boost/detail/utf8_codecvt_facet.hpp

   1 // Copyright (c) 2001 Ronald Garcia, Indiana University (garcia@osl.iu.edu)
   2 // Andrew Lumsdaine, Indiana University (lums@osl.iu.edu).
   3 // Distributed under the Boost Software License, Version 1.0. (See accompany-
   4 // ing file LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
   5
   6 #ifndef BOOST_UTF8_CODECVT_FACET_HPP
   7 #define BOOST_UTF8_CODECVT_FACET_HPP
   8
   9 // MS compatible compilers support #pragma once
  10 #if defined(_MSC_VER) && (_MSC_VER >= 1020)
  11 # pragma once
  12 #endif
  13
  14 /////////1/////////2/////////3/////////4/////////5/////////6/////////7/////////8
  15 // utf8_codecvt_facet.hpp
  16
  17 // This header defines class utf8_codecvt_facet, derived fro
  18 // std::codecvt<wchar_t, char>, which can be used to convert utf8 data in
  19 // files into wchar_t strings in the application.
  20 //
  21 // The header is NOT STANDALONE, and is not to be included by the USER.
  22 // There are at least two libraries which want to use this functionality, and
  23 // we want to avoid code duplication. It would be possible to create utf8
  24 // library, but:
  25 // - this requires review process first
  26 // - in the case, when linking the a library which uses utf8
  27 //   (say 'program_options'), user should also link to the utf8 library.
  28 //   This seems inconvenient, and asking a user to link to an unrevieved
  29 //   library is strange.
  30 // Until the above points are fixed, a library which wants to use utf8 must:
  31 // - include this header from one of it's headers or sources
  32 // - include the corresponding .cpp file from one of the sources
  33 // - before including either file, the library must define
  34 //   - BOOST_UTF8_BEGIN_NAMESPACE to the namespace declaration that must be used
  35 //   - BOOST_UTF8_END_NAMESPACE to the code to close the previous namespace
  36 //   - declaration.
  37 //   - BOOST_UTF8_DECL -- to the code which must be used for all 'exportable'
  38 //     symbols.
  39 //
  40 // For example, program_options library might contain:
  41 //    #define BOOST_UTF8_BEGIN_NAMESPACE <backslash character>
  42 //             namespace boost { namespace program_options {
  43 //    #define BOOST_UTF8_END_NAMESPACE }}
  44 //    #define BOOST_UTF8_DECL BOOST_PROGRAM_OPTIONS_DECL
  45 //    #include "../../detail/utf8/utf8_codecvt.cpp"
  46 //
  47 // Essentially, each library will have its own copy of utf8 code, in
  48 // different namespaces.
  49
  50 // Note:(Robert Ramey).  I have made the following alterations in the original
  51 // code.
  52 // a) Rendered utf8_codecvt<wchar_t, char>  with using templates
  53 // b) Move longer functions outside class definition to prevent inlining
  54 // and make code smaller
  55 // c) added on a derived class to permit translation to/from current
  56 // locale to utf8
  57
  58 //  See http://www.boost.org for updates, documentation, and revision history.
  59
  60 // archives stored as text - note these ar templated on the basic
  61 // stream templates to accommodate wide (and other?) kind of characters
  62 //
  63 // note the fact that on libraries without wide characters, ostream is
  64 // is not a specialization of basic_ostream which in fact is not defined
  65 // in such cases.   So we can't use basic_ostream<OStream::char_type> but rather
  66 // use two template parameters
  67 //
  68 // utf8_codecvt_facet
  69 //   This is an implementation of a std::codecvt facet for translating
  70 //   from UTF-8 externally to UCS-4.  Note that this is not tied to
  71 //   any specific types in order to allow customization on platforms
  72 //   where wchar_t is not big enough.
  73 //
  74 // NOTES:  The current implementation jumps through some unpleasant hoops in
  75 // order to deal with signed character types.  As a std::codecvt_base::result,
  76 // it is necessary  for the ExternType to be convertible to unsigned  char.
  77 // I chose not to tie the extern_type explicitly to char. But if any combination
  78 // of types other than <wchar_t,char_t> is used, then std::codecvt must be
  79 // specialized on those types for this to work.
  80
  81 #include <locale>
  82 #include <cwchar>   // for mbstate_t
  83 #include <cstddef>  // for std::size_t
  84
  85 #include <boost/config.hpp>
  86 #include <boost/detail/workaround.hpp>
  87
  88 #if defined(BOOST_NO_STDC_NAMESPACE)
  89 namespace std {
  90     using ::mbstate_t;
  91     using ::size_t;
  92 }
  93 #endif
  94
  95 #if !defined(__MSL_CPP__) && !defined(__LIBCOMO__)
  96     #define BOOST_CODECVT_DO_LENGTH_CONST const
  97 #else
  98     #define BOOST_CODECVT_DO_LENGTH_CONST
  99 #endif
 100
 101 // maximum lenght of a multibyte string
 102 #define MB_LENGTH_MAX 8
 103
 104 BOOST_UTF8_BEGIN_NAMESPACE
 105
 106 struct BOOST_UTF8_DECL utf8_codecvt_facet :
 107     public std::codecvt<wchar_t, char, std::mbstate_t>
 108 {
 109 public:
 110     explicit utf8_codecvt_facet(std::size_t no_locale_manage=0)
 111         : std::codecvt<wchar_t, char, std::mbstate_t>(no_locale_manage)
 112     {}
 113 protected:
 114     virtual std::codecvt_base::result do_in(
 115         std::mbstate_t& state,
 116         const char * from,
 117         const char * from_end,
 118         const char * & from_next,
 119         wchar_t * to,
 120         wchar_t * to_end,
 121         wchar_t*& to_next
 122     ) const;
 123
 124     virtual std::codecvt_base::result do_out(
 125         std::mbstate_t & state, const wchar_t * from,
 126         const wchar_t * from_end, const wchar_t*  & from_next,
 127         char * to, char * to_end, char * & to_next
 128     ) const;
 129
 130     bool invalid_continuing_octet(unsigned char octet_1) const {
 131         return (octet_1 < 0x80|| 0xbf< octet_1);
 132     }
 133
 134     bool invalid_leading_octet(unsigned char octet_1)   const {
 135         return (0x7f < octet_1 && octet_1 < 0xc0) ||
 136             (octet_1 > 0xfd);
 137     }
 138
 139     // continuing octets = octets except for the leading octet
 140     static unsigned int get_cont_octet_count(unsigned   char lead_octet) {
 141         return get_octet_count(lead_octet) - 1;
 142     }
 143
 144     static unsigned int get_octet_count(unsigned char   lead_octet);
 145
 146     // How many "continuing octets" will be needed for this word
 147     // ==   total octets - 1.
 148     int get_cont_octet_out_count(wchar_t word) const ;
 149
 150     virtual bool do_always_noconv() const throw() { return false; }
 151
 152     // UTF-8 isn't really stateful since we rewind on partial conversions
 153     virtual std::codecvt_base::result do_unshift(
 154         std::mbstate_t&,
 155         char * from,
 156         char * /*to*/,
 157         char * & next
 158     ) const
 159     {
 160         next = from;
 161         return ok;
 162     }
 163
 164     virtual int do_encoding() const throw() {
 165         const int variable_byte_external_encoding=0;
 166         return variable_byte_external_encoding;
 167     }
 168
 169     // How many char objects can I process to get <= max_limit
 170     // wchar_t objects?
 171     virtual int do_length(
 172         BOOST_CODECVT_DO_LENGTH_CONST std::mbstate_t &,
 173         const char * from,
 174         const char * from_end,
 175         std::size_t max_limit
 176 #if BOOST_WORKAROUND(__IBMCPP__, BOOST_TESTED_AT(600))
 177         ) const throw();
 178 #else
 179         ) const;
 180 #endif
 181
 182     // Largest possible value do_length(state,from,from_end,1) could return.
 183     virtual int do_max_length() const throw () {
 184         return 6; // largest UTF-8 encoding of a UCS-4 character
 185     }
 186 };
 187
 188 BOOST_UTF8_END_NAMESPACE
 189
 190 #endif // BOOST_UTF8_CODECVT_FACET_HPP