GSoC 2017 - Coding Period | Week 4

Posted on 25/06/2017 at 2:30 PM by The Vibe.

Software Development Documentation Ruby GSoC 2017

GSoC Banner image

Swiftly like a samurai sword, the fourth week of Google Summer of Code 2017 whizzed past me. In this week, I worked on adding support to import a Daru::DataFrame from a Mongo collection, and also made changes to the Redis Importer as per the suggestions. With the second month of coding to start in another week, the project is going as per schedule.

Mongo Importer

According to my timeline, the importer I had planned to extend support this week, was the Mongo Importer. Mongo is a NoSQL database, which works on the concept of storing data as documents, and grouping all these documents under collections. These collections are then stored in a database and accessed as BSON (Binary JSON) documents when required.

For this Importer, I used the mongo gem. The BSON documents returned from a collection, are parsed to JSON by the Mongo gem, through a to_json method. The best part is that, the JSON Importer built last week has been re-used (inherited) by the Mongo Importer to re-use the feature of JSON-path selectors.

  #! lib/daru/io/importers/mongo.rb

  require 'daru/io/importers/json'

  module Daru
    module IO
      module Importers
        class Mongo < JSON
          def call
            # Connects to a collection
            # Finds matching BSON documents
            # Converts BSON documents to JSON object
            @json_input = @client[@collection]
                          .find(@filter, skip: @skip, limit: @limit)

            # calls the JSON Importer

The client is obtained through a variety of arguments - String , Hash or an existing Mongo::Client connection, through the private instance method get_client() of Daru::IO::Importers::Mongo class.

  def get_client(connection)
    case connection
    when ::Mongo::Client
    when Hash
      hosts = connection.delete :hosts, connection)
    when String
      raise ArgumentError,
        "Expected #{connection} to be either a Mongo instance, "\
        'Mongo connection Hash, or Mongo connection URL String. '\
        "Received #{connection.class} instead."

Progress related to the Mongo Importer can tracked on this Pull Request. Here are some screenshots from the YARD documentation.

Mongo Importer - Example Mongo Importer - Notes Mongo Importer - Parameters

Optional dependencies and Partial requires

daru-io uses a lot of gems that parse various file / database formats. Having all these gems as development dependencies would be costly, especially as daru-io is ideated to support partial requiring of certain importers / exporters. This means that the use-case of daru-io could be any of the below statements -

  require 'daru/io' # => Requires ALL Importers and Exporters
  require 'daru/io/importers' # => Requires ALL Importers
  require 'daru/io/importers/json' # => Requires only JSON Importer

Hence, a better trade-off for using the various parsing these parsing gems is to have them as optional dependencies, rather than development dependencies. By optional dependencies, the parsing gems will directly just be require d by daru-io. In case the require d gem isn't present in the environment, the default LoadError is rescued to print a more helpful message. This paves way for an optional_gem() method, which can be used like

  # Used in Excel Importer (say)
  optional_gem 'spreadsheet', '~> 1.1.1'
  # => true
  # Used in ActiveRecord Importer (say)
  optional_gem 'activerecord', '~> 4.0', requires: 'active_record'
  # => LoadError: Please install the activerecord gem 4.0 version, with
  # `gem install activerecord -v '4.0'` to use the
  # Daru::IO::Importers::ActiveRecord module."

This optional_gem() method could be achieved with something like this, under the hood -

  module Daru
    module IO
      class Base
        def optional_gem(dependency, version=nil, requires: nil,
          gem dependency, version
          require requires || dependency
        rescue LoadError
          statement =
            if version.nil?
              "gem install #{dependency}"
              "gem install #{dependency} -v '#{version}'"
          raise LoadError,
            "Please install the #{dependency} gem #{version} version, with "\
            "#{statement} to use the #{callback} module."

Progress related to this, can be tracked in this Issue and this Pull Request.

Redis Importer

Redis's scan method fetches all matching keys in a random order, and provides the results in a paginated format. This pagination traversal has successfully been implemented in the Redis Importer this week, along with count and offset features. However, the addition of offset doesn't seem to be very significant owing to the random order characteristic of Redis.

Also, discussion is still going on, regarding whether a chunk feature can be added, to enable usage with the Parallel gem, for running same computation on multiple dataframes in parallel.

Progress related to the Redis Importer, can be tracked in this Pull Request.

Integrating Rubocop-rspec check

  • Rubocop is a static code analyzer for Ruby
  • RSpec is a testing ramework
  • rubocop-rspec is a plugin gem to rubocop, that specifies improvements in the written RSpec tests and helps in cleansing them.

Progress related to this, can be tracked in this Pull Request. This Pull Request is intended to be worked on during the next week, after the Redis & Mongo Importers have been merged.