1 // Locale support (codecvt) -*- C++ -*-
3 // Copyright (C) 2000-2013 Free Software Foundation, Inc.
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)
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.
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.
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/>.
26 // ISO C++ 14882: 22.2.1.5 Template class codecvt
29 // Written by Benjamin Kosnik <bkoz@redhat.com>
31 /** @file ext/codecvt_specializations.h
32 * This file is a GNU extension to the Standard C++ Library.
35 #ifndef _EXT_CODECVT_SPECIALIZATIONS_H
36 #define _EXT_CODECVT_SPECIALIZATIONS_H 1
38 #include <bits/c++config.h>
42 namespace __gnu_cxx _GLIBCXX_VISIBILITY(default)
44 _GLIBCXX_BEGIN_NAMESPACE_VERSION
46 /// Extension to use iconv for dealing with character encodings.
47 // This includes conversions and comparisons between various character
48 // sets. This object encapsulates data that may need to be shared between
49 // char_traits, codecvt and ctype.
54 // NB: A conversion descriptor subsumes and enhances the
55 // functionality of a simple state type such as mbstate_t.
56 typedef iconv_t descriptor_type;
59 // Name of internal character set encoding.
60 std::string _M_int_enc;
62 // Name of external character set encoding.
63 std::string _M_ext_enc;
65 // Conversion descriptor between external encoding to internal encoding.
66 descriptor_type _M_in_desc;
68 // Conversion descriptor between internal encoding to external encoding.
69 descriptor_type _M_out_desc;
71 // The byte-order marker for the external encoding, if necessary.
74 // The byte-order marker for the internal encoding, if necessary.
77 // Number of external bytes needed to construct one complete
78 // character in the internal encoding.
79 // NB: -1 indicates variable, or stateful, encodings.
85 : _M_in_desc(0), _M_out_desc(0), _M_ext_bom(0), _M_int_bom(0), _M_bytes(0)
89 encoding_state(const char* __int, const char* __ext,
90 int __ibom = 0, int __ebom = 0, int __bytes = 1)
91 : _M_int_enc(__int), _M_ext_enc(__ext), _M_in_desc(0), _M_out_desc(0),
92 _M_ext_bom(__ebom), _M_int_bom(__ibom), _M_bytes(__bytes)
95 // 21.1.2 traits typedefs
97 // typedef STATE_T state_type
98 // requires: state_type shall meet the requirements of
99 // CopyConstructible types (20.1.3)
100 // NB: This does not preserve the actual state of the conversion
101 // descriptor member, but it does duplicate the encoding
103 encoding_state(const encoding_state& __obj) : _M_in_desc(0), _M_out_desc(0)
104 { construct(__obj); }
106 // Need assignment operator as well.
108 operator=(const encoding_state& __obj)
120 const descriptor_type __err = (iconv_t)(-1);
121 bool __test = _M_in_desc && _M_in_desc != __err;
122 __test &= _M_out_desc && _M_out_desc != __err;
127 character_ratio() const
131 internal_encoding() const
132 { return _M_int_enc; }
136 { return _M_int_bom; }
139 external_encoding() const
140 { return _M_ext_enc; }
144 { return _M_ext_bom; }
146 const descriptor_type&
147 in_descriptor() const
148 { return _M_in_desc; }
150 const descriptor_type&
151 out_descriptor() const
152 { return _M_out_desc; }
158 const descriptor_type __err = (iconv_t)(-1);
159 const bool __have_encodings = _M_int_enc.size() && _M_ext_enc.size();
160 if (!_M_in_desc && __have_encodings)
162 _M_in_desc = iconv_open(_M_int_enc.c_str(), _M_ext_enc.c_str());
163 if (_M_in_desc == __err)
164 std::__throw_runtime_error(__N("encoding_state::_M_init "
165 "creating iconv input descriptor failed"));
167 if (!_M_out_desc && __have_encodings)
169 _M_out_desc = iconv_open(_M_ext_enc.c_str(), _M_int_enc.c_str());
170 if (_M_out_desc == __err)
171 std::__throw_runtime_error(__N("encoding_state::_M_init "
172 "creating iconv output descriptor failed"));
177 construct(const encoding_state& __obj)
180 _M_int_enc = __obj._M_int_enc;
181 _M_ext_enc = __obj._M_ext_enc;
182 _M_ext_bom = __obj._M_ext_bom;
183 _M_int_bom = __obj._M_int_bom;
184 _M_bytes = __obj._M_bytes;
191 const descriptor_type __err = (iconv_t)(-1);
192 if (_M_in_desc && _M_in_desc != __err)
194 iconv_close(_M_in_desc);
197 if (_M_out_desc && _M_out_desc != __err)
199 iconv_close(_M_out_desc);
205 /// encoding_char_traits
206 // Custom traits type with encoding_state for the state type, and the
207 // associated fpos<encoding_state> for the position type, all other
208 // bits equivalent to the required char_traits instantiations.
209 template<typename _CharT>
210 struct encoding_char_traits : public std::char_traits<_CharT>
212 typedef encoding_state state_type;
213 typedef typename std::fpos<state_type> pos_type;
216 _GLIBCXX_END_NAMESPACE_VERSION
220 namespace std _GLIBCXX_VISIBILITY(default)
222 _GLIBCXX_BEGIN_NAMESPACE_VERSION
224 using __gnu_cxx::encoding_state;
226 /// codecvt<InternT, _ExternT, encoding_state> specialization.
227 // This partial specialization takes advantage of iconv to provide
228 // code conversions between a large number of character encodings.
229 template<typename _InternT, typename _ExternT>
230 class codecvt<_InternT, _ExternT, encoding_state>
231 : public __codecvt_abstract_base<_InternT, _ExternT, encoding_state>
235 typedef codecvt_base::result result;
236 typedef _InternT intern_type;
237 typedef _ExternT extern_type;
238 typedef __gnu_cxx::encoding_state state_type;
239 typedef state_type::descriptor_type descriptor_type;
242 static locale::id id;
245 codecvt(size_t __refs = 0)
246 : __codecvt_abstract_base<intern_type, extern_type, state_type>(__refs)
250 codecvt(state_type& __enc, size_t __refs = 0)
251 : __codecvt_abstract_base<intern_type, extern_type, state_type>(__refs)
259 do_out(state_type& __state, const intern_type* __from,
260 const intern_type* __from_end, const intern_type*& __from_next,
261 extern_type* __to, extern_type* __to_end,
262 extern_type*& __to_next) const;
265 do_unshift(state_type& __state, extern_type* __to,
266 extern_type* __to_end, extern_type*& __to_next) const;
269 do_in(state_type& __state, const extern_type* __from,
270 const extern_type* __from_end, const extern_type*& __from_next,
271 intern_type* __to, intern_type* __to_end,
272 intern_type*& __to_next) const;
275 do_encoding() const throw();
278 do_always_noconv() const throw();
281 do_length(state_type&, const extern_type* __from,
282 const extern_type* __end, size_t __max) const;
285 do_max_length() const throw();
288 template<typename _InternT, typename _ExternT>
290 codecvt<_InternT, _ExternT, encoding_state>::id;
292 // This adaptor works around the signature problems of the second
293 // argument to iconv(): SUSv2 and others use 'const char**', but glibc 2.2
294 // uses 'char**', which matches the POSIX 1003.1-2001 standard.
295 // Using this adaptor, g++ will do the work for us.
296 template<typename _Tp>
298 __iconv_adaptor(size_t(*__func)(iconv_t, _Tp, size_t*, char**, size_t*),
299 iconv_t __cd, char** __inbuf, size_t* __inbytes,
300 char** __outbuf, size_t* __outbytes)
301 { return __func(__cd, (_Tp)__inbuf, __inbytes, __outbuf, __outbytes); }
303 template<typename _InternT, typename _ExternT>
305 codecvt<_InternT, _ExternT, encoding_state>::
306 do_out(state_type& __state, const intern_type* __from,
307 const intern_type* __from_end, const intern_type*& __from_next,
308 extern_type* __to, extern_type* __to_end,
309 extern_type*& __to_next) const
311 result __ret = codecvt_base::error;
314 const descriptor_type& __desc = __state.out_descriptor();
315 const size_t __fmultiple = sizeof(intern_type);
316 size_t __fbytes = __fmultiple * (__from_end - __from);
317 const size_t __tmultiple = sizeof(extern_type);
318 size_t __tbytes = __tmultiple * (__to_end - __to);
320 // Argument list for iconv specifies a byte sequence. Thus,
321 // all to/from arrays must be brutally casted to char*.
322 char* __cto = reinterpret_cast<char*>(__to);
326 // Some encodings need a byte order marker as the first item
327 // in the byte stream, to designate endian-ness. The default
328 // value for the byte order marker is NULL, so if this is
329 // the case, it's not necessary and we can just go on our
331 int __int_bom = __state.internal_bom();
334 size_t __size = __from_end - __from;
335 intern_type* __cfixed = static_cast<intern_type*>
336 (__builtin_alloca(sizeof(intern_type) * (__size + 1)));
337 __cfixed[0] = static_cast<intern_type>(__int_bom);
338 char_traits<intern_type>::copy(__cfixed + 1, __from, __size);
339 __cfrom = reinterpret_cast<char*>(__cfixed);
340 __conv = __iconv_adaptor(iconv, __desc, &__cfrom,
341 &__fbytes, &__cto, &__tbytes);
345 intern_type* __cfixed = const_cast<intern_type*>(__from);
346 __cfrom = reinterpret_cast<char*>(__cfixed);
347 __conv = __iconv_adaptor(iconv, __desc, &__cfrom, &__fbytes,
351 if (__conv != size_t(-1))
353 __from_next = reinterpret_cast<const intern_type*>(__cfrom);
354 __to_next = reinterpret_cast<extern_type*>(__cto);
355 __ret = codecvt_base::ok;
359 if (__fbytes < __fmultiple * (__from_end - __from))
361 __from_next = reinterpret_cast<const intern_type*>(__cfrom);
362 __to_next = reinterpret_cast<extern_type*>(__cto);
363 __ret = codecvt_base::partial;
366 __ret = codecvt_base::error;
372 template<typename _InternT, typename _ExternT>
374 codecvt<_InternT, _ExternT, encoding_state>::
375 do_unshift(state_type& __state, extern_type* __to,
376 extern_type* __to_end, extern_type*& __to_next) const
378 result __ret = codecvt_base::error;
381 const descriptor_type& __desc = __state.in_descriptor();
382 const size_t __tmultiple = sizeof(intern_type);
383 size_t __tlen = __tmultiple * (__to_end - __to);
385 // Argument list for iconv specifies a byte sequence. Thus,
386 // all to/from arrays must be brutally casted to char*.
387 char* __cto = reinterpret_cast<char*>(__to);
388 size_t __conv = __iconv_adaptor(iconv,__desc, 0, 0,
391 if (__conv != size_t(-1))
393 __to_next = reinterpret_cast<extern_type*>(__cto);
394 if (__tlen == __tmultiple * (__to_end - __to))
395 __ret = codecvt_base::noconv;
396 else if (__tlen == 0)
397 __ret = codecvt_base::ok;
399 __ret = codecvt_base::partial;
402 __ret = codecvt_base::error;
407 template<typename _InternT, typename _ExternT>
409 codecvt<_InternT, _ExternT, encoding_state>::
410 do_in(state_type& __state, const extern_type* __from,
411 const extern_type* __from_end, const extern_type*& __from_next,
412 intern_type* __to, intern_type* __to_end,
413 intern_type*& __to_next) const
415 result __ret = codecvt_base::error;
418 const descriptor_type& __desc = __state.in_descriptor();
419 const size_t __fmultiple = sizeof(extern_type);
420 size_t __flen = __fmultiple * (__from_end - __from);
421 const size_t __tmultiple = sizeof(intern_type);
422 size_t __tlen = __tmultiple * (__to_end - __to);
424 // Argument list for iconv specifies a byte sequence. Thus,
425 // all to/from arrays must be brutally casted to char*.
426 char* __cto = reinterpret_cast<char*>(__to);
430 // Some encodings need a byte order marker as the first item
431 // in the byte stream, to designate endian-ness. The default
432 // value for the byte order marker is NULL, so if this is
433 // the case, it's not necessary and we can just go on our
435 int __ext_bom = __state.external_bom();
438 size_t __size = __from_end - __from;
439 extern_type* __cfixed = static_cast<extern_type*>
440 (__builtin_alloca(sizeof(extern_type) * (__size + 1)));
441 __cfixed[0] = static_cast<extern_type>(__ext_bom);
442 char_traits<extern_type>::copy(__cfixed + 1, __from, __size);
443 __cfrom = reinterpret_cast<char*>(__cfixed);
444 __conv = __iconv_adaptor(iconv, __desc, &__cfrom,
445 &__flen, &__cto, &__tlen);
449 extern_type* __cfixed = const_cast<extern_type*>(__from);
450 __cfrom = reinterpret_cast<char*>(__cfixed);
451 __conv = __iconv_adaptor(iconv, __desc, &__cfrom,
452 &__flen, &__cto, &__tlen);
456 if (__conv != size_t(-1))
458 __from_next = reinterpret_cast<const extern_type*>(__cfrom);
459 __to_next = reinterpret_cast<intern_type*>(__cto);
460 __ret = codecvt_base::ok;
464 if (__flen < static_cast<size_t>(__from_end - __from))
466 __from_next = reinterpret_cast<const extern_type*>(__cfrom);
467 __to_next = reinterpret_cast<intern_type*>(__cto);
468 __ret = codecvt_base::partial;
471 __ret = codecvt_base::error;
477 template<typename _InternT, typename _ExternT>
479 codecvt<_InternT, _ExternT, encoding_state>::
480 do_encoding() const throw()
483 if (sizeof(_ExternT) <= sizeof(_InternT))
484 __ret = sizeof(_InternT) / sizeof(_ExternT);
488 template<typename _InternT, typename _ExternT>
490 codecvt<_InternT, _ExternT, encoding_state>::
491 do_always_noconv() const throw()
494 template<typename _InternT, typename _ExternT>
496 codecvt<_InternT, _ExternT, encoding_state>::
497 do_length(state_type&, const extern_type* __from,
498 const extern_type* __end, size_t __max) const
499 { return std::min(__max, static_cast<size_t>(__end - __from)); }
501 // _GLIBCXX_RESOLVE_LIB_DEFECTS
502 // 74. Garbled text for codecvt::do_max_length
503 template<typename _InternT, typename _ExternT>
505 codecvt<_InternT, _ExternT, encoding_state>::
506 do_max_length() const throw()
509 _GLIBCXX_END_NAMESPACE_VERSION