Export ActiveRecord Tables to CSV

exporter.rb

require 'csv'

module Exporter
  DEFAULT_EXPORT_TABLES = [ Invoice, InvoiceItem, Item, Merchant, Transaction, User ]
  DESTINATION_FOLDER = "tmp/"

  def self.included(klass)
    klass.extend ClassLevelMethods
  end

  def self.export_tables_to_csv(tables = DEFAULT_EXPORT_TABLES)
    tables.each do |klass|
      klass.send(:include, self)
      klass.export_table_to_csv
    end
  end

  def data
    self.class.column_names.map { |column| send(column) }
  end

  module ClassLevelMethods
    def export_table_to_csv
      CSV.open(filename_for_class, "w") do |output_file|
        output_file << column_names
        data.each{ |row| output_file << row }
      end
    end

    def filename_for_class
      [DESTINATION_FOLDER, to_s.pluralize.underscore, '.csv'].join
    end

    def data
      all.map(&:data)
    end
  end
end

Untested tweaks

I very much like the module approach, I would however suggest to not include it in ActiveRecord::Base class but in a AR::Base subclass that needs the features. This way you keep your features only available where you need them on only there.

Or, take the DCI approach and mix it in only when you need it. That way it's not on the object unless you're exporting.

Not exactly sure if this works since it's a class, but the idea is:

task :export_people do
  person_klass = Person.include(Exportable)
  person_klass.export
end

That way the primary Person class does not include Exportable unless the context is appropriate.

@saturnflyer could probably elaborate :-)

@mattetti and @ngauthier how about this?

it definitely addresses my small complaint, one more pedantic comment would be the naming of def self.export_tables_to_csv(tables = DEFAULT_EXPORT_TABLES). Reading this method, I would expect that the input param to be an array of DB tables. But looking at the constant definition and at the code, the expected param is actually an array of models.
And since I'm at it, I would require activesupport at the top of the file since pluralize and underscore rely on it, I wouldn't call find(:all) but use a batch approach to avoid issues with big DB.

All small details tho.

I chose not to change Jeff's original name on the "tables" variable. Probably better named "models".

activesupport would be required implicitly since we're assuming active record models here. We could also do find_in_batches. My changes were almost entirely stylistic. Ultimately the best approach is to mysql the database's dump utility for any of this :)

(^_-) very true. This code could also work with other ORMs tho but granted using an ORM for that task seems totally overkill.

Yup, that's why there are task-oriented gems like Taps around.