getsentry/raven-ruby

module Raven
  class Processor::UTF8Conversion < Processor
    # Slightly misnamed - actually just removes any bytes with invalid encoding
    # Previously, our JSON backend required UTF-8. Since we now use the built-in
    # JSON, we can use any encoding, but it must be valid anyway so we can do
    # things like call #match and #slice on strings
    REPLACE = "".freeze

    def process(value)
      case value
      when Hash
        !value.frozen? ? value.merge!(value) { |_, v| process v } : value.merge(value) { |_, v| process v }
      when Array
        !value.frozen? ? value.map! { |v| process v } : value.map { |v| process v }
      when Exception
        return value if value.message.valid_encoding?

        clean_exc = value.class.new(remove_invalid_bytes(value.message))
        clean_exc.set_backtrace(value.backtrace)
        clean_exc
      when String
        # Encoding::BINARY / Encoding::ASCII_8BIT is a special binary encoding.
        # valid_encoding? will always return true because it contains all codepoints,
        # so instead we check if it only contains actual ASCII codepoints, and if
        # not we assume it's actually just UTF8 and scrub accordingly.
        if value.encoding == Encoding::BINARY && !value.ascii_only?
          value = value.dup
          value.force_encoding(Encoding::UTF_8)
        end
        return value if value.valid_encoding?

        remove_invalid_bytes(value)
      else
        value
      end
    end

    private

    # Stolen from RSpec
    # https://github.com/rspec/rspec-support/blob/f0af3fd74a94ff7bb700f6ba06dbdc67bba17fbf/lib/rspec/support/encoded_string.rb#L120-L139
    if String.method_defined?(:scrub) # 2.1+
      def remove_invalid_bytes(string)
        string.scrub(REPLACE)
      end
    else
      def remove_invalid_bytes(string)
        string.chars.map do |char|
          char.valid_encoding? ? char : REPLACE
        end.join
      end
    end
  end
end