libstdc++
regex_compiler.tcc
Go to the documentation of this file.
1 // class template regex -*- C++ -*-
2 
3 // Copyright (C) 2013-2017 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 // This compiler refers to "Regular Expression Matching Can Be Simple And Fast"
34 // (http://swtch.com/~rsc/regexp/regexp1.html"),
35 // but doesn't strictly follow it.
36 //
37 // When compiling, states are *chained* instead of tree- or graph-constructed.
38 // It's more like structured programs: there's if statement and loop statement.
39 //
40 // For alternative structure (say "a|b"), aka "if statement", two branches
41 // should be constructed. However, these two shall merge to an "end_tag" at
42 // the end of this operator:
43 //
44 // branch1
45 // / \
46 // => begin_tag end_tag =>
47 // \ /
48 // branch2
49 //
50 // This is the difference between this implementation and that in Russ's
51 // article.
52 //
53 // That's why we introduced dummy node here ------ "end_tag" is a dummy node.
54 // All dummy node will be eliminated at the end of compiling process.
55 
56 namespace std _GLIBCXX_VISIBILITY(default)
57 {
58 namespace __detail
59 {
60 _GLIBCXX_BEGIN_NAMESPACE_VERSION
61 
62  template<typename _TraitsT>
63  _Compiler<_TraitsT>::
64  _Compiler(_IterT __b, _IterT __e,
65  const typename _TraitsT::locale_type& __loc, _FlagT __flags)
66  : _M_flags((__flags
67  & (regex_constants::ECMAScript
68  | regex_constants::basic
69  | regex_constants::extended
70  | regex_constants::grep
71  | regex_constants::egrep
72  | regex_constants::awk))
73  ? __flags
74  : __flags | regex_constants::ECMAScript),
75  _M_scanner(__b, __e, _M_flags, __loc),
76  _M_nfa(make_shared<_RegexT>(__loc, _M_flags)),
77  _M_traits(_M_nfa->_M_traits),
78  _M_ctype(std::use_facet<_CtypeT>(__loc))
79  {
80  _StateSeqT __r(*_M_nfa, _M_nfa->_M_start());
81  __r._M_append(_M_nfa->_M_insert_subexpr_begin());
82  this->_M_disjunction();
83  if (!_M_match_token(_ScannerT::_S_token_eof))
84  __throw_regex_error(regex_constants::error_paren);
85  __r._M_append(_M_pop());
86  __glibcxx_assert(_M_stack.empty());
87  __r._M_append(_M_nfa->_M_insert_subexpr_end());
88  __r._M_append(_M_nfa->_M_insert_accept());
89  _M_nfa->_M_eliminate_dummy();
90  }
91 
92  template<typename _TraitsT>
93  void
94  _Compiler<_TraitsT>::
95  _M_disjunction()
96  {
97  this->_M_alternative();
98  while (_M_match_token(_ScannerT::_S_token_or))
99  {
100  _StateSeqT __alt1 = _M_pop();
101  this->_M_alternative();
102  _StateSeqT __alt2 = _M_pop();
103  auto __end = _M_nfa->_M_insert_dummy();
104  __alt1._M_append(__end);
105  __alt2._M_append(__end);
106  // __alt2 is state._M_next, __alt1 is state._M_alt. The executor
107  // executes _M_alt before _M_next, as well as executing left
108  // alternative before right one.
109  _M_stack.push(_StateSeqT(*_M_nfa,
110  _M_nfa->_M_insert_alt(
111  __alt2._M_start, __alt1._M_start, false),
112  __end));
113  }
114  }
115 
116  template<typename _TraitsT>
117  void
118  _Compiler<_TraitsT>::
119  _M_alternative()
120  {
121  if (this->_M_term())
122  {
123  _StateSeqT __re = _M_pop();
124  this->_M_alternative();
125  __re._M_append(_M_pop());
126  _M_stack.push(__re);
127  }
128  else
129  _M_stack.push(_StateSeqT(*_M_nfa, _M_nfa->_M_insert_dummy()));
130  }
131 
132  template<typename _TraitsT>
133  bool
134  _Compiler<_TraitsT>::
135  _M_term()
136  {
137  if (this->_M_assertion())
138  return true;
139  if (this->_M_atom())
140  {
141  while (this->_M_quantifier());
142  return true;
143  }
144  return false;
145  }
146 
147  template<typename _TraitsT>
148  bool
149  _Compiler<_TraitsT>::
150  _M_assertion()
151  {
152  if (_M_match_token(_ScannerT::_S_token_line_begin))
153  _M_stack.push(_StateSeqT(*_M_nfa, _M_nfa->_M_insert_line_begin()));
154  else if (_M_match_token(_ScannerT::_S_token_line_end))
155  _M_stack.push(_StateSeqT(*_M_nfa, _M_nfa->_M_insert_line_end()));
156  else if (_M_match_token(_ScannerT::_S_token_word_bound))
157  // _M_value[0] == 'n' means it's negative, say "not word boundary".
158  _M_stack.push(_StateSeqT(*_M_nfa, _M_nfa->
159  _M_insert_word_bound(_M_value[0] == 'n')));
160  else if (_M_match_token(_ScannerT::_S_token_subexpr_lookahead_begin))
161  {
162  auto __neg = _M_value[0] == 'n';
163  this->_M_disjunction();
164  if (!_M_match_token(_ScannerT::_S_token_subexpr_end))
165  __throw_regex_error(regex_constants::error_paren,
166  "Parenthesis is not closed.");
167  auto __tmp = _M_pop();
168  __tmp._M_append(_M_nfa->_M_insert_accept());
169  _M_stack.push(
170  _StateSeqT(
171  *_M_nfa,
172  _M_nfa->_M_insert_lookahead(__tmp._M_start, __neg)));
173  }
174  else
175  return false;
176  return true;
177  }
178 
179  template<typename _TraitsT>
180  bool
181  _Compiler<_TraitsT>::
182  _M_quantifier()
183  {
184  bool __neg = (_M_flags & regex_constants::ECMAScript);
185  auto __init = [this, &__neg]()
186  {
187  if (_M_stack.empty())
188  __throw_regex_error(regex_constants::error_badrepeat,
189  "Nothing to repeat before a quantifier.");
190  __neg = __neg && _M_match_token(_ScannerT::_S_token_opt);
191  };
192  if (_M_match_token(_ScannerT::_S_token_closure0))
193  {
194  __init();
195  auto __e = _M_pop();
196  _StateSeqT __r(*_M_nfa,
197  _M_nfa->_M_insert_repeat(_S_invalid_state_id,
198  __e._M_start, __neg));
199  __e._M_append(__r);
200  _M_stack.push(__r);
201  }
202  else if (_M_match_token(_ScannerT::_S_token_closure1))
203  {
204  __init();
205  auto __e = _M_pop();
206  __e._M_append(_M_nfa->_M_insert_repeat(_S_invalid_state_id,
207  __e._M_start, __neg));
208  _M_stack.push(__e);
209  }
210  else if (_M_match_token(_ScannerT::_S_token_opt))
211  {
212  __init();
213  auto __e = _M_pop();
214  auto __end = _M_nfa->_M_insert_dummy();
215  _StateSeqT __r(*_M_nfa,
216  _M_nfa->_M_insert_repeat(_S_invalid_state_id,
217  __e._M_start, __neg));
218  __e._M_append(__end);
219  __r._M_append(__end);
220  _M_stack.push(__r);
221  }
222  else if (_M_match_token(_ScannerT::_S_token_interval_begin))
223  {
224  if (_M_stack.empty())
225  __throw_regex_error(regex_constants::error_badrepeat,
226  "Nothing to repeat before a quantifier.");
227  if (!_M_match_token(_ScannerT::_S_token_dup_count))
228  __throw_regex_error(regex_constants::error_badbrace,
229  "Unexpected token in brace expression.");
230  _StateSeqT __r(_M_pop());
231  _StateSeqT __e(*_M_nfa, _M_nfa->_M_insert_dummy());
232  long __min_rep = _M_cur_int_value(10);
233  bool __infi = false;
234  long __n;
235 
236  // {3
237  if (_M_match_token(_ScannerT::_S_token_comma))
238  if (_M_match_token(_ScannerT::_S_token_dup_count)) // {3,7}
239  __n = _M_cur_int_value(10) - __min_rep;
240  else
241  __infi = true;
242  else
243  __n = 0;
244  if (!_M_match_token(_ScannerT::_S_token_interval_end))
245  __throw_regex_error(regex_constants::error_brace,
246  "Unexpected end of brace expression.");
247 
248  __neg = __neg && _M_match_token(_ScannerT::_S_token_opt);
249 
250  for (long __i = 0; __i < __min_rep; ++__i)
251  __e._M_append(__r._M_clone());
252 
253  if (__infi)
254  {
255  auto __tmp = __r._M_clone();
256  _StateSeqT __s(*_M_nfa,
257  _M_nfa->_M_insert_repeat(_S_invalid_state_id,
258  __tmp._M_start, __neg));
259  __tmp._M_append(__s);
260  __e._M_append(__s);
261  }
262  else
263  {
264  if (__n < 0)
265  __throw_regex_error(regex_constants::error_badbrace,
266  "Invalid range in brace expression.");
267  auto __end = _M_nfa->_M_insert_dummy();
268  // _M_alt is the "match more" branch, and _M_next is the
269  // "match less" one. Switch _M_alt and _M_next of all created
270  // nodes. This is a hack but IMO works well.
271  std::stack<_StateIdT> __stack;
272  for (long __i = 0; __i < __n; ++__i)
273  {
274  auto __tmp = __r._M_clone();
275  auto __alt = _M_nfa->_M_insert_repeat(__tmp._M_start,
276  __end, __neg);
277  __stack.push(__alt);
278  __e._M_append(_StateSeqT(*_M_nfa, __alt, __tmp._M_end));
279  }
280  __e._M_append(__end);
281  while (!__stack.empty())
282  {
283  auto& __tmp = (*_M_nfa)[__stack.top()];
284  __stack.pop();
285  std::swap(__tmp._M_next, __tmp._M_alt);
286  }
287  }
288  _M_stack.push(__e);
289  }
290  else
291  return false;
292  return true;
293  }
294 
295 #define __INSERT_REGEX_MATCHER(__func, args...)\
296  do\
297  if (!(_M_flags & regex_constants::icase))\
298  if (!(_M_flags & regex_constants::collate))\
299  __func<false, false>(args);\
300  else\
301  __func<false, true>(args);\
302  else\
303  if (!(_M_flags & regex_constants::collate))\
304  __func<true, false>(args);\
305  else\
306  __func<true, true>(args);\
307  while (false)
308 
309  template<typename _TraitsT>
310  bool
311  _Compiler<_TraitsT>::
312  _M_atom()
313  {
314  if (_M_match_token(_ScannerT::_S_token_anychar))
315  {
316  if (!(_M_flags & regex_constants::ECMAScript))
317  __INSERT_REGEX_MATCHER(_M_insert_any_matcher_posix);
318  else
319  __INSERT_REGEX_MATCHER(_M_insert_any_matcher_ecma);
320  }
321  else if (_M_try_char())
322  __INSERT_REGEX_MATCHER(_M_insert_char_matcher);
323  else if (_M_match_token(_ScannerT::_S_token_backref))
324  _M_stack.push(_StateSeqT(*_M_nfa, _M_nfa->
325  _M_insert_backref(_M_cur_int_value(10))));
326  else if (_M_match_token(_ScannerT::_S_token_quoted_class))
327  __INSERT_REGEX_MATCHER(_M_insert_character_class_matcher);
328  else if (_M_match_token(_ScannerT::_S_token_subexpr_no_group_begin))
329  {
330  _StateSeqT __r(*_M_nfa, _M_nfa->_M_insert_dummy());
331  this->_M_disjunction();
332  if (!_M_match_token(_ScannerT::_S_token_subexpr_end))
333  __throw_regex_error(regex_constants::error_paren,
334  "Parenthesis is not closed.");
335  __r._M_append(_M_pop());
336  _M_stack.push(__r);
337  }
338  else if (_M_match_token(_ScannerT::_S_token_subexpr_begin))
339  {
340  _StateSeqT __r(*_M_nfa, _M_nfa->_M_insert_subexpr_begin());
341  this->_M_disjunction();
342  if (!_M_match_token(_ScannerT::_S_token_subexpr_end))
343  __throw_regex_error(regex_constants::error_paren,
344  "Parenthesis is not closed.");
345  __r._M_append(_M_pop());
346  __r._M_append(_M_nfa->_M_insert_subexpr_end());
347  _M_stack.push(__r);
348  }
349  else if (!_M_bracket_expression())
350  return false;
351  return true;
352  }
353 
354  template<typename _TraitsT>
355  bool
356  _Compiler<_TraitsT>::
357  _M_bracket_expression()
358  {
359  bool __neg =
360  _M_match_token(_ScannerT::_S_token_bracket_neg_begin);
361  if (!(__neg || _M_match_token(_ScannerT::_S_token_bracket_begin)))
362  return false;
363  __INSERT_REGEX_MATCHER(_M_insert_bracket_matcher, __neg);
364  return true;
365  }
366 #undef __INSERT_REGEX_MATCHER
367 
368  template<typename _TraitsT>
369  template<bool __icase, bool __collate>
370  void
371  _Compiler<_TraitsT>::
372  _M_insert_any_matcher_ecma()
373  {
374  _M_stack.push(_StateSeqT(*_M_nfa,
375  _M_nfa->_M_insert_matcher
376  (_AnyMatcher<_TraitsT, true, __icase, __collate>
377  (_M_traits))));
378  }
379 
380  template<typename _TraitsT>
381  template<bool __icase, bool __collate>
382  void
383  _Compiler<_TraitsT>::
384  _M_insert_any_matcher_posix()
385  {
386  _M_stack.push(_StateSeqT(*_M_nfa,
387  _M_nfa->_M_insert_matcher
388  (_AnyMatcher<_TraitsT, false, __icase, __collate>
389  (_M_traits))));
390  }
391 
392  template<typename _TraitsT>
393  template<bool __icase, bool __collate>
394  void
395  _Compiler<_TraitsT>::
396  _M_insert_char_matcher()
397  {
398  _M_stack.push(_StateSeqT(*_M_nfa,
399  _M_nfa->_M_insert_matcher
400  (_CharMatcher<_TraitsT, __icase, __collate>
401  (_M_value[0], _M_traits))));
402  }
403 
404  template<typename _TraitsT>
405  template<bool __icase, bool __collate>
406  void
407  _Compiler<_TraitsT>::
408  _M_insert_character_class_matcher()
409  {
410  __glibcxx_assert(_M_value.size() == 1);
411  _BracketMatcher<_TraitsT, __icase, __collate> __matcher
412  (_M_ctype.is(_CtypeT::upper, _M_value[0]), _M_traits);
413  __matcher._M_add_character_class(_M_value, false);
414  __matcher._M_ready();
415  _M_stack.push(_StateSeqT(*_M_nfa,
416  _M_nfa->_M_insert_matcher(std::move(__matcher))));
417  }
418 
419  template<typename _TraitsT>
420  template<bool __icase, bool __collate>
421  void
422  _Compiler<_TraitsT>::
423  _M_insert_bracket_matcher(bool __neg)
424  {
425  _BracketMatcher<_TraitsT, __icase, __collate> __matcher(__neg, _M_traits);
426  pair<bool, _CharT> __last_char; // Optional<_CharT>
427  __last_char.first = false;
428  if (!(_M_flags & regex_constants::ECMAScript))
429  {
430  if (_M_try_char())
431  {
432  __last_char.first = true;
433  __last_char.second = _M_value[0];
434  }
435  else if (_M_match_token(_ScannerT::_S_token_bracket_dash))
436  {
437  __last_char.first = true;
438  __last_char.second = '-';
439  }
440  }
441  while (_M_expression_term(__last_char, __matcher));
442  if (__last_char.first)
443  __matcher._M_add_char(__last_char.second);
444  __matcher._M_ready();
445  _M_stack.push(_StateSeqT(
446  *_M_nfa,
447  _M_nfa->_M_insert_matcher(std::move(__matcher))));
448  }
449 
450  template<typename _TraitsT>
451  template<bool __icase, bool __collate>
452  bool
453  _Compiler<_TraitsT>::
454  _M_expression_term(pair<bool, _CharT>& __last_char,
455  _BracketMatcher<_TraitsT, __icase, __collate>& __matcher)
456  {
457  if (_M_match_token(_ScannerT::_S_token_bracket_end))
458  return false;
459 
460  const auto __push_char = [&](_CharT __ch)
461  {
462  if (__last_char.first)
463  __matcher._M_add_char(__last_char.second);
464  else
465  __last_char.first = true;
466  __last_char.second = __ch;
467  };
468  const auto __flush = [&]
469  {
470  if (__last_char.first)
471  {
472  __matcher._M_add_char(__last_char.second);
473  __last_char.first = false;
474  }
475  };
476 
477  if (_M_match_token(_ScannerT::_S_token_collsymbol))
478  {
479  auto __symbol = __matcher._M_add_collate_element(_M_value);
480  if (__symbol.size() == 1)
481  __push_char(__symbol[0]);
482  else
483  __flush();
484  }
485  else if (_M_match_token(_ScannerT::_S_token_equiv_class_name))
486  {
487  __flush();
488  __matcher._M_add_equivalence_class(_M_value);
489  }
490  else if (_M_match_token(_ScannerT::_S_token_char_class_name))
491  {
492  __flush();
493  __matcher._M_add_character_class(_M_value, false);
494  }
495  else if (_M_try_char())
496  __push_char(_M_value[0]);
497  // POSIX doesn't allow '-' as a start-range char (say [a-z--0]),
498  // except when the '-' is the first or last character in the bracket
499  // expression ([--0]). ECMAScript treats all '-' after a range as a
500  // normal character. Also see above, where _M_expression_term gets called.
501  //
502  // As a result, POSIX rejects [-----], but ECMAScript doesn't.
503  // Boost (1.57.0) always uses POSIX style even in its ECMAScript syntax.
504  // Clang (3.5) always uses ECMAScript style even in its POSIX syntax.
505  //
506  // It turns out that no one reads BNFs ;)
507  else if (_M_match_token(_ScannerT::_S_token_bracket_dash))
508  {
509  if (!__last_char.first)
510  {
511  if (!(_M_flags & regex_constants::ECMAScript))
512  {
513  if (_M_match_token(_ScannerT::_S_token_bracket_end))
514  {
515  __push_char('-');
516  return false;
517  }
518  __throw_regex_error(
520  "Unexpected dash in bracket expression. For POSIX syntax, "
521  "a dash is not treated literally only when it is at "
522  "beginning or end.");
523  }
524  __push_char('-');
525  }
526  else
527  {
528  if (_M_try_char())
529  {
530  __matcher._M_make_range(__last_char.second, _M_value[0]);
531  __last_char.first = false;
532  }
533  else if (_M_match_token(_ScannerT::_S_token_bracket_dash))
534  {
535  __matcher._M_make_range(__last_char.second, '-');
536  __last_char.first = false;
537  }
538  else
539  {
540  if (_M_scanner._M_get_token()
541  != _ScannerT::_S_token_bracket_end)
542  __throw_regex_error(
544  "Character is expected after a dash.");
545  __push_char('-');
546  }
547  }
548  }
549  else if (_M_match_token(_ScannerT::_S_token_quoted_class))
550  {
551  __flush();
552  __matcher._M_add_character_class(_M_value,
553  _M_ctype.is(_CtypeT::upper,
554  _M_value[0]));
555  }
556  else
557  __throw_regex_error(regex_constants::error_brack,
558  "Unexpected character in bracket expression.");
559 
560  return true;
561  }
562 
563  template<typename _TraitsT>
564  bool
565  _Compiler<_TraitsT>::
566  _M_try_char()
567  {
568  bool __is_char = false;
569  if (_M_match_token(_ScannerT::_S_token_oct_num))
570  {
571  __is_char = true;
572  _M_value.assign(1, _M_cur_int_value(8));
573  }
574  else if (_M_match_token(_ScannerT::_S_token_hex_num))
575  {
576  __is_char = true;
577  _M_value.assign(1, _M_cur_int_value(16));
578  }
579  else if (_M_match_token(_ScannerT::_S_token_ord_char))
580  __is_char = true;
581  return __is_char;
582  }
583 
584  template<typename _TraitsT>
585  bool
586  _Compiler<_TraitsT>::
587  _M_match_token(_TokenT token)
588  {
589  if (token == _M_scanner._M_get_token())
590  {
591  _M_value = _M_scanner._M_get_value();
592  _M_scanner._M_advance();
593  return true;
594  }
595  return false;
596  }
597 
598  template<typename _TraitsT>
599  int
600  _Compiler<_TraitsT>::
601  _M_cur_int_value(int __radix)
602  {
603  long __v = 0;
604  for (typename _StringT::size_type __i = 0;
605  __i < _M_value.length(); ++__i)
606  __v =__v * __radix + _M_traits.value(_M_value[__i], __radix);
607  return __v;
608  }
609 
610  template<typename _TraitsT, bool __icase, bool __collate>
611  bool
612  _BracketMatcher<_TraitsT, __icase, __collate>::
613  _M_apply(_CharT __ch, false_type) const
614  {
615  return [this, __ch]
616  {
617  if (std::binary_search(_M_char_set.begin(), _M_char_set.end(),
618  _M_translator._M_translate(__ch)))
619  return true;
620  auto __s = _M_translator._M_transform(__ch);
621  for (auto& __it : _M_range_set)
622  if (_M_translator._M_match_range(__it.first, __it.second, __s))
623  return true;
624  if (_M_traits.isctype(__ch, _M_class_set))
625  return true;
626  if (std::find(_M_equiv_set.begin(), _M_equiv_set.end(),
627  _M_traits.transform_primary(&__ch, &__ch+1))
628  != _M_equiv_set.end())
629  return true;
630  for (auto& __it : _M_neg_class_set)
631  if (!_M_traits.isctype(__ch, __it))
632  return true;
633  return false;
634  }() ^ _M_is_non_matching;
635  }
636 
637 _GLIBCXX_END_NAMESPACE_VERSION
638 } // namespace __detail
639 } // namespace
constexpr error_type error_badrepeat(_S_error_badrepeat)
_GLIBCXX17_INLINE constexpr syntax_option_type grep
bool empty() const
Definition: stl_stack.h:185
const _Facet & use_facet(const locale &__loc)
Return a facet.use_facet looks for and returns a reference to a facet of type Facet where Facet is th...
ISO C++ entities toplevel namespace is std.
_GLIBCXX17_INLINE constexpr syntax_option_type awk
_GLIBCXX17_INLINE constexpr syntax_option_type extended
void push(const value_type &__x)
Add data to the top of the stack.
Definition: stl_stack.h:225
shared_ptr< _Tp > make_shared(_Args &&... __args)
Create an object that is owned by a shared_ptr.
void pop()
Removes first element.
Definition: stl_stack.h:258
constexpr error_type error_range(_S_error_range)
constexpr error_type error_badbrace(_S_error_badbrace)
reference top()
Definition: stl_stack.h:198
bool is(mask __m, char_type __c) const
Test char_type classification.
_GLIBCXX17_INLINE constexpr syntax_option_type basic
A standard container giving FILO behavior.
Definition: stl_stack.h:99
constexpr error_type error_brack(_S_error_brack)
integral_constant< bool, false > false_type
The type used as a compile-time boolean with false value.
Definition: type_traits:90
_GLIBCXX17_INLINE constexpr syntax_option_type ECMAScript
constexpr error_type error_brace(_S_error_brace)
_GLIBCXX17_INLINE constexpr syntax_option_type egrep
constexpr error_type error_paren(_S_error_paren)