Ensure that MS Word docx file headers are not corrupted by RubyZip

zip_service.rb

require 'nokogiri'
require 'zip'
require 'zip/zipfilesystem'

  class ZipService
    # Initialize with the directory to zip and the location of the output archive.
    def initialize(input_dir, output_file)
      @input_dir = input_dir
      @output_file = output_file
    end

    # Zip the input directory.
    def write
      entries = Dir.entries(@input_dir) - %w[. ..]

      ::Zip::File.open(@output_file, ::Zip::File::CREATE) do |zipfile|
        write_entries entries, '', zipfile
      end
    end

    private

    # A helper method to make the recursion work.
    def write_entries(entries, path, zipfile)
      entries.each do |e|
        zipfile_path = path == '' ? e : File.join(path, e)
        disk_file_path = File.join(@input_dir, zipfile_path)

        if File.directory? disk_file_path
          recursively_deflate_directory(disk_file_path, zipfile, zipfile_path)
        else
          put_into_archive(disk_file_path, zipfile, zipfile_path, e)
        end
      end
    end

    def recursively_deflate_directory(disk_file_path, zipfile, zipfile_path)
      zipfile.mkdir zipfile_path
      subdir = Dir.entries(disk_file_path) - %w[. ..]
      write_entries subdir, zipfile_path, zipfile
    end

    def put_into_archive(disk_file_path, zipfile, zipfile_path, entry)
      if File.extname(zipfile_path) == ".docx"
        Zip::File.open(disk_file_path) do |zip|
          doc = zip.read("word/document.xml")
          xml = Nokogiri::XML.parse(doc)
          zip.get_output_stream("word/document.xml") {|f| f.write(xml.to_s)}
        end
        zipfile.add(zipfile_path, disk_file_path)
      else
        zipfile.add(zipfile_path, disk_file_path)
      end
    end
  end

This gist zips files recursively, looking for docx files and re-parsing them using Nokogiri. This solution does not use write_buffer.