ActiveRecord: Comparing different ways to load associated data (`includes` vs `preload` vs `eager_load`)

active_record_comparing_different_ways_to_load_associated_data.rb

# Config
require 'active_record'
ActiveRecord::Base.establish_connection adapter: 'sqlite3', database: ':memory:'

# Migrations
ActiveRecord::Schema.define do
  self.verbose = false
  create_table(:users) { |t| t.string :name }
  create_table :posts do |t|
    t.string :name
    t.integer :user_id
  end
end

# Models
class User < ActiveRecord::Base
  has_many :posts
  def self.to_text
    all.map { |u| "#{u.name}: #{u.posts.pluck(:name).join(', ')}" }.join(', ')
  end
end

class Post < ActiveRecord::Base
  belongs_to :user
end

# Helper method so we can see what queries it ran
def self.capture_queries(&block)
  queries = []
  capturer = -> *, data { queries << data.values_at(:name, :sql).select(&:present?).join(": ") }
  ActiveSupport::Notifications.subscribed capturer, "sql.active_record", &block
  queries
end

# Seed data
User.create! name: 'Josh', posts: [Post.new(name: "josh's post")]
User.create! name: 'Becky', posts: [Post.new(name: "becky post")]


# ===== Comparing Different Ways To Access Associated Data =====

# Doing nothing leads to N+1
  capture_queries do
    User.all.to_text # => "Josh: josh's post, Becky: becky post"
  end
  # => ["User Load: SELECT \"users\".* FROM \"users\"",
  #     "Post Pluck: SELECT \"posts\".\"name\" FROM \"posts\" WHERE \"posts\".\"user_id\" = ?",
  #     "Post Pluck: SELECT \"posts\".\"name\" FROM \"posts\" WHERE \"posts\".\"user_id\" = ?"]

# Use `eager_load` to do a left outer join.
  # This may be what you were looking for, it's similar to your example (note that
  # this is not necessarily better, it saves a query, but could potentially lead to
  # many rows of almost entirely duplicated data being returned for each record)
  capture_queries do
    User.eager_load(:posts).to_text # => "Josh: josh's post, Becky: becky post"
  end
  # => ["SQL: SELECT \"users\".\"id\" AS t0_r0, \"users\".\"name\" AS t0_r1, \"posts\".\"id\" AS t1_r0, \"posts\".\"name\" AS t1_r1, \"posts\".\"user_id\" AS t1_r2 FROM \"users\" LEFT OUTER JOIN \"posts\" ON \"posts\".\"user_id\" = \"users\".\"id\""]
  # I'll reformat it so it's easier to see how it works:
  .first[5..].gsub(/(?=FROM|LEFT)/,"\n").gsub(/(?<=,)/, "\n      ")
  # => "SELECT \"users\".\"id\" AS t0_r0,\n" +
  #    "       \"users\".\"name\" AS t0_r1,\n" +
  #    "       \"posts\".\"id\" AS t1_r0,\n" +
  #    "       \"posts\".\"name\" AS t1_r1,\n" +
  #    "       \"posts\".\"user_id\" AS t1_r2 \n" +
  #    "FROM \"users\" \n" +
  #    "LEFT OUTER JOIN \"posts\" ON \"posts\".\"user_id\" = \"users\".\"id\""

# Use `preload` to batch the association queries into a single additional query
  capture_queries do
    User.preload(:posts).to_text # => "Josh: josh's post, Becky: becky post"
  end
  # => ["User Load: SELECT \"users\".* FROM \"users\"",
  #     "Post Load: SELECT \"posts\".* FROM \"posts\" WHERE \"posts\".\"user_id\" IN (?, ?)"]

# Use `includes` most of the time, it will choose between `eager_load` and `preload`, based on simple heuristics.
  # It chooses `preload` by default
  capture_queries do
    User.includes(:posts).to_text # => "Josh: josh's post, Becky: becky post"
  end
  # => ["User Load: SELECT \"users\".* FROM \"users\"",
  #     "Post Load: SELECT \"posts\".* FROM \"posts\" WHERE \"posts\".\"user_id\" IN (?, ?)"]

  # Here, we've implied we need the association in the query (via `references`), so it chooses `eager_load`
  capture_queries do
    User.includes(:posts).references(:posts).to_text # => "Josh: josh's post, Becky: becky post"
  end
  # => ["SQL: SELECT \"users\".\"id\" AS t0_r0, \"users\".\"name\" AS t0_r1, \"posts\".\"id\" AS t1_r0, \"posts\".\"name\" AS t1_r1, \"posts\".\"user_id\" AS t1_r2 FROM \"users\" LEFT OUTER JOIN \"posts\" ON \"posts\".\"user_id\" = \"users\".\"id\""]