3 * This file is part of LyX, the document processor.
4 * Licence details can be found in the file COPYING.
8 * Full author contact details are available in file CREDITS.
13 #include "docstream.h"
22 using lyx::ucs4_codeset;
29 char const * utf8_codeset = "UTF-8";
31 // We use C IO throughout this file, because the facets might be used with
32 // lyxerr in the future.
35 /// codecvt facet for conversion of UCS4 (internal representation) to UTF8
36 /// (external representation) or vice versa
37 class iconv_codecvt_facet : public std::codecvt<lyx::char_type, char, std::mbstate_t>
39 typedef std::codecvt<lyx::char_type, char, std::mbstate_t> base;
41 /// Constructor. You have to specify with \p inout whether you want
42 /// to use this facet only for input, only for output or for both.
43 explicit iconv_codecvt_facet(string const & encoding = "UTF-8",
44 std::ios_base::openmode inout = std::ios_base::in | std::ios_base::out,
46 : base(refs), utf8_(encoding == "UTF-8")
48 if (inout & std::ios_base::in) {
49 in_cd_ = iconv_open(ucs4_codeset, encoding.c_str());
50 if (in_cd_ == (iconv_t)(-1)) {
51 fprintf(stderr, "Error %d returned from iconv_open(in_cd_): %s\n",
52 errno, strerror(errno));
54 throw lyx::iconv_codecvt_facet_exception();
57 in_cd_ = (iconv_t)(-1);
58 if (inout & std::ios_base::out) {
59 out_cd_ = iconv_open(encoding.c_str(), ucs4_codeset);
60 if (out_cd_ == (iconv_t)(-1)) {
61 fprintf(stderr, "Error %d returned from iconv_open(out_cd_): %s\n",
62 errno, strerror(errno));
64 throw lyx::iconv_codecvt_facet_exception();
67 out_cd_ = (iconv_t)(-1);
70 virtual ~iconv_codecvt_facet()
72 if (in_cd_ != (iconv_t)(-1))
73 if (iconv_close(in_cd_) == -1) {
74 fprintf(stderr, "Error %d returned from iconv_close(in_cd_): %s\n",
75 errno, strerror(errno));
78 if (out_cd_ != (iconv_t)(-1))
79 if (iconv_close(out_cd_) == -1) {
80 fprintf(stderr, "Error %d returned from iconv_close(out_cd_): %s\n",
81 errno, strerror(errno));
85 virtual result do_out(state_type &, intern_type const * from,
86 intern_type const * from_end, intern_type const *& from_next,
87 extern_type * to, extern_type * to_end,
88 extern_type *& to_next) const
90 size_t inbytesleft = (from_end - from) * sizeof(intern_type);
91 size_t outbytesleft = (to_end - to) * sizeof(extern_type);
94 return do_iconv(out_cd_, reinterpret_cast<char const **>(&from_next),
95 &inbytesleft, &to_next, &outbytesleft);
97 virtual result do_unshift(state_type &, extern_type * to,
98 extern_type *, extern_type *& to_next) const
100 // utf8 does not use shifting
104 virtual result do_in(state_type &,
105 extern_type const * from, extern_type const * from_end,
106 extern_type const *& from_next,
107 intern_type * to, intern_type * to_end,
108 intern_type *& to_next) const
110 size_t inbytesleft = (from_end - from) * sizeof(extern_type);
111 size_t outbytesleft = (to_end - to) * sizeof(intern_type);
114 return do_iconv(in_cd_, &from_next, &inbytesleft,
115 reinterpret_cast<char **>(&to_next),
118 virtual int do_encoding() const throw()
122 virtual bool do_always_noconv() const throw()
126 virtual int do_length(state_type & /*state*/, extern_type const * from,
127 extern_type const * end, size_t max) const
129 // The docs are a bit unclear about this method.
130 // It seems that we should calculate the actual length of the
131 // converted sequence, but that would not make sense, since
132 // once could just do the conversion directly.
133 // Therefore we just return the number of unconverted
134 // characters, since that is the best guess we can do.
136 intern_type * to = new intern_type[max];
137 intern_type * to_end = to + max;
138 intern_type * to_next = to;
139 extern_type const * from_next = from;
140 do_in(state, from, end, from_next, to, to_end, to_next);
144 size_t const length = end - from;
145 return std::min(length, max);
148 virtual int do_max_length() const throw()
150 // UTF8 uses at most 4 bytes to represent one UCS4 code point
151 // (see RFC 3629). RFC 2279 specifies 6 bytes, but that
152 // information is outdated, and RFC 2279 has been superseded by
154 // All other encodings encode one UCS4 code point in one byte
155 // (and can therefore only encode a subset of UCS4)
156 return utf8_ ? 4 : 1;
159 /// Do the actual conversion. The interface is equivalent to that of
160 /// iconv() (but const correct).
161 inline base::result do_iconv(iconv_t cd, char const ** from,
162 size_t * inbytesleft, char ** to, size_t * outbytesleft) const
164 char const * to_start = *to;
165 size_t converted = iconv(cd, const_cast<char ICONV_CONST **>(from),
166 inbytesleft, to, outbytesleft);
167 if (converted == (size_t)(-1)) {
171 return base::partial;
174 fprintf(stderr, "Error %d returned from iconv: %s\n",
175 errno, strerror(errno));
186 /// Is the narrow encoding UTF8?
196 const char * iconv_codecvt_facet_exception::what() const throw()
198 return "iconv problem in iconv_codecvt_facet initialization";
202 idocfstream::idocfstream() : base()
205 std::locale locale(global, new iconv_codecvt_facet(utf8_codeset, in));
210 idocfstream::idocfstream(const char* s, std::ios_base::openmode mode)
213 // We must imbue the stream before openening the file
215 std::locale locale(global, new iconv_codecvt_facet(utf8_codeset, in));
221 odocfstream::odocfstream(string const & encoding) : base()
224 std::locale locale(global, new iconv_codecvt_facet(encoding, out));
229 odocfstream::odocfstream(const char* s, std::ios_base::openmode mode,
230 string const & encoding)
233 // We must imbue the stream before openening the file
235 std::locale locale(global, new iconv_codecvt_facet(encoding, out));
242 #if (!defined(HAVE_WCHAR_T) || SIZEOF_WCHAR_T != 4) && defined(__GNUC__)
243 // We get undefined references to these virtual methods. This looks like
244 // a bug in gcc. The implementation here does not do anything useful, since
245 // it is overriden in iconv_codecvt_facet.
247 template<> codecvt<lyx::char_type, char, mbstate_t>::result
248 codecvt<lyx::char_type, char, mbstate_t>::do_out(mbstate_t &, const lyx::char_type *, const lyx::char_type *, const lyx::char_type *&,
249 char *, char *, char *&) const { return error; }
250 template<> codecvt<lyx::char_type, char, mbstate_t>::result
251 codecvt<lyx::char_type, char, mbstate_t>::do_unshift(mbstate_t &, char *, char *, char *&) const { return error; }
252 template<> codecvt<lyx::char_type, char, mbstate_t>::result
253 codecvt<lyx::char_type, char, mbstate_t>::do_in(mbstate_t &, const char *, const char *, const char *&,
254 lyx::char_type *, lyx::char_type *, lyx::char_type *&) const { return error; }
255 template<> int codecvt<lyx::char_type, char, mbstate_t>::do_encoding() const throw() { return 0; }
256 template<> bool codecvt<lyx::char_type, char, mbstate_t>::do_always_noconv() const throw() { return true; }
257 #if __GNUC__ == 3 && __GNUC_MINOR__ < 4
258 template<> int codecvt<lyx::char_type, char, mbstate_t>::do_length(mbstate_t const &, const char *, const char *, size_t) const { return 1; }
260 template<> int codecvt<lyx::char_type, char, mbstate_t>::do_length(mbstate_t &, const char *, const char *, size_t) const { return 1; }
262 template<> int codecvt<lyx::char_type, char, mbstate_t>::do_max_length() const throw() { return 4; }