Bear Den Designs

Google-Geocode Gem Woe's

While using the very cool google-geocode gem for ruby, I ran into a small (read: big) problem.

Do a search for “Trinidad” by itself and you get something like:

The problem lies not in the gem but in ruby’s REXML and how it deals with the xml google sends back. This only happens when international characters are involved.

After doing a little googling I saw a patch for rexml which I changed into a monkey patch for google-geocodes helper library rc-rest.

This monkey patch will solve all your accent mark woe’s.

class RCRest

  def get(method, params = {})
    url = make_url method, params

    url.open do |xml|
      body = xml.read

      res = REXML::Document.new Iconv.conv("UTF-8//Ignore", 'UTF-8', body)

      check_error res

      return parse_response(res)
    end
  rescue IOError, SystemCallError, SocketError, Timeout::Error,
         REXML::ParseException => e
    raise CommunicationError.new(e)
  rescue OpenURI::HTTPError => e
    begin
      xml = REXML::Document.new e.io.read
      check_error xml
    rescue REXML::ParseException => e
    end
    new_e = CommunicationError.new e
    new_e.message << "\n\nunhandled error:\n#{xml.to_s}" 
    raise new_e
  end

end

The magical change is the inclusion of Iconv to make REXML happy.