Rewriting munged PDF files.

pdf_fixer.rb

class PdfFixer
  def initialize
    @paths = Dir.glob('in/**/*.pdf').sort
    `rm -rf out && mkdir out`
  end

  def process_files
    start_logging

    @paths.each do |path|
      validate_and_repair(path)
    end

  ensure
    stop_logging
  end

  private

  def start_logging
    @log  = File.open('error.log', 'w')
    @good = File.open('good.log',  'w')
    @bad  = File.open('bad.log',   'w')

    @log.puts('Beginning pdf fix.')
  end

  def good_start?(file)
    first = file.readline
    file.rewind

    first.start_with?('%PDF')
  end

  def validate_and_repair(path)
    file = File.open(path, 'r')

    if good_start?(file)
      @good.puts(path)
    else
      @bad.puts(path)
      try_repair(file)
    end
  end

  def try_repair(file)
    any_pdf_data = nil
    output_path = "out/#{file.path.split('/').slice(1..-2).join('/')}"

    out_file = nil

    file.each_line do |line|
      any_pdf_data = any_pdf_data || line.start_with?('%PDF')

      if any_pdf_data
        `mkdir -p #{output_path}` unless Dir.exists?(output_path)

        out_file ||= File.open(output_path + '/' + file.path.split('/').last, 'w')

        out_file.write(line)
      end
    end

  rescue StandardError => e
    @log.puts("Error in try_repair for file #{file.path}: #{e.message}")
  ensure
    out_file.close if out_file
  end

  def stop_logging
    @log.puts('Finishing pdf fix.')

    @log.close
    @good.close
    @bad.close
  end
end

PdfFixer.new.process_files