d5nguyenvan / remodel

remodel -- a minimal ORM (object-redis-mapper) in less than 300 lines of ruby

remodel your storage layer

use redis instead of mysql to store your application data.

remodel (= redis model) is an ActiveRecord-like mapping layer which offers familiar syntax like has_many, has_one etc. to build your domain model in ruby.

why redis?

redis offers in-memory read and write performance — on the order of 10K to 100K operations per second, comparable to memcached — plus asynchronous persistence to disk. for example, on my macbook (2 ghz):

$ redis-benchmark -d 100 -r 10000 -q
SET: 13864.27 requests per second
GET: 18152.17 requests per second
INCR: 17006.80 requests per second
LPUSH: 17243.99 requests per second
LPOP: 18706.54 requests per second

how to get started

1. install redis and ezras excellent redis-rb ruby client:

    $ brew install redis
    $ gem install redis
   

2. start redis:

    $ ./redis-server
   

3. now the tests should run successfully:

    $ rake
    Started
    .......................................................................................
    Finished in 0.072304 seconds.
    87 tests, 138 assertions, 0 failures, 0 errors
   

example

define your domain model like this:

class Book < Remodel::Entity
  has_many :chapters, :class => 'Chapter', :reverse => :book
  property :title, :class => 'String'
  property :year, :class => 'Integer'
end

class Chapter < Remodel::Entity
  has_one :book, :class => Book, :reverse => :chapters
  property :title, :class => String
end

now you can do:

>> require 'example/book'
=> true
>> book = Book.create :title => 'Moby Dick', :year => 1851
=> #<Book(b:3) title: "Moby Dick", year: 1851>
>> chapter = book.chapters.create :title => 'Ishmael'
=> #<Chapter(c:4) title: "Ishmael">
>> chapter.book
=> #<Book(b:3) title: "Moby Dick", year: 1851>

inspired by

• how to redis — good overview of different mapping options by mattmatt.
• hurl — basically defunkts Hurl::Model is what i started with.
• ohm — object-hash mapping for redis. somewhat similar, but instead of serializing to json, stores each attribute under a separate key.

todo

• better docs
• find_by
• make serializer (json, messagepack, marshal ...) configurable
• benchmarks

status

still pretty alpha — play around at your own risk :)

license

MIT, baby!