Migratrix migrates data from one data storage system to another.
It does this in three phases: Extract, Transform, and Load.
The Extract phase extracts data from a legacy data storage system. Typically this will be a Rails database, but it can be anything that supplies a set of data.
In Migratrix, Extraction objects are usually quite simple, but rely on you writing Legacy ActiveRecord classes that can talk to the legacy database, which can be quite tricky. See the examples section if one ever comes into existence.
The Transform phase converts the data into a format that your system likes. In its simplest form it might grab a legacy object and copy its attributes into a new object. In trickier migrations you have more complicated transforms. For example, your legacy data might have home and business addresses stored as fields on the customer object, while you new system has Customer and Address objects. The Transform would take the extracted Customer and produce the Customer, Address (home) and Address (business) objects.
The transform object will probably always be the most complicated bit of code in your migration, because it does all the heavy lifting.
The Load phase loads this data up into your new database. This name can be tricky because usually we think of "Save" here, but the "ETL" terminology comes from data warehousing, and more importantly you might not actually be saving records to a database. For example, you might write a migration that saves an enum table to a yaml file to be loaded up as a set of constants.
The load object is typically the simplest. In fact, if your
transformed objects respond to the save() method, all you have
to do is specify a generic load and you're done.