mojibake.rb

module Mojibake
  # UTF-8 content, interpreted as latin1
  class Utf8
    CANARY = "æøåÆØÅ".chars.map { |c| c.encode(Encoding::UTF_8).force_encoding(Encoding::ISO_8859_1).encode(Encoding::UTF_8) }.freeze

    def detect?(content)
      CANARY.any? { |c| content.include?(c) }
    end

    def repair(content)
      content.encode(Encoding::ISO_8859_1).force_encoding(Encoding::UTF_8)
    end
  end

  # Latin1 content, interpreted as UTF-8
  class Latin1
    CANARY = "æøåÆØÅ".chars.map { |c| c.encode(Encoding::ISO_8859_1).force_encoding(Encoding::UTF_8) }.freeze

    def detect?(content)
      CANARY.any? { |c| content.include?(c) }
    end

    def repair(content)
      content.encode(Encoding::UTF_8).force_encoding(Encoding::ISO_8859_1)
    end
  end

  class << self
    def sniffers
      [Utf8, Latin1].map(&:new)
    end

    def detect?(content)
      return true if content.encoding == Encoding::ASCII_8BIT

      sniffers.any? do |sniffer|
        sniffer.detect?(content)
      end
    end

    def repair(content)
      content = Mojibake.auto_encode(content)

      sniffers.each do |sniffer|
        return sniffer.repair(content) if sniffer.detect?(content)
      end

      content
    end

    def auto_encode(mixed)
      return mixed.encode(Encoding::UTF_8) unless mixed.encoding == Encoding::ASCII_8BIT

      [Encoding::UTF_8, Encoding::ISO_8859_1].each do |encoding|
        mixed = mixed.dup.force_encoding(encoding)
        return mixed.encode(Encoding::UTF_8) if mixed.valid_encoding?
      end

      raise "Unable to determine encoding of ASCII_8BIT"
    end
  end
end