Remove dependency on "word-to-markdown"

[ronaldtse]

The purpose of the following code was just to convert a Word document into HTML then into Coradoc.

We should supersede this as the gem is introducing unnecessary dependencies we don't want.

module Coradoc
  module Input::Docx
    def self.processor_id
      :docx
    end

    def self.processor_match?(filename)
      %w[.docx .doc].any? { |i| filename.downcase.end_with?(i) }
    end

    def self.processor_execute(input, options = {})
      image_dir = Dir.mktmpdir
      options = options.merge(sourcedir: image_dir)
      doc = WordToMarkdown.new(input, image_dir)
      doc = Coradoc::Input::HTML.cleaner.preprocess_word_html(doc.document.html)
      options = WordToMarkdown::REVERSE_MARKDOWN_OPTIONS.merge(options)
      Coradoc::Input::HTML.to_coradoc(doc, options)
    ensure
      FileUtils.rm_rf(image_dir)
    end

    def self.processor_postprocess(data, options)
      Coradoc::Input::HTML.processor_postprocess(data, options)
    end

    # This processor prefers to work on original files.
    def self.processor_wants_filenames; true; end

    Coradoc::Input.define(self)
  end
end

[hmdne]

WordToMarkdown gem does a little bit more than just run LibreOffice to convert DOCX to HTML. It also postprocesses said HTML by eg. finding headings based on font size:

https://github.com/benbalter/word-to-markdown/blob/main/lib/word-to-markdown/converter.rb
https://github.com/benbalter/word-to-markdown/blob/main/lib/word-to-markdown/document.rb

Should we vendor those?

From my experience, those either improve the resulting conversion or make it a little worse. So there is some inconsistency whether DOCX has been converted manually or not.