Couch Potato

... is a persistence layer written in ruby for CouchDB.

Build Status


Code Climate


The goal of Couch Potato is to create a minimal framework in order to store and retrieve Ruby objects to/from CouchDB and create and query views.

It follows the document/view/querying semantics established by CouchDB and won't try to mimic ActiveRecord behavior in any way as that IS BAD.

Code that uses Couch Potato should be easy to test.

Lastly Couch Potato aims to provide a seamless integration with Ruby on Rails, e.g. routing, form helpers etc.

Core Features

  • persisting objects by including the CouchPotato::Persistence module
  • declarative views with either custom or generated map/reduce functions
  • extensive spec suite

Supported Environments

Check travis.yml for supported Ruby/ActiveSupport versions.


Couch Potato is hosted as a gem which you can install like this:

gem install couch_potato

Using with your ruby application:

require 'rubygems'
require 'couch_potato'

After that you configure the name of the database:

CouchPotato::Config.database_name = 'name_of_the_db'

The server URL will default to http://localhost:5984/ unless specified:

CouchPotato::Config.database_name = ""

But you can also specify the database host separately from the database name:

CouchPotato::Config.database_host = ""
CouchPotato::Config.database_name = "name_of_the_db"

Or with authentication

CouchPotato::Config.database_name = ""

Optionally you can configure the default language for design documents (:javascript (default) or :erlang).

CouchPotato::Config.default_language = :javascript | :erlang

Another switch allows you to store each CouchDB view in its own design document. Otherwise views are grouped by model.

CouchPotato::Config.split_design_documents_per_view = true

If you are using more than one database from your app, you can create aliases:

CouchPotato::Config.additional_databases = {'db1' => 'db1_production', 'db2' => ''}
db1 = CouchPotato.use 'db1'

Using with Rails

Create a config/couchdb.yml:

default: &default
  split_design_documents_per_view: true # optional, default is false
  digest_view_names: true # optional, default is false
  default_language: :erlang # optional, default is javascript
  database_host: ""

  <<: *default
  database: development_db_name
  <<: *default
  database: test_db_name
  <<: *default
  database: <%= ENV['DB_NAME'] %>
    db1: db1_production


Add to your Gemfile:

# gem 'rails' # we don't want to load activerecord so we can't require rails
gem 'railties'
gem 'actionpack'
gem 'actionmailer'
gem 'activemodel'
gem "couch_potato"
gem 'tzinfo'

Note: please make sure that when you run in the Rails console it returns something like 2010/12/10 and not 2010-12-10 - if it does another gem has overwritten Couch Potato's Date patches - in this case move Couch Potato further down in your Gemfile or whereever you load it.


This is a basic tutorial on how to use Couch Potato. If you want to know all the details feel free to read the specs and the rdocs.

Save, load objects

First you need a class.

class User

To make instances of this class persistent include the persistence module:

class User
  include CouchPotato::Persistence

If you want to store any properties you have to declare them:

class User
  include CouchPotato::Persistence

  property :name

Properties can be typed:

class User
  include CouchPotato::Persistence

  property :address, :type => Address

In this case Address also implements CouchPotato::Persistence which means its JSON representation will be added to the user document. Couch Potato also has support for the basic types (right now Integer, Date, Time and :boolean are supported):

class User
  include CouchPotato::Persistence

  property :age, :type => Integer
  property :receive_newsletter, :type => :boolean

With this in place when you set the user's age as a String (e.g. using an HTML form) it will be converted into a Integer automatically.

Properties can have a default value:

class User
  include CouchPotato::Persistence

  property :active, :default => true
  property :signed_up, :default => { }

Now you can save your objects. All database operations are encapsulated in the CouchPotato::Database class. This separates your domain logic from the database access logic which makes it easier to write tests and also keeps you models smaller and cleaner.

user = :name => 'joe'
CouchPotato.database.save_document user # or save_document!

You can of course also retrieve your instance:

CouchPotato.database.load_document "id_of_the_user_document" # => <#User 0x3075>

Handling conflicts

CouchDB uses MVCC to detect write conflicts. If a conflict occurs when trying to update a document CouchDB returns an error. To handle conflicts easily you can save documents like this:

CouchPotato.database.save_document user do |user| = 'joe'

When a conflict occurs Couch Potato automatically reloads the document, runs the block and tries to save it again. Note that the block is also run before initally saving the document.

Caching load reqeusts

You can add a cache to a database instance to enable caching subsequent #load calls to the same id. Any write operation will completely clear the cache.

db = CouchPotato.database
db.cache = {}
db.load '1'
db.load '1' # goes to the cache instead of to the database

In web apps, the idea is to use a per request cache, i.e. set a new cache for every request.

Operations on multiple documents

You can also load a bunch of documents with one request.

CouchPotato.database.load ['user1', 'user2', 'user3'] # => [<#User 0x3075>, <#User 0x3076>, <#User 0x3077>]


You can access the properties you declared above through normal attribute accessors. # => 'joe' = {:first => ['joe', 'joey'], :last => 'doe', :middle => 'J'} # you can set any ruby object that responds_to :to_json (includes all core objects)
user._id # => "02097f33a0046123f1ebc0ebb6937269"
user._rev # => "2769180384"
user.created_at # => Fri Oct 24 19:05:54 +0200 2008
user.updated_at # => Fri Oct 24 19:05:54 +0200 2008 # => false

If you want to have properties that don't map to any JSON type, i.e. other than String, Number, Boolean, Hash or Array you have to define the type like this:

class User
  property :date_of_birth, :type => Date

The date_of_birth property is now automatically serialized to JSON and back when storing/retrieving objects.

If you want to store an Array of objects, just pass the definiton as an Array of Dates:

class User
  property :birthdays, :type => [Date]

Dirty tracking

CouchPotato tracks the dirty state of attributes in the same way ActiveRecord does. Models are always saved though to avoid missing updates when multiple processes update the same documents concurrently.

user = User.create :name => 'joe' # => 'joe'
user.name_changed? # => false
user.name_was # => nil

Object validations

Couch Potato by default uses ActiveModel for validation

class User
  property :name
  validates_presence_of :name

user =
user.valid? # => false
user.errors[:name] # => ['can't be blank']

Finding stuff / views / lists

In order to find data in your CouchDB you have to create a view first. Couch Potato offers you to create and manage those views for you. All you have to do is declare them in your classes:

class User
  include CouchPotato::Persistence
  property :name

  view :all, :key => :created_at

This will create a view called "all" in the "user" design document with a map function that emits "created_at" for every user document.

CouchPotato.database.view User.all

This will load all user documents in your database sorted by created_at.

For large data sets, use batches:

CouchPotato.database.view_in_batches(User.all, batch_size: 100) do |users|

This will query CouchDB with skip/limit until all documents have been yielded.

CouchPotato.database.view User.all(:key => ( 10)..(, :descending => true)

Any options you pass in will be passed onto CouchDB.

Composite keys are also possible:

class User
  property :name

  view :all, :key => [:created_at, :name]

You can let Couch Potato generate these map/reduce functions in Erlang, which reslts in much faster view generation:

class User
  property :name

  view :all, :key => [:created_at, :name], :language => :erlang

So far only very simple views like the above work with Erlang.

You can also pass conditions as a JavaScript string:

class User
  property :name

  view :completed, :key => :name, :conditions => 'doc.completed === true'

The creation of views is based on view specification classes (see CouchPotato::View::BaseViewSpec and its descendants for more detailed documentation). The above code uses the ModelViewSpec class which is used to find models by their properties. For more sophisticated searches you can use other view specifications (either use the built-in or provide your own) by passing a type parameter:

If you have larger structures and you only want to load some attributes you can use the PropertiesViewSpec (the full class name is automatically derived):

class User
  property :name
  property :bio

  view :all, :key => :created_at, :properties => [:name], :type => :properties

CouchPotato.database.view(User.everyone) # => "joe"
CouchPotato.database.view(User.everyone) # => nil

CouchPotato.database.first(User.everyone).name # => "joe" # convenience function, returns nil if nothing found
CouchPotato.database.first!(User.everyone) # would raise CouchPotato::NotFound if nothing was found

If you want Rails to automatically show a 404 page when CouchPotato::NotFound is raised add this to your ApplicationController:

rescue_from CouchPotato::NotFound do
  render(:file => 'public/404.html', :status => :not_found, :layout => false)

You can also pass in custom map/reduce functions with the custom view spec:

class User
  view :all, :map => "function(doc) { emit(doc.created_at, null)}", :include_docs => true, :type => :custom

commonJS modules can also be used in custom views:

class User
  view :all, :map => "function(doc) { emit(null, require("views/lib/test").test)}", :lib => {:test => "exports.test = 'test'"}, :include_docs => true, :type => :custom

If you don't want the results to be converted into models the raw view is your friend:

class User
  view :all, :map => "function(doc) { emit(doc.created_at,}", :type => :raw

When querying this view you will get the raw data returned by CouchDB which looks something like this:

  "total_entries": 2,
  "rows": [
      "value": "alex",
      "key": "2009-01-03 00:02:34 +000",
      "id": "75976rgi7546gi02a"

To process this raw data you can also pass in a results filter:

class User
  view :all, :map => "function(doc) { emit(doc.created_at,}", :type => :raw, :results_filter => lambda {|results| results['rows'].map{|row| row['value']}}

In this case querying the view would only return the emitted value for each row.

You can pass in your own view specifications by passing in :type => MyViewSpecClass. Take a look at the CouchPotato::View::*ViewSpec classes to get an idea of how this works.

Digest view names

If turned on, Couch Potato will append an MD5 digest of the map function to each view name. This makes sure (together with split_design_documents_per_view) that no views/design documents are ever updated. Instead, new ones are created. Since reindexing can take a long time once your database is larger, you want to avoid blocking your app while CouchDB is busy. Instead, you create a new view, warm it up, and only then start using it.


CouchPotato also supports CouchDB lists. With lists you can process the result of a view query with another JavaScript function. This can be useful for example if you want to filter your results, or add some data to each document.

Defining a list works similarly to views:

class User
  include CouchPotato::Persistence

  property :first_name
  view :with_full_name, key: first_namne, list: :add_last_name
  view :all, key: :first_name

  list :add_last_name, <<-JS
    function(head, req) {
      var row;
      send('{"rows": [');
      while(row = getRow()) { = row.doc.first_name + ' doe';
end 'joe')
CouchPotato.database.view(User.with_full_name) # => 'joe doe'

You can also pass in the list at query time:

CouchPotato.database.view(User.all(list: :add_last_name))

And you can pass parameters to the list:

CouchPotato.database.view(User.all(list: :add_last_name, list_params: {filter: '*'}))


Not supported. Not sure if they ever will be. You can implement those yourself using views and custom methods on your models.


Couch Potato supports the usual lifecycle callbacks known from ActiveRecord:

class User
  include CouchPotato::Persistence

  before_create :do_something_before_create
  before_update {|user| user.do_something_on_update}

This will call the method do_something_before_create before creating an object and run the given lambda before updating one. Lambda callbacks get passed the model as their first argument. Method callbacks don't receive any arguments.

Supported callbacks are: :before_validation, :before_validation_on_create, :before_validation_on_update, :before_validation_on_save, :before_create, :after_create, :before_update, :after_update, :before_save, :after_save, :before_destroy, :after_destroy.

If you need access to the database in a callback: Couch Potato automatically assigns a database instance to the model before saving and when loading. It is available as database accessor from within your model instance.


There is basic attachment support: if you want to store any attachments set the _attachments attribute of a model before saving like this:

class User
  include CouchPotato::Persistence

data ='some_file.text') # or from upload
user =
user._attachments = {'photo' => {'data' => data, 'content_type' => 'image/png'}}

When saving this object an attachment with the name photo will be uploaded into CouchDB. It will be available under the url of the user object + /photo. When loading the user at a later time you still have access to the content_type and additionally to the length of the attachment:

user_reloaded = CouchPotato.database.load
user_reloaded._attachments['photo'] # => {'content_type' => 'image/png', 'length' => 37861}

Multi DB Support

Couch Potato supports accessing multiple CouchDBs:

CouchPotato.with_database('couch_customer') do |couch| @customer

Unless configured otherwise this would save the customer model to

You can also first retrieve the database instance:

db = CouchPotato.use('couch_customer') @customer


To make testing easier and faster database logic has been put into its own class, which you can replace and stub out in whatever way you want:

class User
  include CouchPotato::Persistence

# RSpec
describe 'save a user' do
  it 'should save' do
    couchrest_db = stub 'couchrest_db',
    database = couchrest_db
    user =
    database.save_document user

By creating your own instances of CouchPotato::Database and passing them a fake CouchRest database instance you can completely disconnect your unit tests/spec from the database.

For stubbing out the database couch potato offers some helpers via the couch_potato-rspec gem. Use version 2.x of the gem if you are on RSpec 2, use 3.x for RSpec 3.

class Comment
  view :by_commenter_id, :key => :commenter_id

# RSpec
require 'couch_potato/rspec'

db = stub_db # stubs CouchPotato.database
db.stub_view(Comment, :by_commenter_id).with('23').and_return([:comment1, :comment2])

CouchPotato.database.view(Comment.by_commenter_id('23')) # => [:comment1, :comment2]
CouchPotato.database.view_in_batches(Comment.by_commenter_id('23'), batch_size: 1) # => yields [:comment1] and [:comment2]
CouchPotato.database.first(Comment.by_commenter_id('23)) # => :comment1
Testing map/reduce functions

Couch Potato provides custom RSpec matchers for testing the map and reduce functions of your views. For example you can do this:

class User
  include CouchPotato::Persistence

  property :name
  property :age, :type => Integer

  view :by_name, :key => :name
  view :by_age,  :key => :age
  view :oldest_by_name,
    :map => "function(doc) { emit(, doc.age); }",
    :reduce => "function(keys, values, rereduce) { return Math.max.apply(this, values); }"

require 'couch_potato/rspec'

describe User, 'views' do
  it "should map users to their name" do
    User.by_name.should map( => 'bill', :age => 23)).to(['bill', 1])

  it "should reduce the users to the sum of their age" do
    User.by_age.should reduce([], [23, 22]).to(45)

  it "should map/reduce users to the oldest age by name" do
    docs = [ => "John", :age => 25), => "John", :age => 30), => "Jane", :age => 20)]
    User.oldest_by_name.should map_reduce(docs).with_options(:group => true).to(
      {"key" => "John", "value" => 30}, {"key" => "Jane", "value" => 20})

This will actually run your map/reduce functions in a JavaScript interpreter, passing the arguments as JSON and converting the results back to Ruby. map_reduce specs map the input documents, reduce the emitted keys/values, and rereduce the results while also respecting the :group and :group_level couchdb options. For more examples see the spec.

In order for this to work you must have the js executable in your PATH. This is usually part of the spidermonkey package/port. (On MacPorts that's spidemonkey, on Linux it could be one of libjs, libmozjs or libspidermonkey). When you installed CouchDB via your packet manager Spidermonkey should already be there.

Helping out

Please fix bugs, add more specs, implement new features by forking the github repo at

Issues are tracked at github:

There is a mailing list, just write to:

You can run all the specs by calling 'rake spec_unit' and 'rake spec_functional' in the root folder of Couch Potato. The specs require a running CouchDB instance at http://localhost:5984

I will only accept patches that are covered by specs - sorry.


If you have any questions/suggestions etc. please contact me at alex at or @langalex on twitter.