GSoC 2017 - Coding Period | Week 13

Throughout this week, I was working on extending the different methods supported by daru-io, like #from_format , #read_format , #to_format , #to_format_string and #write_format . Along with this, I'm also working on a small bug fix in daru.

Two of my favourite things are coming to an end (kind of) within this week - GSoC (Google Summer of Code) 2017, and GoT (Game of Thrones). With the final evaluation due tomorrow, I'll soon be publishing a comprehensive blog post to submit for the GSoC portal as Work Submission.

• read_format : This method imports from a specific file format. For example, a .xls file.
• from_format : This method imports from a Ruby format instance. For example, a Spreadsheet instance.
• to_format : This method exports a Daru::DataFrame into a Ruby format instance.
• to_format_string : This method exports a Daru::DataFrame into a writable file string.
• write_format : This method writes a Daru::DataFrame into a specific file format.

The main aspect while implementing this in daru-io, is to find a way to automate the linkages between daru and daru-io for all these different methods. Here's the sample logic behind these linkages, that uses Regexp (Regular Expression) matching.

  
  # function : A symbol, like :to_csv
  # instance : A daru-io class, like Daru::IO::Exporters::CSV
  def link(function, instance)
    case function.to_s
    when /\Ato_.*_string\Z/
      define_method(function) { |*args, &io_block| instance.new(self, *args, &io_block).to_s }
    when /\Ato_/
      define_method(function) { |*args, &io_block| instance.new(self, *args, &io_block).to }  
    when /\Awrite_/
      define_method(function) { |*args, &io_block| instance.new(self, *args[1..-1], &io_block).write(*args[0]) }
    when /\Afrom_/
      define_singleton_method(function) { |*args, &io_block| instance.new(*args[1..-1], &io_block).from(*args[0]) }
    when /\Aread_/
      define_singleton_method(function) { |*args, &io_block| instance.new(*args[1..-1], &io_block).read(*args[0]) }
    else
      raise ArgumentError, 'Invalid function name given to monkey-patch into Daru::DataFrame.'
    end
  end
  

This way, the daru-io methods are linked back to daru calls in this way -

  
  importer = Daru::IO::Importers::Format.new(opts, &block)
  exporter = Daru::IO::Exporters::Format.new(df, opts, &block)

  exporter.write(path) #! Linked to df.write(path, opts, &block)
  exporter.to_s #! Linked to df.to_s(opts, &block)
  exporter.to #! Linked to df.to_format(opts, &block)

  importer.read(path) #! Linked to Daru::DataFrame.read_format(path, opts, &block)
  importer.from(instance) #! Linked to Daru::DataFrame.from_format(instance, opts, &block)
  

Progress related to this, can be tracked via this Pull Request.

In this Pull Request I had drafted months ago, I had missed to handle the missing elements like nil . Hence, all the comparisons would fail if the Daru::Vector contains any missing elements. This is currently under progress in this Pull Request and is yet to be merged.