GSoC 2017 - Coding Period | Week 9

This week was pretty different than the previous weeks, as this included a fun-filled round-robin review among the 3 students who were selected by Ruby Science Foundation organization, namely Prasun, Shekhar & myself. So, this is how it works -

• Prasun reviews my work on daru-io
• I review Shekhar's work on daru-view
• Shekhar reviews Prasun's work on Arrayfire

Such a review helps in getting more perspectives about the codebase from someone who hasn't been around when the codebase was being developed.

Sticking to the timeline and working on many branches which are being refined is good.

• Should ActiveRecords Importer / Exporter replace that of SQL?

Clarification : In case of use-cases (say, a non-Rails app) where ActiveRecords isn't used, the SQL IO modules would come in handy. Hence, IMO it'd be better to retain SQL IO modules.

• Timeout exception not handled in Redis & Mongo modules.

Clarification : Acknowledged, this is an Issue which I didn't face during usage. However, it definitely does seem logical to handle Timeout exceptions.

• Can json modules have a simpler use-case than JsonPaths?

Clarification : JsonPaths are optional arguments. By default, '$.{column_name}' is used. However, while creating / accessing from complexly nested hashes (as is the case with most social media APIs), JsonPaths seems to be most easiest way to deal.

• RDoc generation with rdoc leads to error.

Clarification : Ah, this seems to have been left out in documentation. The YARD docs are to generated with yard doc rather than rdoc .

Timeline wise

Most components like nyaplot , highcharts , googlecharts have been adhered to, with the proposed timeline. However, other components like matplotlib and chartkick have been left out.

matplotlib has been dropped, as it has the ability to create only SVG charts which aren't interactive. Meanwhile, chartkick has been dropped as it's dependencies (highcharts , googlecharts ) have already been added support in daru-view .

Code Quality

Shekhar has done a great work with implementation of the adapters architecture. Meanwhile, these are some improvements that could be fixed IMO -

• Couple of Rubocop disables like PerceivedComplexity & CyclomaticComplexity .
• Couple of unused variables on running rspec , that should be taken care of.
• Tidying specs with rubocop, rubocop-rspec and saharspec.
• Update to new YARD doc standards.
• Some pieces of code that guess the datatype seem like they can be DRY-ied with Inheritence.
• Few minor code style enhancements like case..when blocks, as present in view/adapters/googlecharts.rb and view/adapters/highcharts.rb.

In short, blocks like

    
    case
    when var.isa? String
      'String'
    when var.isa? Integer || var.is_a? Float
      'Number'
    else
      'Something else'
    end
    

can be changed to

    
    case var
    when String then 'String'
    when Integer, Float then 'Number'
    else 'Something else'
    end
    

Usability

• Download as SVG feature for charts
• Better documentation with links to respective JS lib should be provided in README, as the whole usage of daru-view depends on the options that can be passed.
• Monkey-patch into Daru::DataFrame / Daru::Vector to allow use-cases like df.plot(opts)

• The Excelx Importer has been added to support reading from .xlsx files along with :skiprows and :skipcols functionality, with this Pull Request.

• The CSV Importer has been added features of :skiprows , empty dataframe and reading from .csv.gz files in this Pull Request.

• The CSV Exporter has been extended support to export to .csv.gz format, with this Pull Request.

• The JSON Exporter has been added to the fray, with JsonPaths feature for complexly nested hashes, :orient option and also block manipulation of the JSON content. Progress can be tracked in this Pull Request.