Integer and String Enum in Rails

TL;DR

For General Use: I would lean towards using strings for enums due to their readability and maintainability benefits.
For Performance-Critical Applications: I would opt for integers, especially if dealing with large datasets or requiring high performance

1. What is an enum and why use it?

Enums (short for enumerators) in Rails are a way to define a set of predefined values for a column in a database table. They provide built-in validation and are highly readable and maintainable. Here's a simple example:

// language: ruby
class User < ApplicationRecord  
  enum status: { active: 0, inactive: 1, cancelled: 2 }
end

It will be shipped build-in methods

// language: ruby
User.active # list all active users
user = User.create status: :active
user.active? # true
user.inactive! # set status to inactive

Enum automatically handles validation, ensuring that only predefined values are accepted.

// language: ruby
user = User.create status: :trial # raise error

2. Integer enum

You can define integer enum by

// language: ruby
class User < ApplicationRecord  
  enum status: { active: 0, inactive: 1, cancelled: 2 }
end

Pros

Performance: Integer enums are generally faster for database operations, including queries and indexing.
Storage Efficiency: Integers take up less space in the database compared to strings.
Consistency: Less prone to human error such as typos.

Cons

Readability: Integer values are not self-explanatory when inspecting the database directly. If you read the database directly or use other frameworks, it can be unclear what the numbers represent.

// language: sql
SELECT * FROM users WHERE status = 0;

Migration Complexity: Changing enum values requires careful database migrations to avoid breaking the application. For example, you have:

// language: ruby
enum program: { summer_2020: 0, winter_2020: 1 }

Next year, you want to add programs for the year 2021. Your updated enum might look like this:

// language: ruby
enum program: { summer_2020: 0, winter_2020: 1, summer_2021: 2, winter_2021: 3 }

Now, suppose you want to group programs by season rather than by year. You might want the enum to look like this:

// language: ruby
enum program: { summer_2020: 0, summer_2021: 2, winter_2020: 1, winter_2021: 3 }

However, this reordering makes the values look disorganized and can create confusion. To properly group them, you would need to create a migration to update existing records:

3. String Enum

You can define string enum by

// language: ruby
class User < ApplicationRecord  
  enum status: { active: 'active', inactive: 'inactive', cancelled: 'cancelled' }
end

Pros

Readability: String values are human-readable and self-explanatory when inspecting the database.

// language: sql
SELECT * FROM users WHERE status = 'active';

Maintainability: It is easier to add new values or change existing ones without worrying about data loss or complex migrations.
Human-Friendly: It is beneficial for teams that frequently read data directly from the database or when multiple systems need to access the enum values.

Cons

Performance: String enums can be slower for database operations compared to integers.
Storage: Strings take up more space in the database, especially with longer names.

4. Space and Speed

You can read this post for details Rails enums integer vs string way | by Michał Rudzki | Medium. I will summarize some points.
Populated with records:

4 kinds of pets and 11 million records
8 kinds of pets and 28.5 million records
8 kinds of pets with long names and 10 million records

Query speed

Integer column is faster by about 14% in first collection and 12% in second one, but this is only for index-less columns.
For columns with indexes the differences are 7% but in favour of strings and 2% in favour of integers respectively. Of course indexes takes a lot of space but depending on usage the absolute values are totally worth it.
For longer pet names differences are more visible and are 22% in index-less columns and 43% for index columns

Storage

Here is the size difference, string takes up: 2297 MB and integer takes up and integer takes up: 2292 MB.
Of course the issue here is that strings are really short so the difference is just 5MB, the longer strings would’ve taken much more space.
So I was a little confused why is it only 5MBs.
I created new branch with longer names and then the difference on 10 million records is: 86MB and whole table takes 802MBs with integer enum and 888MB on string enums, so it is about 10% of difference.

5. Conclusion

Ultimately, the choice depends on your specific use case and the needs of your application and team. By understanding the trade-offs, you can make an informed decision that best suits your project.

Happy coding!