Question

I really like the Rails 3 style migrations, i.e. one change method being smart enough to recognize if the migrations is being installed or rolled back, so I don’t have to write up and down methods mirroring each other. But I have situation that I need to skip some code when the migration is rolled back (updating counter_cache columns that I’m adding).

I looked at http://guides.rubyonrails.org/migrations.html but the examples at the end of section 5 suffer from the same problem:

class AddFuzzToProduct < ActiveRecord::Migration
  class Product < ActiveRecord::Base
  end
  def change
    add_column :products, :fuzz, :string
    Product.reset_column_information
    Product.all.each { |f| f.update_attributes! :fuzz => 'fuzzy' }
  end
end

When this migration is rolled back, the update of fuzz field is unnecessary. Is there a way to prevent it?

I tried looking into Product.column_names but since Rails is smart enough to perform migration in reverse direction, the update is executed before the column is removed. Also, when change method is defined, any up or down methods seem to be ignored. Any other ideas?

Answer 1

In this case I think you’ll have to use up and down methods as usual. Don’t worry, despite the addition of change in Rails 3 those methods aren’t, as far as I know, bound for the chopping block. Continue using them where necessary.

Edit: Here’s an option: Override migrate.

class AddFuzzToProduct < ActiveRecord::Migration
  class Product < ActiveRecord::Base
  end

  def change
    add_column :products, :fuzz, :string
  end

  def migrate(direction)
    super # Let Rails do its thing as usual...

    if direction == :up # ...but then do something extra if we're going 'up.'
      Product.reset_column_information
      Product.all.each { |f| f.update_attributes! :fuzz => 'fuzzy' }
    end
  end
end

Thoughts?