Attachments from Outlook Msgs extractor

pdf_from_msg_extractor.rb

require 'mapi/msg'
require 'fileutils'
require 'pdfkit'
require 'nokogiri'

PDFKit.configure do |config|
  config.default_options[:load_error_handling] = 'ignore'
  config.default_options[:load_media_error_handling] = 'ignore'
  config.default_options[:disable_javascript] = true
end

def extract_pdf_files(path)
  extraction_directory = File.join(File.dirname(path), 'extracted_files')
  FileUtils.mkdir(extraction_directory) unless File.exists?(extraction_directory)

  Dir.glob(File.join(path, '*.msg')).each do |file_path|
    msg = Mapi::Msg.open(file_path)

    pdf_filepath = File.join(extraction_directory, "#{File.basename(file_path, '.*')}.pdf")

    next if File.exists?(pdf_filepath)

    msg.attachments.each do |attachment|
      next unless attachment.filename

      File.open(File.join(extraction_directory, attachment.filename), 'wb') do |output|
        source = attachment.data.io
        source.rewind
        output << source.read
      end
    end

    unless File.exists?(pdf_filepath)
      mime = msg.to_mime

      if mime.parts.nil?
        puts "MIME conversion failed for #{File.basename(file_path)}"

        next
      end

      html_mime_parts(mime).each do |mime_part|
        body = mime_part.body.to_s.dup.encode('UTF-8', invalid: :replace, undef: :replace)
        raw_html = html_from_nokogiri(body)
        raw_html ||= msg.properties.body

        kit = PDFKit.new(raw_html)
        kit.to_file(pdf_filepath)
      end

      unless File.exists?(pdf_filepath)
        text_mime_parts(mime).each do |mime_part|
          body = mime_part.body.to_s.dup.encode('UTF-8', invalid: :replace, undef: :replace)

          kit = PDFKit.new(body)
          kit.to_file(pdf_filepath)
        end
      end
    end
  end
end

def html_mime_parts(mime)
  mime.parts.select do |part|
    part.content_type == 'text/html'
  end
end

def text_mime_parts(mime)
  mime.parts.select do |part|
    part.content_type == 'text/plain'
  end + mime.parts.select do |part|
    part.content_type == 'multipart/alternative'
  end.map do |part|
    part.parts.select do |child_part|
      child_part.content_type == 'text/plain'
    end
  end.flatten.select { |part| part.body =~ /Dear/ }
end

def html_from_nokogiri(html)
  Nokogiri::HTML::Document.new(html).to_html(encoding: 'UTF-8')
end

@alexchumak this is a working prototype for extracting attachments from those MSG files.