Subsetter and object to seed serializer.

gem 'seed_reaper'

!!NOTE!!: This currently removes/replaces EVERYTHING in your db/seeds/ directory. Take care and save anything that you intend on keeping. More customizations to come regarding this functionality.

Lets say you have the following:

class Thing < ActiveRecord::Base
  has_many :associated_things
end

class AssociatedThing < ActiveRecord::Base
  belongs_to :thing
end

This:

SeedReaper::SeedWriter.new(thing: :associated_things).write!

Will produce:

# db/seeds/0_thing.seeds.rb

Thing.new(
  # ...attributes
).save!(validate: false)

AssociatedThing.new(
  # ...attributes
).save!(validate: false)

# ... the rest of the associated_things

The configuration passed to SeedReaper::SeedWriter.new() can be infilitely deep/complex and also accepts a { meta: { count: ... } } hash at any level that will limit the number of seedified records, effectively subsetting a DB through the active record API. The configuration passed must be a list/array at the root level and can be as granular as necessary.

SeedReaper::SeedWriter.new(
  :industry,
  :service_provider_category,
  :service_location,
  :auto_added_service,
  :green_industry,
  {
    base_service_provider: [
      { meta: { count: 10 } },
      :contact,
      :human_resources_contact
    ]
  },
  {
    user: [
      { meta: { count: 50, joins: :admin } },
      :admin,
      {
        contact: [
          :contact_infos
        ]
      }
    ]
  }
).write!

Note that order is significant. The seeds are serialized using .save!(validate: false) but this will obviously not subvert the DB schema constraints in the case of foreign keys etc. I.e. if industries had a foreign key constraint to service_provider_categories, for instance, you would likely want to flip those arround in the list so that the dependency is seeded first (the seed files are prefixed with integers to enforce this during db:seed).

Each root level config element passed to the SeedWriter initializer will get written to a separate seed file and sub configurations for each element will be written inline within the root level seed file. This is necessary in order to corectly interleave the processing of the seeds such that dependencies are not invalidated. This solution automatically handles belongs_to dependencies by serializing the dependency config tree prior to that of the dependent instance and subsequent non-belongs_to associations.

A rails rake task comes stock with this gem, seed_reaper:write, that will take a YAML specification defined in config/seed_reaper.yml (format similar to the hash examples above), and write the seeds based on the specification provided.

You could, for example, do the following to load the seeds from seed reaper in the correct order with the db:seed rake task:

# db/seeds.rb

Dir[File.join(Rails.root, 'db', 'seeds', '*.seeds.rb')].sort.each { |seed| load seed }

Copyright (c) 2022 David Butts. See LICENSE.txt for further details.