[Boost-users] [Spirit] Karma generator for JSON

Subject: [Boost-users] [Spirit] Karma generator for JSON
From: Ovanes Markarian (om_boost_at_[hidden])
Date: 2012-03-09 13:47:36

Hello Hartmut,

as discussed @irc channel I am posting the example of my problem (I use
boost 1.47):

#include <boost/config/warning_disable.hpp>
#include <boost/spirit/include/karma.hpp>
#include <boost/spirit/include/karma_char_class.hpp>

#include <boost/fusion/include/std_pair.hpp>

#include <iterator>
#include <cassert>
#include <utility>
#include <string>

namespace karma = boost::spirit::karma;

typedef std::pair<unsigned char, unsigned char> utf16_t;

template<class OutIter>
struct json_char_encoder2
  : boost::spirit::karma::grammar<OutIter, std::string()>
{
  json_char_encoder2()
    : json_char_encoder2::base_type(encoded_)
  {
    json_sym_.add
      ('"', "\\\"")
      ('/', "\\/")
      ('\\', "\\\\")
      ('\b', "\\b")
      ('\f', "\\f")
      ('\n', "\\n")
      ('\r', "\\r")
      ('\t', "\\t")
    ;

    using boost::spirit::iso8859_1::print;

    encoded_=*(json_sym_ | print | uchar_);
    uchar_ = karma::lit("\\u") << karma::hex << karma::hex;  //also tried
to use unicode_gen_;
  }

  karma::rule<OutIter, std::string()> encoded_;
  karma::symbols<char, char const*> json_sym_;
  karma::rule<OutIter, utf16_t()> uchar_;
  karma::uint_generator<unsigned short, 16> unicode_gen_;
};

int main()
{
  assert(2==sizeof(short));

  typedef std::back_insert_iterator<std::string> iter_t;
  json_char_encoder2<iter_t> encoder2;

  std::string generated;
  iter_t sink(generated);

  std::string sinput="abc";
  sinput.push_back(char(0xAA));
  sinput.push_back(char(0xBB));
  bool result = karma::generate(sink, encoder2, sinput);

  return 0;
}

The problem here is that I would like the generator in case of uchar_ to
retrieve 2 subsequent string bytes and write them out as \uXXXX as
specified at JSON.org. What I get is:
a. IF: replacing utf16_t with unsigned short => and removing 1 of the
karma::hex terminals => 2 calls where each byte is converted to unsigned
short
b. IF: having the source as it is => fails the generation, where the
generated string is empty.

One strange behavior as well: result is true for case b. Why?

Regards,
Ovanes