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"
20 #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 << ")");
70 IconvProcessor::IconvProcessor(char const * tocode, char const * fromcode)
71 : pimpl_(new IconvProcessor::Impl(tocode, fromcode))
76 IconvProcessor::IconvProcessor(IconvProcessor const & other)
77 : pimpl_(new IconvProcessor::Impl(other.pimpl_->tocode_, other.pimpl_->fromcode_))
82 IconvProcessor::~IconvProcessor()
88 void IconvProcessor::operator=(IconvProcessor const & other)
91 pimpl_ = new Impl(other.pimpl_->tocode_, other.pimpl_->fromcode_);
95 bool IconvProcessor::init()
97 if (pimpl_->cd != invalid_cd)
100 pimpl_->cd = iconv_open(pimpl_->tocode_.c_str(), pimpl_->fromcode_.c_str());
101 if (pimpl_->cd != invalid_cd)
104 lyxerr << "Error returned from iconv_open" << endl;
107 lyxerr << "EINVAL The conversion from " << pimpl_->fromcode_
108 << " to " << pimpl_->tocode_
109 << " is not supported by the implementation."
113 lyxerr << "\tSome other error: " << errno << endl;
120 int IconvProcessor::convert(char const * buf, size_t buflen,
121 char * outbuf, size_t maxoutsize)
126 if (pimpl_->cd == invalid_cd) {
131 char ICONV_CONST * inbuf = const_cast<char ICONV_CONST *>(buf);
132 size_t inbytesleft = buflen;
133 size_t outbytesleft = maxoutsize;
135 int res = iconv(pimpl_->cd, &inbuf, &inbytesleft, &outbuf, &outbytesleft);
137 // flush out remaining data. This is needed because iconv sometimes
138 // holds back chars in the stream, waiting for a combination character
139 // (see e.g. http://sources.redhat.com/bugzilla/show_bug.cgi?id=1124)
140 iconv(pimpl_->cd, NULL, NULL, &outbuf, &outbytesleft);
143 //lyxerr << "Inbytesleft: " << inbytesleft << endl;
144 //lyxerr << "Outbytesleft: " << outbytesleft << endl;
147 // Everything went well.
148 return maxoutsize - outbytesleft;
150 // There are some errors in the conversion
151 lyxerr << "Error returned from iconv" << endl;
154 lyxerr << "E2BIG There is not sufficient room at *outbuf." << endl;
157 lyxerr << "EILSEQ An invalid multibyte sequence"
158 << " has been encountered in the input.\n"
159 << "When converting from " << pimpl_->fromcode_
160 << " to " << pimpl_->tocode_ << ".\n";
161 lyxerr << "Input:" << hex;
162 for (size_t i = 0; i < buflen; ++i) {
163 // char may be signed, avoid output of
164 // something like 0xffffffc2
165 boost::uint32_t const b =
166 *reinterpret_cast<unsigned char const *>(buf + i);
167 lyxerr << " 0x" << (unsigned int)b;
169 lyxerr << dec << endl;
172 lyxerr << "EINVAL An incomplete multibyte sequence"
173 << " has been encountered in the input.\n"
174 << "When converting from " << pimpl_->fromcode_
175 << " to " << pimpl_->tocode_ << ".\n";
176 lyxerr << "Input:" << hex;
177 for (size_t i = 0; i < buflen; ++i) {
178 // char may be signed, avoid output of
179 // something like 0xffffffc2
180 boost::uint32_t const b =
181 *reinterpret_cast<unsigned char const *>(buf + i);
182 lyxerr << " 0x" << (unsigned int)b;
184 lyxerr << dec << endl;
187 lyxerr << "\tSome other error: " << errno << endl;
190 // We got an error so we close down the conversion engine
191 if (iconv_close(pimpl_->cd) == -1) {
192 lyxerr << "Error returned from iconv_close("
193 << errno << ")" << endl;
195 pimpl_->cd = invalid_cd;
200 std::string IconvProcessor::from() const
202 return pimpl_->fromcode_;
206 std::string IconvProcessor::to() const
208 return pimpl_->tocode_;
215 template<typename RetType, typename InType>
217 iconv_convert(IconvProcessor & processor, InType const * buf, size_t buflen)
220 return vector<RetType>();
222 char const * inbuf = reinterpret_cast<char const *>(buf);
223 size_t inbytesleft = buflen * sizeof(InType);
225 static std::vector<char> outbuf(32768);
226 // The number of UCS4 code points in buf is at most inbytesleft.
227 // The output encoding will use at most
228 // max_encoded_bytes(pimpl_->tocode_) per UCS4 code point.
229 size_t maxoutbufsize = max_encoded_bytes(processor.to()) * inbytesleft;
230 if (outbuf.size() < maxoutbufsize)
231 outbuf.resize(maxoutbufsize);
233 int bytes = processor.convert(inbuf, inbytesleft, &outbuf[0], outbuf.size());
236 // FIXME Maybe throw an exception and handle that in the caller?
237 return vector<RetType>();
239 RetType const * tmp = reinterpret_cast<RetType const *>(&outbuf[0]);
240 return vector<RetType>(tmp, tmp + bytes / sizeof(RetType));
246 vector<char_type> utf8_to_ucs4(vector<char> const & utf8str)
249 return vector<char_type>();
251 return utf8_to_ucs4(&utf8str[0], utf8str.size());
256 utf8_to_ucs4(char const * utf8str, size_t ls)
258 static IconvProcessor processor(ucs4_codeset, "UTF-8");
259 return iconv_convert<char_type>(processor, utf8str, ls);
264 utf16_to_ucs4(unsigned short const * s, size_t ls)
266 static IconvProcessor processor(ucs4_codeset, utf16_codeset);
267 return iconv_convert<char_type>(processor, s, ls);
271 vector<unsigned short>
272 ucs4_to_utf16(char_type const * s, size_t ls)
274 static IconvProcessor processor(utf16_codeset, ucs4_codeset);
275 return iconv_convert<unsigned short>(processor, s, ls);
280 ucs4_to_utf8(char_type c)
282 static IconvProcessor processor("UTF-8", ucs4_codeset);
283 return iconv_convert<char>(processor, &c, 1);
288 ucs4_to_utf8(vector<char_type> const & ucs4str)
291 return vector<char>();
293 return ucs4_to_utf8(&ucs4str[0], ucs4str.size());
298 ucs4_to_utf8(char_type const * ucs4str, size_t ls)
300 static IconvProcessor processor("UTF-8", ucs4_codeset);
301 return iconv_convert<char>(processor, ucs4str, ls);
306 eightbit_to_ucs4(char const * s, size_t ls, string const & encoding)
308 static map<string, IconvProcessor> processors;
309 if (processors.find(encoding) == processors.end()) {
310 IconvProcessor processor(ucs4_codeset, encoding.c_str());
311 processors.insert(make_pair(encoding, processor));
313 return iconv_convert<char_type>(processors[encoding], s, ls);
318 ucs4_to_eightbit(char_type const * ucs4str, size_t ls, string const & encoding)
320 static map<string, IconvProcessor> processors;
321 if (processors.find(encoding) == processors.end()) {
322 IconvProcessor processor(encoding.c_str(), ucs4_codeset);
323 processors.insert(make_pair(encoding, processor));
325 return iconv_convert<char>(processors[encoding], ucs4str, ls);
329 char ucs4_to_eightbit(char_type ucs4, string const & encoding)
331 static map<string, IconvProcessor> processors;
332 map<string, IconvProcessor>::iterator it = processors.find(encoding);
333 if (it == processors.end()) {
334 IconvProcessor processor(encoding.c_str(), ucs4_codeset);
335 it = processors.insert(make_pair(encoding, processor)).first;
339 int const bytes = it->second.convert((char *)(&ucs4), 4, &out, 1);
346 void ucs4_to_multibytes(char_type ucs4, vector<char> & out,
347 string const & encoding)
349 static map<string, IconvProcessor> processors;
350 map<string, IconvProcessor>::iterator it = processors.find(encoding);
351 if (it == processors.end()) {
352 IconvProcessor processor(encoding.c_str(), ucs4_codeset);
353 it = processors.insert(make_pair(encoding, processor)).first;
357 int bytes = it->second.convert((char *)(&ucs4), 4, &out[0], 4);
364 int max_encoded_bytes(std::string const & encoding)
366 // FIXME: this information should be transferred to lib/encodings
367 // UTF8 uses at most 4 bytes to represent one UCS4 code point
368 // (see RFC 3629). RFC 2279 specifies 6 bytes, but that
369 // information is outdated, and RFC 2279 has been superseded by
371 // The CJK encodings use (different) multibyte representation as well.
372 // All other encodings encode one UCS4 code point in one byte
373 // (and can therefore only encode a subset of UCS4)
374 // Note that BIG5 and SJIS do not work with LaTeX (see lib/encodings).
375 // Furthermore, all encodings that use shifting (like SJIS) do not work with
376 // iconv_codecvt_facet.
377 if (encoding == "UTF-8" ||
379 encoding == "EUC-TW")
381 else if (encoding == "EUC-JP")
383 else if (encoding == "ISO-2022-JP")
385 else if (encoding == "BIG5" ||
386 encoding == "EUC-KR" ||
387 encoding == "EUC-CN" ||
388 encoding == "SJIS" ||