Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
242 lines (192 sloc) 8.72 KB

Changelog (Current version)

  • Fix a bug where rated items would show up in recommended sets
  • Fix support for versions of Redis < 2.4. Redis 1.x is untested.
  • Fix a bug where workers would not queue up jobs.


IMPORTANT: This is a MAJOR version bump that should greatly improve the performance of Recommendable. Most of the library has been rewritten from the ground up and, as such, there are steps you must take to make your application compatibile with this version.

  1. Flush your Redis database. Yes, you'll have to regenerate any recommendations at the end. Sorry.
  2. Please make a migration like the following. This migration will be irreversible if you uncomment the bit to drop the tables. If you do want to drop the tables, you should probably make a backup of your database first. I am not responsible for lost data, angry users, broken marriages, et cetera.
# Require the model that receives recommendations so Recommendable detects it
require Rails.root.join('app', 'models', 'user')

class UpdateRecommendable < ActiveRecord::Migration
  def up
    Recommendable.redis.flushdb # Step 1

    connection = ActiveRecord::Base.connection
    # Transfer likes
    result = connection.execute('SELECT * FROM recommendable_likes')
    result.each do |row|
      liked_set = Recommendable::Helpers::RedisKeyMapper.liked_set_for(row['likeable_type'], row['user_id'])
      liked_by_set = Recommendable::Helpers::RedisKeyMapper.liked_by_set_for(row['likeable_type'], row['likeable_id'])
      Recommendable.redis.sadd(liked_set, row['likeable_id'])
      Recommendable.redis.sadd(liked_by_set, row['user_id'])

    # Transfer dislikes
    result = connection.execute('SELECT * FROM recommendable_dislikes')
    result.each do |row|
      disliked_set = Recommendable::Helpers::RedisKeyMapper.disliked_set_for(row['dislikeable_type'], row['user_id'])
      disliked_by_set = Recommendable::Helpers::RedisKeyMapper.disliked_by_set_for(row['dislikeable_type'], row['dislikeable_id'])
      Recommendable.redis.sadd(disliked_set, row['dislikeable_id'])
      Recommendable.redis.sadd(disliked_by_set, row['user_id'])

    # Transfer hidden items
    result = connection.execute('SELECT * FROM recommendable_ignores')
    result.each do |row|
      set = Recommendable::Helpers::RedisKeyMapper.hidden_set_for(row['ignorable_type'], row['user_id'])
      Recommendable.redis.sadd(set, row['ignorable_id'])

    # Transfer bookmarks
    result = connection.execute('SELECT * FROM recommendable_stashes')
    result.each do |row|
      set = Recommendable::Helpers::RedisKeyMapper.bookmarked_set_for(row['stashable_type'], row['user_id'])
      Recommendable.redis.sadd(set, row['stashable_id'])

    # Recalculate scores
    Recommendable.config.ratable_classes.each do |klass|
      klass.pluck(:id).each { |id| Recommendable::Helpers::Calculations.update_score_for(klass, id) }

    # Remove old tables. Uncomment this if you're feeling confident. Please
    # back up your database before doing this. Thanks.
    # drop_table :recommendable_likes
    # drop_table :recommendable_dislikes
    # drop_table :recommendable_ignores
    # drop_Table :recommendable_stashes

  def down
    # Recreate the tables from previous migrations, if you must.
  • Recommendable no longer requires Rails (it does, however, still require ActiveSupport)
  • Add support for more ORMs: DataMapper, Mongoid, and MongoMapper
  • Renamed the concept of Ignoring to Hiding
  • Renamed the concept of Stashing to Bookmarking
  • Likes, Dislikes, Hidden Items, and Bookmarked items are no longer stored as models. They are, instead, permanently stored in Redis.
  • Added a Configuration class
  • Enable Ruby 1.8.7 support

1.1.7 (Current version)

  • Fix #50, a method that was forgotten to time during the ignoreable => ignorable typo update


  • Minor code cleanup for my benefit


  • Fix #47, a problem where models could not recommend themselves


  • Fix #41, a problem where Resque rake tasks were required regardless of whether or not it was bundled
  • Fix #46 by adding UniqueJob middleware for Sidekiq.
  • Added caches for an item's number of likes and dislikes received


  • Update Redis.


  • Fix #38, a problem with enqueueing users based on updating the score of a recommendable record


  • Support for Sidekiq, Resque, DelayedJob and Rails::Queueing (issue #28)
    • You must manually bundle Sidekiq, Resque, or DelayedJob. Rails::Queueing is available as a fallback for Rails 4.x
  • Use apotonick/hooks to implement callbacks (issue #25). See the detailed README for more info on usage.


  • Dynamic finders now return ActiveRecord::Relations! This means you can chain other ActiveRecord query methods like so:
current_user.recommended_posts.where(:category => "technology")
current_user.stashed_books.where(:author => "Cormac McCarthy")
current_user.disliked_shows.joins(:cast_members).where(' = Kim Kardashian')
  • You can now specify a count for User#recommendations:
  • Bug fixes


  • NOTE: This release is NOT backwards compatible. Please migrate your databases:
rename_column :recommendable_ignores, :ignoreable_id, :ignorable_id
rename_column :recommendable_ignores, :ignoreable_type, :ignorable_type
rename_table  :recommendable_stashed_items, :recommendable_stashes
  • Fix an issue with recommendable models implemented via STI
  • Fix a library-wide typo of "ignoreable" to "ignorable"


  • Yanked due to breaking bug


  • Revert changes made in 0.1.7 due to licensing


  • Yanked and no longer available.
  • Attempted switch from Resque to Sidekiq


  • Dynamic finders for your User class:

current_user.liked_movies current_user.disliked_shows current_user.recommended_movies

  • Implement sorting of recommendable models:
=> [#<Movie id: 14>, #<Movie id: 15>, #<Movie id: 13>, #<Movie id: 12>, #<Movie id: 11>]
=> #<Movie id: 14>
  • Bugfix: users/recommendable objects will now be removed from Redis upon being destroyed


  • Major bugfix: similarity values were, incorrectly, being calculated as 0.0 for every user pair. Sorry!
  • The tables for all models are now all prepended with "recommendable_" to avoid possible collision. If upgrading from a previous version, please do the following:
$ rails g migration RenameRecommendableTables

And paste this into your new migration:

class RenameRecommendableTables < ActiveRecord::Migration
  def up
    rename_table :likes,         :recommendable_likes
    rename_table :dislikes,      :recommendable_dislikes
    rename_table :ignores,       :recommendable_ignores
    rename_table :stashed_items, :recommendable_stashed_items

  def down
    rename_table :recommendable_likes,         :likes
    rename_table :recommendable_dislikes,      :dislikes
    rename_table :recommendable_ignores,       :ignores
    rename_table :recommendable_stashed_items, :stashed_items


  • acts_as_recommendable is no longer needed in your models
  • Instead of declaring acts_as_recommended_to in your User class, please use recommends instead, passing in a list of your recommendable models as a list of symbols (e.g. recommends :movies, :books)
  • Your initializer should no longer declare the user class. This is no longer necessary and is deprecated.
  • Fix an issue that caused the unnecessary need for eager loading models in development
  • Removed aliases: liked_records, liked_records_for, disliked_records, and disliked_records_for
  • Renamed methods:
    • has_ignored? => ignored?
    • has_stashed? => stashed?
  • Code quality tweaks

0.1.3 (current version)

  • Improvements to speed of similarity calculations.
  • Added an instance method to items that act_as_recommendable, rated_by. This returns an array of users that like or dislike the item.


  • Fix an issue that could cause similarity values between users to be incorrect.
  • User#common_likes_with and User#common_dislikes_with now return the actual Model instances by default. Accordingly, these methods are no longer private.


  • Introduce the "stashed item" feature. This gives the ability for users to save an item in a list to try later, while at the same time removing it from their list of recommendations. This lets your users keep getting new recommendations without necessarily having to rate what they know they want to try.


  • Welcome to recommendable!