Chris@16: // Copyright (c) 2001 Ronald Garcia, Indiana University (garcia@osl.iu.edu)
Chris@16: // Andrew Lumsdaine, Indiana University (lums@osl.iu.edu).
Chris@16: // Distributed under the Boost Software License, Version 1.0. (See accompany-
Chris@16: // ing file LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
Chris@16: 
Chris@16: #ifndef BOOST_UTF8_CODECVT_FACET_HPP
Chris@16: #define BOOST_UTF8_CODECVT_FACET_HPP
Chris@16: 
Chris@16: // MS compatible compilers support #pragma once
Chris@16: #if defined(_MSC_VER) && (_MSC_VER >= 1020)
Chris@16: # pragma once
Chris@16: #endif
Chris@16: 
Chris@16: /////////1/////////2/////////3/////////4/////////5/////////6/////////7/////////8
Chris@16: // utf8_codecvt_facet.hpp
Chris@16: 
Chris@101: // This header defines class utf8_codecvt_facet, derived from 
Chris@16: // std::codecvt<wchar_t, char>, which can be used to convert utf8 data in
Chris@16: // files into wchar_t strings in the application.
Chris@16: //
Chris@16: // The header is NOT STANDALONE, and is not to be included by the USER.
Chris@16: // There are at least two libraries which want to use this functionality, and
Chris@16: // we want to avoid code duplication. It would be possible to create utf8
Chris@16: // library, but:
Chris@16: // - this requires review process first
Chris@16: // - in the case, when linking the a library which uses utf8 
Chris@16: //   (say 'program_options'), user should also link to the utf8 library.
Chris@16: //   This seems inconvenient, and asking a user to link to an unrevieved 
Chris@16: //   library is strange. 
Chris@16: // Until the above points are fixed, a library which wants to use utf8 must:
Chris@101: // - include this header in one of it's headers or sources
Chris@101: // - include the corresponding boost/detail/utf8_codecvt_facet.ipp file in one
Chris@101: //   of its sources
Chris@16: // - before including either file, the library must define
Chris@16: //   - BOOST_UTF8_BEGIN_NAMESPACE to the namespace declaration that must be used
Chris@16: //   - BOOST_UTF8_END_NAMESPACE to the code to close the previous namespace
Chris@101: //     declaration.
Chris@16: //   - BOOST_UTF8_DECL -- to the code which must be used for all 'exportable'
Chris@16: //     symbols.
Chris@16: //
Chris@16: // For example, program_options library might contain:
Chris@16: //    #define BOOST_UTF8_BEGIN_NAMESPACE <backslash character> 
Chris@16: //             namespace boost { namespace program_options {
Chris@16: //    #define BOOST_UTF8_END_NAMESPACE }}
Chris@16: //    #define BOOST_UTF8_DECL BOOST_PROGRAM_OPTIONS_DECL
Chris@101: //    #include <boost/detail/utf8_codecvt_facet.ipp>
Chris@16: //
Chris@16: // Essentially, each library will have its own copy of utf8 code, in
Chris@16: // different namespaces. 
Chris@16: 
Chris@16: // Note:(Robert Ramey).  I have made the following alterations in the original
Chris@16: // code.
Chris@16: // a) Rendered utf8_codecvt<wchar_t, char>  with using templates
Chris@16: // b) Move longer functions outside class definition to prevent inlining
Chris@16: // and make code smaller
Chris@16: // c) added on a derived class to permit translation to/from current
Chris@16: // locale to utf8
Chris@16: 
Chris@16: //  See http://www.boost.org for updates, documentation, and revision history.
Chris@16: 
Chris@16: // archives stored as text - note these ar templated on the basic
Chris@16: // stream templates to accommodate wide (and other?) kind of characters
Chris@16: //
Chris@16: // note the fact that on libraries without wide characters, ostream is
Chris@16: // is not a specialization of basic_ostream which in fact is not defined
Chris@16: // in such cases.   So we can't use basic_ostream<OStream::char_type> but rather
Chris@16: // use two template parameters
Chris@16: //
Chris@16: // utf8_codecvt_facet
Chris@16: //   This is an implementation of a std::codecvt facet for translating 
Chris@16: //   from UTF-8 externally to UCS-4.  Note that this is not tied to
Chris@16: //   any specific types in order to allow customization on platforms
Chris@16: //   where wchar_t is not big enough.
Chris@16: //
Chris@16: // NOTES:  The current implementation jumps through some unpleasant hoops in
Chris@16: // order to deal with signed character types.  As a std::codecvt_base::result,
Chris@16: // it is necessary  for the ExternType to be convertible to unsigned  char.
Chris@16: // I chose not to tie the extern_type explicitly to char. But if any combination
Chris@16: // of types other than <wchar_t,char_t> is used, then std::codecvt must be
Chris@16: // specialized on those types for this to work.
Chris@16: 
Chris@16: #include <locale>
Chris@16: #include <cwchar>   // for mbstate_t
Chris@16: #include <cstddef>  // for std::size_t
Chris@16: 
Chris@16: #include <boost/config.hpp>
Chris@16: #include <boost/detail/workaround.hpp>
Chris@16: 
Chris@16: #if defined(BOOST_NO_STDC_NAMESPACE)
Chris@16: namespace std {
Chris@16:     using ::mbstate_t;
Chris@16:     using ::size_t;
Chris@16: }
Chris@16: #endif
Chris@16: 
Chris@16: // maximum lenght of a multibyte string
Chris@16: #define MB_LENGTH_MAX 8
Chris@16: 
Chris@16: BOOST_UTF8_BEGIN_NAMESPACE
Chris@16: 
Chris@101: //----------------------------------------------------------------------------//
Chris@101: //                                                                            //
Chris@101: //                          utf8_codecvt_facet                                //
Chris@101: //                                                                            //
Chris@101: //            See utf8_codecvt_facet.ipp for the implementation.              //
Chris@101: //----------------------------------------------------------------------------//
Chris@101: 
Chris@101: 
Chris@16: struct BOOST_UTF8_DECL utf8_codecvt_facet :
Chris@16:     public std::codecvt<wchar_t, char, std::mbstate_t>  
Chris@16: {
Chris@16: public:
Chris@16:     explicit utf8_codecvt_facet(std::size_t no_locale_manage=0)
Chris@16:         : std::codecvt<wchar_t, char, std::mbstate_t>(no_locale_manage) 
Chris@16:     {}
Chris@16: protected:
Chris@16:     virtual std::codecvt_base::result do_in(
Chris@16:         std::mbstate_t& state, 
Chris@16:         const char * from,
Chris@16:         const char * from_end, 
Chris@16:         const char * & from_next,
Chris@16:         wchar_t * to, 
Chris@16:         wchar_t * to_end, 
Chris@16:         wchar_t*& to_next
Chris@16:     ) const;
Chris@16: 
Chris@16:     virtual std::codecvt_base::result do_out(
Chris@101:         std::mbstate_t & state,
Chris@101:         const wchar_t * from,
Chris@101:         const wchar_t * from_end,
Chris@101:         const wchar_t*  & from_next,
Chris@101:         char * to,
Chris@101:         char * to_end,
Chris@101:         char * & to_next
Chris@16:     ) const;
Chris@16: 
Chris@16:     bool invalid_continuing_octet(unsigned char octet_1) const {
Chris@16:         return (octet_1 < 0x80|| 0xbf< octet_1);
Chris@16:     }
Chris@16: 
Chris@16:     bool invalid_leading_octet(unsigned char octet_1)   const {
Chris@16:         return (0x7f < octet_1 && octet_1 < 0xc0) ||
Chris@16:             (octet_1 > 0xfd);
Chris@16:     }
Chris@16: 
Chris@16:     // continuing octets = octets except for the leading octet
Chris@101:     static unsigned int get_cont_octet_count(unsigned char lead_octet) {
Chris@16:         return get_octet_count(lead_octet) - 1;
Chris@16:     }
Chris@16: 
Chris@101:     static unsigned int get_octet_count(unsigned char lead_octet);
Chris@16: 
Chris@16:     // How many "continuing octets" will be needed for this word
Chris@16:     // ==   total octets - 1.
Chris@16:     int get_cont_octet_out_count(wchar_t word) const ;
Chris@16: 
Chris@101:     virtual bool do_always_noconv() const BOOST_NOEXCEPT_OR_NOTHROW {
Chris@101:         return false;
Chris@101:     }
Chris@16: 
Chris@16:     // UTF-8 isn't really stateful since we rewind on partial conversions
Chris@16:     virtual std::codecvt_base::result do_unshift(
Chris@16:         std::mbstate_t&,
Chris@16:         char * from,
Chris@16:         char * /*to*/,
Chris@16:         char * & next
Chris@101:     ) const {
Chris@16:         next = from;
Chris@16:         return ok;
Chris@16:     }
Chris@16: 
Chris@101:     virtual int do_encoding() const BOOST_NOEXCEPT_OR_NOTHROW {
Chris@16:         const int variable_byte_external_encoding=0;
Chris@16:         return variable_byte_external_encoding;
Chris@16:     }
Chris@16: 
Chris@16:     // How many char objects can I process to get <= max_limit
Chris@16:     // wchar_t objects?
Chris@16:     virtual int do_length(
Chris@101:         const std::mbstate_t &,
Chris@16:         const char * from,
Chris@16:         const char * from_end, 
Chris@16:         std::size_t max_limit
Chris@101:     ) const
Chris@16: #if BOOST_WORKAROUND(__IBMCPP__, BOOST_TESTED_AT(600))
Chris@101:     throw()
Chris@16: #endif
Chris@101:     ;
Chris@101:     virtual int do_length(
Chris@101:         std::mbstate_t & s,
Chris@101:         const char * from,
Chris@101:         const char * from_end, 
Chris@101:         std::size_t max_limit
Chris@101:     ) const
Chris@101: #if BOOST_WORKAROUND(__IBMCPP__, BOOST_TESTED_AT(600))
Chris@101:     throw()
Chris@101: #endif
Chris@101:     {
Chris@101:         return do_length(
Chris@101:             const_cast<const std::mbstate_t &>(s),
Chris@101:             from,
Chris@101:             from_end,
Chris@101:             max_limit
Chris@101:         );
Chris@101:     }
Chris@16:     // Largest possible value do_length(state,from,from_end,1) could return.
Chris@101:     virtual int do_max_length() const BOOST_NOEXCEPT_OR_NOTHROW {
Chris@16:         return 6; // largest UTF-8 encoding of a UCS-4 character
Chris@16:     }
Chris@16: };
Chris@16: 
Chris@16: BOOST_UTF8_END_NAMESPACE
Chris@16: 
Chris@16: #endif // BOOST_UTF8_CODECVT_FACET_HPP