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 "support/docstream.h"
14 #include "support/unicode.h"
25 using lyx::ucs4_codeset;
28 #if defined(_MSC_VER) && (_MSC_VER >= 1600)
29 std::locale::id numpunct<lyx::char_type>::id;
35 // We use C IO throughout this file, because the facets might be used with
36 // lyxerr in the future.
39 /// codecvt facet for conversion of UCS4 (internal representation) to UTF8
40 /// (external representation) or vice versa
41 class iconv_codecvt_facet : public codecvt<lyx::char_type, char, mbstate_t>
43 typedef codecvt<lyx::char_type, char, mbstate_t> base;
45 /// Constructor. You have to specify with \p inout whether you want
46 /// to use this facet only for input, only for output or for both.
47 explicit iconv_codecvt_facet(string const & encoding = "UTF-8",
48 ios_base::openmode inout = ios_base::in | ios_base::out,
50 : base(refs), encoding_(encoding)
52 if (inout & ios_base::in) {
53 in_cd_ = iconv_open(ucs4_codeset, encoding.c_str());
54 if (in_cd_ == (iconv_t)(-1)) {
55 fprintf(stderr, "Error %d returned from iconv_open(in_cd_): %s\n",
56 errno, strerror(errno));
58 throw lyx::iconv_codecvt_facet_exception();
61 in_cd_ = (iconv_t)(-1);
62 if (inout & ios_base::out) {
63 out_cd_ = iconv_open(encoding.c_str(), ucs4_codeset);
64 if (out_cd_ == (iconv_t)(-1)) {
65 fprintf(stderr, "Error %d returned from iconv_open(out_cd_): %s\n",
66 errno, strerror(errno));
68 throw lyx::iconv_codecvt_facet_exception();
71 out_cd_ = (iconv_t)(-1);
74 virtual ~iconv_codecvt_facet()
76 if (in_cd_ != (iconv_t)(-1))
77 if (iconv_close(in_cd_) == -1) {
78 fprintf(stderr, "Error %d returned from iconv_close(in_cd_): %s\n",
79 errno, strerror(errno));
82 if (out_cd_ != (iconv_t)(-1))
83 if (iconv_close(out_cd_) == -1) {
84 fprintf(stderr, "Error %d returned from iconv_close(out_cd_): %s\n",
85 errno, strerror(errno));
89 virtual result do_out(state_type &, intern_type const * from,
90 intern_type const * from_end, intern_type const *& from_next,
91 extern_type * to, extern_type * to_end,
92 extern_type *& to_next) const
94 #define WORKAROUND_ICONV_BUG 1
95 #if WORKAROUND_ICONV_BUG
96 // Due to a bug in some iconv versions, when the last char
97 // in the buffer is a wide char and the output encoding is
98 // ISO-2022-JP and we are going to switch to another encoding,
99 // the appropriate escape sequence for changing the character
100 // set is not output (see bugs 5216, 5280, and also 5489).
101 // As a workaround, we append a nul char in order to force
102 // a switch to ASCII, and then remove it from output after
104 intern_type * from_new = 0;
105 intern_type const * from_old = from;
107 if (*(from_end - 1) >= 0x80 && encoding_ == "ISO-2022-JP") {
108 size_t len = from_end - from;
109 from_new = new intern_type[len + 1];
110 memcpy(from_new, from, len * sizeof(intern_type));
112 from_end = from_new + len + 1;
117 size_t inbytesleft = (from_end - from) * sizeof(intern_type);
118 size_t outbytesleft = (to_end - to) * sizeof(extern_type);
119 #if WORKAROUND_ICONV_BUG
120 outbytesleft += extra * sizeof(extern_type);
124 result const retval = do_iconv(out_cd_,
125 reinterpret_cast<char const **>(&from_next),
126 &inbytesleft, &to_next, &outbytesleft);
127 #if WORKAROUND_ICONV_BUG
128 // Remove from output the nul char that we inserted at the end
129 // of the input buffer in order to circumvent an iconv bug.
133 from_next = from_old + (from_next - from);
138 if (retval == base::error) {
140 "Error %d returned from iconv when converting from %s to %s: %s\n",
141 errno, ucs4_codeset, encoding_.c_str(),
143 fputs("Converted input:", stderr);
144 for (intern_type const * i = from; i < from_next; ++i) {
145 unsigned int const c = *i;
146 fprintf(stderr, " 0x%04x", c);
148 unsigned int const c = *from_next;
149 fprintf(stderr, "\nStopped at: 0x%04x\n", c);
150 fputs("Unconverted input:", stderr);
151 for (intern_type const * i = from_next + 1; i < from_end; ++i) {
152 unsigned int const c = *i;
153 fprintf(stderr, " 0x%04x", c);
155 fputs("\nConverted output:", stderr);
156 for (extern_type const * i = to; i < to_next; ++i) {
157 // extern_type may be signed, avoid output of
158 // something like 0xffffffc2
159 unsigned int const c =
160 *reinterpret_cast<unsigned char const *>(i);
161 fprintf(stderr, " 0x%02x", c);
168 virtual result do_unshift(state_type &, extern_type * to,
169 extern_type *, extern_type *& to_next) const
171 // utf8 does not use shifting
175 virtual result do_in(state_type &,
176 extern_type const * from, extern_type const * from_end,
177 extern_type const *& from_next,
178 intern_type * to, intern_type * to_end,
179 intern_type *& to_next) const
181 size_t inbytesleft = (from_end - from) * sizeof(extern_type);
182 size_t outbytesleft = (to_end - to) * sizeof(intern_type);
185 result const retval = do_iconv(in_cd_, &from_next, &inbytesleft,
186 reinterpret_cast<char **>(&to_next),
188 if (retval == base::error) {
190 "Error %d returned from iconv when converting from %s to %s: %s\n",
191 errno, encoding_.c_str(), ucs4_codeset,
193 fputs("Converted input:", stderr);
194 for (extern_type const * i = from; i < from_next; ++i) {
195 // extern_type may be signed, avoid output of
196 // something like 0xffffffc2
197 unsigned int const c =
198 *reinterpret_cast<unsigned char const *>(i);
199 fprintf(stderr, " 0x%02x", c);
201 unsigned int const c =
202 *reinterpret_cast<unsigned char const *>(from_next);
203 fprintf(stderr, "\nStopped at: 0x%02x\n", c);
204 fputs("Unconverted input:", stderr);
205 for (extern_type const * i = from_next + 1; i < from_end; ++i) {
206 unsigned int const c =
207 *reinterpret_cast<unsigned char const *>(i);
208 fprintf(stderr, " 0x%02x", c);
210 fputs("\nConverted output:", stderr);
211 for (intern_type const * i = to; i < to_next; ++i) {
212 unsigned int const c = *i;
213 fprintf(stderr, " 0x%02x", c);
220 virtual int do_encoding() const throw()
224 virtual bool do_always_noconv() const throw()
228 virtual int do_length(state_type & /*state*/, extern_type const * from,
229 extern_type const * end, size_t max) const
231 // The docs are a bit unclear about this method.
232 // It seems that we should calculate the actual length of the
233 // converted sequence, but that would not make sense, since
234 // once could just do the conversion directly.
235 // Therefore we just return the number of unconverted
236 // characters, since that is the best guess we can do.
238 intern_type * to = new intern_type[max];
239 intern_type * to_end = to + max;
240 intern_type * to_next = to;
241 extern_type const * from_next = from;
242 do_in(state, from, end, from_next, to, to_end, to_next);
246 size_t const length = end - from;
247 return min(length, max);
250 virtual int do_max_length() const throw()
252 return lyx::max_encoded_bytes(encoding_);
255 /// Do the actual conversion. The interface is equivalent to that of
256 /// iconv() (but const correct).
257 inline base::result do_iconv(iconv_t cd, char const ** from,
258 size_t * inbytesleft, char ** to, size_t * outbytesleft) const
260 char const * const to_start = *to;
261 size_t converted = iconv(cd, const_cast<char ICONV_CONST **>(from),
262 inbytesleft, to, outbytesleft);
263 if (converted == (size_t)(-1)) {
266 // As strange as it may seem, this
267 // does happen on windows when parsing
268 // comments with accented chars in
269 // tex2lyx. See the following thread
271 // http://thread.gmane.org/gmane.editors.lyx.devel/117636
275 return base::partial;
287 /// The narrow encoding
297 void setEncoding(Ios & ios, string const & encoding, ios_base::openmode mode)
299 // We must imbue the stream before openening the file
301 locale locale(global, new iconv_codecvt_facet(encoding, mode));
306 const char * iconv_codecvt_facet_exception::what() const throw()
308 return "iconv problem in iconv_codecvt_facet initialization";
312 ifdocstream::ifdocstream() : base()
314 setEncoding(*this, "UTF-8", in);
318 ifdocstream::ifdocstream(const char* s, ios_base::openmode mode,
319 string const & encoding)
322 setEncoding(*this, encoding, in);
327 ofdocstream::ofdocstream(): base()
329 setEncoding(*this, "UTF-8", out);
333 ofdocstream::ofdocstream(const char* s, ios_base::openmode mode,
334 string const & encoding)
337 setEncoding(*this, encoding, out);
342 void ofdocstream::reset(string const & encoding)
344 setEncoding(*this, encoding, out);
349 SetEnc setEncoding(string const & encoding)
351 return SetEnc(encoding);
355 odocstream & operator<<(odocstream & os, SetEnc e)
357 if (has_facet<iconv_codecvt_facet>(os.rdbuf()->getloc())) {
358 // This stream must be a file stream, since we never imbue
359 // any other stream with a locale having a iconv_codecvt_facet.
360 // Flush the stream so that all pending output is written
361 // with the old encoding.
363 locale locale(os.rdbuf()->getloc(),
364 new iconv_codecvt_facet(e.encoding, ios_base::out));
365 // FIXME Does changing the codecvt facet of an open file
366 // stream always work? It does with gcc 4.1, but I have read
367 // somewhere that it does not with MSVC.
368 // What does the standard say?
375 //CHECKME: I just copied the code above, and have no idea whether it
376 //is correct... (JMarc)
377 idocstream & operator<<(idocstream & is, SetEnc e)
379 if (has_facet<iconv_codecvt_facet>(is.rdbuf()->getloc())) {
380 // This stream must be a file stream, since we never imbue
381 // any other stream with a locale having a iconv_codecvt_facet.
382 // Flush the stream so that all pending output is written
383 // with the old encoding.
385 locale locale(is.rdbuf()->getloc(),
386 new iconv_codecvt_facet(e.encoding, ios_base::in));
387 // FIXME Does changing the codecvt facet of an open file
388 // stream always work? It does with gcc 4.1, but I have read
389 // somewhere that it does not with MSVC.
390 // What does the standard say?
397 #if ! defined(USE_WCHAR_T)
398 odocstream & operator<<(odocstream & os, char c)
406 void otexstream::put(char_type const & c)
409 if (!canbreakline_ && c == ' ')
411 protectspace_ = false;
416 canbreakline_ = false;
418 canbreakline_ = true;
423 SafeBreakLine safebreakln;
426 otexstream & operator<<(otexstream & ots, BreakLine)
428 if (ots.canBreakLine()) {
430 ots.canBreakLine(false);
431 ots.texrow().newline();
433 ots.protectSpace(false);
438 otexstream & operator<<(otexstream & ots, SafeBreakLine)
440 if (ots.canBreakLine()) {
442 ots.canBreakLine(false);
443 ots.texrow().newline();
445 ots.protectSpace(false);
450 otexstream & operator<<(otexstream & ots, odocstream_manip pf)
453 if (pf == static_cast<odocstream_manip>(endl))
454 ots.texrow().newline();
459 otexstream & operator<<(otexstream & ots, docstring const & s)
461 size_t const len = s.length();
463 // Check whether there's something to output
467 if (ots.protectSpace()) {
468 if (!ots.canBreakLine() && s[0] == ' ')
470 ots.protectSpace(false);
473 ots.texrow().newlines(count(s.begin(), s.end(), '\n'));
474 ots.canBreakLine(s[len - 1] != '\n');
479 otexstream & operator<<(otexstream & ots, char const * s)
481 size_t const len = strlen(s);
483 // Check whether there's something to output
487 if (ots.protectSpace()) {
488 if (!ots.canBreakLine() && s[0] == ' ')
490 ots.protectSpace(false);
493 ots.texrow().newlines(count(s, s + len, '\n'));
494 ots.canBreakLine(s[len - 1] != '\n');
499 otexstream & operator<<(otexstream & ots, char c)
501 if (ots.protectSpace()) {
502 if (!ots.canBreakLine() && c == ' ')
504 ots.protectSpace(false);
508 ots.texrow().newline();
509 ots.canBreakLine(c != '\n');
514 template <typename Type>
515 otexstream & operator<<(otexstream & ots, Type value)
518 ots.canBreakLine(true);
519 ots.protectSpace(false);
523 template otexstream & operator<< <SetEnc>(otexstream & os, SetEnc);
524 template otexstream & operator<< <double>(otexstream &, double);
525 template otexstream & operator<< <int>(otexstream &, int);
526 template otexstream & operator<< <unsigned int>(otexstream &, unsigned int);
527 template otexstream & operator<< <unsigned long>(otexstream &, unsigned long);
531 #if ! defined(USE_WCHAR_T) && defined(__GNUC__)
532 // We get undefined references to these virtual methods. This looks like
533 // a bug in gcc. The implementation here does not do anything useful, since
534 // it is overriden in iconv_codecvt_facet.
537 template<> codecvt<lyx::char_type, char, mbstate_t>::result
538 codecvt<lyx::char_type, char, mbstate_t>::do_out(
539 mbstate_t &, const lyx::char_type *, const lyx::char_type *,
540 const lyx::char_type *&, char *, char *, char *&) const
546 template<> codecvt<lyx::char_type, char, mbstate_t>::result
547 codecvt<lyx::char_type, char, mbstate_t>::do_unshift(
548 mbstate_t &, char *, char *, char *&) const
554 template<> codecvt<lyx::char_type, char, mbstate_t>::result
555 codecvt<lyx::char_type, char, mbstate_t>::do_in(
556 mbstate_t &, const char *, const char *, const char *&,
557 lyx::char_type *, lyx::char_type *, lyx::char_type *&) const
564 int codecvt<lyx::char_type, char, mbstate_t>::do_encoding() const throw()
571 bool codecvt<lyx::char_type, char, mbstate_t>::do_always_noconv() const throw()
576 #if __GNUC__ == 3 && __GNUC_MINOR__ < 4
579 int codecvt<lyx::char_type, char, mbstate_t>::do_length(
580 mbstate_t const &, const char *, const char *, size_t) const
588 int codecvt<lyx::char_type, char, mbstate_t>::do_length(
589 mbstate_t &, const char *, const char *, size_t) const
597 int codecvt<lyx::char_type, char, mbstate_t>::do_max_length() const throw()