gistfile1.rb

# coding: utf-8

class ApplicationController < ActionController::Base
  before_filter :normalise_param_encodings

  # On M17N aware VMs, ensure params from the user are marked with an appropriate encoding.
  #
  # As of Rails 2.3, Rack returns all params with an ASCII-8BIT encoding, which causes an
  # exception if a param is mixed with a UTF-8 string or ERB template. Hopefully that will be
  # fixed at some point and this won't be necessary any more.
  #
  # I've read in a few places that most browsers seem to submit data to the server in the same
  # encoding as the last page it received from that server. My brief testing on FF 3.0.x
  # confirmed this (for FF at least). FF also doesn't seem to explicitly specify the charset
  # on either GET or POST requests (unless they're via AJAX).
  #
  # Since we always serve UTF-8, I'm going to assume all data we get is the same. If it isn't,
  # I sanitise it.
  #
  # In *theory*, request.content_charset would contain the charset of the request, but it
  # never seems to.
  #
  # As well as marking the strings as UTF-8, I also ensure they contain valid utf-8 data. The
  # iconv technique for doing this is based on
  # http://po-ru.com/diary/fixing-invalid-utf-8-in-ruby-revisited/
  #
  def normalise_param_encodings
    return unless String.method_defined?(:force_encoding)

    normalise_object_encoding(params)
  end

  def normalise_object_encoding(obj)
    case obj
    when String
      unless obj.frozen?
        obj.force_encoding(Encoding::UTF_8)
        ic = Iconv.new('UTF-8//IGNORE', 'UTF-8')
        obj.replace(ic.iconv(obj + ' ')[0..-2])
      end
    when Array
      obj.each { |o| normalise_object_encoding(o) }
    when Hash
      obj.each { |k,v| normalise_object_encoding(v) }
    end
  end
end