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;
23 using lyx::ucs2_codeset;
30 char const * utf8_codeset = "UTF-8";
32 // We use C IO throughout this file, because the facets might be used with
33 // lyxerr in the future.
36 /// codecvt facet for conversion of UCS4 (internal representation) to UTF8
37 /// (external representation) or vice versa
38 class iconv_codecvt_facet : public std::codecvt<lyx::char_type, char, std::mbstate_t>
40 typedef std::codecvt<lyx::char_type, char, std::mbstate_t> base;
42 /// Constructor. You have to specify with \p inout whether you want
43 /// to use this facet only for input, only for output or for both.
44 explicit iconv_codecvt_facet(string const & encoding = "UTF-8",
45 std::ios_base::openmode inout = std::ios_base::in | std::ios_base::out,
47 : base(refs), utf8_(encoding == "UTF-8")
49 if (inout & std::ios_base::in) {
50 in_cd_ = iconv_open(ucs4_codeset, encoding.c_str());
51 if (in_cd_ == (iconv_t)(-1)) {
52 fprintf(stderr, "Error %d returned from iconv_open(in_cd_): %s\n",
53 errno, strerror(errno));
55 throw lyx::iconv_codecvt_facet_exception();
58 in_cd_ = (iconv_t)(-1);
59 if (inout & std::ios_base::out) {
60 out_cd_ = iconv_open(encoding.c_str(), ucs4_codeset);
61 if (out_cd_ == (iconv_t)(-1)) {
62 fprintf(stderr, "Error %d returned from iconv_open(out_cd_): %s\n",
63 errno, strerror(errno));
65 throw lyx::iconv_codecvt_facet_exception();
68 out_cd_ = (iconv_t)(-1);
71 virtual ~iconv_codecvt_facet()
73 if (in_cd_ != (iconv_t)(-1))
74 if (iconv_close(in_cd_) == -1) {
75 fprintf(stderr, "Error %d returned from iconv_close(in_cd_): %s\n",
76 errno, strerror(errno));
79 if (out_cd_ != (iconv_t)(-1))
80 if (iconv_close(out_cd_) == -1) {
81 fprintf(stderr, "Error %d returned from iconv_close(out_cd_): %s\n",
82 errno, strerror(errno));
86 virtual result do_out(state_type &, intern_type const * from,
87 intern_type const * from_end, intern_type const *& from_next,
88 extern_type * to, extern_type * to_end,
89 extern_type *& to_next) const
91 size_t inbytesleft = (from_end - from) * sizeof(intern_type);
92 size_t outbytesleft = (to_end - to) * sizeof(extern_type);
95 return do_iconv(out_cd_, reinterpret_cast<char const **>(&from_next),
96 &inbytesleft, &to_next, &outbytesleft);
98 virtual result do_unshift(state_type &, extern_type * to,
99 extern_type *, extern_type *& to_next) const
101 // utf8 does not use shifting
105 virtual result do_in(state_type &,
106 extern_type const * from, extern_type const * from_end,
107 extern_type const *& from_next,
108 intern_type * to, intern_type * to_end,
109 intern_type *& to_next) const
111 size_t inbytesleft = (from_end - from) * sizeof(extern_type);
112 size_t outbytesleft = (to_end - to) * sizeof(intern_type);
115 return do_iconv(in_cd_, &from_next, &inbytesleft,
116 reinterpret_cast<char **>(&to_next),
119 virtual int do_encoding() const throw()
123 virtual bool do_always_noconv() const throw()
127 virtual int do_length(state_type & /*state*/, extern_type const * from,
128 extern_type const * end, size_t max) const
130 // The docs are a bit unclear about this method.
131 // It seems that we should calculate the actual length of the
132 // converted sequence, but that would not make sense, since
133 // once could just do the conversion directly.
134 // Therefore we just return the number of unconverted
135 // characters, since that is the best guess we can do.
137 intern_type * to = new intern_type[max];
138 intern_type * to_end = to + max;
139 intern_type * to_next = to;
140 extern_type const * from_next = from;
141 do_in(state, from, end, from_next, to, to_end, to_next);
145 size_t const length = end - from;
146 return std::min(length, max);
149 virtual int do_max_length() const throw()
151 // UTF8 uses at most 6 bytes to represent one UCS4 code point.
152 // All other encodings encode one UCS4 code point in one byte
153 // (and can therefore only encode a subset of UCS4)
154 return utf8_ ? 6 : 1;
157 /// Do the actual conversion. The interface is equivalent to that of
158 /// iconv() (but const correct).
159 inline base::result do_iconv(iconv_t cd, char const ** from,
160 size_t * inbytesleft, char ** to, size_t * outbytesleft) const
162 char const * to_start = *to;
163 size_t converted = iconv(cd, const_cast<char ICONV_CONST **>(from),
164 inbytesleft, to, outbytesleft);
165 if (converted == (size_t)(-1)) {
169 return base::partial;
172 fprintf(stderr, "Error %d returned from iconv: %s\n",
173 errno, strerror(errno));
184 /// Is the narrow encoding UTF8?
194 const char * iconv_codecvt_facet_exception::what() const throw()
196 return "iconv problem in iconv_codecvt_facet initialization";
200 idocfstream::idocfstream() : base()
203 std::locale locale(global, new iconv_codecvt_facet(utf8_codeset, in));
208 idocfstream::idocfstream(const char* s, std::ios_base::openmode mode)
211 // We must imbue the stream before openening the file
213 std::locale locale(global, new iconv_codecvt_facet(utf8_codeset, in));
219 odocfstream::odocfstream(string const & encoding) : base()
222 std::locale locale(global, new iconv_codecvt_facet(encoding, out));
227 odocfstream::odocfstream(const char* s, std::ios_base::openmode mode,
228 string const & encoding)
231 // We must imbue the stream before openening the file
233 std::locale locale(global, new iconv_codecvt_facet(encoding, out));
240 #if (!defined(HAVE_WCHAR_T) || SIZEOF_WCHAR_T != 4) && defined(__GNUC__)
241 // We get undefined references to these virtual methods. This looks like
242 // a bug in gcc. The implementation here does not do anything useful, since
243 // it is overriden in iconv_codecvt_facet.
245 template<> codecvt<lyx::char_type, char, mbstate_t>::result
246 codecvt<lyx::char_type, char, mbstate_t>::do_out(mbstate_t &, const lyx::char_type *, const lyx::char_type *, const lyx::char_type *&,
247 char *, char *, char *&) const { return error; }
248 template<> codecvt<lyx::char_type, char, mbstate_t>::result
249 codecvt<lyx::char_type, char, mbstate_t>::do_unshift(mbstate_t &, 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_in(mbstate_t &, const char *, const char *, const char *&,
252 lyx::char_type *, lyx::char_type *, lyx::char_type *&) const { return error; }
253 template<> int codecvt<lyx::char_type, char, mbstate_t>::do_encoding() const throw() { return 0; }
254 template<> bool codecvt<lyx::char_type, char, mbstate_t>::do_always_noconv() const throw() { return true; }
255 #if __GNUC__ == 3 && __GNUC_MINOR__ < 4
256 template<> int codecvt<lyx::char_type, char, mbstate_t>::do_length(mbstate_t const &, const char *, const char *, size_t) const { return 1; }
258 template<> int codecvt<lyx::char_type, char, mbstate_t>::do_length(mbstate_t &, const char *, const char *, size_t) const { return 1; }
260 template<> int codecvt<lyx::char_type, char, mbstate_t>::do_max_length() const throw() { return 4; }