapter(Jobber::LastRunTime, Foo)
Foo.last_run_time('hello', t = Time.now) # => true
Foo.last_run_time('hello') # => approximately t

# 3. Adapt Jobber.cache and Jobber to recognize #last_run_time.
Jobber.cache_adapter(Jobber::LastRunTime, Jobber)
Jobber.last_run_time('hello', t = Time.now) # => true
Jobber.last_run_time('hello') # => approximately t

# 4. Adapt Jobber.cache and Jobber::LastRunTime to recognize #last_run_time.
Jobber.cache_adapter!(Jobber::LastRunTime)
# or Jobber.cache_adapter(Jobber::LastRunTime, Jobber::LastRunTime)
Jobber::LastRunTime.last_run_time('hello', t = Time.now) # => true
Jobber::LastRunTime.last_run_time('hello') # => approximately t

Background

Stockpile is the evolution of concepts I have applied to Rails applications over the last few years when working with Redis, and avoids the following common but suboptimal patterns:

  • Developers use REDIS or $redis to initialize and access their Redis instances. This could be fixed by using Redis.current, but that still exposes implementation details unnecessarily.

  • Redis methods are often exposed directly in controllers or models, as

    render json: $redis.hget('last_run_time', params[:method])
    

    We don’t like seeing direct database access methods in our controllers, so why do we put up with this for Redis?

  • Each Redis client manages its own connections, and at least one client reconnection is forgotten when using a forking server like Unicorn.

  • Some providers of Redis services restrict the number of simultaneous connections to a given Redis instance. With Rails caching, an application cache, and Resque there are at least three simultaneous connections to Redis for a given Rails server instancne, unless the same connection is reused.

Sample Rails Application

I will be adapting a sample Rails application to demonstrate how Stockpile can be used in Rails. A link to it will be provided here when it is complete.

Install

Stockpile is not intended to be installed by itself, as it does not implement a key-value store specific connection manager. Instead, install an store-specific gem which depends on Stockpile.

gem 'stockpile-redis', '~> 1.0'

Or manually install:

% gem install stockpile-redis

and require Stockpile in your code:

require 'stockpile/redis'

Stockpile Semantic Versioning

Stockpile uses a Semantic Versioning scheme with one change:

  • When PATCH is zero (0), it will be omitted from version references.

:include: Contributing.rdoc

:include: Licence.rdoc