1 /**
2 * \file unicode.cpp
3 * This file is part of LyX, the document processor.
4 * Licence details can be found in the file COPYING.
5 *
6 * \author Lars Gullik Bjønnes
7 *
8 * Full author contact details are available in file CREDITS.
9 *
10 * A collection of unicode conversion functions, using iconv.
11 */
12
13 #include <config.h>
14
15 #include "support/unicode.h"
16 #include "support/debug.h"
17
18 #include <QThreadStorage>
19
20 #include <iconv.h>
21
22 #include <boost/cstdint.hpp>
23
24 #include <cerrno>
25 #include <map>
26 #include <ostream>
27 //Needed in MSVC
28 #include <string>
29
30
31 using namespace std;
32
33 namespace {
34
35 #ifdef WORDS_BIGENDIAN
36 char const * utf16_codeset = "UTF16-BE";
37 #else
38 char const * utf16_codeset = "UTF16-LE";
39 #endif
40
41 }
42
43
44 namespace lyx {
45
46 #ifdef WORDS_BIGENDIAN
47 char const * ucs4_codeset = "UCS-4BE";
48 #else
49 char const * ucs4_codeset = "UCS-4LE";
50 #endif
51
52
53 struct IconvProcessor::Handler {
54 // assumes cd is valid
Handlerlyx::IconvProcessor::Handler55 Handler(iconv_t const cd) : cd(cd) {}
~Handlerlyx::IconvProcessor::Handler56 ~Handler() {
57 if (iconv_close(cd) == -1)
58 LYXERR0("Error returned from iconv_close(" << errno << ')');
59 }
60 iconv_t const cd;
61 };
62
63
IconvProcessor(string tocode,string fromcode)64 IconvProcessor::IconvProcessor(string tocode, string fromcode)
65 : tocode_(move(tocode)), fromcode_(move(fromcode))
66 {}
67
68
69 // for gcc 4.6
IconvProcessor(IconvProcessor && other)70 IconvProcessor::IconvProcessor(IconvProcessor && other)
71 : tocode_(move(other.tocode_)), fromcode_(move(other.fromcode_)),
72 h_(move(other.h_))
73 {}
74
75
init()76 bool IconvProcessor::init()
77 {
78 if (h_)
79 return true;
80 iconv_t cd = iconv_open(tocode_.c_str(), fromcode_.c_str());
81 if (cd != (iconv_t)(-1)) {
82 h_ = make_unique<Handler>(cd);
83 return true;
84 }
85 lyxerr << "Error returned from iconv_open" << endl;
86 switch (errno) {
87 case EINVAL:
88 lyxerr << "EINVAL The conversion from " << fromcode_ << " to "
89 << tocode_ << " is not supported by the implementation."
90 << endl;
91 break;
92 default:
93 lyxerr << "\tSome other error: " << errno << endl;
94 break;
95 }
96 return false;
97 }
98
99
convert(char const * buf,size_t buflen,char * outbuf,size_t maxoutsize)100 int IconvProcessor::convert(char const * buf, size_t buflen,
101 char * outbuf, size_t maxoutsize)
102 {
103 if (buflen == 0)
104 return 0;
105
106 if (!h_ && !init())
107 return -1;
108
109 char ICONV_CONST * inbuf = const_cast<char ICONV_CONST *>(buf);
110 size_t inbytesleft = buflen;
111 size_t outbytesleft = maxoutsize;
112
113 int res = iconv(h_->cd, &inbuf, &inbytesleft, &outbuf, &outbytesleft);
114
115 // flush out remaining data. This is needed because iconv sometimes
116 // holds back chars in the stream, waiting for a combination character
117 // (see e.g. http://sources.redhat.com/bugzilla/show_bug.cgi?id=1124)
118 iconv(h_->cd, NULL, NULL, &outbuf, &outbytesleft);
119
120 //lyxerr << dec;
121 //lyxerr << "Inbytesleft: " << inbytesleft << endl;
122 //lyxerr << "Outbytesleft: " << outbytesleft << endl;
123
124 if (res != -1)
125 // Everything went well.
126 return maxoutsize - outbytesleft;
127
128 // There are some errors in the conversion
129 lyxerr << "Error returned from iconv" << endl;
130 switch (errno) {
131 case E2BIG:
132 lyxerr << "E2BIG There is not sufficient room at *outbuf." << endl;
133 break;
134 case EILSEQ:
135 case EINVAL:
136 lyxerr << (errno == EINVAL
137 ? "EINVAL An incomplete "
138 : "EILSEQ An invalid ")
139 << "multibyte sequence has been encountered in the input.\n"
140 << "When converting from " << fromcode_
141 << " to " << tocode_ << ".\n";
142 lyxerr << "Input:" << hex;
143 for (size_t i = 0; i < buflen; ++i) {
144 // char may be signed, avoid output of
145 // something like 0xffffffc2
146 boost::uint32_t const b =
147 *reinterpret_cast<unsigned char const *>(buf + i);
148 lyxerr << " 0x" << (unsigned int)b;
149 }
150 lyxerr << dec << endl;
151 break;
152 default:
153 lyxerr << "\tSome other error: " << errno << endl;
154 break;
155 }
156 // We got an error so we close down the conversion engine
157 h_.reset();
158 return -1;
159 }
160
161
162 namespace {
163
164
165 template<typename RetType, typename InType>
166 vector<RetType>
iconv_convert(IconvProcessor & processor,InType const * buf,size_t buflen)167 iconv_convert(IconvProcessor & processor, InType const * buf, size_t buflen)
168 {
169 if (buflen == 0)
170 return vector<RetType>();
171
172 char const * inbuf = reinterpret_cast<char const *>(buf);
173 size_t inbytesleft = buflen * sizeof(InType);
174
175 static QThreadStorage<std::vector<char> *> static_outbuf;
176 if (!static_outbuf.hasLocalData())
177 static_outbuf.setLocalData(new std::vector<char>(32768));
178 std::vector<char> & outbuf = *static_outbuf.localData();
179 // The number of UCS4 code points in buf is at most inbytesleft.
180 // The output encoding will use at most
181 // max_encoded_bytes(pimpl_->tocode_) per UCS4 code point.
182 size_t maxoutbufsize = max_encoded_bytes(processor.to()) * inbytesleft;
183 if (outbuf.size() < maxoutbufsize)
184 outbuf.resize(maxoutbufsize);
185
186 int bytes = processor.convert(inbuf, inbytesleft, &outbuf[0], outbuf.size());
187 if (bytes <= 0)
188 // Conversion failed
189 // FIXME Maybe throw an exception and handle that in the caller?
190 return vector<RetType>();
191
192 RetType const * tmp = reinterpret_cast<RetType const *>(&outbuf[0]);
193 return vector<RetType>(tmp, tmp + bytes / sizeof(RetType));
194 }
195
196 } // namespace
197
198
utf8ToUcs4()199 IconvProcessor & utf8ToUcs4()
200 {
201 static QThreadStorage<IconvProcessor *> processor;
202 if (!processor.hasLocalData())
203 processor.setLocalData(new IconvProcessor(ucs4_codeset, "UTF-8"));
204 return *processor.localData();
205 }
206
207
utf8_to_ucs4(vector<char> const & utf8str)208 vector<char_type> utf8_to_ucs4(vector<char> const & utf8str)
209 {
210 if (utf8str.empty())
211 return vector<char_type>();
212
213 return utf8_to_ucs4(&utf8str[0], utf8str.size());
214 }
215
216
217 vector<char_type>
utf8_to_ucs4(char const * utf8str,size_t ls)218 utf8_to_ucs4(char const * utf8str, size_t ls)
219 {
220 return iconv_convert<char_type>(utf8ToUcs4(), utf8str, ls);
221 }
222
223
224 vector<char_type>
utf16_to_ucs4(unsigned short const * s,size_t ls)225 utf16_to_ucs4(unsigned short const * s, size_t ls)
226 {
227 static QThreadStorage<IconvProcessor *> processor;
228 if (!processor.hasLocalData())
229 processor.setLocalData(new IconvProcessor(ucs4_codeset, utf16_codeset));
230 return iconv_convert<char_type>(*processor.localData(), s, ls);
231 }
232
233
234 vector<unsigned short>
ucs4_to_utf16(char_type const * s,size_t ls)235 ucs4_to_utf16(char_type const * s, size_t ls)
236 {
237 static QThreadStorage<IconvProcessor *> processor;
238 if (!processor.hasLocalData())
239 processor.setLocalData(new IconvProcessor(utf16_codeset, ucs4_codeset));
240 return iconv_convert<unsigned short>(*processor.localData(), s, ls);
241 }
242
243
ucs4ToUtf8()244 IconvProcessor & ucs4ToUtf8()
245 {
246 static QThreadStorage<IconvProcessor *> processor;
247 if (!processor.hasLocalData())
248 processor.setLocalData(new IconvProcessor("UTF-8", ucs4_codeset));
249 return *processor.localData();
250 }
251
252 namespace {
253
getProc(map<string,IconvProcessor> & processors,string const & encoding,bool to)254 IconvProcessor & getProc(map<string, IconvProcessor> & processors,
255 string const & encoding, bool to)
256 {
257 string const & fromcode = to ? ucs4_codeset : encoding;
258 string const & tocode = to ? encoding : ucs4_codeset;
259 map<string, IconvProcessor>::iterator const it = processors.find(encoding);
260 if (it == processors.end()) {
261 IconvProcessor p(fromcode, tocode);
262 return processors.insert(make_pair(encoding, move(p))).first->second;
263 } else
264 return it->second;
265 }
266
267 } // namespace
268
269
270 vector<char>
ucs4_to_utf8(char_type c)271 ucs4_to_utf8(char_type c)
272 {
273 return iconv_convert<char>(ucs4ToUtf8(), &c, 1);
274 }
275
276
277 vector<char>
ucs4_to_utf8(vector<char_type> const & ucs4str)278 ucs4_to_utf8(vector<char_type> const & ucs4str)
279 {
280 if (ucs4str.empty())
281 return vector<char>();
282
283 return ucs4_to_utf8(&ucs4str[0], ucs4str.size());
284 }
285
286
287 vector<char>
ucs4_to_utf8(char_type const * ucs4str,size_t ls)288 ucs4_to_utf8(char_type const * ucs4str, size_t ls)
289 {
290 return iconv_convert<char>(ucs4ToUtf8(), ucs4str, ls);
291 }
292
293
294 vector<char_type>
eightbit_to_ucs4(char const * s,size_t ls,string const & encoding)295 eightbit_to_ucs4(char const * s, size_t ls, string const & encoding)
296 {
297 static QThreadStorage<map<string, IconvProcessor> *> static_processors;
298 if (!static_processors.hasLocalData())
299 static_processors.setLocalData(new map<string, IconvProcessor>);
300 map<string, IconvProcessor> & processors = *static_processors.localData();
301 IconvProcessor & processor = getProc(processors, encoding, true);
302 return iconv_convert<char_type>(processor, s, ls);
303 }
304
305
306 namespace {
307
ucs4To8bitProcessors()308 map<string, IconvProcessor> & ucs4To8bitProcessors()
309 {
310 static QThreadStorage<map<string, IconvProcessor> *> processors;
311 if (!processors.hasLocalData())
312 processors.setLocalData(new map<string, IconvProcessor>);
313 return *processors.localData();
314 }
315
316 } // namespace
317
318 vector<char>
ucs4_to_eightbit(char_type const * ucs4str,size_t ls,string const & encoding)319 ucs4_to_eightbit(char_type const * ucs4str, size_t ls, string const & encoding)
320 {
321 map<string, IconvProcessor> & processors(ucs4To8bitProcessors());
322 IconvProcessor & processor = getProc(processors, encoding, false);
323 return iconv_convert<char>(processor, ucs4str, ls);
324 }
325
326
ucs4_to_eightbit(char_type ucs4,string const & encoding)327 char ucs4_to_eightbit(char_type ucs4, string const & encoding)
328 {
329 map<string, IconvProcessor> & processors(ucs4To8bitProcessors());
330 IconvProcessor & processor = getProc(processors, encoding, false);
331 char out;
332 int const bytes = processor.convert((char *)(&ucs4), 4, &out, 1);
333 if (bytes > 0)
334 return out;
335 return 0;
336 }
337
338
ucs4_to_multibytes(char_type ucs4,vector<char> & out,string const & encoding)339 void ucs4_to_multibytes(char_type ucs4, vector<char> & out,
340 string const & encoding)
341 {
342 static QThreadStorage<map<string, IconvProcessor> *> static_processors;
343 if (!static_processors.hasLocalData())
344 static_processors.setLocalData(new map<string, IconvProcessor>);
345 map<string, IconvProcessor> & processors = *static_processors.localData();
346 IconvProcessor & processor = getProc(processors, encoding, false);
347 out.resize(4);
348 int bytes = processor.convert((char *)(&ucs4), 4, &out[0], 4);
349 if (bytes > 0)
350 out.resize(bytes);
351 else
352 out.clear();
353 }
354
max_encoded_bytes(std::string const & encoding)355 int max_encoded_bytes(std::string const & encoding)
356 {
357 // FIXME: this information should be transferred to lib/encodings
358 // UTF8 uses at most 4 bytes to represent one UCS4 code point
359 // (see RFC 3629). RFC 2279 specifies 6 bytes, but that
360 // information is outdated, and RFC 2279 has been superseded by
361 // RFC 3629.
362 // The CJK encodings use (different) multibyte representation as well.
363 // All other encodings encode one UCS4 code point in one byte
364 // (and can therefore only encode a subset of UCS4)
365 // Furthermore, all encodings that use shifting (like SJIS) do not work with
366 // iconv_codecvt_facet.
367 if (encoding == "UTF-8" ||
368 encoding == "GB" ||
369 encoding == "EUC-TW")
370 return 4;
371 else if (encoding == "EUC-JP")
372 return 3;
373 else if (encoding == "ISO-2022-JP")
374 return 8;
375 else if (encoding == "BIG5" ||
376 encoding == "EUC-KR" ||
377 encoding == "EUC-CN" ||
378 encoding == "SJIS" ||
379 encoding == "GBK")
380 return 2;
381 else
382 return 1;
383 }
384
385 } // namespace lyx
386