active-record-associations

Active Record Associations Expert

Define correct, performant, and maintainable associations between Rails models.

Philosophy

Core Principles:

1. The FK lives on the belongs_to side — always. If you're confused about where it goes, ask "which table has the _id column?" That model gets belongs_to.
2. Always set dependent: — orphaned records are bugs waiting to happen.
3. Prefer has_many :through over has_and_belongs_to_many — every time. HABTM is a dead end you'll regret.
4. Bi-directional by default — declare both sides. Use inverse_of when Rails can't infer it.
5. Eager load aggressively — N+1 queries are the #1 performance killer in Rails apps.
6. Database constraints back up model associations — foreign keys, indexes, and unique constraints belong in migrations, not just models.

When To Use This Skill

• Defining relationships between models (one-to-one, one-to-many, many-to-many)
• Creating migrations with foreign keys and join tables
• Fixing N+1 query problems
• Setting up polymorphic or self-referential associations
• Choosing between has_many :through and HABTM
• Adding counter caches
• Debugging association-related errors

Instructions

Step 1: Identify the Relationship Type

Before writing any code, determine the relationship:

Relationship	Parent Model	Child Model	FK Location
One-to-many	has_many :children	belongs_to :parent	Child table
One-to-one	has_one :child	belongs_to :parent	Child table
Many-to-many	has_many :others, through: :join	has_many :others, through: :join	Join table
Polymorphic	has_many :things, as: :thingable	belongs_to :thingable, polymorphic: true	Child table (_id + _type)
Self-referential	has_many :children, class_name: "Self"	belongs_to :parent, class_name: "Self"	Same table

The golden rule: belongs_to goes on the model whose table has the foreign key column.

Step 2: Check Existing Patterns

Look at the existing codebase first — existing patterns tell you what conventions to follow:

# Find existing associations
rg "has_many\|has_one\|belongs_to\|has_and_belongs_to_many" app/models/

# Check existing migrations for FK patterns
rg "add_reference\|t.belongs_to\|t.references\|foreign_key" db/migrate/

# Look at schema
cat db/schema.rb | grep -A5 "create_table"

Match existing project conventions for naming, dependent options, and index patterns.

Step 3: Write the Association

One-to-Many (most common)

# app/models/author.rb
class Author < ApplicationRecord
  has_many :books, dependent: :destroy
end

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

Migration:

class CreateBooks < ActiveRecord::Migration[8.1]
  def change
    create_table :books do |t|
      t.references :author, null: false, foreign_key: true
      t.string :title
      t.timestamps
    end
  end
end

One-to-One

class User < ApplicationRecord
  has_one :profile, dependent: :destroy
end

class Profile < ApplicationRecord
  belongs_to :user
end

Migration — add a unique index:

create_table :profiles do |t|
  t.references :user, null: false, foreign_key: true, index: { unique: true }
  t.timestamps
end

Many-to-Many (always use has_many :through)

class Doctor < ApplicationRecord
  has_many :appointments, dependent: :destroy
  has_many :patients, through: :appointments
end

class Appointment < ApplicationRecord
  belongs_to :doctor
  belongs_to :patient
end

class Patient < ApplicationRecord
  has_many :appointments, dependent: :destroy
  has_many :doctors, through: :appointments
end

Migration — the join model gets its own table with a primary key:

create_table :appointments do |t|
  t.references :doctor, null: false, foreign_key: true
  t.references :patient, null: false, foreign_key: true
  t.datetime :scheduled_at
  t.timestamps
end

Step 4: Set dependent: on Every Association

Every has_many and has_one needs a dependent: option — without one, deleting a parent leaves orphaned child records with dangling foreign keys.

Option	When to Use
dependent: :destroy	Default choice. Runs callbacks on children. Use when children have their own associations or callbacks.
dependent: :delete_all	Performance optimization. Skips callbacks. Use for leaf nodes with no further associations.
dependent: :nullify	Keep the child records but remove the association. FK must allow NULL.
dependent: :restrict_with_error	Prevent deletion if children exist. Business rule enforcement.
dependent: :destroy_async	Large datasets. Enqueues background job. Requires Active Job. Don't use with DB-level FK constraints.

# Missing dependent — orphaned records when author is destroyed
has_many :books

# Fixed — children cleaned up properly
has_many :books, dependent: :destroy

On belongs_to: Don't set dependent: on belongs_to — it causes confusion and can lead to circular destruction.

Step 5: Handle Eager Loading (Kill N+1)

The Problem:

# N+1 — fires 1 query for authors + N queries for books
Author.all.each { |a| puts a.books.count }

The Fix — use includes:

# 2 queries total — always
Author.includes(:books).each { |a| puts a.books.count }

When to use which:

Method	SQL Strategy	Use When
includes	Smart default. Uses preload or eager_load depending on whether you reference the association in conditions.	Most cases. Start here.
preload	Separate queries (SELECT * FROM books WHERE author_id IN (...))	You want separate queries. Can't use for filtering.
eager_load	Single LEFT JOIN	You need to filter/order by associated columns in WHERE or ORDER BY.
strict_loading	Raises if lazy-loaded	Prevent N+1 at the model level during development.

# Filter by association column — must use eager_load (or includes handles it)
Author.includes(:books).where(books: { published: true })

# Nested eager loading
Author.includes(books: :reviews)

# Strict loading on a model (development safety net)
class Author < ApplicationRecord
  self.strict_loading_by_default = true  # Rails 7+
  has_many :books, dependent: :destroy
end

# Strict loading on a query
Author.strict_loading.all

Step 6: Use inverse_of When Needed

Rails auto-detects inverse associations in simple cases. You need inverse_of when:

• You use :foreign_key or :class_name
• You have scoped associations
• You use :through associations

# Rails CAN'T auto-detect this — specify inverse_of
class Author < ApplicationRecord
  has_many :books, inverse_of: :writer
end

class Book < ApplicationRecord
  belongs_to :writer, class_name: "Author", foreign_key: "author_id"
end

Why it matters: Without correct inverse_of, you get:

• Extra queries (N+1 for already-loaded data)
• Inconsistent in-memory objects
• Failed presence validations on new records

Step 7: Write the Migration

Every association needs a corresponding migration. Associations don't create database columns.

Always include:

1. Foreign key column — use t.references or add_reference
2. Database-level foreign key constraint — foreign_key: true
3. Index — automatic with t.references, but verify
4. NOT NULL constraint — unless the association is optional: true

# New table
create_table :books do |t|
  t.references :author, null: false, foreign_key: true
  t.timestamps
end

# Adding to existing table
add_reference :books, :author, null: false, foreign_key: true

For polymorphic:

t.references :commentable, polymorphic: true, null: false
# Creates: commentable_id (bigint) + commentable_type (string) + composite index

Common Mistakes (and Fixes)

1. FK on the Wrong Table

# Wrong — puts FK on authors table (authors don't have book_id!)
class Author < ApplicationRecord
  belongs_to :book  # ← backwards
end

# Correct — books table has author_id
class Book < ApplicationRecord
  belongs_to :author
end

Rule: The table with the _id column gets belongs_to.

2. Missing dependent: on has_many

# Destroys author, leaves orphaned books with dangling author_id
class Author < ApplicationRecord
  has_many :books  # ← missing dependent
end

# Fixed
class Author < ApplicationRecord
  has_many :books, dependent: :destroy
end

3. Using HABTM Instead of has_many :through

# Avoid — can't add attributes, callbacks, or validations to the join
class Student < ApplicationRecord
  has_and_belongs_to_many :courses
end

# Better — flexible, extensible, future-proof
class Student < ApplicationRecord
  has_many :enrollments, dependent: :destroy
  has_many :courses, through: :enrollments
end

class Enrollment < ApplicationRecord
  belongs_to :student
  belongs_to :course
  # Now you can add: grade, enrolled_at, status, etc.
end

4. Polymorphic When You Shouldn't

Use polymorphic when:

• Multiple unrelated models need the same child (comments, attachments, tags)
• You're building a framework/engine feature

Don't use polymorphic when:

• Only 2-3 specific models — use separate FKs instead
• You need database-level referential integrity (polymorphic FKs can't have DB constraints)
• You need to join across the polymorphic boundary efficiently

# Polymorphic is overkill here — just use two belongs_to
class Comment < ApplicationRecord
  belongs_to :post     # Better than polymorphic if only commenting on posts and articles
  belongs_to :article
end

5. N+1 in Views

# Controller — forgot to eager load
def index
  @posts = Post.all  # N+1 when view calls post.author.name
end

# Fixed
def index
  @posts = Post.includes(:author, :comments)
end

6. optional: true Without Reason

# belongs_to validates presence by default (Rails 5+)
# Don't add optional: true unless the FK is genuinely nullable
belongs_to :author                    # Validates author exists ✓
belongs_to :author, optional: true    # Only if author_id can be NULL

Quick Reference

Association Cheat Sheet

# One-to-many
has_many :posts, dependent: :destroy
belongs_to :user

# One-to-one
has_one :profile, dependent: :destroy
belongs_to :user

# Many-to-many (through)
has_many :taggings, dependent: :destroy
has_many :tags, through: :taggings

# Polymorphic
has_many :comments, as: :commentable, dependent: :destroy
belongs_to :commentable, polymorphic: true

# Self-referential
has_many :subordinates, class_name: "Employee", foreign_key: "manager_id", dependent: :nullify
belongs_to :manager, class_name: "Employee", optional: true

# Counter cache (declare on belongs_to side, column on has_many side)
belongs_to :author, counter_cache: true
# Requires: add_column :authors, :books_count, :integer, default: 0, null: false

# Scoped association
has_many :published_books, -> { where(published: true) }, class_name: "Book", dependent: :destroy

# With ordering
has_many :books, -> { order(created_at: :desc) }, dependent: :destroy

Migration Patterns

# Standard FK
t.references :author, null: false, foreign_key: true

# Polymorphic FK
t.references :commentable, polymorphic: true, null: false

# Self-referential FK
t.references :manager, foreign_key: { to_table: :employees }

# Custom FK name
t.bigint :creator_id
add_foreign_key :posts, :users, column: :creator_id

# Join table for has_many :through
create_table :enrollments do |t|
  t.references :student, null: false, foreign_key: true
  t.references :course, null: false, foreign_key: true
  t.timestamps
end
add_index :enrollments, [:student_id, :course_id], unique: true

Eager Loading Decision Tree

Need to filter/order by associated columns?
  YES → includes (auto-detects) or eager_load (explicit LEFT JOIN)
  NO  → includes (auto-detects) or preload (explicit separate queries)

Need nested associations?
  → Author.includes(books: [:reviews, :publisher])

Want to prevent N+1 entirely?
  → strict_loading (model or query level)

belongs_to Options Quick Reference

Option	Default	Purpose
optional: true	false	Allow NULL foreign key
counter_cache: true	false	Maintain count column on parent
touch: true	false	Update parent's updated_at on save
polymorphic: true	false	Polymorphic association
class_name:	Inferred	Specify non-standard class
foreign_key:	"#{name}_id"	Specify non-standard FK column
inverse_of:	Auto-detected	Specify inverse association name

has_many / has_one Options Quick Reference

Option	Default	Purpose
dependent:	None	Always set this. See Step 4.
class_name:	Inferred	Non-standard class name
foreign_key:	"#{model}_id"	Non-standard FK column
inverse_of:	Auto-detected	Specify inverse association
through:	None	Through association
source:	Inferred	Source association name for :through
as:	None	Polymorphic interface name
counter_cache:	None	(has_many only) read counter cache column

Debugging Associations

# Check what associations a model has
Author.reflect_on_all_associations.map { |a| [a.macro, a.name] }

# Check a specific association
Author.reflect_on_association(:books)

# See the SQL an association generates
Author.first.books.to_sql

# Check if inverse is set
Author.reflect_on_association(:books).inverse_of

# Find orphaned records
Book.left_joins(:author).where(authors: { id: nil })

# Reset counter cache
Author.reset_counters(author_id, :books)

Anti-Patterns to Avoid

1. No dependent: on has_many/has_one — orphaned records cause subtle data bugs
2. HABTM — use has_many :through instead; HABTM can't hold attributes, validations, or callbacks on the join
3. Missing database constraints — foreign_key: true in migrations
4. Lazy loading in loops — use includes / preload / eager_load
5. Polymorphic for 2 models — just use separate foreign keys
6. optional: true by default — only when the FK is genuinely nullable
7. Missing inverse_of — when using custom :foreign_key or :class_name
8. No unique index on has_one FK — allows duplicate records without it
9. Forgetting the migration — associations don't create columns
10. dependent: :destroy on huge collections — use :delete_all or :destroy_async

For detailed patterns, options reference, and advanced examples, see the references/ directory:

• references/association-types.md — Deep dive into belongs_to, has_one, has_many, has_many :through, HABTM
• references/polymorphic.md — Polymorphic associations, STI, and delegated types
• references/self-referential.md — Tree structures, manager-employee, social follows
• references/eager-loading.md — includes/preload/eager_load strategies, inverse_of
• references/counter-caches.md — Counter caches, scoped associations, callbacks, extensions
• references/testing.md — Testing patterns, migration examples, performance tips, troubleshooting