Make irreversible migrations work in change

In Rails 2, migrations had only 2 methods: up and down. They are called when running migrations up and down, respectively, with the Rake tasks rake db:migrate and rake db:rollback. In the migration’s up method, you had to write code called only when running the migration forward, and in down, the code called only when rolling the migration back. For example, if you created a table in up, you had to drop it manually on rollback. Rails 3 gave us a great method, change, which allowed us to write code that creates a table and drops it automatically when rolling the migration back. This was a really great step forward, and thankfully it still exists in Rails 4. But unfortunately, there was a problem: you didn’t have the opportunity to say, “Don’t run this piece of code in down; run it only in up.” So the solution was just to use the old syntax. Since Rails 4 was released, there has been a feature to fix this situation.

Reversible

How would you write a migration that should add a new column in up and fill it for all records in the table only in up, then only delete this column in down? I bet that in Rails 3 you would write it like this:

def up
  add_column :users, :location, :string
  User.update_all(location: 'Minsk')
end

def down
  remove_column :users, :location
end

You had to avoid using the change method, which saves some time. For example, if you didn’t need to update column value immediately after adding it, you could cut this code down to something like this:

def change
  add_column :users, :location, :string
end

And that’s it. In up, it will add the column to the table and remove it in down. Much less code, and that’s a benefit.

Rails 4 provides one more useful way to write what we need in one place:

def change
  add_column :users, :location, :string
  reversible do |direction|
    direction.up { User.update_all(location: 'Minsk') }
  end
end

Check out the original documentation.