Switch branches/tags
Find file
Fetching contributors…
Cannot retrieve contributors at this time
197 lines (144 sloc) 6.34 KB

GitModel: distributed, versioned NoSQL for Ruby

GitModel is an ActiveModel-compliant persistence framework for Ruby that uses Git for versioning and remote syncing.

GitModel persists Ruby objects using Git as a data storage engine. It's an ActiveModel implementation so it works stand-alone or in Rails 3 as a drop-in replacement for ActiveRecord or DataMapper.

Because the database is a Git repository it can be synced across multiple machines, manipulated with standard Git client tools, can be branched and merged, and of course keeps the history of all changes.

Why it's awesome

  • Schema-less NoSQL data store
  • Each record is a normal Ruby object, attributes are any Ruby type or large chunks of binary data
  • Never lose data, history is kept forever and can be restored simply using standard Git tools
  • Branch and merge your production data
    • GitModel can actually work with different branches
    • Branch or tag snapshots of your data
    • Experiment on production data using branches, for example to test a migration
  • Distributed (synced using standard Git push/pull)
  • All ActiveModel
  • Transactions
  • Metadata for all database changes (Git commit messages, date & time, etc.)
  • In order to be easily human-editable, the database is simply files and directores stored in a Git repository. GitModel uses the Git repo directly (rather than Git's checked-out "working copy") but you can do a "git checkout" to view and manipulate the database contents, and then "git commit"
  • Test-driven development and excellent test coverage
  • Clean and easy-to-use API


It is not yet production ready but I'm working on it. Please feel free to contribute tests and/or code to help!

I will attempt to follow Semantic Versioning so 1.0.0 will be considered the first stable release, until then the API may change at any time.

See the "To do" section below for details, but the main thing that needs finishing is support for querying. Right now you can find an instance by it's id, but there is incomplete support (90% complete) for querying, e.g.:

Post.find(:category => 'ruby', :date => lambda{|d| d > 1.month.ago} :order_by => :date, :order => :asc, :limit => 5)

This includes support for indexing all attributes so that queries don't need to load every object.


It's available as a RubyGem:

> gem install gitmodel


GitModel.db_root = '/tmp/gitmodel-data'

class Post
  include GitModel::Persistable

  attribute :title
  attribute :body
  attribute :categories, :default => []
  attribute :allow_comments, :default => true

  blob :image

p1 = => 'lessons-learned', :title => 'Lessons learned', :body => '...')
p1.image = some_binary_data!

p = Post.find('lessons-learned')

p2 = => 'hotdog-eating-contest', :title => 'I won!')
p2.body = 'This weekend I won a hotdog eating contest!'
p2.image = some_binary_data
p2.blobs['hotdogs.jpg'] = some_binary_data
p2.blobs['the-aftermath.jpg'] = some_binary_data!

p3 = Post.create!(:id => 'running-with-scissors', :title => 'Running with scissors', :body => '...')

p4 = Post.find('running-with-scissors')

class Comment
  include GitModel::Persistable
  attribute :text

c1 = Comment.create!(:id => '2010-01-03-328', :text => '...')
c2 = Comment.create!(:id => '2010-05-29-742', :text => '...')

An example of a project that uses GitModel is Balisong, a blogging app for coders (but it doesn't save objects to the data store. It's read-only so far, assuming that posts will be edited with a text editor).

Database file structure

The database is stored in a human-editable format. Simply do "git checkout -f" and you'll see directories and files.

Each type of object is stored in a top-level directory (this is analogous to ActiveRecord tables), and each object is stored in a subdirectory which is named using the object's id (i.e. the primary key). Attributes that are Ruby types (strings, numbers, hashes, arrays, whatever) are stored in a file named attributes.json and binary attributes ("blobs") are stored in their own files.

For example, the database for the example above would have a directory structure that looks like this:

  • db-root
    • comments
      • 2010-01-03-328
        • attributes.json
      • 2010-05-29-742
        • attributes.json
    • posts
      • hotdog-eating-contest
        • attributes.json
        • hotdogs.jpg
        • image
        • the-aftermath.jpg
      • lessons-learned
        • attributes.json
        • image
      • running-with-scissors
        • attributes.json


Do you have an improvement to make? Please submit a pull request on GitHub or a patch, including a test written with RSpec. To run all tests simply run autospec.

The main author is Paul Dowman (@pauldowman).

Thanks to everyone who has contributed so far:

To do

  • Finish Query support
    • Update index (efficiently) when Persistable objects are saved
    • Add Rake task to generate index
    • Update README
  • Add validations and other feature examples to sample code in README
  • Finish some pending specs
  • API documentation
  • Rails integration
    • Generators
    • Rake tasks
  • Performance
    • Haven't optimized for performance yet.
    • Use Rugged instead of Grit
    • Remove the transaction lock (see transaction.rb line 19)
    • Ability to iterate over result set without eager loading of all instances
  • Persistable.find/find_all/etc could be based on staged files so that queries reflect uncommitted changes
  • Better query support
    • Associations
    • Use AREL?