Line data Source code
1 : // class template regex -*- C++ -*-
2 :
3 : // Copyright (C) 2013-2019 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_executor.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 : namespace std _GLIBCXX_VISIBILITY(default)
32 : {
33 : _GLIBCXX_BEGIN_NAMESPACE_VERSION
34 :
35 : namespace __detail
36 : {
37 : template<typename _BiIter, typename _Alloc, typename _TraitsT,
38 : bool __dfs_mode>
39 : bool _Executor<_BiIter, _Alloc, _TraitsT, __dfs_mode>::
40 : _M_search()
41 : {
42 : if (_M_search_from_first())
43 : return true;
44 : if (_M_flags & regex_constants::match_continuous)
45 : return false;
46 : _M_flags |= regex_constants::match_prev_avail;
47 : while (_M_begin != _M_end)
48 : {
49 : ++_M_begin;
50 : if (_M_search_from_first())
51 : return true;
52 : }
53 : return false;
54 : }
55 :
56 : // The _M_main function operates in different modes, DFS mode or BFS mode,
57 : // indicated by template parameter __dfs_mode, and dispatches to one of the
58 : // _M_main_dispatch overloads.
59 : //
60 : // ------------------------------------------------------------
61 : //
62 : // DFS mode:
63 : //
64 : // It applies a Depth-First-Search (aka backtracking) on given NFA and input
65 : // string.
66 : // At the very beginning the executor stands in the start state, then it
67 : // tries every possible state transition in current state recursively. Some
68 : // state transitions consume input string, say, a single-char-matcher or a
69 : // back-reference matcher; some don't, like assertion or other anchor nodes.
70 : // When the input is exhausted and/or the current state is an accepting
71 : // state, the whole executor returns true.
72 : //
73 : // TODO: This approach is exponentially slow for certain input.
74 : // Try to compile the NFA to a DFA.
75 : //
76 : // Time complexity: \Omega(match_length), O(2^(_M_nfa.size()))
77 : // Space complexity: \theta(match_results.size() + match_length)
78 : //
79 : template<typename _BiIter, typename _Alloc, typename _TraitsT,
80 : bool __dfs_mode>
81 : bool _Executor<_BiIter, _Alloc, _TraitsT, __dfs_mode>::
82 : _M_main_dispatch(_Match_mode __match_mode, __dfs)
83 : {
84 : _M_has_sol = false;
85 : *_M_states._M_get_sol_pos() = _BiIter();
86 : _M_cur_results = _M_results;
87 : _M_dfs(__match_mode, _M_states._M_start);
88 : return _M_has_sol;
89 : }
90 :
91 : // ------------------------------------------------------------
92 : //
93 : // BFS mode:
94 : //
95 : // Russ Cox's article (http://swtch.com/~rsc/regexp/regexp1.html)
96 : // explained this algorithm clearly.
97 : //
98 : // It first computes epsilon closure (states that can be achieved without
99 : // consuming characters) for every state that's still matching,
100 : // using the same DFS algorithm, but doesn't re-enter states (using
101 : // _M_states._M_visited to check), nor follow _S_opcode_match.
102 : //
103 : // Then apply DFS using every _S_opcode_match (in _M_states._M_match_queue)
104 : // as the start state.
105 : //
106 : // It significantly reduces potential duplicate states, so has a better
107 : // upper bound; but it requires more overhead.
108 : //
109 : // Time complexity: \Omega(match_length * match_results.size())
110 : // O(match_length * _M_nfa.size() * match_results.size())
111 : // Space complexity: \Omega(_M_nfa.size() + match_results.size())
112 : // O(_M_nfa.size() * match_results.size())
113 : template<typename _BiIter, typename _Alloc, typename _TraitsT,
114 : bool __dfs_mode>
115 0 : bool _Executor<_BiIter, _Alloc, _TraitsT, __dfs_mode>::
116 : _M_main_dispatch(_Match_mode __match_mode, __bfs)
117 : {
118 0 : _M_states._M_queue(_M_states._M_start, _M_results);
119 0 : bool __ret = false;
120 : while (1)
121 : {
122 0 : _M_has_sol = false;
123 0 : if (_M_states._M_match_queue.empty())
124 : break;
125 0 : std::fill_n(_M_states._M_visited_states.get(), _M_nfa.size(), false);
126 0 : auto __old_queue = std::move(_M_states._M_match_queue);
127 0 : for (auto& __task : __old_queue)
128 : {
129 0 : _M_cur_results = std::move(__task.second);
130 0 : _M_dfs(__match_mode, __task.first);
131 : }
132 0 : if (__match_mode == _Match_mode::_Prefix)
133 0 : __ret |= _M_has_sol;
134 0 : if (_M_current == _M_end)
135 : break;
136 0 : ++_M_current;
137 : }
138 0 : if (__match_mode == _Match_mode::_Exact)
139 0 : __ret = _M_has_sol;
140 0 : _M_states._M_match_queue.clear();
141 0 : return __ret;
142 : }
143 :
144 : // Return whether now match the given sub-NFA.
145 : template<typename _BiIter, typename _Alloc, typename _TraitsT,
146 : bool __dfs_mode>
147 0 : bool _Executor<_BiIter, _Alloc, _TraitsT, __dfs_mode>::
148 : _M_lookahead(_StateIdT __next)
149 : {
150 : // Backreferences may refer to captured content.
151 : // We may want to make this faster by not copying,
152 : // but let's not be clever prematurely.
153 0 : _ResultsVec __what(_M_cur_results);
154 0 : _Executor __sub(_M_current, _M_end, __what, _M_re, _M_flags);
155 0 : __sub._M_states._M_start = __next;
156 0 : if (__sub._M_search_from_first())
157 : {
158 0 : for (size_t __i = 0; __i < __what.size(); __i++)
159 0 : if (__what[__i].matched)
160 0 : _M_cur_results[__i] = __what[__i];
161 : return true;
162 : }
163 : return false;
164 : }
165 :
166 : // __rep_count records how many times (__rep_count.second)
167 : // this node is visited under certain input iterator
168 : // (__rep_count.first). This prevent the executor from entering
169 : // infinite loop by refusing to continue when it's already been
170 : // visited more than twice. It's `twice` instead of `once` because
171 : // we need to spare one more time for potential group capture.
172 : template<typename _BiIter, typename _Alloc, typename _TraitsT,
173 : bool __dfs_mode>
174 0 : void _Executor<_BiIter, _Alloc, _TraitsT, __dfs_mode>::
175 : _M_rep_once_more(_Match_mode __match_mode, _StateIdT __i)
176 : {
177 0 : const auto& __state = _M_nfa[__i];
178 0 : auto& __rep_count = _M_rep_count[__i];
179 0 : if (__rep_count.second == 0 || __rep_count.first != _M_current)
180 : {
181 0 : auto __back = __rep_count;
182 0 : __rep_count.first = _M_current;
183 0 : __rep_count.second = 1;
184 0 : _M_dfs(__match_mode, __state._M_alt);
185 0 : __rep_count = __back;
186 : }
187 : else
188 : {
189 0 : if (__rep_count.second < 2)
190 : {
191 0 : __rep_count.second++;
192 0 : _M_dfs(__match_mode, __state._M_alt);
193 0 : __rep_count.second--;
194 : }
195 : }
196 0 : }
197 :
198 : // _M_alt branch is "match once more", while _M_next is "get me out
199 : // of this quantifier". Executing _M_next first or _M_alt first don't
200 : // mean the same thing, and we need to choose the correct order under
201 : // given greedy mode.
202 : template<typename _BiIter, typename _Alloc, typename _TraitsT,
203 : bool __dfs_mode>
204 0 : void _Executor<_BiIter, _Alloc, _TraitsT, __dfs_mode>::
205 : _M_handle_repeat(_Match_mode __match_mode, _StateIdT __i)
206 : {
207 0 : const auto& __state = _M_nfa[__i];
208 :
209 : // Greedy.
210 0 : if (!__state._M_neg)
211 : {
212 0 : _M_rep_once_more(__match_mode, __i);
213 : // If it's DFS executor and already accepted, we're done.
214 0 : if (!__dfs_mode || !_M_has_sol)
215 0 : _M_dfs(__match_mode, __state._M_next);
216 : }
217 : else // Non-greedy mode
218 : {
219 : if (__dfs_mode)
220 : {
221 : // vice-versa.
222 0 : _M_dfs(__match_mode, __state._M_next);
223 0 : if (!_M_has_sol)
224 0 : _M_rep_once_more(__match_mode, __i);
225 : }
226 : else
227 : {
228 : // DON'T attempt anything, because there's already another
229 : // state with higher priority accepted. This state cannot
230 : // be better by attempting its next node.
231 0 : if (!_M_has_sol)
232 : {
233 0 : _M_dfs(__match_mode, __state._M_next);
234 : // DON'T attempt anything if it's already accepted. An
235 : // accepted state *must* be better than a solution that
236 : // matches a non-greedy quantifier one more time.
237 0 : if (!_M_has_sol)
238 0 : _M_rep_once_more(__match_mode, __i);
239 : }
240 : }
241 : }
242 0 : }
243 :
244 : template<typename _BiIter, typename _Alloc, typename _TraitsT,
245 : bool __dfs_mode>
246 955 : void _Executor<_BiIter, _Alloc, _TraitsT, __dfs_mode>::
247 : _M_handle_subexpr_begin(_Match_mode __match_mode, _StateIdT __i)
248 : {
249 955 : const auto& __state = _M_nfa[__i];
250 :
251 955 : auto& __res = _M_cur_results[__state._M_subexpr];
252 955 : auto __back = __res.first;
253 955 : __res.first = _M_current;
254 955 : _M_dfs(__match_mode, __state._M_next);
255 955 : __res.first = __back;
256 955 : }
257 :
258 : template<typename _BiIter, typename _Alloc, typename _TraitsT,
259 : bool __dfs_mode>
260 114 : void _Executor<_BiIter, _Alloc, _TraitsT, __dfs_mode>::
261 : _M_handle_subexpr_end(_Match_mode __match_mode, _StateIdT __i)
262 : {
263 114 : const auto& __state = _M_nfa[__i];
264 :
265 114 : auto& __res = _M_cur_results[__state._M_subexpr];
266 114 : auto __back = __res;
267 114 : __res.second = _M_current;
268 114 : __res.matched = true;
269 114 : _M_dfs(__match_mode, __state._M_next);
270 114 : __res = __back;
271 114 : }
272 :
273 : template<typename _BiIter, typename _Alloc, typename _TraitsT,
274 : bool __dfs_mode>
275 0 : inline void _Executor<_BiIter, _Alloc, _TraitsT, __dfs_mode>::
276 : _M_handle_line_begin_assertion(_Match_mode __match_mode, _StateIdT __i)
277 : {
278 0 : const auto& __state = _M_nfa[__i];
279 0 : if (_M_at_begin())
280 0 : _M_dfs(__match_mode, __state._M_next);
281 0 : }
282 :
283 : template<typename _BiIter, typename _Alloc, typename _TraitsT,
284 : bool __dfs_mode>
285 0 : inline void _Executor<_BiIter, _Alloc, _TraitsT, __dfs_mode>::
286 : _M_handle_line_end_assertion(_Match_mode __match_mode, _StateIdT __i)
287 : {
288 0 : const auto& __state = _M_nfa[__i];
289 0 : if (_M_at_end())
290 0 : _M_dfs(__match_mode, __state._M_next);
291 0 : }
292 :
293 : template<typename _BiIter, typename _Alloc, typename _TraitsT,
294 : bool __dfs_mode>
295 0 : inline void _Executor<_BiIter, _Alloc, _TraitsT, __dfs_mode>::
296 : _M_handle_word_boundary(_Match_mode __match_mode, _StateIdT __i)
297 : {
298 0 : const auto& __state = _M_nfa[__i];
299 0 : if (_M_word_boundary() == !__state._M_neg)
300 0 : _M_dfs(__match_mode, __state._M_next);
301 0 : }
302 :
303 : // Here __state._M_alt offers a single start node for a sub-NFA.
304 : // We recursively invoke our algorithm to match the sub-NFA.
305 : template<typename _BiIter, typename _Alloc, typename _TraitsT,
306 : bool __dfs_mode>
307 0 : void _Executor<_BiIter, _Alloc, _TraitsT, __dfs_mode>::
308 : _M_handle_subexpr_lookahead(_Match_mode __match_mode, _StateIdT __i)
309 : {
310 0 : const auto& __state = _M_nfa[__i];
311 0 : if (_M_lookahead(__state._M_alt) == !__state._M_neg)
312 0 : _M_dfs(__match_mode, __state._M_next);
313 0 : }
314 :
315 : template<typename _BiIter, typename _Alloc, typename _TraitsT,
316 : bool __dfs_mode>
317 1411 : void _Executor<_BiIter, _Alloc, _TraitsT, __dfs_mode>::
318 : _M_handle_match(_Match_mode __match_mode, _StateIdT __i)
319 : {
320 1411 : const auto& __state = _M_nfa[__i];
321 :
322 1411 : if (_M_current == _M_end)
323 : return;
324 : if (__dfs_mode)
325 : {
326 2822 : if (__state._M_matches(*_M_current))
327 : {
328 570 : ++_M_current;
329 570 : _M_dfs(__match_mode, __state._M_next);
330 570 : --_M_current;
331 : }
332 : }
333 : else
334 0 : if (__state._M_matches(*_M_current))
335 0 : _M_states._M_queue(__state._M_next, _M_cur_results);
336 : }
337 :
338 : template<typename _BiIter, typename _TraitsT>
339 : struct _Backref_matcher
340 : {
341 : _Backref_matcher(bool __icase, const _TraitsT& __traits)
342 : : _M_traits(__traits) { }
343 :
344 : bool
345 : _M_apply(_BiIter __expected_begin,
346 : _BiIter __expected_end, _BiIter __actual_begin,
347 : _BiIter __actual_end)
348 : {
349 : return _M_traits.transform(__expected_begin, __expected_end)
350 : == _M_traits.transform(__actual_begin, __actual_end);
351 : }
352 :
353 : const _TraitsT& _M_traits;
354 : };
355 :
356 : template<typename _BiIter, typename _CharT>
357 : struct _Backref_matcher<_BiIter, std::regex_traits<_CharT>>
358 : {
359 : using _TraitsT = std::regex_traits<_CharT>;
360 0 : _Backref_matcher(bool __icase, const _TraitsT& __traits)
361 0 : : _M_icase(__icase), _M_traits(__traits) { }
362 :
363 : bool
364 0 : _M_apply(_BiIter __expected_begin,
365 : _BiIter __expected_end, _BiIter __actual_begin,
366 : _BiIter __actual_end)
367 : {
368 0 : if (!_M_icase)
369 0 : return _GLIBCXX_STD_A::__equal4(__expected_begin, __expected_end,
370 : __actual_begin, __actual_end);
371 : typedef std::ctype<_CharT> __ctype_type;
372 0 : const auto& __fctyp = use_facet<__ctype_type>(_M_traits.getloc());
373 0 : return _GLIBCXX_STD_A::__equal4(__expected_begin, __expected_end,
374 : __actual_begin, __actual_end,
375 0 : [this, &__fctyp](_CharT __lhs, _CharT __rhs)
376 : {
377 : return __fctyp.tolower(__lhs)
378 0 : == __fctyp.tolower(__rhs);
379 0 : });
380 : }
381 :
382 : bool _M_icase;
383 : const _TraitsT& _M_traits;
384 : };
385 :
386 : // First fetch the matched result from _M_cur_results as __submatch;
387 : // then compare it with
388 : // (_M_current, _M_current + (__submatch.second - __submatch.first)).
389 : // If matched, keep going; else just return and try another state.
390 : template<typename _BiIter, typename _Alloc, typename _TraitsT,
391 : bool __dfs_mode>
392 0 : void _Executor<_BiIter, _Alloc, _TraitsT, __dfs_mode>::
393 : _M_handle_backref(_Match_mode __match_mode, _StateIdT __i)
394 : {
395 : __glibcxx_assert(__dfs_mode);
396 :
397 0 : const auto& __state = _M_nfa[__i];
398 0 : auto& __submatch = _M_cur_results[__state._M_backref_index];
399 0 : if (!__submatch.matched)
400 0 : return;
401 0 : auto __last = _M_current;
402 0 : for (auto __tmp = __submatch.first;
403 0 : __last != _M_end && __tmp != __submatch.second;
404 0 : ++__tmp)
405 0 : ++__last;
406 0 : if (_Backref_matcher<_BiIter, _TraitsT>(
407 0 : _M_re.flags() & regex_constants::icase,
408 0 : _M_re._M_automaton->_M_traits)._M_apply(
409 : __submatch.first, __submatch.second, _M_current, __last))
410 : {
411 0 : if (__last != _M_current)
412 : {
413 0 : auto __backup = _M_current;
414 0 : _M_current = __last;
415 0 : _M_dfs(__match_mode, __state._M_next);
416 0 : _M_current = __backup;
417 : }
418 : else
419 0 : _M_dfs(__match_mode, __state._M_next);
420 : }
421 : }
422 :
423 : template<typename _BiIter, typename _Alloc, typename _TraitsT,
424 : bool __dfs_mode>
425 114 : void _Executor<_BiIter, _Alloc, _TraitsT, __dfs_mode>::
426 : _M_handle_accept(_Match_mode __match_mode, _StateIdT __i)
427 : {
428 : if (__dfs_mode)
429 : {
430 : __glibcxx_assert(!_M_has_sol);
431 114 : if (__match_mode == _Match_mode::_Exact)
432 114 : _M_has_sol = _M_current == _M_end;
433 : else
434 0 : _M_has_sol = true;
435 114 : if (_M_current == _M_begin
436 114 : && (_M_flags & regex_constants::match_not_null))
437 0 : _M_has_sol = false;
438 114 : if (_M_has_sol)
439 : {
440 114 : if (_M_nfa._M_flags & regex_constants::ECMAScript)
441 114 : _M_results = _M_cur_results;
442 : else // POSIX
443 : {
444 : __glibcxx_assert(_M_states._M_get_sol_pos());
445 : // Here's POSIX's logic: match the longest one. However
446 : // we never know which one (lhs or rhs of "|") is longer
447 : // unless we try both of them and compare the results.
448 : // The member variable _M_sol_pos records the end
449 : // position of the last successful match. It's better
450 : // to be larger, because POSIX regex is always greedy.
451 : // TODO: This could be slow.
452 0 : if (*_M_states._M_get_sol_pos() == _BiIter()
453 0 : || std::distance(_M_begin,
454 0 : *_M_states._M_get_sol_pos())
455 0 : < std::distance(_M_begin, _M_current))
456 : {
457 0 : *_M_states._M_get_sol_pos() = _M_current;
458 0 : _M_results = _M_cur_results;
459 : }
460 : }
461 : }
462 : }
463 : else
464 : {
465 0 : if (_M_current == _M_begin
466 0 : && (_M_flags & regex_constants::match_not_null))
467 : return;
468 0 : if (__match_mode == _Match_mode::_Prefix || _M_current == _M_end)
469 0 : if (!_M_has_sol)
470 : {
471 0 : _M_has_sol = true;
472 0 : _M_results = _M_cur_results;
473 : }
474 : }
475 : }
476 :
477 : template<typename _BiIter, typename _Alloc, typename _TraitsT,
478 : bool __dfs_mode>
479 0 : void _Executor<_BiIter, _Alloc, _TraitsT, __dfs_mode>::
480 : _M_handle_alternative(_Match_mode __match_mode, _StateIdT __i)
481 : {
482 0 : const auto& __state = _M_nfa[__i];
483 :
484 0 : if (_M_nfa._M_flags & regex_constants::ECMAScript)
485 : {
486 : // TODO: Fix BFS support. It is wrong.
487 0 : _M_dfs(__match_mode, __state._M_alt);
488 : // Pick lhs if it matches. Only try rhs if it doesn't.
489 0 : if (!_M_has_sol)
490 0 : _M_dfs(__match_mode, __state._M_next);
491 : }
492 : else
493 : {
494 : // Try both and compare the result.
495 : // See "case _S_opcode_accept:" handling above.
496 0 : _M_dfs(__match_mode, __state._M_alt);
497 0 : auto __has_sol = _M_has_sol;
498 0 : _M_has_sol = false;
499 0 : _M_dfs(__match_mode, __state._M_next);
500 0 : _M_has_sol |= __has_sol;
501 : }
502 0 : }
503 :
504 : template<typename _BiIter, typename _Alloc, typename _TraitsT,
505 : bool __dfs_mode>
506 2594 : void _Executor<_BiIter, _Alloc, _TraitsT, __dfs_mode>::
507 : _M_dfs(_Match_mode __match_mode, _StateIdT __i)
508 : {
509 2594 : if (_M_states._M_visited(__i))
510 : return;
511 :
512 2594 : switch (_M_nfa[__i]._M_opcode())
513 : {
514 0 : case _S_opcode_repeat:
515 0 : _M_handle_repeat(__match_mode, __i); break;
516 955 : case _S_opcode_subexpr_begin:
517 955 : _M_handle_subexpr_begin(__match_mode, __i); break;
518 114 : case _S_opcode_subexpr_end:
519 114 : _M_handle_subexpr_end(__match_mode, __i); break;
520 0 : case _S_opcode_line_begin_assertion:
521 0 : _M_handle_line_begin_assertion(__match_mode, __i); break;
522 0 : case _S_opcode_line_end_assertion:
523 0 : _M_handle_line_end_assertion(__match_mode, __i); break;
524 0 : case _S_opcode_word_boundary:
525 0 : _M_handle_word_boundary(__match_mode, __i); break;
526 0 : case _S_opcode_subexpr_lookahead:
527 0 : _M_handle_subexpr_lookahead(__match_mode, __i); break;
528 1411 : case _S_opcode_match:
529 1411 : _M_handle_match(__match_mode, __i); break;
530 0 : case _S_opcode_backref:
531 0 : _M_handle_backref(__match_mode, __i); break;
532 114 : case _S_opcode_accept:
533 114 : _M_handle_accept(__match_mode, __i); break;
534 0 : case _S_opcode_alternative:
535 0 : _M_handle_alternative(__match_mode, __i); break;
536 2594 : default:
537 : __glibcxx_assert(false);
538 : }
539 : }
540 :
541 : // Return whether now is at some word boundary.
542 : template<typename _BiIter, typename _Alloc, typename _TraitsT,
543 : bool __dfs_mode>
544 0 : bool _Executor<_BiIter, _Alloc, _TraitsT, __dfs_mode>::
545 : _M_word_boundary() const
546 : {
547 0 : if (_M_current == _M_begin && (_M_flags & regex_constants::match_not_bow))
548 : return false;
549 0 : if (_M_current == _M_end && (_M_flags & regex_constants::match_not_eow))
550 : return false;
551 :
552 0 : bool __left_is_word = false;
553 0 : if (_M_current != _M_begin
554 0 : || (_M_flags & regex_constants::match_prev_avail))
555 : {
556 0 : auto __prev = _M_current;
557 0 : if (_M_is_word(*std::prev(__prev)))
558 0 : __left_is_word = true;
559 : }
560 0 : bool __right_is_word =
561 0 : _M_current != _M_end && _M_is_word(*_M_current);
562 :
563 0 : return __left_is_word != __right_is_word;
564 : }
565 : } // namespace __detail
566 :
567 : _GLIBCXX_END_NAMESPACE_VERSION
568 : } // namespace
|
