hyperion

Hyperion is a lightning fast Ruby object model backed with Redis. It was created in 48 hours to back Leatherbound.me, a ebook comparison shopping tool that was built for Rails Rumble. We withstood 15,000 uniques on Monday morning, and Hyperion stayed super quick even with tons of concurrent queries on a small Linode.

Key features

CRUD on Ruby classes, any attributes get stuffed into a Redis key.
Indexing of objects on any attributes or collections of attributes.

A quickie

adrian$ gem install hyperion require ‘hyperion’ class Person < Hyperion # First we set up the attributes on this object. attr_accessor :name, :email, :password, :other_amazing_info, :car_color, :state # Let’s use the email as our primary key. # Another option would be to not define our own, in which case :id will be created & used. hyperion_key :email # Lets add some indexes! # We can search for all Persons with a given name. hyperion_index :name # We can also search for all people with a given car color and state. hyperion_index [:car_color, :state] end

p1 = Person.new(=> ‘Adrian Pike’, :email => ‘[email protected]’, …) p1.save Person.find(‘[email protected]’) #=> #<Person @name=“Adrian Pike”, @email=“[email protected]”> Person.find(:name => ‘Adrian Pike’) #=> [#<Person @name=“Adrian Pike”, @email=“[email protected]”>] Person.find(:car_color => ‘purple’, :state => ‘denial’) #=> [#<Person @name=“Adrian Pike”, @email=“[email protected]”>]

Requirements

Ruby 1.9
activesupport
Redis

Roadmap

Change the indexes to use ZSETs.
Delete. (ha!)
Schema versioning and migrations.
Faster faster faster! Let’s utilize more of Redis’ exposed data structures to make it better.
Clever auto-sharding based on primary keyspace
- Figure out a way to auto-balance shards
Slow full-table searches.
Partial index searches

Note on Patches/Pull Requests