libstdc++
regex_compiler.tcc
Go to the documentation of this file.
1// class template regex -*- C++ -*-
2
3// Copyright (C) 2013-2024 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 * @file bits/regex_compiler.tcc
27 * This is an internal header file, included by other library headers.
28 * Do not attempt to use it directly. @headername{regex}
29 */
30
31// FIXME make comments doxygen format.
32
33/*
34// This compiler refers to "Regular Expression Matching Can Be Simple And Fast"
35// (http://swtch.com/~rsc/regexp/regexp1.html),
36// but doesn't strictly follow it.
37//
38// When compiling, states are *chained* instead of tree- or graph-constructed.
39// It's more like structured programs: there's if statement and loop statement.
40//
41// For alternative structure (say "a|b"), aka "if statement", two branches
42// should be constructed. However, these two shall merge to an "end_tag" at
43// the end of this operator:
44//
45// branch1
46// / \
47// => begin_tag end_tag =>
48// \ /
49// branch2
50//
51// This is the difference between this implementation and that in Russ's
52// article.
53//
54// That's why we introduced dummy node here ------ "end_tag" is a dummy node.
55// All dummy nodes will be eliminated at the end of compilation.
56*/
57
58namespace std _GLIBCXX_VISIBILITY(default)
59{
60_GLIBCXX_BEGIN_NAMESPACE_VERSION
61
62namespace __detail
63{
64 template<typename _TraitsT>
65 _Compiler<_TraitsT>::
66 _Compiler(const _CharT* __b, const _CharT* __e,
67 const typename _TraitsT::locale_type& __loc, _FlagT __flags)
68 : _M_flags(_S_validate(__flags)),
69 _M_scanner(__b, __e, _M_flags, __loc),
70 _M_nfa(make_shared<_RegexT>(__loc, _M_flags)),
71 _M_traits(_M_nfa->_M_traits),
72 _M_ctype(std::use_facet<_CtypeT>(__loc))
73 {
74 _StateSeqT __r(*_M_nfa, _M_nfa->_M_start());
75 __r._M_append(_M_nfa->_M_insert_subexpr_begin());
76 this->_M_disjunction();
77 if (!_M_match_token(_ScannerT::_S_token_eof))
78 __throw_regex_error(regex_constants::error_paren);
79 __r._M_append(_M_pop());
80 __glibcxx_assert(_M_stack.empty());
81 __r._M_append(_M_nfa->_M_insert_subexpr_end());
82 __r._M_append(_M_nfa->_M_insert_accept());
83 _M_nfa->_M_eliminate_dummy();
84 }
85
86 template<typename _TraitsT>
87 void
88 _Compiler<_TraitsT>::
89 _M_disjunction()
90 {
91 this->_M_alternative();
92 while (_M_match_token(_ScannerT::_S_token_or))
93 {
94 _StateSeqT __alt1 = _M_pop();
95 this->_M_alternative();
96 _StateSeqT __alt2 = _M_pop();
97 auto __end = _M_nfa->_M_insert_dummy();
98 __alt1._M_append(__end);
99 __alt2._M_append(__end);
100 // __alt2 is state._M_next, __alt1 is state._M_alt. The executor
101 // executes _M_alt before _M_next, as well as executing left
102 // alternative before right one.
103 _M_stack.push(_StateSeqT(*_M_nfa,
104 _M_nfa->_M_insert_alt(
105 __alt2._M_start, __alt1._M_start, false),
106 __end));
107 }
108 }
109
110 template<typename _TraitsT>
111 void
112 _Compiler<_TraitsT>::
113 _M_alternative()
114 {
115 if (this->_M_term())
116 {
117 _StateSeqT __re = _M_pop();
118 this->_M_alternative();
119 __re._M_append(_M_pop());
120 _M_stack.push(__re);
121 }
122 else
123 _M_stack.push(_StateSeqT(*_M_nfa, _M_nfa->_M_insert_dummy()));
124 }
125
126 template<typename _TraitsT>
127 bool
128 _Compiler<_TraitsT>::
129 _M_term()
130 {
131 if (this->_M_assertion())
132 return true;
133 if (this->_M_atom())
134 {
135 while (this->_M_quantifier())
136 ;
137 return true;
138 }
139 return false;
140 }
141
142 template<typename _TraitsT>
143 bool
144 _Compiler<_TraitsT>::
145 _M_assertion()
146 {
147 if (_M_match_token(_ScannerT::_S_token_line_begin))
148 _M_stack.push(_StateSeqT(*_M_nfa, _M_nfa->_M_insert_line_begin()));
149 else if (_M_match_token(_ScannerT::_S_token_line_end))
150 _M_stack.push(_StateSeqT(*_M_nfa, _M_nfa->_M_insert_line_end()));
151 else if (_M_match_token(_ScannerT::_S_token_word_bound))
152 // _M_value[0] == 'n' means it's negative, say "not word boundary".
153 _M_stack.push(_StateSeqT(*_M_nfa, _M_nfa->
154 _M_insert_word_bound(_M_value[0] == 'n')));
155 else if (_M_match_token(_ScannerT::_S_token_subexpr_lookahead_begin))
156 {
157 auto __neg = _M_value[0] == 'n';
158 this->_M_disjunction();
159 if (!_M_match_token(_ScannerT::_S_token_subexpr_end))
160 __throw_regex_error(regex_constants::error_paren);
161 auto __tmp = _M_pop();
162 __tmp._M_append(_M_nfa->_M_insert_accept());
163 _M_stack.push(
164 _StateSeqT(
165 *_M_nfa,
166 _M_nfa->_M_insert_lookahead(__tmp._M_start, __neg)));
167 }
168 else
169 return false;
170 return true;
171 }
172
173 template<typename _TraitsT>
174 bool
175 _Compiler<_TraitsT>::
176 _M_quantifier()
177 {
178 bool __neg = (_M_flags & regex_constants::ECMAScript);
179 auto __init = [this, &__neg]()
180 {
181 if (_M_stack.empty())
182 __throw_regex_error(regex_constants::error_badrepeat);
183 __neg = __neg && _M_match_token(_ScannerT::_S_token_opt);
184 };
185 if (_M_match_token(_ScannerT::_S_token_closure0))
186 {
187 __init();
188 auto __e = _M_pop();
189 _StateSeqT __r(*_M_nfa,
190 _M_nfa->_M_insert_repeat(_S_invalid_state_id,
191 __e._M_start, __neg));
192 __e._M_append(__r);
193 _M_stack.push(__r);
194 }
195 else if (_M_match_token(_ScannerT::_S_token_closure1))
196 {
197 __init();
198 auto __e = _M_pop();
199 __e._M_append(_M_nfa->_M_insert_repeat(_S_invalid_state_id,
200 __e._M_start, __neg));
201 _M_stack.push(__e);
202 }
203 else if (_M_match_token(_ScannerT::_S_token_opt))
204 {
205 __init();
206 auto __e = _M_pop();
207 auto __end = _M_nfa->_M_insert_dummy();
208 _StateSeqT __r(*_M_nfa,
209 _M_nfa->_M_insert_repeat(_S_invalid_state_id,
210 __e._M_start, __neg));
211 __e._M_append(__end);
212 __r._M_append(__end);
213 _M_stack.push(__r);
214 }
215 else if (_M_match_token(_ScannerT::_S_token_interval_begin))
216 {
217 if (_M_stack.empty())
218 __throw_regex_error(regex_constants::error_badrepeat);
219 if (!_M_match_token(_ScannerT::_S_token_dup_count))
220 __throw_regex_error(regex_constants::error_badbrace);
221 _StateSeqT __r(_M_pop());
222 _StateSeqT __e(*_M_nfa, _M_nfa->_M_insert_dummy());
223 long __min_rep = _M_cur_int_value(10);
224 bool __infi = false;
225 long __n = 0;
226
227 // {3
228 if (_M_match_token(_ScannerT::_S_token_comma))
229 {
230 if (_M_match_token(_ScannerT::_S_token_dup_count)) // {3,7}
231 __n = _M_cur_int_value(10) - __min_rep;
232 else
233 __infi = true;
234 }
235 if (!_M_match_token(_ScannerT::_S_token_interval_end))
236 __throw_regex_error(regex_constants::error_brace);
237
238 __neg = __neg && _M_match_token(_ScannerT::_S_token_opt);
239
240 for (long __i = 0; __i < __min_rep; ++__i)
241 __e._M_append(__r._M_clone());
242
243 if (__infi)
244 {
245 auto __tmp = __r._M_clone();
246 _StateSeqT __s(*_M_nfa,
247 _M_nfa->_M_insert_repeat(_S_invalid_state_id,
248 __tmp._M_start, __neg));
249 __tmp._M_append(__s);
250 __e._M_append(__s);
251 }
252 else
253 {
254 if (__n < 0)
255 __throw_regex_error(regex_constants::error_badbrace);
256 auto __end = _M_nfa->_M_insert_dummy();
257 // _M_alt is the "match more" branch, and _M_next is the
258 // "match less" one. Switch _M_alt and _M_next of all created
259 // nodes. This is a hack but IMO works well.
260 std::stack<_StateIdT> __stack;
261 for (long __i = 0; __i < __n; ++__i)
262 {
263 auto __tmp = __r._M_clone();
264 auto __alt = _M_nfa->_M_insert_repeat(__tmp._M_start,
265 __end, __neg);
266 __stack.push(__alt);
267 __e._M_append(_StateSeqT(*_M_nfa, __alt, __tmp._M_end));
268 }
269 __e._M_append(__end);
270 while (!__stack.empty())
271 {
272 auto& __tmp = (*_M_nfa)[__stack.top()];
273 __stack.pop();
274 std::swap(__tmp._M_next, __tmp._M_alt);
275 }
276 }
277 _M_stack.push(__e);
278 }
279 else
280 return false;
281 return true;
282 }
283
284#define __INSERT_REGEX_MATCHER(__func, ...)\
285 do {\
286 if (!(_M_flags & regex_constants::icase))\
287 if (!(_M_flags & regex_constants::collate))\
288 __func<false, false>(__VA_ARGS__);\
289 else\
290 __func<false, true>(__VA_ARGS__);\
291 else\
292 if (!(_M_flags & regex_constants::collate))\
293 __func<true, false>(__VA_ARGS__);\
294 else\
295 __func<true, true>(__VA_ARGS__);\
296 } while (false)
297
298 template<typename _TraitsT>
299 bool
300 _Compiler<_TraitsT>::
301 _M_atom()
302 {
303 if (_M_match_token(_ScannerT::_S_token_anychar))
304 {
305 if (!(_M_flags & regex_constants::ECMAScript))
306 __INSERT_REGEX_MATCHER(_M_insert_any_matcher_posix);
307 else
308 __INSERT_REGEX_MATCHER(_M_insert_any_matcher_ecma);
309 }
310 else if (_M_try_char())
311 __INSERT_REGEX_MATCHER(_M_insert_char_matcher);
312 else if (_M_match_token(_ScannerT::_S_token_backref))
313 _M_stack.push(_StateSeqT(*_M_nfa, _M_nfa->
314 _M_insert_backref(_M_cur_int_value(10))));
315 else if (_M_match_token(_ScannerT::_S_token_quoted_class))
316 __INSERT_REGEX_MATCHER(_M_insert_character_class_matcher);
317 else if (_M_match_token(_ScannerT::_S_token_subexpr_no_group_begin))
318 {
319 _StateSeqT __r(*_M_nfa, _M_nfa->_M_insert_dummy());
320 this->_M_disjunction();
321 if (!_M_match_token(_ScannerT::_S_token_subexpr_end))
322 __throw_regex_error(regex_constants::error_paren);
323 __r._M_append(_M_pop());
324 _M_stack.push(__r);
325 }
326 else if (_M_match_token(_ScannerT::_S_token_subexpr_begin))
327 {
328 _StateSeqT __r(*_M_nfa, _M_nfa->_M_insert_subexpr_begin());
329 this->_M_disjunction();
330 if (!_M_match_token(_ScannerT::_S_token_subexpr_end))
331 __throw_regex_error(regex_constants::error_paren);
332 __r._M_append(_M_pop());
333 __r._M_append(_M_nfa->_M_insert_subexpr_end());
334 _M_stack.push(__r);
335 }
336 else if (!_M_bracket_expression())
337 return false;
338 return true;
339 }
340
341 template<typename _TraitsT>
342 bool
343 _Compiler<_TraitsT>::
344 _M_bracket_expression()
345 {
346 bool __neg =
347 _M_match_token(_ScannerT::_S_token_bracket_neg_begin);
348 if (!(__neg || _M_match_token(_ScannerT::_S_token_bracket_begin)))
349 return false;
350 __INSERT_REGEX_MATCHER(_M_insert_bracket_matcher, __neg);
351 return true;
352 }
353#undef __INSERT_REGEX_MATCHER
354
355 template<typename _TraitsT>
356 template<bool __icase, bool __collate>
357 void
358 _Compiler<_TraitsT>::
359 _M_insert_any_matcher_ecma()
360 {
361 _M_stack.push(_StateSeqT(*_M_nfa,
362 _M_nfa->_M_insert_matcher
363 (_AnyMatcher<_TraitsT, true, __icase, __collate>
364 (_M_traits))));
365 }
366
367 template<typename _TraitsT>
368 template<bool __icase, bool __collate>
369 void
370 _Compiler<_TraitsT>::
371 _M_insert_any_matcher_posix()
372 {
373 _M_stack.push(_StateSeqT(*_M_nfa,
374 _M_nfa->_M_insert_matcher
375 (_AnyMatcher<_TraitsT, false, __icase, __collate>
376 (_M_traits))));
377 }
378
379 template<typename _TraitsT>
380 template<bool __icase, bool __collate>
381 void
382 _Compiler<_TraitsT>::
383 _M_insert_char_matcher()
384 {
385 _M_stack.push(_StateSeqT(*_M_nfa,
386 _M_nfa->_M_insert_matcher
387 (_CharMatcher<_TraitsT, __icase, __collate>
388 (_M_value[0], _M_traits))));
389 }
390
391 template<typename _TraitsT>
392 template<bool __icase, bool __collate>
393 void
394 _Compiler<_TraitsT>::
395 _M_insert_character_class_matcher()
396 {
397 __glibcxx_assert(_M_value.size() == 1);
398 _BracketMatcher<__icase, __collate> __matcher
399 (_M_ctype.is(_CtypeT::upper, _M_value[0]), _M_traits);
400 __matcher._M_add_character_class(_M_value, false);
401 __matcher._M_ready();
402 _M_stack.push(_StateSeqT(*_M_nfa,
403 _M_nfa->_M_insert_matcher(std::move(__matcher))));
404 }
405
406 template<typename _TraitsT>
407 template<bool __icase, bool __collate>
408 void
409 _Compiler<_TraitsT>::
410 _M_insert_bracket_matcher(bool __neg)
411 {
412 _BracketMatcher<__icase, __collate> __matcher(__neg, _M_traits);
413 _BracketState __last_char;
414 if (_M_try_char())
415 __last_char.set(_M_value[0]);
416 else if (_M_match_token(_ScannerT::_S_token_bracket_dash))
417 // Dash as first character is a normal character.
418 __last_char.set('-');
419 while (_M_expression_term(__last_char, __matcher))
420 ;
421 if (__last_char._M_is_char())
422 __matcher._M_add_char(__last_char.get());
423 __matcher._M_ready();
424 _M_stack.push(_StateSeqT(
425 *_M_nfa,
426 _M_nfa->_M_insert_matcher(std::move(__matcher))));
427 }
428
429 template<typename _TraitsT>
430 template<bool __icase, bool __collate>
431 bool
432 _Compiler<_TraitsT>::
433 _M_expression_term(_BracketState& __last_char,
434 _BracketMatcher<__icase, __collate>& __matcher)
435 {
436 if (_M_match_token(_ScannerT::_S_token_bracket_end))
437 return false;
438
439 // Add any previously cached char into the matcher and update cache.
440 const auto __push_char = [&](_CharT __ch)
441 {
442 if (__last_char._M_is_char())
443 __matcher._M_add_char(__last_char.get());
444 __last_char.set(__ch);
445 };
446 // Add any previously cached char into the matcher and update cache.
447 const auto __push_class = [&]
448 {
449 if (__last_char._M_is_char())
450 __matcher._M_add_char(__last_char.get());
451 // We don't cache anything here, just record that the last thing
452 // processed was a character class (or similar).
453 __last_char.reset(_BracketState::_Type::_Class);
454 };
455
456 if (_M_match_token(_ScannerT::_S_token_collsymbol))
457 {
458 auto __symbol = __matcher._M_add_collate_element(_M_value);
459 if (__symbol.size() == 1)
460 __push_char(__symbol[0]);
461 else
462 __push_class();
463 }
464 else if (_M_match_token(_ScannerT::_S_token_equiv_class_name))
465 {
466 __push_class();
467 __matcher._M_add_equivalence_class(_M_value);
468 }
469 else if (_M_match_token(_ScannerT::_S_token_char_class_name))
470 {
471 __push_class();
472 __matcher._M_add_character_class(_M_value, false);
473 }
474 else if (_M_try_char())
475 __push_char(_M_value[0]);
476 // POSIX doesn't allow '-' as a start-range char (say [a-z--0]),
477 // except when the '-' is the first or last character in the bracket
478 // expression ([--0]). ECMAScript treats all '-' after a range as a
479 // normal character. Also see above, where _M_expression_term gets called.
480 //
481 // As a result, POSIX rejects [-----], but ECMAScript doesn't.
482 // Boost (1.57.0) always uses POSIX style even in its ECMAScript syntax.
483 // Clang (3.5) always uses ECMAScript style even in its POSIX syntax.
484 //
485 // It turns out that no one reads BNFs ;)
486 else if (_M_match_token(_ScannerT::_S_token_bracket_dash))
487 {
488 if (_M_match_token(_ScannerT::_S_token_bracket_end))
489 {
490 // For "-]" the dash is a literal character.
491 __push_char('-');
492 return false;
493 }
494 else if (__last_char._M_is_class())
495 {
496 // "\\w-" is invalid, start of range must be a single char.
497 __throw_regex_error(regex_constants::error_range,
498 "Invalid start of '[x-x]' range in "
499 "regular expression");
500 }
501 else if (__last_char._M_is_char())
502 {
503 if (_M_try_char())
504 {
505 // "x-y"
506 __matcher._M_make_range(__last_char.get(), _M_value[0]);
507 __last_char.reset();
508 }
509 else if (_M_match_token(_ScannerT::_S_token_bracket_dash))
510 {
511 // "x--"
512 __matcher._M_make_range(__last_char.get(), '-');
513 __last_char.reset();
514 }
515 else
516 __throw_regex_error(regex_constants::error_range,
517 "Invalid end of '[x-x]' range in "
518 "regular expression");
519 }
520 else if (_M_flags & regex_constants::ECMAScript)
521 {
522 // A dash that is not part of an existing range. Might be the
523 // start of a new range, or might just be a literal '-' char.
524 // Only ECMAScript allows that in the middle of a bracket expr.
525 __push_char('-');
526 }
527 else
528 __throw_regex_error(regex_constants::error_range,
529 "Invalid location of '-' within '[...]' in "
530 "POSIX regular expression");
531 }
532 else if (_M_match_token(_ScannerT::_S_token_quoted_class))
533 {
534 __push_class();
535 __matcher._M_add_character_class(_M_value,
536 _M_ctype.is(_CtypeT::upper,
537 _M_value[0]));
538 }
539 else
540 __throw_regex_error(regex_constants::error_brack,
541 "Unexpected character within '[...]' in "
542 "regular expression");
543 return true;
544 }
545
546 template<typename _TraitsT>
547 bool
548 _Compiler<_TraitsT>::
549 _M_try_char()
550 {
551 bool __is_char = false;
552 if (_M_match_token(_ScannerT::_S_token_oct_num))
553 {
554 __is_char = true;
555 _M_value.assign(1, _M_cur_int_value(8));
556 }
557 else if (_M_match_token(_ScannerT::_S_token_hex_num))
558 {
559 __is_char = true;
560 _M_value.assign(1, _M_cur_int_value(16));
561 }
562 else if (_M_match_token(_ScannerT::_S_token_ord_char))
563 __is_char = true;
564 return __is_char;
565 }
566
567 template<typename _TraitsT>
568 bool
569 _Compiler<_TraitsT>::
570 _M_match_token(_TokenT __token)
571 {
572 if (__token == _M_scanner._M_get_token())
573 {
574 _M_value = _M_scanner._M_get_value();
575 _M_scanner._M_advance();
576 return true;
577 }
578 return false;
579 }
580
581 template<typename _TraitsT>
582 int
583 _Compiler<_TraitsT>::
584 _M_cur_int_value(int __radix)
585 {
586 int __v = 0;
587 for (_CharT __c : _M_value)
588 if (__builtin_mul_overflow(__v, __radix, &__v)
589 || __builtin_add_overflow(__v, _M_traits.value(__c, __radix), &__v))
590 std::__throw_regex_error(regex_constants::error_backref,
591 "invalid back reference");
592 return __v;
593 }
594
595 template<typename _TraitsT, bool __icase, bool __collate>
596 bool
597 _BracketMatcher<_TraitsT, __icase, __collate>::
598 _M_apply(_CharT __ch, false_type) const
599 {
600 return [this, __ch]
601 {
602 if (std::binary_search(_M_char_set.begin(), _M_char_set.end(),
603 _M_translator._M_translate(__ch)))
604 return true;
605 auto __s = _M_translator._M_transform(__ch);
606 for (auto& __it : _M_range_set)
607 if (_M_translator._M_match_range(__it.first, __it.second, __s))
608 return true;
609 if (_M_traits.isctype(__ch, _M_class_set))
610 return true;
611 if (std::find(_M_equiv_set.begin(), _M_equiv_set.end(),
612 _M_traits.transform_primary(&__ch, &__ch+1))
613 != _M_equiv_set.end())
614 return true;
615 for (auto& __it : _M_neg_class_set)
616 if (!_M_traits.isctype(__ch, __it))
617 return true;
618 return false;
619 }() ^ _M_is_non_matching;
620 }
621} // namespace __detail
622
623_GLIBCXX_END_NAMESPACE_VERSION
624} // namespace
constexpr std::remove_reference< _Tp >::type && move(_Tp &&__t) noexcept
Convert a value to an rvalue.
Definition move.h:127
const _Facet & use_facet(const locale &__loc)
Return a facet.
ISO C++ entities toplevel namespace is std.
constexpr error_type error_backref(_S_error_backref)
A standard container giving FILO behavior.
Definition stl_stack.h:100
void pop()
Removes first element.
Definition stl_stack.h:291
void push(const value_type &__x)
Add data to the top of the stack.
Definition stl_stack.h:258
bool empty() const
Definition stl_stack.h:215
reference top()
Definition stl_stack.h:230