1 // Locale support (codecvt) -*- C++ -*-
2
3 // Copyright (C) 2000-2023 Free Software Foundation, Inc.
4 //
5 // This file is part of the GNU ISO C++ Library. This library is free
6 // software; you can redistribute it and/or modify it under the
7 // terms of the GNU General Public License as published by the
8 // Free Software Foundation; either version 3, or (at your option)
9 // any later version.
10
11 // This library is distributed in the hope that it will be useful,
12 // but WITHOUT ANY WARRANTY; without even the implied warranty of
13 // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 // GNU General Public License for more details.
15
16 // Under Section 7 of GPL version 3, you are granted additional
17 // permissions described in the GCC Runtime Library Exception, version
18 // 3.1, as published by the Free Software Foundation.
19
20 // You should have received a copy of the GNU General Public License and
21 // a copy of the GCC Runtime Library Exception along with this program;
22 // see the files COPYING3 and COPYING.RUNTIME respectively. If not, see
23 // <http://www.gnu.org/licenses/>.
24
25 //
26 // ISO C++ 14882: 22.2.1.5 Template class codecvt
27 //
28
29 // Written by Benjamin Kosnik <bkoz@redhat.com>
30
31 /** @file ext/codecvt_specializations.h
32 * This file is a GNU extension to the Standard C++ Library.
33 */
34
35 #ifndef _EXT_CODECVT_SPECIALIZATIONS_H
36 #define _EXT_CODECVT_SPECIALIZATIONS_H 1
37
38 #include <bits/requires_hosted.h> // GNU extensions are currently omitted
39
40 #include <bits/c++config.h>
41 #include <locale>
42 #include <iconv.h>
43
44 namespace __gnu_cxx _GLIBCXX_VISIBILITY(default)
45 {
46 _GLIBCXX_BEGIN_NAMESPACE_VERSION
47 _GLIBCXX_BEGIN_NAMESPACE_CXX11
48
49 /// Extension to use iconv for dealing with character encodings.
50 // This includes conversions and comparisons between various character
51 // sets. This object encapsulates data that may need to be shared between
52 // char_traits, codecvt and ctype.
53 class encoding_state
54 {
55 public:
56 // Types:
57 // NB: A conversion descriptor subsumes and enhances the
58 // functionality of a simple state type such as mbstate_t.
59 typedef iconv_t descriptor_type;
60
61 protected:
62 // Name of internal character set encoding.
63 std::string _M_int_enc;
64
65 // Name of external character set encoding.
66 std::string _M_ext_enc;
67
68 // Conversion descriptor between external encoding to internal encoding.
69 descriptor_type _M_in_desc;
70
71 // Conversion descriptor between internal encoding to external encoding.
72 descriptor_type _M_out_desc;
73
74 // The byte-order marker for the external encoding, if necessary.
75 int _M_ext_bom;
76
77 // The byte-order marker for the internal encoding, if necessary.
78 int _M_int_bom;
79
80 // Number of external bytes needed to construct one complete
81 // character in the internal encoding.
82 // NB: -1 indicates variable, or stateful, encodings.
83 int _M_bytes;
84
85 public:
86 explicit
87 encoding_state()
88 : _M_in_desc(0), _M_out_desc(0), _M_ext_bom(0), _M_int_bom(0), _M_bytes(0)
89 { }
90
91 explicit
92 encoding_state(const char* __int, const char* __ext,
93 int __ibom = 0, int __ebom = 0, int __bytes = 1)
94 : _M_int_enc(__int), _M_ext_enc(__ext), _M_in_desc(0), _M_out_desc(0),
95 _M_ext_bom(__ebom), _M_int_bom(__ibom), _M_bytes(__bytes)
96 { init(); }
97
98 // 21.1.2 traits typedefs
99 // p4
100 // typedef STATE_T state_type
101 // requires: state_type shall meet the requirements of
102 // CopyConstructible types (20.1.3)
103 // NB: This does not preserve the actual state of the conversion
104 // descriptor member, but it does duplicate the encoding
105 // information.
106 encoding_state(const encoding_state& __obj) : _M_in_desc(0), _M_out_desc(0)
107 { construct(__obj); }
108
109 // Need assignment operator as well.
110 encoding_state&
111 operator=(const encoding_state& __obj)
112 {
113 construct(__obj);
114 return *this;
115 }
116
117 ~encoding_state()
118 { destroy(); }
119
120 bool
121 good() const throw()
122 {
123 const descriptor_type __err = (iconv_t)(-1);
124 bool __test = _M_in_desc && _M_in_desc != __err;
125 __test &= _M_out_desc && _M_out_desc != __err;
126 return __test;
127 }
128
129 int
130 character_ratio() const
131 { return _M_bytes; }
132
133 const std::string
134 internal_encoding() const
135 { return _M_int_enc; }
136
137 int
138 internal_bom() const
139 { return _M_int_bom; }
140
141 const std::string
142 external_encoding() const
143 { return _M_ext_enc; }
144
145 int
146 external_bom() const
147 { return _M_ext_bom; }
148
149 const descriptor_type&
150 in_descriptor() const
151 { return _M_in_desc; }
152
153 const descriptor_type&
154 out_descriptor() const
155 { return _M_out_desc; }
156
157 protected:
158 void
159 init()
160 {
161 const descriptor_type __err = (iconv_t)(-1);
162 const bool __have_encodings = _M_int_enc.size() && _M_ext_enc.size();
163 if (!_M_in_desc && __have_encodings)
164 {
165 _M_in_desc = iconv_open(_M_int_enc.c_str(), _M_ext_enc.c_str());
166 if (_M_in_desc == __err)
167 std::__throw_runtime_error(__N("encoding_state::_M_init "
168 "creating iconv input descriptor failed"));
169 }
170 if (!_M_out_desc && __have_encodings)
171 {
172 _M_out_desc = iconv_open(_M_ext_enc.c_str(), _M_int_enc.c_str());
173 if (_M_out_desc == __err)
174 std::__throw_runtime_error(__N("encoding_state::_M_init "
175 "creating iconv output descriptor failed"));
176 }
177 }
178
179 void
180 construct(const encoding_state& __obj)
181 {
182 destroy();
183 _M_int_enc = __obj._M_int_enc;
184 _M_ext_enc = __obj._M_ext_enc;
185 _M_ext_bom = __obj._M_ext_bom;
186 _M_int_bom = __obj._M_int_bom;
187 _M_bytes = __obj._M_bytes;
188 init();
189 }
190
191 void
192 destroy() throw()
193 {
194 const descriptor_type __err = (iconv_t)(-1);
195 if (_M_in_desc && _M_in_desc != __err)
196 {
197 iconv_close(_M_in_desc);
198 _M_in_desc = 0;
199 }
200 if (_M_out_desc && _M_out_desc != __err)
201 {
202 iconv_close(_M_out_desc);
203 _M_out_desc = 0;
204 }
205 }
206 };
207
208 /// encoding_char_traits
209 // Custom traits type with encoding_state for the state type, and the
210 // associated fpos<encoding_state> for the position type, all other
211 // bits equivalent to the required char_traits instantiations.
212 template<typename _CharT>
213 struct encoding_char_traits
214 : public std::char_traits<_CharT>
215 {
216 typedef encoding_state state_type;
217 typedef typename std::fpos<state_type> pos_type;
218 };
219
220 _GLIBCXX_END_NAMESPACE_CXX11
221 _GLIBCXX_END_NAMESPACE_VERSION
222 } // namespace
223
224
225 namespace std _GLIBCXX_VISIBILITY(default)
226 {
227 _GLIBCXX_BEGIN_NAMESPACE_VERSION
228
229 using __gnu_cxx::encoding_state;
230
231 /// codecvt<InternT, _ExternT, encoding_state> specialization.
232 // This partial specialization takes advantage of iconv to provide
233 // code conversions between a large number of character encodings.
234 template<typename _InternT, typename _ExternT>
235 class codecvt<_InternT, _ExternT, encoding_state>
236 : public __codecvt_abstract_base<_InternT, _ExternT, encoding_state>
237 {
238 public:
239 // Types:
240 typedef codecvt_base::result result;
241 typedef _InternT intern_type;
242 typedef _ExternT extern_type;
243 typedef __gnu_cxx::encoding_state state_type;
244 typedef state_type::descriptor_type descriptor_type;
245
246 // Data Members:
247 static locale::id id;
248
249 explicit
250 codecvt(size_t __refs = 0)
251 : __codecvt_abstract_base<intern_type, extern_type, state_type>(__refs)
252 { }
253
254 explicit
255 codecvt(state_type& __enc, size_t __refs = 0)
256 : __codecvt_abstract_base<intern_type, extern_type, state_type>(__refs)
257 { }
258
259 protected:
260 virtual
261 ~codecvt() { }
262
263 virtual result
264 do_out(state_type& __state, const intern_type* __from,
265 const intern_type* __from_end, const intern_type*& __from_next,
266 extern_type* __to, extern_type* __to_end,
267 extern_type*& __to_next) const;
268
269 virtual result
270 do_unshift(state_type& __state, extern_type* __to,
271 extern_type* __to_end, extern_type*& __to_next) const;
272
273 virtual result
274 do_in(state_type& __state, const extern_type* __from,
275 const extern_type* __from_end, const extern_type*& __from_next,
276 intern_type* __to, intern_type* __to_end,
277 intern_type*& __to_next) const;
278
279 virtual int
280 do_encoding() const throw();
281
282 virtual bool
283 do_always_noconv() const throw();
284
285 virtual int
286 do_length(state_type&, const extern_type* __from,
287 const extern_type* __end, size_t __max) const;
288
289 virtual int
290 do_max_length() const throw();
291 };
292
293 template<typename _InternT, typename _ExternT>
294 locale::id
295 codecvt<_InternT, _ExternT, encoding_state>::id;
296
297 // This adaptor works around the signature problems of the second
298 // argument to iconv(): SUSv2 and others use 'const char**', but glibc 2.2
299 // uses 'char**', which matches the POSIX 1003.1-2001 standard.
300 // Using this adaptor, g++ will do the work for us.
301 template<typename _Tp>
302 inline size_t
303 __iconv_adaptor(size_t(*__func)(iconv_t, _Tp, size_t*, char**, size_t*),
304 iconv_t __cd, char** __inbuf, size_t* __inbytes,
305 char** __outbuf, size_t* __outbytes)
306 { return __func(__cd, (_Tp)__inbuf, __inbytes, __outbuf, __outbytes); }
307
308 template<typename _InternT, typename _ExternT>
309 codecvt_base::result
310 codecvt<_InternT, _ExternT, encoding_state>::
311 do_out(state_type& __state, const intern_type* __from,
312 const intern_type* __from_end, const intern_type*& __from_next,
313 extern_type* __to, extern_type* __to_end,
314 extern_type*& __to_next) const
315 {
316 result __ret = codecvt_base::error;
317 if (__state.good())
318 {
319 const descriptor_type& __desc = __state.out_descriptor();
320 const size_t __fmultiple = sizeof(intern_type);
321 size_t __fbytes = __fmultiple * (__from_end - __from);
322 const size_t __tmultiple = sizeof(extern_type);
323 size_t __tbytes = __tmultiple * (__to_end - __to);
324
325 // Argument list for iconv specifies a byte sequence. Thus,
326 // all to/from arrays must be brutally casted to char*.
327 char* __cto = reinterpret_cast<char*>(__to);
328 char* __cfrom;
329 size_t __conv;
330
331 // Some encodings need a byte order marker as the first item
332 // in the byte stream, to designate endian-ness. The default
333 // value for the byte order marker is NULL, so if this is
334 // the case, it's not necessary and we can just go on our
335 // merry way.
336 int __int_bom = __state.internal_bom();
337 if (__int_bom)
338 {
339 size_t __size = __from_end - __from;
340 intern_type* __cfixed = static_cast<intern_type*>
341 (__builtin_alloca(sizeof(intern_type) * (__size + 1)));
342 __cfixed[0] = static_cast<intern_type>(__int_bom);
343 char_traits<intern_type>::copy(__cfixed + 1, __from, __size);
344 __cfrom = reinterpret_cast<char*>(__cfixed);
345 __conv = __iconv_adaptor(iconv, __desc, &__cfrom,
346 &__fbytes, &__cto, &__tbytes);
347 }
348 else
349 {
350 intern_type* __cfixed = const_cast<intern_type*>(__from);
351 __cfrom = reinterpret_cast<char*>(__cfixed);
352 __conv = __iconv_adaptor(iconv, __desc, &__cfrom, &__fbytes,
353 &__cto, &__tbytes);
354 }
355
356 if (__conv != size_t(-1))
357 {
358 __from_next = reinterpret_cast<const intern_type*>(__cfrom);
359 __to_next = reinterpret_cast<extern_type*>(__cto);
360 __ret = codecvt_base::ok;
361 }
362 else
363 {
364 if (__fbytes < __fmultiple * (__from_end - __from))
365 {
366 __from_next = reinterpret_cast<const intern_type*>(__cfrom);
367 __to_next = reinterpret_cast<extern_type*>(__cto);
368 __ret = codecvt_base::partial;
369 }
370 else
371 __ret = codecvt_base::error;
372 }
373 }
374 return __ret;
375 }
376
377 template<typename _InternT, typename _ExternT>
378 codecvt_base::result
379 codecvt<_InternT, _ExternT, encoding_state>::
380 do_unshift(state_type& __state, extern_type* __to,
381 extern_type* __to_end, extern_type*& __to_next) const
382 {
383 result __ret = codecvt_base::error;
384 if (__state.good())
385 {
386 const descriptor_type& __desc = __state.in_descriptor();
387 const size_t __tmultiple = sizeof(intern_type);
388 size_t __tlen = __tmultiple * (__to_end - __to);
389
390 // Argument list for iconv specifies a byte sequence. Thus,
391 // all to/from arrays must be brutally casted to char*.
392 char* __cto = reinterpret_cast<char*>(__to);
393 size_t __conv = __iconv_adaptor(iconv,__desc, 0, 0,
394 &__cto, &__tlen);
395
396 if (__conv != size_t(-1))
397 {
398 __to_next = reinterpret_cast<extern_type*>(__cto);
399 if (__tlen == __tmultiple * (__to_end - __to))
400 __ret = codecvt_base::noconv;
401 else if (__tlen == 0)
402 __ret = codecvt_base::ok;
403 else
404 __ret = codecvt_base::partial;
405 }
406 else
407 __ret = codecvt_base::error;
408 }
409 return __ret;
410 }
411
412 template<typename _InternT, typename _ExternT>
413 codecvt_base::result
414 codecvt<_InternT, _ExternT, encoding_state>::
415 do_in(state_type& __state, const extern_type* __from,
416 const extern_type* __from_end, const extern_type*& __from_next,
417 intern_type* __to, intern_type* __to_end,
418 intern_type*& __to_next) const
419 {
420 result __ret = codecvt_base::error;
421 if (__state.good())
422 {
423 const descriptor_type& __desc = __state.in_descriptor();
424 const size_t __fmultiple = sizeof(extern_type);
425 size_t __flen = __fmultiple * (__from_end - __from);
426 const size_t __tmultiple = sizeof(intern_type);
427 size_t __tlen = __tmultiple * (__to_end - __to);
428
429 // Argument list for iconv specifies a byte sequence. Thus,
430 // all to/from arrays must be brutally casted to char*.
431 char* __cto = reinterpret_cast<char*>(__to);
432 char* __cfrom;
433 size_t __conv;
434
435 // Some encodings need a byte order marker as the first item
436 // in the byte stream, to designate endian-ness. The default
437 // value for the byte order marker is NULL, so if this is
438 // the case, it's not necessary and we can just go on our
439 // merry way.
440 int __ext_bom = __state.external_bom();
441 if (__ext_bom)
442 {
443 size_t __size = __from_end - __from;
444 extern_type* __cfixed = static_cast<extern_type*>
445 (__builtin_alloca(sizeof(extern_type) * (__size + 1)));
446 __cfixed[0] = static_cast<extern_type>(__ext_bom);
447 char_traits<extern_type>::copy(__cfixed + 1, __from, __size);
448 __cfrom = reinterpret_cast<char*>(__cfixed);
449 __conv = __iconv_adaptor(iconv, __desc, &__cfrom,
450 &__flen, &__cto, &__tlen);
451 }
452 else
453 {
454 extern_type* __cfixed = const_cast<extern_type*>(__from);
455 __cfrom = reinterpret_cast<char*>(__cfixed);
456 __conv = __iconv_adaptor(iconv, __desc, &__cfrom,
457 &__flen, &__cto, &__tlen);
458 }
459
460
461 if (__conv != size_t(-1))
462 {
463 __from_next = reinterpret_cast<const extern_type*>(__cfrom);
464 __to_next = reinterpret_cast<intern_type*>(__cto);
465 __ret = codecvt_base::ok;
466 }
467 else
468 {
469 if (__flen < static_cast<size_t>(__from_end - __from))
470 {
471 __from_next = reinterpret_cast<const extern_type*>(__cfrom);
472 __to_next = reinterpret_cast<intern_type*>(__cto);
473 __ret = codecvt_base::partial;
474 }
475 else
476 __ret = codecvt_base::error;
477 }
478 }
479 return __ret;
480 }
481
482 template<typename _InternT, typename _ExternT>
483 int
484 codecvt<_InternT, _ExternT, encoding_state>::
485 do_encoding() const throw()
486 {
487 int __ret = 0;
488 if (sizeof(_ExternT) <= sizeof(_InternT))
489 __ret = sizeof(_InternT) / sizeof(_ExternT);
490 return __ret;
491 }
492
493 template<typename _InternT, typename _ExternT>
494 bool
495 codecvt<_InternT, _ExternT, encoding_state>::
496 do_always_noconv() const throw()
497 { return false; }
498
499 template<typename _InternT, typename _ExternT>
500 int
501 codecvt<_InternT, _ExternT, encoding_state>::
502 do_length(state_type&, const extern_type* __from,
503 const extern_type* __end, size_t __max) const
504 { return std::min(__max, static_cast<size_t>(__end - __from)); }
505
506 // _GLIBCXX_RESOLVE_LIB_DEFECTS
507 // 74. Garbled text for codecvt::do_max_length
508 template<typename _InternT, typename _ExternT>
509 int
510 codecvt<_InternT, _ExternT, encoding_state>::
511 do_max_length() const throw()
512 { return 1; }
513
514 _GLIBCXX_END_NAMESPACE_VERSION
515 } // namespace
516
517 #endif