More at rubyonrails.org:

Active Record and PostgreSQL

This guide covers PostgreSQL specific usage of Active Record.

After reading this guide, you will know:

  • How to use PostgreSQL's datatypes.
  • How to use UUID primary keys.
  • How to include non-key columns in indexes.
  • How to use deferrable foreign keys.
  • How to use unique constraints.
  • How to implement exclusion constraints.
  • How to implement full text search with PostgreSQL.
  • How to back your Active Record models with database views.

In order to use the PostgreSQL adapter you need to have at least version 9.3 installed. Older versions are not supported.

To get started with PostgreSQL have a look at the configuring Rails guide. It describes how to properly set up Active Record for PostgreSQL.

1 Datatypes

PostgreSQL offers a number of specific datatypes. Following is a list of types, that are supported by the PostgreSQL adapter.

1.1 Bytea

# db/migrate/20140207133952_create_documents.rb
create_table :documents do |t|
  t.binary "payload"
end

# app/models/document.rb
class Document < ApplicationRecord
end

# Usage
data = File.read(Rails.root + "tmp/output.pdf")
Document.create payload: data

1.2 Array

# db/migrate/20140207133952_create_books.rb
create_table :books do |t|
  t.string "title"
  t.string "tags", array: true
  t.integer "ratings", array: true
end
add_index :books, :tags, using: "gin"
add_index :books, :ratings, using: "gin"

# app/models/book.rb
class Book < ApplicationRecord
end

# Usage
Book.create title: "Brave New World",
            tags: ["fantasy", "fiction"],
            ratings: [4, 5]

## Books for a single tag
Book.where("'fantasy' = ANY (tags)")

## Books for multiple tags
Book.where("tags @> ARRAY[?]::varchar[]", ["fantasy", "fiction"])

## Books with 3 or more ratings
Book.where("array_length(ratings, 1) >= 3")

1.3 Hstore

You need to enable the hstore extension to use hstore.

# db/migrate/20131009135255_create_profiles.rb
class CreateProfiles < ActiveRecord::Migration[8.1]
  enable_extension "hstore" unless extension_enabled?("hstore")
  create_table :profiles do |t|
    t.hstore "settings"
  end
end

# app/models/profile.rb
class Profile < ApplicationRecord
end

irb> Profile.create(settings: { "color" => "blue", "resolution" => "800x600" })

irb> profile = Profile.first
irb> profile.settings
=> {"color"=>"blue", "resolution"=>"800x600"}

irb> profile.settings = {"color" => "yellow", "resolution" => "1280x1024"}
irb> profile.save!

irb> Profile.where("settings->'color' = ?", "yellow")
=> #<ActiveRecord::Relation [#<Profile id: 1, settings: {"color"=>"yellow", "resolution"=>"1280x1024"}>]>

1.4 JSON and JSONB

# db/migrate/20131220144913_create_events.rb
# ... for json datatype:
create_table :events do |t|
  t.json "payload"
end
# ... or for jsonb datatype:
create_table :events do |t|
  t.jsonb "payload"
end

# app/models/event.rb
class Event < ApplicationRecord
end

irb> Event.create(payload: { kind: "user_renamed", change: ["jack", "john"]})

irb> event = Event.first
irb> event.payload
=> {"kind"=>"user_renamed", "change"=>["jack", "john"]}

## Query based on JSON document
# The -> operator returns the original JSON type (which might be an object), whereas ->> returns text
irb> Event.where("payload->>'kind' = ?", "user_renamed")

1.5 Range Types

This type is mapped to Ruby Range objects.

# db/migrate/20130923065404_create_events.rb
create_table :events do |t|
  t.daterange "duration"
end

# app/models/event.rb
class Event < ApplicationRecord
end

irb> Event.create(duration: Date.new(2014, 2, 11)..Date.new(2014, 2, 12))

irb> event = Event.first
irb> event.duration
=> Tue, 11 Feb 2014...Thu, 13 Feb 2014

## All Events on a given date
irb> Event.where("duration @> ?::date", Date.new(2014, 2, 12))

## Working with range bounds
irb> event = Event.select("lower(duration) AS starts_at").select("upper(duration) AS ends_at").first

irb> event.starts_at
=> Tue, 11 Feb 2014
irb> event.ends_at
=> Thu, 13 Feb 2014

1.6 Composite Types

Currently there is no special support for composite types. They are mapped to normal text columns:

CREATE TYPE full_address AS
(
  city VARCHAR(90),
  street VARCHAR(90)
);

# db/migrate/20140207133952_create_contacts.rb
execute <<-SQL
  CREATE TYPE full_address AS
  (
    city VARCHAR(90),
    street VARCHAR(90)
  );
SQL
create_table :contacts do |t|
  t.column :address, :full_address
end

# app/models/contact.rb
class Contact < ApplicationRecord
end

irb> Contact.create address: "(Paris,Champs-Élysées)"
irb> contact = Contact.first
irb> contact.address
=> "(Paris,Champs-Élysées)"
irb> contact.address = "(Paris,Rue Basse)"
irb> contact.save!

1.7 Enumerated Types

The type can be mapped as a normal text column, or to an ActiveRecord::Enum.

# db/migrate/20131220144913_create_articles.rb
def change
  create_enum :article_status, ["draft", "published", "archived"]

  create_table :articles do |t|
    t.enum :status, enum_type: :article_status, default: "draft", null: false
  end
end

You can also create an enum type and add an enum column to an existing table:

# db/migrate/20230113024409_add_status_to_articles.rb
def change
  create_enum :article_status, ["draft", "published", "archived"]

  add_column :articles, :status, :enum, enum_type: :article_status, default: "draft", null: false
end

The above migrations are both reversible, but you can define separate #up and #down methods if required. Make sure you remove any columns or tables that depend on the enum type before dropping it:

def down
  drop_table :articles

  # OR: remove_column :articles, :status
  drop_enum :article_status
end

Declaring an enum attribute in the model adds helper methods and prevents invalid values from being assigned to instances of the class:

# app/models/article.rb
class Article < ApplicationRecord
  enum :status, {
    draft: "draft", published: "published", archived: "archived"
  }, prefix: true
end

irb> article = Article.create
irb> article.status
=> "draft" # default status from PostgreSQL, as defined in migration above

irb> article.status_published!
irb> article.status
=> "published"

irb> article.status_archived?
=> false

irb> article.status = "deleted"
ArgumentError: 'deleted' is not a valid status

To rename the enum you can use rename_enum along with updating any model usage:

# db/migrate/20150718144917_rename_article_status.rb
def change
  rename_enum :article_status, :article_state
end

To add a new value you can use add_enum_value:

# db/migrate/20150720144913_add_new_state_to_articles.rb
def up
  add_enum_value :article_state, "archived" # will be at the end after published
  add_enum_value :article_state, "in review", before: "published"
  add_enum_value :article_state, "approved", after: "in review"
  add_enum_value :article_state, "rejected", if_not_exists: true # won't raise an error if the value already exists
end

Enum values can't be dropped, which also means add_enum_value is irreversible. You can read why here.

To rename a value you can use rename_enum_value:

# db/migrate/20150722144915_rename_article_state.rb
def change
  rename_enum_value :article_state, from: "archived", to: "deleted"
end

Hint: to show all the values of the all enums you have, you can call this query in bin/rails db or psql console:

SELECT n.nspname AS enum_schema,
       t.typname AS enum_name,
       e.enumlabel AS enum_value
  FROM pg_type t
      JOIN pg_enum e ON t.oid = e.enumtypid
      JOIN pg_catalog.pg_namespace n ON n.oid = t.typnamespace

1.8 UUID

If you're using PostgreSQL earlier than version 13.0 you may need to enable special extensions to use UUIDs. Enable the pgcrypto extension (PostgreSQL >= 9.4) or uuid-ossp extension (for even earlier releases).

# db/migrate/20131220144913_create_revisions.rb
create_table :revisions do |t|
  t.uuid :identifier
end

# app/models/revision.rb
class Revision < ApplicationRecord
end

irb> Revision.create identifier: "A0EEBC99-9C0B-4EF8-BB6D-6BB9BD380A11"

irb> revision = Revision.first
irb> revision.identifier
=> "a0eebc99-9c0b-4ef8-bb6d-6bb9bd380a11"

You can use uuid type to define references in migrations:

# db/migrate/20150418012400_create_blog.rb
enable_extension "pgcrypto" unless extension_enabled?("pgcrypto")
create_table :posts, id: :uuid

create_table :comments, id: :uuid do |t|
  # t.belongs_to :post, type: :uuid
  t.references :post, type: :uuid
end

# app/models/post.rb
class Post < ApplicationRecord
  has_many :comments
end

# app/models/comment.rb
class Comment < ApplicationRecord
  belongs_to :post
end

See this section for more details on using UUIDs as primary key.

1.9 Bit String Types

# db/migrate/20131220144913_create_users.rb
create_table :users, force: true do |t|
  t.column :settings, "bit(8)"
end

# app/models/user.rb
class User < ApplicationRecord
end

irb> User.create settings: "01010011"
irb> user = User.first
irb> user.settings
=> "01010011"
irb> user.settings = "0xAF"
irb> user.settings
=> "10101111"
irb> user.save!

1.10 Network Address Types

The types inet and cidr are mapped to Ruby IPAddr objects. The macaddr type is mapped to normal text.

# db/migrate/20140508144913_create_devices.rb
create_table(:devices, force: true) do |t|
  t.inet "ip"
  t.cidr "network"
  t.macaddr "address"
end

# app/models/device.rb
class Device < ApplicationRecord
end

irb> macbook = Device.create(ip: "192.168.1.12", network: "192.168.2.0/24", address: "32:01:16:6d:05:ef")

irb> macbook.ip
=> #<IPAddr: IPv4:192.168.1.12/255.255.255.255>

irb> macbook.network
=> #<IPAddr: IPv4:192.168.2.0/255.255.255.0>

irb> macbook.address
=> "32:01:16:6d:05:ef"

1.11 Geometric Types

All geometric types, with the exception of points are mapped to normal text. A point is cast to an array containing x and y coordinates.

1.12 Interval

This type is mapped to ActiveSupport::Duration objects.

# db/migrate/20200120000000_create_events.rb
create_table :events do |t|
  t.interval "duration"
end

# app/models/event.rb
class Event < ApplicationRecord
end

irb> Event.create(duration: 2.days)

irb> event = Event.first
irb> event.duration
=> 2 days

2 UUID Primary Keys

You need to enable the pgcrypto (only PostgreSQL >= 9.4) or uuid-ossp extension to generate random UUIDs.

# db/migrate/20131220144913_create_devices.rb
enable_extension "pgcrypto" unless extension_enabled?("pgcrypto")
create_table :devices, id: :uuid do |t|
  t.string :kind
end

# app/models/device.rb
class Device < ApplicationRecord
end

irb> device = Device.create
irb> device.id
=> "814865cd-5a1d-4771-9306-4268f188fe9e"

gen_random_uuid() (from pgcrypto) is assumed if no :default option was passed to create_table.

To use the Rails model generator for a table using UUID as the primary key, pass --primary-key-type=uuid to the model generator.

For example:

$ rails generate model Device --primary-key-type=uuid kind:string

When building a model with a foreign key that will reference this UUID, treat uuid as the native field type, for example:

$ rails generate model Case device_id:uuid

3 Indexing

PostgreSQL includes a variety of index options. The following options are supported by the PostgreSQL adapter in addition to the common index options

3.1 Include

When creating a new index, non-key columns can be included with the :include option. These keys are not used in index scans for searching, but can be read during an index only scan without having to visit the associated table.

# db/migrate/20131220144913_add_index_users_on_email_include_id.rb

add_index :users, :email, include: :id

Multiple columns are supported:

# db/migrate/20131220144913_add_index_users_on_email_include_id_and_created_at.rb

add_index :users, :email, include: [:id, :created_at]

4 Generated Columns

Generated columns are supported since version 12.0 of PostgreSQL.

# db/migrate/20131220144913_create_users.rb
create_table :users do |t|
  t.string :name
  t.virtual :name_upcased, type: :string, as: "upper(name)", stored: true
end

# app/models/user.rb
class User < ApplicationRecord
end

# Usage
user = User.create(name: "John")
User.last.name_upcased # => "JOHN"

5 Deferrable Foreign Keys

By default, table constraints in PostgreSQL are checked immediately after each statement. It intentionally does not allow creating records where the referenced record is not yet in the referenced table. It is possible to run this integrity check later on when the transaction is committed by adding DEFERRABLE to the foreign key definition though. To defer all checks by default it can be set to DEFERRABLE INITIALLY DEFERRED. Rails exposes this PostgreSQL feature by adding the :deferrable key to the foreign_key options in the add_reference and add_foreign_key methods.

One example of this is creating circular dependencies in a transaction even if you have created foreign keys:

add_reference :person, :alias, foreign_key: { deferrable: :deferred }
add_reference :alias, :person, foreign_key: { deferrable: :deferred }

If the reference was created with the foreign_key: true option, the following transaction would fail when executing the first INSERT statement. It does not fail when the deferrable: :deferred option is set though.

ActiveRecord::Base.lease_connection.transaction do
  person = Person.create(id: SecureRandom.uuid, alias_id: SecureRandom.uuid, name: "John Doe")
  Alias.create(id: person.alias_id, person_id: person.id, name: "jaydee")
end

When the :deferrable option is set to :immediate, let the foreign keys keep the default behavior of checking the constraint immediately, but allow manually deferring the checks using set_constraints within a transaction. This will cause the foreign keys to be checked when the transaction is committed:

ActiveRecord::Base.lease_connection.transaction do
  ActiveRecord::Base.lease_connection.set_constraints(:deferred)
  person = Person.create(alias_id: SecureRandom.uuid, name: "John Doe")
  Alias.create(id: person.alias_id, person_id: person.id, name: "jaydee")
end

By default :deferrable is false and the constraint is always checked immediately.

6 Unique Constraint

# db/migrate/20230422225213_create_items.rb
create_table :items do |t|
  t.integer :position, null: false
  t.unique_constraint [:position], deferrable: :immediate
end

If you want to change an existing unique index to deferrable, you can use :using_index to create deferrable unique constraints.

add_unique_constraint :items, deferrable: :deferred, using_index: "index_items_on_position"

Like foreign keys, unique constraints can be deferred by setting :deferrable to either :immediate or :deferred. By default, :deferrable is false and the constraint is always checked immediately.

7 Exclusion Constraints

# db/migrate/20131220144913_create_products.rb
create_table :products do |t|
  t.integer :price, null: false
  t.daterange :availability_range, null: false

  t.exclusion_constraint "price WITH =, availability_range WITH &&", using: :gist, name: "price_check"
end

Like foreign keys, exclusion constraints can be deferred by setting :deferrable to either :immediate or :deferred. By default, :deferrable is false and the constraint is always checked immediately.

8 Full Text Search

# db/migrate/20131220144913_create_documents.rb
create_table :documents do |t|
  t.string :title
  t.string :body
end

add_index :documents, "to_tsvector('english', title || ' ' || body)", using: :gin, name: "documents_idx"

# app/models/document.rb
class Document < ApplicationRecord
end

# Usage
Document.create(title: "Cats and Dogs", body: "are nice!")

## all documents matching 'cat & dog'
Document.where("to_tsvector('english', title || ' ' || body) @@ to_tsquery(?)",
                 "cat & dog")

Optionally, you can store the vector as automatically generated column (from PostgreSQL 12.0):

# db/migrate/20131220144913_create_documents.rb
create_table :documents do |t|
  t.string :title
  t.string :body

  t.virtual :textsearchable_index_col,
            type: :tsvector, as: "to_tsvector('english', title || ' ' || body)", stored: true
end

add_index :documents, :textsearchable_index_col, using: :gin, name: "documents_idx"

# Usage
Document.create(title: "Cats and Dogs", body: "are nice!")

## all documents matching 'cat & dog'
Document.where("textsearchable_index_col @@ to_tsquery(?)", "cat & dog")

9 Database Views

Imagine you need to work with a legacy database containing the following table:

rails_pg_guide=# \d "TBL_ART"
                                        Table "public.TBL_ART"
   Column   |            Type             |                         Modifiers
------------+-----------------------------+------------------------------------------------------------
 INT_ID     | integer                     | not null default nextval('"TBL_ART_INT_ID_seq"'::regclass)
 STR_TITLE  | character varying           |
 STR_STAT   | character varying           | default 'draft'::character varying
 DT_PUBL_AT | timestamp without time zone |
 BL_ARCH    | boolean                     | default false
Indexes:
    "TBL_ART_pkey" PRIMARY KEY, btree ("INT_ID")

This table does not follow the Rails conventions at all. Because simple PostgreSQL views are updateable by default, we can wrap it as follows:

# db/migrate/20131220144913_create_articles_view.rb
execute <<-SQL
CREATE VIEW articles AS
  SELECT "INT_ID" AS id,
         "STR_TITLE" AS title,
         "STR_STAT" AS status,
         "DT_PUBL_AT" AS published_at,
         "BL_ARCH" AS archived
  FROM "TBL_ART"
  WHERE "BL_ARCH" = 'f'
SQL

# app/models/article.rb
class Article < ApplicationRecord
  self.primary_key = "id"
  def archive!
    update_attribute :archived, true
  end
end

irb> first = Article.create! title: "Winter is coming", status: "published", published_at: 1.year.ago
irb> second = Article.create! title: "Brace yourself", status: "draft", published_at: 1.month.ago

irb> Article.count
=> 2
irb> first.archive!
irb> Article.count
=> 1

This application only cares about non-archived Articles. A view also allows for conditions so we can exclude the archived Articles directly.

10 Structure Dumps

If your config.active_record.schema_format is :sql, Rails will call pg_dump to generate a structure dump.

You can use ActiveRecord::Tasks::DatabaseTasks.structure_dump_flags to configure pg_dump. For example, to exclude comments from your structure dump, add this to an initializer:

ActiveRecord::Tasks::DatabaseTasks.structure_dump_flags = ["--no-comments"]

11 Explain

Along with the standard explain options, the PostgreSQL adapter supports buffers.

Company.where(id: owning_companies_ids).explain(:analyze, :buffers)
#=> EXPLAIN (ANALYZE, BUFFERS) SELECT "companies".* FROM "companies"
# ...
# Seq Scan on companies  (cost=0.00..2.21 rows=3 width=64)
# ...

See their documentation for more details.

Back to top