module Unbreakable
  # Processors are {http://markevans.github.com/dragonfly/file.Processing.html
  # Dragonfly} processors. For example:
  #
  #     class MyProcessor
  #       def coolify(temp_object, opts = {})
  #         SomeLib.coolify(temp_object.data, opts)
  #       end
  #
  #       def uglify(temp_object, ugliness)
  #         `uglify -i #{temp_object.path} -u #{ugliness}`
  #       end
  #
  #       def conditional(temp_object, format, pages)
  #         throw :unable_to_handle unless format == :pdf
  #         # do stuff
  #       end
  #
  #     private
  #
  #       def my_helper_method
  #         # do stuff
  #       end
  #     end
  #     MyScraper.processor.register MyProcessor
  #
  # Public methods must return an object with which a +TempObject+ may be
  # initialized (+String+, +File+, +Tempfile+, +Pathname+ or +TempObject+).
  #
  # You can raise +Dragonfly::Configurable::NotConfigured+ if a configurable
  # variable is required but missing. If a variable is invalid, you can raise
  # +Dragonfly::Configurable::BadConfigAttribute+.
  #
  # If a process has dependencies or conditions, then you can test for these
  # conditions and throw +:unable_to_handle+ to skip processing.
  #
  # If multiple processors define a public method by the same name, the methods
  # will be run in reverse order from the last processor to define the method
  # until one fails to throw +:unable_to_handle+. If all raise an error, then
  # +Dragonfly::FunctionManager::UnableToHandle+ will be thrown.
  #
  # As such, if you are writing a document to plain-text converter, you can
  # write a pdftotext processor, a doctopdf processor, etc. which all define
  # a +to_text+ public method, and use +:unable_to_handle+ to make sure the
  # correct processor runs.
  module Processors
    # If you are writing a simple scraper and only need one processor, you may
    # implement a single +transform+ processor method by subclassing this class:
    #
    #     require 'nokogiri'
    #     class MyProcessor < Unbreakable::Processors::Transform
    #       # Extracts the page title from an HTML page.
    #       def perform
    #         Nokogiri::HTML(temp_object.data).at_css('title')
    #       end
    #
    #       # Saves the page title to an external database.
    #       def persist(arg)
    #         MyModel.create(:title => arg)
    #       end
    #     end
    #     MyScraper.processor.register MyProcessor
    #
    # The following instance methods must be implemented in sub-classes:
    #
    # * +perform+
    # * +persist+
    #
    # +transform+ calls +persist+ with the output of +perform+. This makes it
    # easy for others to subclass your processor and just change the +persist+
    # method to change the external database, for example, while still taking
    # advantage of the hard work done by +perform+. 
    class Transform
      include Dragonfly::Configurable
      include Dragonfly::Loggable

      attr_reader :temp_object, :opts

      # +#transform+ must be defined on the subclass for Dragonfly to see it.
      # @param [Class] subclass a subclass
      def self.inherited(subclass)
        subclass.class_eval do
          # @param [Dragonfly::TempObject] temp_object
          # @param [Hash] opts
          # @return [Dragonfly::TempObject] the same object
          def transform(temp_object, opts = {})
            @temp_object, @opts = temp_object, opts
            persist perform
            temp_object
          end
        end
      end

    private

      # Transforms a record.
      # @return [Hash] the transformed record
      def perform
        raise NotImplementedError
      end

      # Persists a transformed record.
      # @param arg a transformed record
      def persist(arg)
        raise NotImplementedError
      end
    end
  end
end