BLOG
Solving The N+1 Counters Problem
3d render in cartoon style of a computer surrounded by red diamonds

Juan Aparicio

Engineering Manager

Mar 19, 2024

8 min read

Ruby

Rails

Backend

The N+1 problem is a common performance issue that’s present in most rails apps I’ve worked in. It occurs when listing a collection of objects and, for each object, fetching associations from the database is required.

In this post, I’ll focus on a specific variation: displaying the count of associated objects.

The Problem

Consider having a User model and a Post model, where a User has many Posts. The goal is to display the number of posts each user has in a list of users.


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

# app/models/post.rb
class Post < ApplicationRecord
  belongs_to :user
end

# app/controllers/users_controller.rb
class UsersController < ApplicationController
  def index
    @users = User.all
  end
end
 
<!-- app/views/users/index.html.erb -->
<% @users.each do |user| %>
  <p><%= user.name %>, <%= user.posts.count %> posts</p>
<% end %>

Solving with counter caches

Rails’ counter_cache feature is a straightforward solution for storing a running count of posts for each user, eliminating the need to query the database for this information.


# app/models/post.rb
class Post < ApplicationRecord
  belongs_to :user, counter_cache: true
end

<!-- app/views/users/index.html.erb -->
<% @users.each do |user| %>
  <p><%= user.name %>, <%= user.posts_count %> posts</p>
<% end %>

How this works

  1. Add a posts_count column to the users table.
  2. Add the counter_cache: true option to the belongs_to association in the Post model.
  3. Rails will automatically update the posts_count column in the users table when posts are created, updated, or destroyed.

This method is often the first solution considered for addressing this issue.

The problem with counter caches

While counter caches are effective in many scenarios, they have limitations, such as:

  • Difficulty filtering the count of associated objects.
  • Challenges updating records in bulk without triggering callbacks. (remember, counter_cache uses a callback in the model to update the count)
  • The potential need for manual count updates due to the above issues.

Solving with includes and size

An alternative solution involves using includes to eager load associated objects and then counting them in memory with the size method.


class User < ApplicationRecord
  has_many :posts

  delegate :size, to: :posts, prefix: true
end

# app/controllers/users_controller.rb
class UsersController < ApplicationController
  def index
    @users = User.includes(:posts)
  end
end

<!-- app/views/users/index.html.erb -->
<% @users.each do |user| %>
  <p><%= user.name %>, <%= user.posts_size %> posts</p>
<% end %>

How this works

  1. Eager load the posts for each user.
  2. Use the delegate method to create a posts_size method that returns the count of posts for each user.
  3. Use the posts_size method in the view to display the count.

This solution is great because it doesn’t require any extra database columns, and it’s more flexible than counter_cache. But it has a downside: it’s not as performant as counter_cache because it needs to load all the associated objects into memory.

The problem with includes and size

The potential downside of this approach, is when there are many associated records, which leads to a large amount of data being loaded into memory.

Solving the filtered counter problem

For situations where counter_cache falls short, such as needing to filter counts of associated objects, the counter_culture gem offers a robust solution. However, embracing simplicity and minimizing dependencies is a valuable strategy. A Rails-centric approach can be just as effective, counter_culture solves this problem really well. It provides a counter_culture method that takes a block to define the conditions for the counter. It also provides a reset_counters method to update the counters in bulk without triggering callbacks.

After re-reading the rails ActiveRecord documentation, I figured, that solving this is also dead-simple:


class User < ApplicationRecord
  has_many :posts
  has_many :published_posts, -> { where.not(published_at: nil) }, class_name: 'Post'

  delegate :size, to: :posts, prefix: true
  delegate :size, to: :published_posts, prefix: true
end

# app/controllers/users_controller.rb
class UsersController < ApplicationController
  def index
    @users = User.all.includes(:posts, :published_posts)
  end
end

<!-- app/views/users/index.html.erb -->
<% @users.each do |user| %>
  <p><%= user.name %>, <%= user.posts_size %> posts, <%= user.published_posts_size %> published posts</p>
<% end %>

How this works

  1. Define a new association called published_posts that filters the posts that have a published_at date.
  2. Use the delegate method to create a published_posts_size method that returns the count of published posts for each user.
  3. Use the published_posts_count method in the view to display the count.

Why is this not triggering N+1 queries?

Well, for a long time, I had a misconception about size method in ActiveRecord::Relation. I thought that calling size on an association would always load the associated objects into memory and count them there everytime.

However, this is not the case. When you call size on an association, it will check if the association has already been loaded into memory (in this case, by the includes method), and if it has, it will count the elements in memory. If it hasn’t, it will execute a SELECT COUNT(*) ... query to the database.

So, in this case, the code is not triggering N+1 queries because the associated objects are eager-loaded into memory with includes(:posts, :published_posts).

Conclusion

The N+1 counters problem is a common issue when working with associations in rails. There are different solutions to the problem, and the best one depends on the specific requirements of your application.

  • counter_cache is the go-to solution for the problem, but it has some limitations. Also, if you need to filter the count of associated objects or update records in bulk without triggering callbacks, it may not be the best solution.
  • counter_culture gem is a great solution for the filtered counter problem, but it adds a dependency to your project, and also it will still be affected by the callback limitations.
  • includes + size is a great solution for the problem, but it’s not as performant as counter_cache (for reads!) because it needs to load all the associated objects into memory. If your collection is small, this is not a problem.

Maybe try to keep things simple in the beginning, use the tools that Rails provides, like counter_cache or includes + size. You can always refactor later if the need arises. Don’t overcomplicate things from the start.

Juan Aparicio

Engineering Manager

Mar 19, 2024

8 min read

Ruby

Rails

Backend