Coding Rails with Data Integrity, Part 2

Jay Hayes

4 min read

Aug 5, 2013

Coding Rails with Data Integrity, Part 2

Last time, we discussed how database null constraints and default values can increase confidence in your app’s data.

This time, I want to take a look at uniqueness constraints. Rails provides uniqueness validations, but as we learned in the previous post, validations aren’t necessarily the right tool to ensure data integrity. And Rails’ uniqueness validation in particular is not without its perils.

Uniqueness by Example

Eventually you’ll realize that correctness demands certain data be unique. One of the most common examples for a uniqueness constraint is for users. If you allowed multiple users to have the same email address or username, users would be indistinguishable. Such information serves as identification for individuals.This example has been driven into the ground, so let’s talk about something more interesting.

Team Membership

Consider an app that has teams and users. Say we want users to be a member of any number of teams. This is an example of a many-to-many relationship and will require a join table. Let’s call this table memberships:

    class CreateMemberships < ActiveRecord::Migration
      def change
        create_table :memberships do |t|
          t.belongs_to :team, null: false
          t.belongs_to :user, null: false

We’ve intentionally left out the uniqueness constraint in this example to illustrate its pitfalls. Also, don’t miss the null constraints! These join records don’t make much sense without pointing in both directions.

Let’s we have the respective models set up as needed for such a relationship. Time to take them for a spin:

    t = Team.create!    # => #<Team id: 1>
    t.users             # => []
    u = User.create!    # => #<User id: 1>
    t.users << u        # => [#<User id: 1>], nice!
    t.users << u        # => [#<User id: 1>, #<User id: 1>], WHOOPS!
    Membership.all      # => [#<Membership id: 1, team_id: 1, user_id: 1>, #<Membership id: 2, team_id: 1, user_id: 1>]

As you can see, without the uniqueness constraint we allow users to be a member of a team more than once. That doesn’t make much sense (I recently experienced this first hand). Now we’ll lock this table down by adding the constraint. In SQL, uniqueness is enfored by creating a “unique” index. This index may span multiple columns to ensure uniqueness with respect to multiple attributes:

    class CreateMemberships < ActiveRecord::Migration
      def change
        create_table :memberships do |t|
          t.belongs_to :team, null: false
          t.belongs_to :user, null: false
          t.index [:team_id, :user_id], unique: true

That’s pretty much it! Now the database won’t allow multiple membership records to exist that reference the same team and user.

    t = Team.create!  # => #<Team id: 1>
    t.users           # => []
    u = User.create!  # => #<User id: 1>
    t.users << u      # => [#<User id: 1>], nice!
    t.users << u      # => [#<User id: 1>], thanks for no errors, Rails!

Note: You won’t be able to adjust the migration as I did in the example without rolling it back first.

Adding Constraints to Existing Tables

As always, it is tricky to add constraints to existing tables because we are unable to assume the state of the data will allow such constraints to be added. For uniqueness we need to make sure the target column is unique before adding the constraint. Here is an idea to get your wheels turning:

    # Assume widgets have a `foo` string column with no constraints
    class AddUniquenessConstraintToWidgetFoo < ActiveRecord::Migration
      def up
        execute "UPDATE widgets SET foo = coalesce(foo,'') || id"
        add_index :widgets, :foo, unique: true
      def down
        # For simplicity we won't allow the rollback of this migration as it is
        # difficult to get the data in exactly the same state as before the
        # migration.
        raise ActiveRecord::IrreversibleMigration

Since we’re assuming there are no constraints on foo we first need to make sure it is at least an empty string, so we coalesce its value to an empty string. Then we concatenate that value with the record’s id since we know it as a unique value. The resulting value will always be unique to the record!

You may also be interested in my solution to the team membership problem mentioned above.

Until next time…

I hope that you now have an understanding of the uniqueness database constraint and why it can really help improve your app’s data integrity. There is one other constraint that was overlooked in the memberships example. That’ll be our next target: foreign keys.

What other ways have you come up with to ensure data integrity in your apps? We’d love to hear what you think!

Juan Pablo Claude

Reviewer Big Nerd Ranch

During his tenure at BNR, Juan Pablo has taught bootcamps on macOS development, iOS development, Python, and Django. He has also participated in consulting projects in those areas. Juan Pablo is currently a Director of Technology focusing mainly on managing engineers and his interests include Machine Learning and Data Science.

Speak with a Nerd

Schedule a call today! Our team of Nerds are ready to help

Let's Talk

Related Posts

We are ready to discuss your needs.

Stay in Touch WITH Big Nerd Ranch News