Rails Migrations Tricks [Guide + Code] — Cheatsheet included

Rails Migrations were introduced to be a convenient way to alter the database in a structured and organized manner, and effectively bring great consistency over time on a project.

Migrating with style

As Ruby on Rails developers (I’m assuming you are too 😘), we should be pretty familiar with Rails Migrations and the Rails generator. But while the most common migrations are no-brainers, some of them are not -or not clearly- covered in the manual. At work at ForestAdmin it’s our job to handle every possible DB schema, in Rails, NodeJS and others, which means we’ve come across a few variations already. So let’s walk through 9 tricks for generators and migrations gathered from experience (and Stack Overflow 🙊).

List of tips (Level of expertise):

  • Drop DB columns ★☆☆
  • Drop a table ★☆☆
  • Roll back a specific migration ★☆☆
  • Default values ★★☆
  • :text V/S :string ★★☆
  • Rename a column ★★☆
  • Rename a model ★★☆
  • Reset your database ★★★
  • Replicate records safely ★★★
  • Retro-update existing records’ when adding a new column ★★★
  • Migrate safely JSON stores ★★★

I’m starting with a basic that’s actually in the docs before getting to the tougher ones:

1- One line migration to drop DB columns:

Rails automatically generates this kind of migrations thanks to the command line generator:

$ rails generate migration RemoveFieldNameFromTableName field_name:datatype

The migration file generated will contain the following line:

remove_column :table_name, :column_name

For instance:

$ rails generate migration RemoveNameFromMerchant name:string

Gives:

class RemoveNameFromMerchant < ActiveRecord::Migration
def change
remove_column :merchants, :name, :string
end
end

Which would remove the field/column “name” from the Merchant table after running rake db:migrate.
🌲 When using ForestAdmin, for every new migration the server should be restarted and the webpage refreshed (🍎+R) for the new schema to be displayed correcly.

2- The right way to write a migration to drop a DB table

The Rails magic isn’t always enough: since this brutally throws away quite a lot of data at a time, there is no one line generator to drop entire tables. We need to generate a migration (important to get the timestamp that keeps everyone’s database synced), and manually add the deletion inside:

$ rails generate migration DropMerchantsTable

This will generate the empty .rb file in /db/migrate/ that still needs to be filled to drop the “Merchant” table in this case.

A Quick-and-Dirty™ implementation would look like this:

class DropMerchantsTable < ActiveRecord::Migration
def up
drop_table :merchants
end
  def down
raise ActiveRecord::IrreversibleMigration
end
end

It’s “correct” as it shows that the migration is one-way only and should not/cannot be reversed. But in order to make a truly clean job in case these modifications were to be reversed we need to have a symmetrical migration (assuming we could recover the lost data), which we can do by declaring all the fields of our table in the migration file:

class DropMerchants < ActiveRecord::Migration
def change
drop_table :merchants do |t|
t.string :name, null: false
t.timestamps null: false
end
end
end

This can be long if the model is complex, but it ensures full reversibility.

Here again, changes will enter effect as usual after running rake db:migrate.

3- Rollback a specific migration

It’s generally not a good idea to do that as migrations are made to ensure consistency over projects, environments, and time, and reverting one migration breaks the chain. However in some cases and for debugging, we can rollback the migration with the corresponding timestamp with this command:

$ rake db:migrate:down VERSION=20170815201547

Which will revert the corresponding migration file: db\migrate\20170815201547_create_merchants.rb (where “create_merchants” does not play a role, as the only identifier is the timestamp).

Once again: ⚠️ This migrates down only this specific file, and not up to this specific file.

A better idea is to revert all migrations up to a certain point in time. In this case we can use the same command with the “STEP” argument defining the number of migration files we want to chronologically roll back:

$ rake db:rollback STEP=3

As you probably know, to rollback only the last migration we can simply ommit the STEP argument:

$ rake db:rollback

4- Set default values for new records

Edit: as reported by MCAU, as of Rails 5.2 (and maybe some earlier versions), this is not actual anymore. Default values set in the active record migrations will apply not only to the database columns, but also when instantiating with Model.new(). The hereunder recommendation can still apply in the case you would want to have some computed “virtual” attributes being setup upon instantiation.

Defining default values in ActiveRecord only works part of the time, in particular it won’t work when calling Model.new(), as it won’t call anything from ActiveRecord.

Instead of “enforcing” the default values with a migration, writing an after_initialize callback in our model will let us set the default values we need, and even associations:

class Merchant < ActiveRecord::Base
has_one :client
after_initialize :init
  def init
self.name ||= "SuperStore" # will set the default value only if it's nil
self.client ||= Client.last # also setting a default association
self.in_business = true if self.in_business.nil? # ⚠️ only default booleans to 'true' if they're 'nil' to avoid overwriting a 'false'
end
end

This allows to have one flexible and easily editable place to initialize our models.

Overriding initialize can also work, but let’s not forget to call super to avoid breaking the init chain.

5- When should we use ‘text’ or ‘string’ in Rails?

What’s the difference between stringand text? And when should each one be used? Quite simply, the difference relies in how the symbol is converted into its respective column type in query language: for instance with MySQL :string is mapped to VARCHAR(255).

When should each be used?

As a general rule of thumb, use :string for short text input (username, email, password, titles, etc.) and use :text for longer expected input such as descriptions, comment content, etc. However, performance requirements can come into play.

With MySQL, we can have indexes on varchars, but we can’t have any on text. ➡️ Use :string when indexes are needed.

With POSTGRES, we should use :text wherever we can, unless there is a size constraint, since there is no performance penalty for text V/S varchar.

6- Rename a database column with a one line Ruby on Rails migration

While creating a Migration as for renaming a column, Rails 4 generates a change method instead of up and down as mentioned in the above answer. The generated change method is as below :

$ rails g migration ChangeColumnName

which will create a migration file similar to this :

class ChangeColumnName < ActiveRecord::Migration
def change
rename_column :table_name, :old_column, :new_column
end
end

7- Rename an entire ActiveRecord model with a Rails migration

It’s not always easy to find the right name for each model in advance. When I choose a poor name for a model, I sometimes resolve to changing it for the greater good, despite the work involved. Here is how to proceed;

Rails doesn’t quite have a ready to fire command line for that, but we can write it ourselves quickly. Let’s create a migration:

$ rails generate migration RenameOldTableToNewTable

Which we can then fill in with “rename_table”:

class RenameOldTableToNewTable < ActiveRecord::Migration
def change
rename_table :old_table_name, :new_table_name #use the pluralized name of tables here
end
end

We will still have to go through and manually rename the model in all our files (And not forget to check for capitalized and pluralized versions of the model).

8- Purge, recreate, drop a Ruby on Rails database

To fully delete and rebuild our database, we can either:

$ rake db:reset db:migrate

Which will reset the database and reload the current schema, or:

$ rake db:drop db:create db:migrate

Which will destroy the DB, create a new one and then migrate the current schema.

⚠️ All data will be lost in both cases.

Just to compare the different dropping and migrating commands:

  • rake db:schema:load - Creates tables and columns within the (existing) database following schema.rb. db:schema:load is used when you setup a new instance of your app.
  • rake db:reset Clears the database (presumably does a rake db:drop + rake db:create + rake db:migrate) and runs migration on a fresh database.
  • rake db:migrate runs (single) migrations that have not run yet. Typically, you would use db:migrate after having made changes to the schema of an existing DB via new migration files.
  • rake db:create creates the database
  • rake db:drop deletes the database
  • rake db:setup does db:create, db:schema:load, db:seed
  • rake db:reset does db:drop, db:setup
  • rake db:seed runs the seeds task to populate the database with preliminary data

Extra tip: Replicate/Duplicate an ActiveRecord record

Sometimes you need to make a perfect copy of a record, changing only one or a few values, such as for an association.

⛔️ Using the clone method will result in a shallow copy of the record, including copying the frozen state and the ID. No callbacks will be called either, which is probably not the effect we’ll be expecting for practical purposes.

new_record = old_record.clone # Shallow copy, without callbacks

✅ The dup method will duplicate the record calling after initialize hooks, clearing associations and IDs. A duped record will return true for new_record? and be ‘saveable’:

new_record = old_record.dup # For practical matters

Extra Tip 2: Updating existing records with a newly added default value

(Asked in the comments)

Example: If you have a User Model and you Add a status field

rails g migration add_status_to_user

How would existing User record be updated with a value of status after the migration?

(Response)

First, if you need all your users to have a status value, you would better specify a default value for new users to ensure it is not nil. This can be added as a parameter in the migration that adds the column, or later in another migration, as such:

def up
change_column :users, :status, :string, default: 'active'
end
def down
change_column :users, :status, :string, default: nil
end

Or with the one-liner:

change_column_default :users:status, from: nil, to: 'active'

To update the existing records, it’s a best practice to do it too through a migration, so that all the versions of the database will be up to date automatically (as opposed to running some code manually on each one of them).

The right way to do that is to create a one-way new migration where you will write the update code:

$ rails g migration update_status_of_existing_users

def up
User.update_all({ :status => 'active'})
end
def down
end

The migration should be one-way and not blindly set values back to nil in the down direction to avoid losing data if you were to roll it back.

Depending on your project, you could merge both migrations into one.

⚠️⚠️ Remember to only use raw Ruby+Rails logic in your migrations and never use anything except attributes coming from your own models.

If you need to use any kind of more advanced logic in your migration, define the methods you need in your migration and call them from there.

Using methods coming directly from your models can seem like a good idea at first to DRY your code, but with time and commits your model will change. The method you are calling in this migration might not run the same way, or exist anymore, which will prevent any new-comer to your project to run the migration set of the project smoothly.

Extra Tip 3: Migrating with Rails’ JSON Store

Working with Rails’ JSON store is remarkably easy and convenient, but requires great care to make migration reversible to prevent frustration.

Consider the following case, where the chronology is important:

  • We successfully create an hstore (as :text) in the DB, and migrated.
  • We then add in the model the rails line allowing this hstore to store “JSON attributes” i.e.: store :hstore_column_name, accessors: [ :attribute_one_name, ...]
  • For some reason, we decide to revert the migration to change something.
  • We make our modifications in the migration, and try to re-run migrate.

Our migration will then fail with an unintuitive error message:

NoMethodError: undefined method `accessor' for #<ActiveRecord::xxx>

What is happening is that the line declaring the hstore is still present in our model at this point. When Rails tries to run the migration, the model is called, and this line throws an error, as this same DB column we’re trying to create is supposed to be already available for use by the hstore; but does not exist yet.

Quick Fix

A quick fix is to remove the hstore declaration in the model and to migrate, then re-add it.

Better solution

In general, making hstore migrations fully reversible is not obvious. A migration allowing for model evolutions (if this hstore disappears a year from now and a newcomer needs to migrate their DB to the latest), and for data population at the same time (for instance if you are renaming an hstore or some of its attributes) will require 3 successive migration files, where hstore declarations should happen inside the migration files.

🚀 Now What?

→ 👋🏻 Thanks for reading. If you liked it, please ‘clap’ several times or share it so that more people can see it, and I keep writing on similar topics.

Subscribe to get my essays here, in which I try to give a helicopter view on technology related topics.

More about me here. You can also check out some of my other top articles:

Additional Documentation:
http://edgeguides.rubyonrails.org/active_record_migrations.html

This mini-guide is intented for Rails 4+, but many of the commands and much of the code will also work for Rails 3+ and Rails 5.