3 * This file is part of LyX, the document processor.
4 * Licence details can be found in the file COPYING.
6 * \author Lars Gullik Bjønnes
8 * Full author contact details are available in file CREDITS.
10 * A collection of unicode conversion functions, using iconv.
15 #include "support/unicode.h"
16 #include "support/debug.h"
17 #include "support/mutex.h"
21 #include <boost/cstdint.hpp>
32 #ifdef WORDS_BIGENDIAN
33 char const * utf16_codeset = "UTF16-BE";
35 char const * utf16_codeset = "UTF16-LE";
43 #ifdef WORDS_BIGENDIAN
44 char const * ucs4_codeset = "UCS-4BE";
46 char const * ucs4_codeset = "UCS-4LE";
49 static const iconv_t invalid_cd = (iconv_t)(-1);
52 struct IconvProcessor::Impl
54 Impl(string const & to, string const & from)
55 : cd(invalid_cd), tocode_(to), fromcode_(from)
60 if (cd != invalid_cd && iconv_close(cd) == -1)
61 LYXERR0("Error returned from iconv_close(" << errno << ")");
68 Mutex mutex_; // iconv() is not thread save, see #7240
72 IconvProcessor::IconvProcessor(char const * tocode, char const * fromcode)
73 : pimpl_(new IconvProcessor::Impl(tocode, fromcode))
78 IconvProcessor::IconvProcessor(IconvProcessor const & other)
79 : pimpl_(new IconvProcessor::Impl(other.pimpl_->tocode_, other.pimpl_->fromcode_))
84 IconvProcessor::~IconvProcessor()
90 void IconvProcessor::operator=(IconvProcessor const & other)
93 pimpl_ = new Impl(other.pimpl_->tocode_, other.pimpl_->fromcode_);
97 bool IconvProcessor::init()
99 if (pimpl_->cd != invalid_cd)
102 pimpl_->cd = iconv_open(pimpl_->tocode_.c_str(), pimpl_->fromcode_.c_str());
103 if (pimpl_->cd != invalid_cd)
106 lyxerr << "Error returned from iconv_open" << endl;
109 lyxerr << "EINVAL The conversion from " << pimpl_->fromcode_
110 << " to " << pimpl_->tocode_
111 << " is not supported by the implementation."
115 lyxerr << "\tSome other error: " << errno << endl;
122 int IconvProcessor::convert(char const * buf, size_t buflen,
123 char * outbuf, size_t maxoutsize)
125 Mutex::Locker lock(&pimpl_->mutex_);
130 if (pimpl_->cd == invalid_cd) {
135 char ICONV_CONST * inbuf = const_cast<char ICONV_CONST *>(buf);
136 size_t inbytesleft = buflen;
137 size_t outbytesleft = maxoutsize;
139 int res = iconv(pimpl_->cd, &inbuf, &inbytesleft, &outbuf, &outbytesleft);
141 // flush out remaining data. This is needed because iconv sometimes
142 // holds back chars in the stream, waiting for a combination character
143 // (see e.g. http://sources.redhat.com/bugzilla/show_bug.cgi?id=1124)
144 iconv(pimpl_->cd, NULL, NULL, &outbuf, &outbytesleft);
147 //lyxerr << "Inbytesleft: " << inbytesleft << endl;
148 //lyxerr << "Outbytesleft: " << outbytesleft << endl;
151 // Everything went well.
152 return maxoutsize - outbytesleft;
154 // There are some errors in the conversion
155 lyxerr << "Error returned from iconv" << endl;
158 lyxerr << "E2BIG There is not sufficient room at *outbuf." << endl;
161 lyxerr << "EILSEQ An invalid multibyte sequence"
162 << " has been encountered in the input.\n"
163 << "When converting from " << pimpl_->fromcode_
164 << " to " << pimpl_->tocode_ << ".\n";
165 lyxerr << "Input:" << hex;
166 for (size_t i = 0; i < buflen; ++i) {
167 // char may be signed, avoid output of
168 // something like 0xffffffc2
169 boost::uint32_t const b =
170 *reinterpret_cast<unsigned char const *>(buf + i);
171 lyxerr << " 0x" << (unsigned int)b;
173 lyxerr << dec << endl;
176 lyxerr << "EINVAL An incomplete multibyte sequence"
177 << " has been encountered in the input.\n"
178 << "When converting from " << pimpl_->fromcode_
179 << " to " << pimpl_->tocode_ << ".\n";
180 lyxerr << "Input:" << hex;
181 for (size_t i = 0; i < buflen; ++i) {
182 // char may be signed, avoid output of
183 // something like 0xffffffc2
184 boost::uint32_t const b =
185 *reinterpret_cast<unsigned char const *>(buf + i);
186 lyxerr << " 0x" << (unsigned int)b;
188 lyxerr << dec << endl;
191 lyxerr << "\tSome other error: " << errno << endl;
194 // We got an error so we close down the conversion engine
195 if (iconv_close(pimpl_->cd) == -1) {
196 lyxerr << "Error returned from iconv_close("
197 << errno << ")" << endl;
199 pimpl_->cd = invalid_cd;
204 std::string IconvProcessor::from() const
206 return pimpl_->fromcode_;
210 std::string IconvProcessor::to() const
212 return pimpl_->tocode_;
219 template<typename RetType, typename InType>
221 iconv_convert(IconvProcessor & processor, InType const * buf, size_t buflen)
224 return vector<RetType>();
226 char const * inbuf = reinterpret_cast<char const *>(buf);
227 size_t inbytesleft = buflen * sizeof(InType);
229 static std::vector<char> outbuf(32768);
230 // The number of UCS4 code points in buf is at most inbytesleft.
231 // The output encoding will use at most
232 // max_encoded_bytes(pimpl_->tocode_) per UCS4 code point.
233 size_t maxoutbufsize = max_encoded_bytes(processor.to()) * inbytesleft;
234 if (outbuf.size() < maxoutbufsize)
235 outbuf.resize(maxoutbufsize);
237 int bytes = processor.convert(inbuf, inbytesleft, &outbuf[0], outbuf.size());
240 // FIXME Maybe throw an exception and handle that in the caller?
241 return vector<RetType>();
243 RetType const * tmp = reinterpret_cast<RetType const *>(&outbuf[0]);
244 return vector<RetType>(tmp, tmp + bytes / sizeof(RetType));
250 vector<char_type> utf8_to_ucs4(vector<char> const & utf8str)
253 return vector<char_type>();
255 return utf8_to_ucs4(&utf8str[0], utf8str.size());
260 utf8_to_ucs4(char const * utf8str, size_t ls)
262 static IconvProcessor processor(ucs4_codeset, "UTF-8");
263 return iconv_convert<char_type>(processor, utf8str, ls);
268 utf16_to_ucs4(unsigned short const * s, size_t ls)
270 static IconvProcessor processor(ucs4_codeset, utf16_codeset);
271 return iconv_convert<char_type>(processor, s, ls);
275 vector<unsigned short>
276 ucs4_to_utf16(char_type const * s, size_t ls)
278 static IconvProcessor processor(utf16_codeset, ucs4_codeset);
279 return iconv_convert<unsigned short>(processor, s, ls);
284 ucs4_to_utf8(char_type c)
286 static IconvProcessor processor("UTF-8", ucs4_codeset);
287 return iconv_convert<char>(processor, &c, 1);
292 ucs4_to_utf8(vector<char_type> const & ucs4str)
295 return vector<char>();
297 return ucs4_to_utf8(&ucs4str[0], ucs4str.size());
302 ucs4_to_utf8(char_type const * ucs4str, size_t ls)
304 static IconvProcessor processor("UTF-8", ucs4_codeset);
305 return iconv_convert<char>(processor, ucs4str, ls);
310 eightbit_to_ucs4(char const * s, size_t ls, string const & encoding)
312 static map<string, IconvProcessor> processors;
313 if (processors.find(encoding) == processors.end()) {
314 IconvProcessor processor(ucs4_codeset, encoding.c_str());
315 processors.insert(make_pair(encoding, processor));
317 return iconv_convert<char_type>(processors[encoding], s, ls);
322 ucs4_to_eightbit(char_type const * ucs4str, size_t ls, string const & encoding)
324 static map<string, IconvProcessor> processors;
325 if (processors.find(encoding) == processors.end()) {
326 IconvProcessor processor(encoding.c_str(), ucs4_codeset);
327 processors.insert(make_pair(encoding, processor));
329 return iconv_convert<char>(processors[encoding], ucs4str, ls);
333 char ucs4_to_eightbit(char_type ucs4, string const & encoding)
335 static map<string, IconvProcessor> processors;
336 map<string, IconvProcessor>::iterator it = processors.find(encoding);
337 if (it == processors.end()) {
338 IconvProcessor processor(encoding.c_str(), ucs4_codeset);
339 it = processors.insert(make_pair(encoding, processor)).first;
343 int const bytes = it->second.convert((char *)(&ucs4), 4, &out, 1);
350 void ucs4_to_multibytes(char_type ucs4, vector<char> & out,
351 string const & encoding)
353 static map<string, IconvProcessor> processors;
354 map<string, IconvProcessor>::iterator it = processors.find(encoding);
355 if (it == processors.end()) {
356 IconvProcessor processor(encoding.c_str(), ucs4_codeset);
357 it = processors.insert(make_pair(encoding, processor)).first;
361 int bytes = it->second.convert((char *)(&ucs4), 4, &out[0], 4);
368 int max_encoded_bytes(std::string const & encoding)
370 // FIXME: this information should be transferred to lib/encodings
371 // UTF8 uses at most 4 bytes to represent one UCS4 code point
372 // (see RFC 3629). RFC 2279 specifies 6 bytes, but that
373 // information is outdated, and RFC 2279 has been superseded by
375 // The CJK encodings use (different) multibyte representation as well.
376 // All other encodings encode one UCS4 code point in one byte
377 // (and can therefore only encode a subset of UCS4)
378 // Note that BIG5 and SJIS do not work with LaTeX (see lib/encodings).
379 // Furthermore, all encodings that use shifting (like SJIS) do not work with
380 // iconv_codecvt_facet.
381 if (encoding == "UTF-8" ||
383 encoding == "EUC-TW")
385 else if (encoding == "EUC-JP")
387 else if (encoding == "ISO-2022-JP")
389 else if (encoding == "BIG5" ||
390 encoding == "EUC-KR" ||
391 encoding == "EUC-CN" ||
392 encoding == "SJIS" ||