]> git.proxmox.com Git - ceph.git/blob - ceph/src/boost/boost/locale/generic_codecvt.hpp
update sources to v12.2.3
[ceph.git] / ceph / src / boost / boost / locale / generic_codecvt.hpp
1 //
2 // Copyright (c) 2015 Artyom Beilis (Tonkikh)
3 //
4 // Distributed under the Boost Software License, Version 1.0. (See
5 // accompanying file LICENSE_1_0.txt or copy at
6 // http://www.boost.org/LICENSE_1_0.txt)
7 //
8 #ifndef BOOST_LOCALE_GENERIC_CODECVT_HPP
9 #define BOOST_LOCALE_GENERIC_CODECVT_HPP
10
11 #include <boost/locale/utf.hpp>
12 #include <boost/cstdint.hpp>
13 #include <boost/static_assert.hpp>
14 #include <locale>
15
16 namespace boost {
17 namespace locale {
18
19 #ifndef BOOST_LOCALE_DOXYGEN
20 //
21 // Make sure that mbstate can keep 16 bit of UTF-16 sequence
22 //
23 BOOST_STATIC_ASSERT(sizeof(std::mbstate_t)>=2);
24 #endif
25
26 #if defined(_MSC_VER) && _MSC_VER < 1700
27 // up to MSVC 11 (2012) do_length is non-standard it counts wide characters instead of narrow and does not change mbstate
28 #define BOOST_LOCALE_DO_LENGTH_MBSTATE_CONST
29 #endif
30
31 ///
32 /// \brief A base class that used to define constants for generic_codecvt
33 ///
34 class generic_codecvt_base {
35 public:
36 ///
37 /// Initail state for converting to or from unicode code points, used by initial_state in derived classes
38 ///
39 enum initial_convertion_state {
40 to_unicode_state, ///< The state would be used by to_unicode functions
41 from_unicode_state ///< The state would be used by from_unicode functions
42 };
43 };
44
45 ///
46 /// \brief Geneneric generic codecvt facet, various stateless encodings to UTF-16 and UTF-32 using wchar_t, char32_t and char16_t
47 ///
48 /// Implementations should dervide from this class defining itself as CodecvtImpl and provide following members
49 ///
50 /// - `state_type` - a type of special object that allows to store intermediate cached data, for example `iconv_t` descriptor
51 /// - `state_type initial_state(generic_codecvt_base::initial_convertion_state direction) const` - member function that creates initial state
52 /// - `int max_encoding_length() const` - a maximal length that one Unicode code point is represented, for UTF-8 for example it is 4 from ISO-8859-1 it is 1
53 /// - `utf::code_point to_unicode(state_type &state,char const *&begin,char const *end)` - extract first code point from the text in range [begin,end), in case of success begin would point to the next character sequence to be encoded to next code point, in case of incomplete sequence - utf::incomplete shell be returned, and in case of invalid input sequence utf::illegal shell be returned and begin would remain unmodified
54 /// - `utf::code_point from_unicode(state_type &state,utf::code_point u,char *begin,char const *end)` - convert a unicode code point `u` into a character seqnece at [begin,end). Return the length of the sequence in case of success, utf::incomplete in case of not enough room to encode the code point of utf::illegal in case conversion can not be performed
55 ///
56 ///
57 /// For example implementaion of codecvt for latin1/ISO-8859-1 character set
58 ///
59 /// \code
60 ///
61 /// template<typename CharType>
62 /// class latin1_codecvt :boost::locale::generic_codecvt<CharType,latin1_codecvt<CharType> >
63 /// {
64 /// public:
65 ///
66 /// /* Standard codecvt constructor */
67 /// latin1_codecvt(size_t refs = 0) : boost::locale::generic_codecvt<CharType,latin1_codecvt<CharType> >(refs)
68 /// {
69 /// }
70 ///
71 /// /* State is unused but required by generic_codecvt */
72 /// struct state_type {};
73 ///
74 /// state_type initial_state(generic_codecvt_base::initial_convertion_state /*unused*/) const
75 /// {
76 /// return state_type();
77 /// }
78 ///
79 /// int max_encoding_length() const
80 /// {
81 /// return 1;
82 /// }
83 ///
84 /// boost::locale::utf::code_point to_unicode(state_type &,char const *&begin,char const *end) const
85 /// {
86 /// if(begin == end)
87 /// return boost::locale::utf::incomplete;
88 /// return *begin++;
89 /// }
90 ///
91 /// boost::locale::utf::code_point from_unicode(state_type &,boost::locale::utf::code_point u,char *begin,char const *end) const
92 /// {
93 /// if(u >= 256)
94 /// return boost::locale::utf::illegal;
95 /// if(begin == end)
96 /// return boost::locale::utf::incomplete;
97 /// *begin = u;
98 /// return 1;
99 /// }
100 /// };
101 ///
102 /// \endcode
103 ///
104 /// When external tools used for encoding conversion, the `state_type` is useful to save objects used for conversions. For example,
105 /// icu::UConverter can be saved in such a state for an efficient use:
106 ///
107 /// \code
108 /// template<typename CharType>
109 /// class icu_codecvt :boost::locale::generic_codecvt<CharType,icu_codecvt<CharType> >
110 /// {
111 /// public:
112 ///
113 /// /* Standard codecvt constructor */
114 /// icu_codecvt(std::string const &name,refs = 0) :
115 /// boost::locale::generic_codecvt<CharType,latin1_codecvt<CharType> >(refs)
116 /// { ... }
117 ///
118 /// /* State is unused but required by generic_codecvt */
119 /// struct std::unique_ptr<UConverter,void (*)(UConverter*)> state_type;
120 ///
121 /// state_type &&initial_state(generic_codecvt_base::initial_convertion_state /*unused*/) const
122 /// {
123 /// UErrorCode err = U_ZERO_ERROR;
124 /// state_type ptr(ucnv_safeClone(converter_,0,0,&err,ucnv_close);
125 /// return std::move(ptr);
126 /// }
127 ///
128 /// boost::locale::utf::code_point to_unicode(state_type &ptr,char const *&begin,char const *end) const
129 /// {
130 /// UErrorCode err = U_ZERO_ERROR;
131 /// boost::locale::utf::code_point cp = ucnv_getNextUChar(ptr.get(),&begin,end,&err);
132 /// ...
133 /// }
134 /// ...
135 /// };
136 /// \endcode
137 ///
138 ///
139 template<typename CharType,typename CodecvtImpl,int CharSize=sizeof(CharType)>
140 class generic_codecvt;
141
142 ///
143 /// \brief UTF-16 to/from UTF-8 codecvt facet to use with char16_t or wchar_t on Windows
144 ///
145 /// Note in order to fit the requirements of usability by std::wfstream it uses mbstate_t
146 /// to handle intermediate states in handling of variable length UTF-16 sequences
147 ///
148 /// Its member functions implement standard virtual functions of basic codecvt
149 ///
150 template<typename CharType,typename CodecvtImpl>
151 class generic_codecvt<CharType,CodecvtImpl,2> : public std::codecvt<CharType,char,std::mbstate_t>, public generic_codecvt_base
152 {
153 public:
154
155 typedef CharType uchar;
156
157 generic_codecvt(size_t refs = 0) :
158 std::codecvt<CharType,char,std::mbstate_t>(refs)
159 {
160 }
161 CodecvtImpl const &implementation() const
162 {
163 return *static_cast<CodecvtImpl const *>(this);
164 }
165
166 protected:
167
168
169 virtual std::codecvt_base::result do_unshift(std::mbstate_t &s,char *from,char * /*to*/,char *&next) const
170 {
171 boost::uint16_t &state = *reinterpret_cast<boost::uint16_t *>(&s);
172 #ifdef DEBUG_CODECVT
173 std::cout << "Entering unshift " << std::hex << state << std::dec << std::endl;
174 #endif
175 if(state != 0)
176 return std::codecvt_base::error;
177 next=from;
178 return std::codecvt_base::ok;
179 }
180 virtual int do_encoding() const throw()
181 {
182 return 0;
183 }
184 virtual int do_max_length() const throw()
185 {
186 return implementation().max_encoding_length();
187 }
188 virtual bool do_always_noconv() const throw()
189 {
190 return false;
191 }
192
193 virtual int
194 do_length( std::mbstate_t
195 #ifdef BOOST_LOCALE_DO_LENGTH_MBSTATE_CONST
196 const
197 #endif
198 &std_state,
199 char const *from,
200 char const *from_end,
201 size_t max) const
202 {
203 #ifndef BOOST_LOCALE_DO_LENGTH_MBSTATE_CONST
204 char const *save_from = from;
205 boost::uint16_t &state = *reinterpret_cast<boost::uint16_t *>(&std_state);
206 #else
207 size_t save_max = max;
208 boost::uint16_t state = *reinterpret_cast<boost::uint16_t const *>(&std_state);
209 #endif
210
211 typedef typename CodecvtImpl::state_type state_type;
212 state_type cvt_state = implementation().initial_state(generic_codecvt_base::to_unicode_state);
213 while(max > 0 && from < from_end){
214 char const *prev_from = from;
215 boost::uint32_t ch=implementation().to_unicode(cvt_state,from,from_end);
216 if(ch==boost::locale::utf::incomplete || ch==boost::locale::utf::illegal) {
217 from = prev_from;
218 break;
219 }
220 max --;
221 if(ch > 0xFFFF) {
222 if(state == 0) {
223 from = prev_from;
224 state = 1;
225 }
226 else {
227 state = 0;
228 }
229 }
230 }
231 #ifndef BOOST_LOCALE_DO_LENGTH_MBSTATE_CONST
232 return from - save_from;
233 #else
234 return save_max - max;
235 #endif
236 }
237
238
239 virtual std::codecvt_base::result
240 do_in( std::mbstate_t &std_state,
241 char const *from,
242 char const *from_end,
243 char const *&from_next,
244 uchar *to,
245 uchar *to_end,
246 uchar *&to_next) const
247 {
248 std::codecvt_base::result r=std::codecvt_base::ok;
249
250 // mbstate_t is POD type and should be initialized to 0 (i.a. state = stateT())
251 // according to standard. We use it to keep a flag 0/1 for surrogate pair writing
252 //
253 // if 0 no code above >0xFFFF observed, of 1 a code above 0xFFFF observerd
254 // and first pair is written, but no input consumed
255 boost::uint16_t &state = *reinterpret_cast<boost::uint16_t *>(&std_state);
256 typedef typename CodecvtImpl::state_type state_type;
257 state_type cvt_state = implementation().initial_state(generic_codecvt_base::to_unicode_state);
258 while(to < to_end && from < from_end)
259 {
260 #ifdef DEBUG_CODECVT
261 std::cout << "Entering IN--------------" << std::endl;
262 std::cout << "State " << std::hex << state <<std::endl;
263 std::cout << "Left in " << std::dec << from_end - from << " out " << to_end -to << std::endl;
264 #endif
265 char const *from_saved = from;
266
267 uint32_t ch=implementation().to_unicode(cvt_state,from,from_end);
268
269 if(ch==boost::locale::utf::illegal) {
270 from = from_saved;
271 r=std::codecvt_base::error;
272 break;
273 }
274 if(ch==boost::locale::utf::incomplete) {
275 from = from_saved;
276 r=std::codecvt_base::partial;
277 break;
278 }
279 // Normal codepoints go direcly to stream
280 if(ch <= 0xFFFF) {
281 *to++=ch;
282 }
283 else {
284 // for other codepoints we do following
285 //
286 // 1. We can't consume our input as we may find ourselfs
287 // in state where all input consumed but not all output written,i.e. only
288 // 1st pair is written
289 // 2. We only write first pair and mark this in the state, we also revert back
290 // the from pointer in order to make sure this codepoint would be read
291 // once again and then we would consume our input together with writing
292 // second surrogate pair
293 ch-=0x10000;
294 boost::uint16_t vh = ch >> 10;
295 boost::uint16_t vl = ch & 0x3FF;
296 boost::uint16_t w1 = vh + 0xD800;
297 boost::uint16_t w2 = vl + 0xDC00;
298 if(state == 0) {
299 from = from_saved;
300 *to++ = w1;
301 state = 1;
302 }
303 else {
304 *to++ = w2;
305 state = 0;
306 }
307 }
308 }
309 from_next=from;
310 to_next=to;
311 if(r == std::codecvt_base::ok && (from!=from_end || state!=0))
312 r = std::codecvt_base::partial;
313 #ifdef DEBUG_CODECVT
314 std::cout << "Returning ";
315 switch(r) {
316 case std::codecvt_base::ok:
317 std::cout << "ok" << std::endl;
318 break;
319 case std::codecvt_base::partial:
320 std::cout << "partial" << std::endl;
321 break;
322 case std::codecvt_base::error:
323 std::cout << "error" << std::endl;
324 break;
325 default:
326 std::cout << "other" << std::endl;
327 break;
328 }
329 std::cout << "State " << std::hex << state <<std::endl;
330 std::cout << "Left in " << std::dec << from_end - from << " out " << to_end -to << std::endl;
331 #endif
332 return r;
333 }
334
335 virtual std::codecvt_base::result
336 do_out( std::mbstate_t &std_state,
337 uchar const *from,
338 uchar const *from_end,
339 uchar const *&from_next,
340 char *to,
341 char *to_end,
342 char *&to_next) const
343 {
344 std::codecvt_base::result r=std::codecvt_base::ok;
345 // mbstate_t is POD type and should be initialized to 0 (i.a. state = stateT())
346 // according to standard. We assume that sizeof(mbstate_t) >=2 in order
347 // to be able to store first observerd surrogate pair
348 //
349 // State: state!=0 - a first surrogate pair was observerd (state = first pair),
350 // we expect the second one to come and then zero the state
351 ///
352 boost::uint16_t &state = *reinterpret_cast<boost::uint16_t *>(&std_state);
353 typedef typename CodecvtImpl::state_type state_type;
354 state_type cvt_state = implementation().initial_state(generic_codecvt_base::from_unicode_state);
355 while(to < to_end && from < from_end)
356 {
357 #ifdef DEBUG_CODECVT
358 std::cout << "Entering OUT --------------" << std::endl;
359 std::cout << "State " << std::hex << state <<std::endl;
360 std::cout << "Left in " << std::dec << from_end - from << " out " << to_end -to << std::endl;
361 #endif
362 boost::uint32_t ch=0;
363 if(state != 0) {
364 // if the state idecates that 1st surrogate pair was written
365 // we should make sure that the second one that comes is actually
366 // second surrogate
367 boost::uint16_t w1 = state;
368 boost::uint16_t w2 = *from;
369 // we don't forward from as writing may fail to incomplete or
370 // partial conversion
371 if(0xDC00 <= w2 && w2<=0xDFFF) {
372 boost::uint16_t vh = w1 - 0xD800;
373 boost::uint16_t vl = w2 - 0xDC00;
374 ch=((uint32_t(vh) << 10) | vl) + 0x10000;
375 }
376 else {
377 // Invalid surrogate
378 r=std::codecvt_base::error;
379 break;
380 }
381 }
382 else {
383 ch = *from;
384 if(0xD800 <= ch && ch<=0xDBFF) {
385 // if this is a first surrogate pair we put
386 // it into the state and consume it, note we don't
387 // go forward as it should be illegal so we increase
388 // the from pointer manually
389 state = ch;
390 from++;
391 continue;
392 }
393 else if(0xDC00 <= ch && ch<=0xDFFF) {
394 // if we observe second surrogate pair and
395 // first only may be expected we should break from the loop with error
396 // as it is illegal input
397 r=std::codecvt_base::error;
398 break;
399 }
400 }
401 if(!boost::locale::utf::is_valid_codepoint(ch)) {
402 r=std::codecvt_base::error;
403 break;
404 }
405 boost::uint32_t len = implementation().from_unicode(cvt_state,ch,to,to_end);
406 if(len == boost::locale::utf::incomplete) {
407 r=std::codecvt_base::partial;
408 break;
409 }
410 else if(len == boost::locale::utf::illegal) {
411 r=std::codecvt_base::error;
412 break;
413 }
414 else
415 to+= len;
416 state = 0;
417 from++;
418 }
419 from_next=from;
420 to_next=to;
421 if(r==std::codecvt_base::ok && from!=from_end)
422 r = std::codecvt_base::partial;
423 #ifdef DEBUG_CODECVT
424 std::cout << "Returning ";
425 switch(r) {
426 case std::codecvt_base::ok:
427 std::cout << "ok" << std::endl;
428 break;
429 case std::codecvt_base::partial:
430 std::cout << "partial" << std::endl;
431 break;
432 case std::codecvt_base::error:
433 std::cout << "error" << std::endl;
434 break;
435 default:
436 std::cout << "other" << std::endl;
437 break;
438 }
439 std::cout << "State " << std::hex << state <<std::endl;
440 std::cout << "Left in " << std::dec << from_end - from << " out " << to_end -to << std::endl;
441 #endif
442 return r;
443 }
444
445 };
446
447 ///
448 /// \brief UTF-32 to/from UTF-8 codecvt facet to use with char32_t or wchar_t on POSIX platforms
449 ///
450 /// Its member functions implement standard virtual functions of basic codecvt.
451 /// mbstate_t is not used for UTF-32 handling due to fixed length encoding
452 ///
453 template<typename CharType,typename CodecvtImpl>
454 class generic_codecvt<CharType,CodecvtImpl,4> : public std::codecvt<CharType,char,std::mbstate_t>, public generic_codecvt_base
455 {
456 public:
457 typedef CharType uchar;
458
459 generic_codecvt(size_t refs = 0) :
460 std::codecvt<CharType,char,std::mbstate_t>(refs)
461 {
462 }
463
464 CodecvtImpl const &implementation() const
465 {
466 return *static_cast<CodecvtImpl const *>(this);
467 }
468
469 protected:
470
471 virtual std::codecvt_base::result do_unshift(std::mbstate_t &/*s*/,char *from,char * /*to*/,char *&next) const
472 {
473 next=from;
474 return std::codecvt_base::ok;
475 }
476 virtual int do_encoding() const throw()
477 {
478 return 0;
479 }
480 virtual int do_max_length() const throw()
481 {
482 return implementation().max_encoding_length();
483 }
484 virtual bool do_always_noconv() const throw()
485 {
486 return false;
487 }
488
489 virtual int
490 do_length( std::mbstate_t
491 #ifdef BOOST_LOCALE_DO_LENGTH_MBSTATE_CONST
492 const
493 #endif
494 &/*state*/,
495 char const *from,
496 char const *from_end,
497 size_t max) const
498 {
499 #ifndef BOOST_LOCALE_DO_LENGTH_MBSTATE_CONST
500 char const *start_from = from;
501 #else
502 size_t save_max = max;
503 #endif
504 typedef typename CodecvtImpl::state_type state_type;
505 state_type cvt_state = implementation().initial_state(generic_codecvt_base::to_unicode_state);
506 while(max > 0 && from < from_end){
507 char const *save_from = from;
508 boost::uint32_t ch=implementation().to_unicode(cvt_state,from,from_end);
509 if(ch==boost::locale::utf::incomplete || ch==boost::locale::utf::illegal) {
510 from = save_from;
511 break;
512 }
513 max--;
514 }
515 #ifndef BOOST_LOCALE_DO_LENGTH_MBSTATE_CONST
516 return from - start_from;
517 #else
518 return save_max - max;
519 #endif
520 }
521
522
523 virtual std::codecvt_base::result
524 do_in( std::mbstate_t &/*state*/,
525 char const *from,
526 char const *from_end,
527 char const *&from_next,
528 uchar *to,
529 uchar *to_end,
530 uchar *&to_next) const
531 {
532 std::codecvt_base::result r=std::codecvt_base::ok;
533
534 // mbstate_t is POD type and should be initialized to 0 (i.a. state = stateT())
535 // according to standard. We use it to keep a flag 0/1 for surrogate pair writing
536 //
537 // if 0 no code above >0xFFFF observed, of 1 a code above 0xFFFF observerd
538 // and first pair is written, but no input consumed
539 typedef typename CodecvtImpl::state_type state_type;
540 state_type cvt_state = implementation().initial_state(generic_codecvt_base::to_unicode_state);
541 while(to < to_end && from < from_end)
542 {
543 #ifdef DEBUG_CODECVT
544 std::cout << "Entering IN--------------" << std::endl;
545 std::cout << "State " << std::hex << state <<std::endl;
546 std::cout << "Left in " << std::dec << from_end - from << " out " << to_end -to << std::endl;
547 #endif
548 char const *from_saved = from;
549
550 uint32_t ch=implementation().to_unicode(cvt_state,from,from_end);
551
552 if(ch==boost::locale::utf::illegal) {
553 r=std::codecvt_base::error;
554 from = from_saved;
555 break;
556 }
557 if(ch==boost::locale::utf::incomplete) {
558 r=std::codecvt_base::partial;
559 from=from_saved;
560 break;
561 }
562 *to++=ch;
563 }
564 from_next=from;
565 to_next=to;
566 if(r == std::codecvt_base::ok && from!=from_end)
567 r = std::codecvt_base::partial;
568 #ifdef DEBUG_CODECVT
569 std::cout << "Returning ";
570 switch(r) {
571 case std::codecvt_base::ok:
572 std::cout << "ok" << std::endl;
573 break;
574 case std::codecvt_base::partial:
575 std::cout << "partial" << std::endl;
576 break;
577 case std::codecvt_base::error:
578 std::cout << "error" << std::endl;
579 break;
580 default:
581 std::cout << "other" << std::endl;
582 break;
583 }
584 std::cout << "State " << std::hex << state <<std::endl;
585 std::cout << "Left in " << std::dec << from_end - from << " out " << to_end -to << std::endl;
586 #endif
587 return r;
588 }
589
590 virtual std::codecvt_base::result
591 do_out( std::mbstate_t &/*std_state*/,
592 uchar const *from,
593 uchar const *from_end,
594 uchar const *&from_next,
595 char *to,
596 char *to_end,
597 char *&to_next) const
598 {
599 std::codecvt_base::result r=std::codecvt_base::ok;
600 typedef typename CodecvtImpl::state_type state_type;
601 state_type cvt_state = implementation().initial_state(generic_codecvt_base::from_unicode_state);
602 while(to < to_end && from < from_end)
603 {
604 #ifdef DEBUG_CODECVT
605 std::cout << "Entering OUT --------------" << std::endl;
606 std::cout << "State " << std::hex << state <<std::endl;
607 std::cout << "Left in " << std::dec << from_end - from << " out " << to_end -to << std::endl;
608 #endif
609 boost::uint32_t ch=0;
610 ch = *from;
611 if(!boost::locale::utf::is_valid_codepoint(ch)) {
612 r=std::codecvt_base::error;
613 break;
614 }
615 boost::uint32_t len = implementation().from_unicode(cvt_state,ch,to,to_end);
616 if(len == boost::locale::utf::incomplete) {
617 r=std::codecvt_base::partial;
618 break;
619 }
620 else if(len == boost::locale::utf::illegal) {
621 r=std::codecvt_base::error;
622 break;
623 }
624 to+=len;
625 from++;
626 }
627 from_next=from;
628 to_next=to;
629 if(r==std::codecvt_base::ok && from!=from_end)
630 r = std::codecvt_base::partial;
631 #ifdef DEBUG_CODECVT
632 std::cout << "Returning ";
633 switch(r) {
634 case std::codecvt_base::ok:
635 std::cout << "ok" << std::endl;
636 break;
637 case std::codecvt_base::partial:
638 std::cout << "partial" << std::endl;
639 break;
640 case std::codecvt_base::error:
641 std::cout << "error" << std::endl;
642 break;
643 default:
644 std::cout << "other" << std::endl;
645 break;
646 }
647 std::cout << "State " << std::hex << state <<std::endl;
648 std::cout << "Left in " << std::dec << from_end - from << " out " << to_end -to << std::endl;
649 #endif
650 return r;
651 }
652 };
653
654
655 template<typename CharType,typename CodecvtImpl>
656 class generic_codecvt<CharType,CodecvtImpl,1> : public std::codecvt<CharType,char,std::mbstate_t>, public generic_codecvt_base
657 {
658 public:
659 typedef CharType uchar;
660
661 CodecvtImpl const &implementation() const
662 {
663 return *static_cast<CodecvtImpl const *>(this);
664 }
665
666 generic_codecvt(size_t refs = 0) : std::codecvt<char,char,std::mbstate_t>(refs)
667 {
668 }
669 };
670
671 } // locale
672 } // namespace boost
673
674 #endif
675 ///
676 // vim: tabstop=4 expandtab shiftwidth=4 softtabstop=4