Question

Say I have an existing dataset in which something is an embedded document but we want to turn it into a reference instead. Is there any kind of automatic or semi-automatic way to do this refactoring.

Answer 1

Refactoring itself is simple. Just swap “embeds_one” for “has_one” (or substitute appropriate terms for your mapper library). It’s the data migration that will cause some pain. Or maybe it won’t. Here’s a small ruby script I whipped up in under 10 minutes. It should cover what I think you need.

source_collection = 'users'
field_to_expand = 'address'
parent_field = 'user_id'
expanded_collection = 'addresses'

require 'mongo'

db = Mongo::Connection.new.db('test')
users = db.collection(source_collection)
addresses = db.collection(expanded_collection)


# prepare test data
users.remove()
addresses.remove()

users.insert({name: 'Joe', address: {city: 'Rio de Janeiro'}})
users.find().to_a # => [{"_id"=>BSON::ObjectId('50614e910ed4c08a6a000001'), "name"=>"Joe", "address"=>{"city"=>"Rio de Janeiro"}}]


users.find().each do |u|
  # move subdocument to a separate collection
  addr = u[field_to_expand]
  addr[parent_field] = u['_id']
  addresses.insert(addr)

  # erase from original document
  users.update({_id: u['_id']}, {'$unset' => {field_to_expand => 1}})
end

users.find().to_a # => [{"_id"=>BSON::ObjectId('50614e910ed4c08a6a000001'), "name"=>"Joe"}]
addresses.find().to_a # => [{"_id"=>BSON::ObjectId('50614e910ed4c08a6a000002'), "city"=>"Rio de Janeiro", "user_id"=>BSON::ObjectId('50614e910ed4c08a6a000001')}]