This gem offers a solution for safely writing code in migrations, ensuring no regressions or data corruption in production.
In some cases, Rails migrations affect not only the database schema but also the data itself. However, the code used to modify data during migrations may become outdated and fail. While there are techniques, like defining model classes within migrations or using raw SQL, these methods are not foolproof and can't always ensure data integrity on the migration run.
Beyond outdated code issues, data migrations in production can sometimes lead to data corruption. How can this be prevented? This gem addresses the problem by offering a safeguard at no additional cost.
This gem essentially encourages writing tests for data migrations by allowing developers to write the data-related code in separate methods. By testing the data migration logic independently, it helps prevent problems related to outdated migrations or data corruption.
If the gem purpose is still not clear please check out this blog post.
Add this line to your application's Gemfile:
gem 'migration_data'
And then execute:
$ bundle
Or install it yourself as:
$ gem install migration_data
In your migration define a #data
method:
class CreateUsers < ActiveRecord::Migration
def change
# Database schema changes as usual
end
def data
User.create!(name: 'Andrey', email: 'ka8725@gmail.com')
end
def rollback
User.find_by(name: 'Andrey', email: 'ka8725@gmail.com').destroy
end
end
Now when migrations run with rake db:migrate
command the #data
method is executed right after the standard #change
or #up
method.
When migrations roll back with rake db:rollback
command the #rollback
is executed right after the standard #change
or #down
method.
It might appear that a data migration should run before the standard #change
, #up
, and #down
methods. Define #data_before
and #rollback_before
methods for "up" and "down" directions correspondingly. There are also #data_after
and #rollback_after
methods for symmetry in that case, but basically they play the same role as #data
and #rollback
methods.
All these methods can be defined in one migration. They are executed in the following order when migration is run on up:
#data_before
#change/up
#data
#data_after
and on down:
#rollback_before
#change/down
#rollback
#rollback_after
.
Note: in some circumstances, the
reset_column_information
method should be called on a model which table is changed in the migration. Especially when you are certain that there should be present some column for a model but it's absent for some reason. Read more about this in the official Rails docs.
At some point, one might realize that data migrations should not run on particular environments, e.g. test
.
On performing migrations in test
environment, a data migration might try to add the same data that has already been added by seeds. In that case, migrations might fail with a duplication error.
Use MigrationData.config.skip = true
to skip data migrations execution. One might put this code in an initializer, e.g. config/initializers/migration_data.rb
:
if Rails.env.test?
MigrationData.config.skip = true
end
To keep your migrations working don't forget to write tests for them. It's preferably to put the tests for migrations into spec/db/migrations
folder, but actually it's up to you. Possible RSpec
test (spec/db/migrations/create_user.rb
) for the migration looks like this:
require 'spec_helper'
require 'migration_data/testing'
require_migration 'create_users'
describe CreateUsers do
describe '#data' do
it 'works' do
expect { described_class.new.data }.to_not raise_exception
end
end
describe '#rollback' do
before do
described_class.new.data
end
it 'works' do
expect { described_class.new.rollback }.to_not raise_exception
end
end
end
The helper to load migrations require_migration
is defined in the migration_data/testing
. So you should to require it to have access to this convenient require extension.
- Fork it ( http://github.com/ka8725/migration_data/fork )
- Create your feature branch (
git checkout -b my-new-feature
) - Commit your changes (
git commit -am 'Add some feature'
) - Push to the branch (
git push origin my-new-feature
) - Create new Pull Request