ActiveRecordUnion

Use unions on ActiveRecord scopes without ugliness.

If you find yourself writing pluck(:id) and then feeding that into another query, you may be able to reduce the number of database requests by using a nested query or a UNION without writing crazy JOIN statements.

Quick usage examples:

current_user.posts.union(Post.published)
current_user.posts.union(Post.published).where(id: [6, 7])
current_user.posts.union("published_at < ?", Time.now)
user_1.posts.union(user_2.posts).union(Post.published)
user_1.posts.union_all(user_2.posts)

ActiveRecordUnion is tested against Rails 6.0, 6.1, 7.0, 7.1, 7.2 and 8.0.

If you are using Postgres, you might alternatively check out ActiveRecordExtended which includes support for unions as well as other goodies.

Installation

Add this line to your application's Gemfile:

gem 'active_record_union'

And then execute:

$ bundle

Or install it yourself as:

$ gem install active_record_union

Usage

ActiveRecordUnion adds union and union_all methods to ActiveRecord::Relation so we can easily gather together queries on mutiple scopes.

Consider some users with posts:

class User < ActiveRecord::Base
  has_many :posts
end

class Post < ActiveRecord::Base
  belongs_to :user

  scope :published, -> { where("published_at < ?", Time.now) }
end

With ActiveRecordUnion, we can do:

# the current user's (draft) posts and all published posts from anyone
current_user.posts.union(Post.published)

Which is equivalent to the following SQL:

SELECT "posts".* FROM (
  SELECT "posts".* FROM "posts"  WHERE "posts"."user_id" = 1
  UNION
  SELECT "posts".* FROM "posts"  WHERE (published_at < '2014-07-19 16:04:21.918366')
) "posts"

Because the union method returns another ActiveRecord::Relation, we can run further queries on the union.

current_user.posts.union(Post.published).where(id: [6, 7])

SELECT "posts".* FROM (
  SELECT "posts".* FROM "posts"  WHERE "posts"."user_id" = 1
  UNION
  SELECT "posts".* FROM "posts"  WHERE (published_at < '2014-07-19 16:06:04.460771')
) "posts"  WHERE "posts"."id" IN (6, 7)

The union method can also accept anything that where does.

current_user.posts.union("published_at < ?", Time.now)
# equivalent to...
current_user.posts.union(Post.where("published_at < ?", Time.now))

We can also chain union calls to UNION more than two scopes, though the UNIONs will be nested which may not be the prettiest SQL.

user_1.posts.union(user_2.posts).union(Post.published)
# equivalent to...
[user_1.posts, user_2.posts, Post.published].inject(:union)

SELECT "posts".* FROM (
  SELECT "posts".* FROM (
    SELECT "posts".* FROM "posts"  WHERE "posts"."user_id" = 1
    UNION
    SELECT "posts".* FROM "posts"  WHERE "posts"."user_id" = 2
  ) "posts"
  UNION
  SELECT "posts".* FROM "posts"  WHERE (published_at < '2014-07-19 16:12:45.882648')
) "posts"

UNION ALL

By default, UNION will remove any duplicates from the result set. If you don't care about duplicates or you know that the two queries you are combining will not have duplicates, you call use UNION ALL to tell the database to skip its deduplication step. In some cases this can give significant performance improvements.

user_1.posts.union_all(user_2.posts)

SELECT "posts".* FROM (
  SELECT "posts".* FROM "posts" WHERE "posts"."user_id" = 1
  UNION ALL
  SELECT "posts".* FROM "posts" WHERE "posts"."user_id" = 2
) "posts"

Caveats

There's a couple things to be aware of when using ActiveRecordUnion:

ActiveRecordUnion will raise an error if you try to UNION any relations that do any preloading/eager-loading. There's no sensible way to do the preloading in the subselects. If enough people complain, maybe, we can change ActiveRecordUnion to let the queries run anyway but without preloading any records.
There's no easy way to get SQLite to allow ORDER BY in the UNION subselects. If you get a syntax error, you can either write my_relation.reorder(nil).union(other.reorder(nil)) or switch to Postgres.

Another nifty way to reduce extra queries

ActiveRecord already supports turning scopes into nested queries in WHERE clauses. The nested relation defaults to selecting id by default.

For example, if a user has_and_belongs_to_many :favorited_posts, we can quickly find which of the current user's posts are liked by a certain other user.

current_user.posts.where(id: other_user.favorited_posts)

SELECT "posts".* FROM "posts"
  WHERE "posts"."user_id" = 1
  AND "posts"."id" IN (
    SELECT "posts"."id"
      FROM "posts" INNER JOIN "user_favorited_posts" ON "posts"."id" = "user_favorited_posts"."post_id"
      WHERE "user_favorited_posts"."user_id" = 2
  )

If we want to select something other than id, we use select to specify. The following is equivalent to the above, but the query is done against the join table.

current_user.posts.where(id: UserFavoritedPost.where(user_id: other_user.id).select(:post_id))

SELECT "posts".* FROM "posts"
  WHERE "posts"."user_id" = 1
  AND "posts"."id" IN (
    SELECT "user_favorited_posts"."post_id"
      FROM "user_favorited_posts"
      WHERE "user_favorited_posts"."user_id" = 2
  )

(The above example is illustrative only. It might be better with a JOIN.)

State of the Union in ActiveRecord

Why does this gem exist?

Right now in ActiveRecord, if we call scope.union we get an Arel::Nodes::Union object instead of an ActiveRecord::Relation.

We could call to_sql on the Arel object and then use find_by_sql, but that's not super clean. Also, on Rails 4.0 and 4.1 if the original scopes included an association then the to_sql may produce a query with values that need to be bound (represented by ?s in the SQL) and we have to provide those ourselves. (E.g. user.posts.to_sql produces SELECT "posts".* FROM "posts" WHERE "posts"."user_id" = ?.) Rails 4.2's to_sql replaces the bind values before showing the SQL string and thus can more readily be used with find_by_sql. (E.g. Rails 4.2 to_sql would say WHERE "posts"."user_id" = 1 instead of WHERE "posts"."user_id" = ?.)

While ActiveRecord may eventually have the ability to cleanly perform UNIONs, it's currently stalled. If you're interested, the relevant URLs as of July 2014 are:

rails/rails#939 and rails/arel#239 and https://github.com/yakaz/rails/commit/29b8ebd187e0888d5e71b2e1e4a12334860bc76c

This is a gem not a Rails pull request because the standard of code quality for a PR is a bit higher, and we'd have to wait for the PR to be merged and relased to use UNIONs. That said, the code here is fairly clean and it may end up in a PR sometime.

Changelog

1.3.0 - January 14, 2018

Ready for Rails 5.2! Updates provided by @glebm.

1.2.0 - June 26, 2016

Ready for Rails 5.0! Updates provided by @glebm.

1.1.1 - Mar 19, 2016

Fix broken polymorphic associations and joins due to improper handling of bind values. Fix by @efradelos, reported by @Machiaweliczny and @seandougall.
Quote table name aliases properly. Reported by @odedniv.

1.1.0 - Mar 29, 2015 - Add UNION ALL support, courtesy of @pic.

1.0.1 - Sept 2, 2014 - Allow ORDER BY in UNION subselects for databases that support it (not SQLite).

1.0.0 - July 24, 2014 - Initial release.

License

ActiveRecordUnion is dedicated to the public domain by its author, Brian Hempel. No rights are reserved. No restrictions are placed on the use of ActiveRecordUnion. That freedom also means, of course, that no warrenty of fitness is claimed; use ActiveRecordUnion at your own risk.

This public domain dedication follows the the CC0 1.0 at https://creativecommons.org/publicdomain/zero/1.0/

Contributing

Fork it ( https://github.com/brianhempel/active_record_union/fork )
Create your feature branch (git checkout -b my-new-feature)
Run the tests:
Install MySQL and PostgreSQL.
You need to be able to connect to a local MySQL and Postgres database as the default user, so the specs can create a test_active_record_union database. To set up the users this test expects, execute bin/create-db-users (or set the environment variables referenced in spec/support/databases.rb).
Run rake to test with all supported Rails versions. All needed dependencies will be installed via Bundler (gem install bundler if you happen not to have Bundler yet).
Run rake test_rails_8_0 or rake test_rails_7_2 etc. to test a specific Rails version.
There is also a bin/console command to load up a REPL for playing around
Commit your changes (git commit -am 'Add some feature')
Push to the branch (git push origin my-new-feature)
Create a new Pull Request

brianhempel/active_record_union