The Better Primary Key: A Guide to ULIDs for Rails Developers

In a previous article, I talked about Snowflake IDs. They are great, but they require a bit of configuration because you need to manage "Worker IDs."

If you want something simpler that gives you the same benefits - secure URLs and fast database sorting—you should look at ULIDs (Universally Unique Lexicographically Sortable Identifiers).

Why not just use UUIDs?

Most developers switch from auto-incrementing integers to UUIDs because they want to hide their business volume. If your order ID is order/550e8400-e29b..., no one knows if you have 1 customer or 1 million.

But UUIDs have a major problem: they are completely random.
When you insert millions of random UUIDs into a database, the "B-Tree" index gets fragmented and slow. Your database has to jump all over the hard drive to find where to put the new data.

ULIDs fix this.
A ULID is 128 bits (just like a UUID), but the first part of the ID is a timestamp. This means ULIDs are chronological. They are unique like a UUID, but they sort like an Integer.

Here is how to implement ULIDs in your Rails 8 app in 4 easy steps.

STEP 1: Install the Gem

We will use the ulid gem to generate the strings. Add this to your Gemfile:

gem "ulid"

Run bundle install in your terminal.

STEP 2: The ULID Concern

We want to make it very easy to add ULIDs to any model. The best way to do this is with a Concern. This code will ensure that every time we create a new record, a ULID is generated and assigned to the ID.

Create a new file at app/models/concerns/has_ulid.rb:

# app/models/concerns/has_ulid.rb
module HasUlid
  extend ActiveSupport::Concern

  included do
    # Before we save to the DB, generate the ULID
    before_create :set_ulid
  end

  private

  def set_ulid
    # ULID.generate creates a string like: 01ARZ3NDEKTSV4RRFFQ69G5FAV
    self.id ||= ULID.generate
  end
end

STEP 3: The Migration

When you create a new model, you need to tell Rails that the id is a string, and you must disable the default auto-increment logic.

rails g model Product name:string

Open the migration file and modify it like this:

# db/migrate/XXXXXXXXXXXXXX_create_products.rb
class CreateProducts < ActiveRecord::Migration[8.0]
  def change
    # id: false stops the automatic integer ID
    create_table :products, id: false do |t|
      # We use string for ULID primary key
      t.string :id, primary_key: true
      t.string :name

      t.timestamps
    end
  end
end

STEP 4: Update the Model

Now, just include the concern we wrote in Step 2.

# app/models/product.rb
class Product < ApplicationRecord
  include HasUlid
end

STEP 5: See it in action

Open your Rails console (bin/rails c) and create a few products:

Product.create(name: "Laptop")
Product.create(name: "Monitor")
Product.create(name: "Keyboard")

# Check the IDs
Product.pluck(:id)
# => ["01HQV...", "01HQV...", "01HQV..."]

If you look closely, the IDs all start with the same characters because they were created in the same minute. Because they are sortable, you can still run Product.order(:id) and they will be in the correct chronological order!

Why I like ULIDs for Rails

1. Better Performance: Because the IDs are sortable, PostgreSQL (or SQLite) can append them to the end of the index. This is much faster for "Write-Heavy" apps than random UUIDs.
2. Readability: ULIDs use a special alphabet (Crockford's Base32) that excludes confusing letters like "I", "L", and "O". This makes them easier to read if a user has to type one in.
3. No Setup: Unlike Snowflake IDs, you don't need to configure server IDs or worker nodes. You just install the gem and go.

That's pretty much it. It’s a small architectural change that makes your app feel much more professional and scalable.