1 // Locale support (codecvt) -*- C++ -*-
2
3 // Copyright (C) 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007,
4 // 2008, 2009, 2010
5 // Free Software Foundation, Inc.
6 //
7 // This file is part of the GNU ISO C++ Library. This library is free
8 // software; you can redistribute it and/or modify it under the
9 // terms of the GNU General Public License as published by the
10 // Free Software Foundation; either version 3, or (at your option)
11 // any later version.
12
13 // This library is distributed in the hope that it will be useful,
14 // but WITHOUT ANY WARRANTY; without even the implied warranty of
15 // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 // GNU General Public License for more details.
17
18 // Under Section 7 of GPL version 3, you are granted additional
19 // permissions described in the GCC Runtime Library Exception, version
20 // 3.1, as published by the Free Software Foundation.
21
22 // You should have received a copy of the GNU General Public License and
23 // a copy of the GCC Runtime Library Exception along with this program;
24 // see the files COPYING3 and COPYING.RUNTIME respectively. If not, see
25 // <http://www.gnu.org/licenses/>.
26
27 //
28 // ISO C++ 14882: 22.2.1.5 Template class codecvt
29 //
30
31 // Written by Benjamin Kosnik <bkoz@redhat.com>
32
33 /** @file ext/codecvt_specializations.h
34 * This file is a GNU extension to the Standard C++ Library.
35 */
36
37 #ifndef _EXT_CODECVT_SPECIALIZATIONS_H
38 #define _EXT_CODECVT_SPECIALIZATIONS_H 1
39
40 #include <bits/c++config.h>
41 #include <locale>
42 #include <iconv.h>
43
_GLIBCXX_VISIBILITY(default)44 namespace __gnu_cxx _GLIBCXX_VISIBILITY(default)
45 {
46 _GLIBCXX_BEGIN_NAMESPACE_VERSION
47
48 /// Extension to use iconv for dealing with character encodings.
49 // This includes conversions and comparisons between various character
50 // sets. This object encapsulates data that may need to be shared between
51 // char_traits, codecvt and ctype.
52 class encoding_state
53 {
54 public:
55 // Types:
56 // NB: A conversion descriptor subsumes and enhances the
57 // functionality of a simple state type such as mbstate_t.
58 typedef iconv_t descriptor_type;
59
60 protected:
61 // Name of internal character set encoding.
62 std::string _M_int_enc;
63
64 // Name of external character set encoding.
65 std::string _M_ext_enc;
66
67 // Conversion descriptor between external encoding to internal encoding.
68 descriptor_type _M_in_desc;
69
70 // Conversion descriptor between internal encoding to external encoding.
71 descriptor_type _M_out_desc;
72
73 // The byte-order marker for the external encoding, if necessary.
74 int _M_ext_bom;
75
76 // The byte-order marker for the internal encoding, if necessary.
77 int _M_int_bom;
78
79 // Number of external bytes needed to construct one complete
80 // character in the internal encoding.
81 // NB: -1 indicates variable, or stateful, encodings.
82 int _M_bytes;
83
84 public:
85 explicit
86 encoding_state()
87 : _M_in_desc(0), _M_out_desc(0), _M_ext_bom(0), _M_int_bom(0), _M_bytes(0)
88 { }
89
90 explicit
91 encoding_state(const char* __int, const char* __ext,
92 int __ibom = 0, int __ebom = 0, int __bytes = 1)
93 : _M_int_enc(__int), _M_ext_enc(__ext), _M_in_desc(0), _M_out_desc(0),
94 _M_ext_bom(__ebom), _M_int_bom(__ibom), _M_bytes(__bytes)
95 { init(); }
96
97 // 21.1.2 traits typedefs
98 // p4
99 // typedef STATE_T state_type
100 // requires: state_type shall meet the requirements of
101 // CopyConstructible types (20.1.3)
102 // NB: This does not preserve the actual state of the conversion
103 // descriptor member, but it does duplicate the encoding
104 // information.
105 encoding_state(const encoding_state& __obj) : _M_in_desc(0), _M_out_desc(0)
106 { construct(__obj); }
107
108 // Need assignment operator as well.
109 encoding_state&
110 operator=(const encoding_state& __obj)
111 {
112 construct(__obj);
113 return *this;
114 }
115
116 ~encoding_state()
117 { destroy(); }
118
119 bool
120 good() const throw()
121 {
122 const descriptor_type __err = (iconv_t)(-1);
123 bool __test = _M_in_desc && _M_in_desc != __err;
124 __test &= _M_out_desc && _M_out_desc != __err;
125 return __test;
126 }
127
128 int
129 character_ratio() const
130 { return _M_bytes; }
131
132 const std::string
133 internal_encoding() const
134 { return _M_int_enc; }
135
136 int
137 internal_bom() const
138 { return _M_int_bom; }
139
140 const std::string
141 external_encoding() const
142 { return _M_ext_enc; }
143
144 int
145 external_bom() const
146 { return _M_ext_bom; }
147
148 const descriptor_type&
149 in_descriptor() const
150 { return _M_in_desc; }
151
152 const descriptor_type&
153 out_descriptor() const
154 { return _M_out_desc; }
155
156 protected:
157 void
158 init()
159 {
160 const descriptor_type __err = (iconv_t)(-1);
161 const bool __have_encodings = _M_int_enc.size() && _M_ext_enc.size();
162 if (!_M_in_desc && __have_encodings)
163 {
164 _M_in_desc = iconv_open(_M_int_enc.c_str(), _M_ext_enc.c_str());
165 if (_M_in_desc == __err)
166 std::__throw_runtime_error(__N("encoding_state::_M_init "
167 "creating iconv input descriptor failed"));
168 }
169 if (!_M_out_desc && __have_encodings)
170 {
171 _M_out_desc = iconv_open(_M_ext_enc.c_str(), _M_int_enc.c_str());
172 if (_M_out_desc == __err)
173 std::__throw_runtime_error(__N("encoding_state::_M_init "
174 "creating iconv output descriptor failed"));
175 }
176 }
177
178 void
179 construct(const encoding_state& __obj)
180 {
181 destroy();
182 _M_int_enc = __obj._M_int_enc;
183 _M_ext_enc = __obj._M_ext_enc;
184 _M_ext_bom = __obj._M_ext_bom;
185 _M_int_bom = __obj._M_int_bom;
186 _M_bytes = __obj._M_bytes;
187 init();
188 }
189
190 void
191 destroy() throw()
192 {
193 const descriptor_type __err = (iconv_t)(-1);
194 if (_M_in_desc && _M_in_desc != __err)
195 {
196 iconv_close(_M_in_desc);
197 _M_in_desc = 0;
198 }
199 if (_M_out_desc && _M_out_desc != __err)
200 {
201 iconv_close(_M_out_desc);
202 _M_out_desc = 0;
203 }
204 }
205 };
206
207 /// encoding_char_traits
208 // Custom traits type with encoding_state for the state type, and the
209 // associated fpos<encoding_state> for the position type, all other
210 // bits equivalent to the required char_traits instantiations.
211 template<typename _CharT>
212 struct encoding_char_traits : public std::char_traits<_CharT>
213 {
214 typedef encoding_state state_type;
215 typedef typename std::fpos<state_type> pos_type;
216 };
217
218 _GLIBCXX_END_NAMESPACE_VERSION
219 } // namespace
220
221
_GLIBCXX_VISIBILITY(default)222 namespace std _GLIBCXX_VISIBILITY(default)
223 {
224 _GLIBCXX_BEGIN_NAMESPACE_VERSION
225
226 using __gnu_cxx::encoding_state;
227
228 /// codecvt<InternT, _ExternT, encoding_state> specialization.
229 // This partial specialization takes advantage of iconv to provide
230 // code conversions between a large number of character encodings.
231 template<typename _InternT, typename _ExternT>
232 class codecvt<_InternT, _ExternT, encoding_state>
233 : public __codecvt_abstract_base<_InternT, _ExternT, encoding_state>
234 {
235 public:
236 // Types:
237 typedef codecvt_base::result result;
238 typedef _InternT intern_type;
239 typedef _ExternT extern_type;
240 typedef __gnu_cxx::encoding_state state_type;
241 typedef state_type::descriptor_type descriptor_type;
242
243 // Data Members:
244 static locale::id id;
245
246 explicit
247 codecvt(size_t __refs = 0)
248 : __codecvt_abstract_base<intern_type, extern_type, state_type>(__refs)
249 { }
250
251 explicit
252 codecvt(state_type& __enc, size_t __refs = 0)
253 : __codecvt_abstract_base<intern_type, extern_type, state_type>(__refs)
254 { }
255
256 protected:
257 virtual
258 ~codecvt() { }
259
260 virtual result
261 do_out(state_type& __state, const intern_type* __from,
262 const intern_type* __from_end, const intern_type*& __from_next,
263 extern_type* __to, extern_type* __to_end,
264 extern_type*& __to_next) const;
265
266 virtual result
267 do_unshift(state_type& __state, extern_type* __to,
268 extern_type* __to_end, extern_type*& __to_next) const;
269
270 virtual result
271 do_in(state_type& __state, const extern_type* __from,
272 const extern_type* __from_end, const extern_type*& __from_next,
273 intern_type* __to, intern_type* __to_end,
274 intern_type*& __to_next) const;
275
276 virtual int
277 do_encoding() const throw();
278
279 virtual bool
280 do_always_noconv() const throw();
281
282 virtual int
283 do_length(state_type&, const extern_type* __from,
284 const extern_type* __end, size_t __max) const;
285
286 virtual int
287 do_max_length() const throw();
288 };
289
290 template<typename _InternT, typename _ExternT>
291 locale::id
292 codecvt<_InternT, _ExternT, encoding_state>::id;
293
294 // This adaptor works around the signature problems of the second
295 // argument to iconv(): SUSv2 and others use 'const char**', but glibc 2.2
296 // uses 'char**', which matches the POSIX 1003.1-2001 standard.
297 // Using this adaptor, g++ will do the work for us.
298 template<typename _Tp>
299 inline size_t
300 __iconv_adaptor(size_t(*__func)(iconv_t, _Tp, size_t*, char**, size_t*),
301 iconv_t __cd, char** __inbuf, size_t* __inbytes,
302 char** __outbuf, size_t* __outbytes)
303 { return __func(__cd, (_Tp)__inbuf, __inbytes, __outbuf, __outbytes); }
304
305 template<typename _InternT, typename _ExternT>
306 codecvt_base::result
307 codecvt<_InternT, _ExternT, encoding_state>::
308 do_out(state_type& __state, const intern_type* __from,
309 const intern_type* __from_end, const intern_type*& __from_next,
310 extern_type* __to, extern_type* __to_end,
311 extern_type*& __to_next) const
312 {
313 result __ret = codecvt_base::error;
314 if (__state.good())
315 {
316 const descriptor_type& __desc = __state.out_descriptor();
317 const size_t __fmultiple = sizeof(intern_type);
318 size_t __fbytes = __fmultiple * (__from_end - __from);
319 const size_t __tmultiple = sizeof(extern_type);
320 size_t __tbytes = __tmultiple * (__to_end - __to);
321
322 // Argument list for iconv specifies a byte sequence. Thus,
323 // all to/from arrays must be brutally casted to char*.
324 char* __cto = reinterpret_cast<char*>(__to);
325 char* __cfrom;
326 size_t __conv;
327
328 // Some encodings need a byte order marker as the first item
329 // in the byte stream, to designate endian-ness. The default
330 // value for the byte order marker is NULL, so if this is
331 // the case, it's not necessary and we can just go on our
332 // merry way.
333 int __int_bom = __state.internal_bom();
334 if (__int_bom)
335 {
336 size_t __size = __from_end - __from;
337 intern_type* __cfixed = static_cast<intern_type*>
338 (__builtin_alloca(sizeof(intern_type) * (__size + 1)));
339 __cfixed[0] = static_cast<intern_type>(__int_bom);
340 char_traits<intern_type>::copy(__cfixed + 1, __from, __size);
341 __cfrom = reinterpret_cast<char*>(__cfixed);
342 __conv = __iconv_adaptor(iconv, __desc, &__cfrom,
343 &__fbytes, &__cto, &__tbytes);
344 }
345 else
346 {
347 intern_type* __cfixed = const_cast<intern_type*>(__from);
348 __cfrom = reinterpret_cast<char*>(__cfixed);
349 __conv = __iconv_adaptor(iconv, __desc, &__cfrom, &__fbytes,
350 &__cto, &__tbytes);
351 }
352
353 if (__conv != size_t(-1))
354 {
355 __from_next = reinterpret_cast<const intern_type*>(__cfrom);
356 __to_next = reinterpret_cast<extern_type*>(__cto);
357 __ret = codecvt_base::ok;
358 }
359 else
360 {
361 if (__fbytes < __fmultiple * (__from_end - __from))
362 {
363 __from_next = reinterpret_cast<const intern_type*>(__cfrom);
364 __to_next = reinterpret_cast<extern_type*>(__cto);
365 __ret = codecvt_base::partial;
366 }
367 else
368 __ret = codecvt_base::error;
369 }
370 }
371 return __ret;
372 }
373
374 template<typename _InternT, typename _ExternT>
375 codecvt_base::result
376 codecvt<_InternT, _ExternT, encoding_state>::
377 do_unshift(state_type& __state, extern_type* __to,
378 extern_type* __to_end, extern_type*& __to_next) const
379 {
380 result __ret = codecvt_base::error;
381 if (__state.good())
382 {
383 const descriptor_type& __desc = __state.in_descriptor();
384 const size_t __tmultiple = sizeof(intern_type);
385 size_t __tlen = __tmultiple * (__to_end - __to);
386
387 // Argument list for iconv specifies a byte sequence. Thus,
388 // all to/from arrays must be brutally casted to char*.
389 char* __cto = reinterpret_cast<char*>(__to);
390 size_t __conv = __iconv_adaptor(iconv,__desc, 0, 0,
391 &__cto, &__tlen);
392
393 if (__conv != size_t(-1))
394 {
395 __to_next = reinterpret_cast<extern_type*>(__cto);
396 if (__tlen == __tmultiple * (__to_end - __to))
397 __ret = codecvt_base::noconv;
398 else if (__tlen == 0)
399 __ret = codecvt_base::ok;
400 else
401 __ret = codecvt_base::partial;
402 }
403 else
404 __ret = codecvt_base::error;
405 }
406 return __ret;
407 }
408
409 template<typename _InternT, typename _ExternT>
410 codecvt_base::result
411 codecvt<_InternT, _ExternT, encoding_state>::
412 do_in(state_type& __state, const extern_type* __from,
413 const extern_type* __from_end, const extern_type*& __from_next,
414 intern_type* __to, intern_type* __to_end,
415 intern_type*& __to_next) const
416 {
417 result __ret = codecvt_base::error;
418 if (__state.good())
419 {
420 const descriptor_type& __desc = __state.in_descriptor();
421 const size_t __fmultiple = sizeof(extern_type);
422 size_t __flen = __fmultiple * (__from_end - __from);
423 const size_t __tmultiple = sizeof(intern_type);
424 size_t __tlen = __tmultiple * (__to_end - __to);
425
426 // Argument list for iconv specifies a byte sequence. Thus,
427 // all to/from arrays must be brutally casted to char*.
428 char* __cto = reinterpret_cast<char*>(__to);
429 char* __cfrom;
430 size_t __conv;
431
432 // Some encodings need a byte order marker as the first item
433 // in the byte stream, to designate endian-ness. The default
434 // value for the byte order marker is NULL, so if this is
435 // the case, it's not necessary and we can just go on our
436 // merry way.
437 int __ext_bom = __state.external_bom();
438 if (__ext_bom)
439 {
440 size_t __size = __from_end - __from;
441 extern_type* __cfixed = static_cast<extern_type*>
442 (__builtin_alloca(sizeof(extern_type) * (__size + 1)));
443 __cfixed[0] = static_cast<extern_type>(__ext_bom);
444 char_traits<extern_type>::copy(__cfixed + 1, __from, __size);
445 __cfrom = reinterpret_cast<char*>(__cfixed);
446 __conv = __iconv_adaptor(iconv, __desc, &__cfrom,
447 &__flen, &__cto, &__tlen);
448 }
449 else
450 {
451 extern_type* __cfixed = const_cast<extern_type*>(__from);
452 __cfrom = reinterpret_cast<char*>(__cfixed);
453 __conv = __iconv_adaptor(iconv, __desc, &__cfrom,
454 &__flen, &__cto, &__tlen);
455 }
456
457
458 if (__conv != size_t(-1))
459 {
460 __from_next = reinterpret_cast<const extern_type*>(__cfrom);
461 __to_next = reinterpret_cast<intern_type*>(__cto);
462 __ret = codecvt_base::ok;
463 }
464 else
465 {
466 if (__flen < static_cast<size_t>(__from_end - __from))
467 {
468 __from_next = reinterpret_cast<const extern_type*>(__cfrom);
469 __to_next = reinterpret_cast<intern_type*>(__cto);
470 __ret = codecvt_base::partial;
471 }
472 else
473 __ret = codecvt_base::error;
474 }
475 }
476 return __ret;
477 }
478
479 template<typename _InternT, typename _ExternT>
480 int
481 codecvt<_InternT, _ExternT, encoding_state>::
482 do_encoding() const throw()
483 {
484 int __ret = 0;
485 if (sizeof(_ExternT) <= sizeof(_InternT))
486 __ret = sizeof(_InternT) / sizeof(_ExternT);
487 return __ret;
488 }
489
490 template<typename _InternT, typename _ExternT>
491 bool
492 codecvt<_InternT, _ExternT, encoding_state>::
493 do_always_noconv() const throw()
494 { return false; }
495
496 template<typename _InternT, typename _ExternT>
497 int
498 codecvt<_InternT, _ExternT, encoding_state>::
499 do_length(state_type&, const extern_type* __from,
500 const extern_type* __end, size_t __max) const
501 { return std::min(__max, static_cast<size_t>(__end - __from)); }
502
503 // _GLIBCXX_RESOLVE_LIB_DEFECTS
504 // 74. Garbled text for codecvt::do_max_length
505 template<typename _InternT, typename _ExternT>
506 int
507 codecvt<_InternT, _ExternT, encoding_state>::
508 do_max_length() const throw()
509 { return 1; }
510
511 _GLIBCXX_END_NAMESPACE_VERSION
512 } // namespace
513
514 #endif
515