libstdc++
regex_scanner.tcc
Go to the documentation of this file.
00001 // class template regex -*- C++ -*-
00002 
00003 // Copyright (C) 2013-2015 Free Software Foundation, Inc.
00004 //
00005 // This file is part of the GNU ISO C++ Library.  This library is free
00006 // software; you can redistribute it and/or modify it under the
00007 // terms of the GNU General Public License as published by the
00008 // Free Software Foundation; either version 3, or (at your option)
00009 // any later version.
00010 
00011 // This library is distributed in the hope that it will be useful,
00012 // but WITHOUT ANY WARRANTY; without even the implied warranty of
00013 // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
00014 // GNU General Public License for more details.
00015 
00016 // Under Section 7 of GPL version 3, you are granted additional
00017 // permissions described in the GCC Runtime Library Exception, version
00018 // 3.1, as published by the Free Software Foundation.
00019 
00020 // You should have received a copy of the GNU General Public License and
00021 // a copy of the GCC Runtime Library Exception along with this program;
00022 // see the files COPYING3 and COPYING.RUNTIME respectively.  If not, see
00023 // <http://www.gnu.org/licenses/>.
00024 
00025 /**
00026  *  @file bits/regex_scanner.tcc
00027  *  This is an internal header file, included by other library headers.
00028  *  Do not attempt to use it directly. @headername{regex}
00029  */
00030 
00031 // FIXME make comments doxygen format.
00032 
00033 // N3376 specified 6 regex styles: ECMAScript, basic, extended, grep, egrep
00034 // and awk
00035 // 1) grep is basic except '\n' is treated as '|'
00036 // 2) egrep is extended except '\n' is treated as '|'
00037 // 3) awk is extended except special escaping rules, and there's no
00038 //    back-reference.
00039 //
00040 // References:
00041 //
00042 // ECMAScript: ECMA-262 15.10
00043 //
00044 // basic, extended:
00045 // http://pubs.opengroup.org/onlinepubs/009695399/basedefs/xbd_chap09.html
00046 //
00047 // awk: http://pubs.opengroup.org/onlinepubs/000095399/utilities/awk.html
00048 
00049 namespace std _GLIBCXX_VISIBILITY(default)
00050 {
00051 namespace __detail
00052 {
00053 _GLIBCXX_BEGIN_NAMESPACE_VERSION
00054 
00055   template<typename _CharT>
00056     _Scanner<_CharT>::
00057     _Scanner(typename _Scanner::_IterT __begin,
00058              typename _Scanner::_IterT __end,
00059              _FlagT __flags, std::locale __loc)
00060     : _ScannerBase(__flags),
00061       _M_current(__begin), _M_end(__end),
00062       _M_ctype(std::use_facet<_CtypeT>(__loc)),
00063       _M_eat_escape(_M_is_ecma()
00064                     ? &_Scanner::_M_eat_escape_ecma
00065                     : &_Scanner::_M_eat_escape_posix)
00066     { _M_advance(); }
00067 
00068   template<typename _CharT>
00069     void
00070     _Scanner<_CharT>::
00071     _M_advance()
00072     {
00073       if (_M_current == _M_end)
00074         {
00075           _M_token = _S_token_eof;
00076           return;
00077         }
00078 
00079       if (_M_state == _S_state_normal)
00080         _M_scan_normal();
00081       else if (_M_state == _S_state_in_bracket)
00082         _M_scan_in_bracket();
00083       else if (_M_state == _S_state_in_brace)
00084         _M_scan_in_brace();
00085       else
00086         {
00087           _GLIBCXX_DEBUG_ASSERT(false);
00088         }
00089     }
00090 
00091   // Differences between styles:
00092   // 1) "\(", "\)", "\{" in basic. It's not escaping.
00093   // 2) "(?:", "(?=", "(?!" in ECMAScript.
00094   template<typename _CharT>
00095     void
00096     _Scanner<_CharT>::
00097     _M_scan_normal()
00098     {
00099       auto __c = *_M_current++;
00100 
00101       if (std::strchr(_M_spec_char, _M_ctype.narrow(__c, ' ')) == nullptr)
00102         {
00103           _M_token = _S_token_ord_char;
00104           _M_value.assign(1, __c);
00105           return;
00106         }
00107       if (__c == '\\')
00108         {
00109           if (_M_current == _M_end)
00110             __throw_regex_error(regex_constants::error_escape);
00111 
00112           if (!_M_is_basic()
00113               || (*_M_current != '('
00114                   && *_M_current != ')'
00115                   && *_M_current != '{'))
00116             {
00117               (this->*_M_eat_escape)();
00118               return;
00119             }
00120           __c = *_M_current++;
00121         }
00122       if (__c == '(')
00123         {
00124           if (_M_is_ecma() && *_M_current == '?')
00125             {
00126               if (++_M_current == _M_end)
00127                 __throw_regex_error(regex_constants::error_paren);
00128 
00129               if (*_M_current == ':')
00130                 {
00131                   ++_M_current;
00132                   _M_token = _S_token_subexpr_no_group_begin;
00133                 }
00134               else if (*_M_current == '=')
00135                 {
00136                   ++_M_current;
00137                   _M_token = _S_token_subexpr_lookahead_begin;
00138                   _M_value.assign(1, 'p');
00139                 }
00140               else if (*_M_current == '!')
00141                 {
00142                   ++_M_current;
00143                   _M_token = _S_token_subexpr_lookahead_begin;
00144                   _M_value.assign(1, 'n');
00145                 }
00146               else
00147                 __throw_regex_error(regex_constants::error_paren);
00148             }
00149           else if (_M_flags & regex_constants::nosubs)
00150             _M_token = _S_token_subexpr_no_group_begin;
00151           else
00152             _M_token = _S_token_subexpr_begin;
00153         }
00154       else if (__c == ')')
00155         _M_token = _S_token_subexpr_end;
00156       else if (__c == '[')
00157         {
00158           _M_state = _S_state_in_bracket;
00159           _M_at_bracket_start = true;
00160           if (_M_current != _M_end && *_M_current == '^')
00161             {
00162               _M_token = _S_token_bracket_neg_begin;
00163               ++_M_current;
00164             }
00165           else
00166             _M_token = _S_token_bracket_begin;
00167         }
00168       else if (__c == '{')
00169         {
00170           _M_state = _S_state_in_brace;
00171           _M_token = _S_token_interval_begin;
00172         }
00173       else if (__c != ']' && __c != '}')
00174         {
00175           auto __it = _M_token_tbl;
00176           auto __narrowc = _M_ctype.narrow(__c, '\0');
00177           for (; __it->first != '\0'; ++__it)
00178             if (__it->first == __narrowc)
00179               {
00180                 _M_token = __it->second;
00181                 return;
00182               }
00183           _GLIBCXX_DEBUG_ASSERT(false);
00184         }
00185       else
00186         {
00187           _M_token = _S_token_ord_char;
00188           _M_value.assign(1, __c);
00189         }
00190     }
00191 
00192   // Differences between styles:
00193   // 1) different semantics of "[]" and "[^]".
00194   // 2) Escaping in bracket expr.
00195   template<typename _CharT>
00196     void
00197     _Scanner<_CharT>::
00198     _M_scan_in_bracket()
00199     {
00200       if (_M_current == _M_end)
00201         __throw_regex_error(regex_constants::error_brack);
00202 
00203       auto __c = *_M_current++;
00204 
00205       if (__c == '[')
00206         {
00207           if (_M_current == _M_end)
00208             __throw_regex_error(regex_constants::error_brack);
00209 
00210           if (*_M_current == '.')
00211             {
00212               _M_token = _S_token_collsymbol;
00213               _M_eat_class(*_M_current++);
00214             }
00215           else if (*_M_current == ':')
00216             {
00217               _M_token = _S_token_char_class_name;
00218               _M_eat_class(*_M_current++);
00219             }
00220           else if (*_M_current == '=')
00221             {
00222               _M_token = _S_token_equiv_class_name;
00223               _M_eat_class(*_M_current++);
00224             }
00225           else
00226             {
00227               _M_token = _S_token_ord_char;
00228               _M_value.assign(1, __c);
00229             }
00230         }
00231       // In POSIX, when encountering "[]" or "[^]", the ']' is interpreted
00232       // literally. So "[]]" and "[^]]" are valid regexes. See the testcases
00233       // `*/empty_range.cc`.
00234       else if (__c == ']' && (_M_is_ecma() || !_M_at_bracket_start))
00235         {
00236           _M_token = _S_token_bracket_end;
00237           _M_state = _S_state_normal;
00238         }
00239       // ECMAScript and awk permits escaping in bracket.
00240       else if (__c == '\\' && (_M_is_ecma() || _M_is_awk()))
00241         (this->*_M_eat_escape)();
00242       else
00243         {
00244           _M_token = _S_token_ord_char;
00245           _M_value.assign(1, __c);
00246         }
00247       _M_at_bracket_start = false;
00248     }
00249 
00250   // Differences between styles:
00251   // 1) "\}" in basic style.
00252   template<typename _CharT>
00253     void
00254     _Scanner<_CharT>::
00255     _M_scan_in_brace()
00256     {
00257       if (_M_current == _M_end)
00258         __throw_regex_error(regex_constants::error_brace);
00259 
00260       auto __c = *_M_current++;
00261 
00262       if (_M_ctype.is(_CtypeT::digit, __c))
00263         {
00264           _M_token = _S_token_dup_count;
00265           _M_value.assign(1, __c);
00266           while (_M_current != _M_end
00267                  && _M_ctype.is(_CtypeT::digit, *_M_current))
00268             _M_value += *_M_current++;
00269         }
00270       else if (__c == ',')
00271         _M_token = _S_token_comma;
00272       // basic use \}.
00273       else if (_M_is_basic())
00274         {
00275           if (__c == '\\' && _M_current != _M_end && *_M_current == '}')
00276             {
00277               _M_state = _S_state_normal;
00278               _M_token = _S_token_interval_end;
00279               ++_M_current;
00280             }
00281           else
00282             __throw_regex_error(regex_constants::error_badbrace);
00283         }
00284       else if (__c == '}')
00285         {
00286           _M_state = _S_state_normal;
00287           _M_token = _S_token_interval_end;
00288         }
00289       else
00290         __throw_regex_error(regex_constants::error_badbrace);
00291     }
00292 
00293   template<typename _CharT>
00294     void
00295     _Scanner<_CharT>::
00296     _M_eat_escape_ecma()
00297     {
00298       if (_M_current == _M_end)
00299         __throw_regex_error(regex_constants::error_escape);
00300 
00301       auto __c = *_M_current++;
00302       auto __pos = _M_find_escape(_M_ctype.narrow(__c, '\0'));
00303 
00304       if (__pos != nullptr && (__c != 'b' || _M_state == _S_state_in_bracket))
00305         {
00306           _M_token = _S_token_ord_char;
00307           _M_value.assign(1, *__pos);
00308         }
00309       else if (__c == 'b')
00310         {
00311           _M_token = _S_token_word_bound;
00312           _M_value.assign(1, 'p');
00313         }
00314       else if (__c == 'B')
00315         {
00316           _M_token = _S_token_word_bound;
00317           _M_value.assign(1, 'n');
00318         }
00319       // N3376 28.13
00320       else if (__c == 'd'
00321                || __c == 'D'
00322                || __c == 's'
00323                || __c == 'S'
00324                || __c == 'w'
00325                || __c == 'W')
00326         {
00327           _M_token = _S_token_quoted_class;
00328           _M_value.assign(1, __c);
00329         }
00330       else if (__c == 'c')
00331         {
00332           if (_M_current == _M_end)
00333             __throw_regex_error(regex_constants::error_escape);
00334           _M_token = _S_token_ord_char;
00335           _M_value.assign(1, *_M_current++);
00336         }
00337       else if (__c == 'x' || __c == 'u')
00338         {
00339           _M_value.erase();
00340           for (int __i = 0; __i < (__c == 'x' ? 2 : 4); __i++)
00341             {
00342               if (_M_current == _M_end
00343                   || !_M_ctype.is(_CtypeT::xdigit, *_M_current))
00344                 __throw_regex_error(regex_constants::error_escape);
00345               _M_value += *_M_current++;
00346             }
00347           _M_token = _S_token_hex_num;
00348         }
00349       // ECMAScript recognizes multi-digit back-references.
00350       else if (_M_ctype.is(_CtypeT::digit, __c))
00351         {
00352           _M_value.assign(1, __c);
00353           while (_M_current != _M_end
00354                  && _M_ctype.is(_CtypeT::digit, *_M_current))
00355             _M_value += *_M_current++;
00356           _M_token = _S_token_backref;
00357         }
00358       else
00359         {
00360           _M_token = _S_token_ord_char;
00361           _M_value.assign(1, __c);
00362         }
00363     }
00364 
00365   // Differences between styles:
00366   // 1) Extended doesn't support backref, but basic does.
00367   template<typename _CharT>
00368     void
00369     _Scanner<_CharT>::
00370     _M_eat_escape_posix()
00371     {
00372       if (_M_current == _M_end)
00373         __throw_regex_error(regex_constants::error_escape);
00374 
00375       auto __c = *_M_current;
00376       auto __pos = std::strchr(_M_spec_char, _M_ctype.narrow(__c, '\0'));
00377 
00378       if (__pos != nullptr && *__pos != '\0')
00379         {
00380           _M_token = _S_token_ord_char;
00381           _M_value.assign(1, __c);
00382         }
00383       // We MUST judge awk before handling backrefs. There's no backref in awk.
00384       else if (_M_is_awk())
00385         {
00386           _M_eat_escape_awk();
00387           return;
00388         }
00389       else if (_M_is_basic() && _M_ctype.is(_CtypeT::digit, __c) && __c != '0')
00390         {
00391           _M_token = _S_token_backref;
00392           _M_value.assign(1, __c);
00393         }
00394       else
00395         {
00396 #ifdef __STRICT_ANSI__
00397           // POSIX says it is undefined to escape ordinary characters
00398           __throw_regex_error(regex_constants::error_escape);
00399 #else
00400           _M_token = _S_token_ord_char;
00401           _M_value.assign(1, __c);
00402 #endif
00403         }
00404       ++_M_current;
00405     }
00406 
00407   template<typename _CharT>
00408     void
00409     _Scanner<_CharT>::
00410     _M_eat_escape_awk()
00411     {
00412       auto __c = *_M_current++;
00413       auto __pos = _M_find_escape(_M_ctype.narrow(__c, '\0'));
00414 
00415       if (__pos != nullptr)
00416         {
00417           _M_token = _S_token_ord_char;
00418           _M_value.assign(1, *__pos);
00419         }
00420       // \ddd for oct representation
00421       else if (_M_ctype.is(_CtypeT::digit, __c)
00422                && __c != '8'
00423                && __c != '9')
00424         {
00425           _M_value.assign(1,  __c);
00426           for (int __i = 0;
00427                __i < 2
00428                && _M_current != _M_end
00429                && _M_ctype.is(_CtypeT::digit, *_M_current)
00430                && *_M_current != '8'
00431                && *_M_current != '9';
00432                __i++)
00433             _M_value += *_M_current++;
00434           _M_token = _S_token_oct_num;
00435           return;
00436         }
00437       else
00438         __throw_regex_error(regex_constants::error_escape);
00439     }
00440 
00441   // Eats a character class or throws an exception.
00442   // __ch could be ':', '.' or '=', _M_current is the char after ']' when
00443   // returning.
00444   template<typename _CharT>
00445     void
00446     _Scanner<_CharT>::
00447     _M_eat_class(char __ch)
00448     {
00449       for (_M_value.clear(); _M_current != _M_end && *_M_current != __ch;)
00450         _M_value += *_M_current++;
00451       if (_M_current == _M_end
00452           || *_M_current++ != __ch
00453           || _M_current == _M_end // skip __ch
00454           || *_M_current++ != ']') // skip ']'
00455         {
00456           if (__ch == ':')
00457             __throw_regex_error(regex_constants::error_ctype);
00458           else
00459             __throw_regex_error(regex_constants::error_collate);
00460         }
00461     }
00462 
00463 #ifdef _GLIBCXX_DEBUG
00464   template<typename _CharT>
00465     std::ostream&
00466     _Scanner<_CharT>::
00467     _M_print(std::ostream& ostr)
00468     {
00469       switch (_M_token)
00470       {
00471       case _S_token_anychar:
00472         ostr << "any-character\n";
00473         break;
00474       case _S_token_backref:
00475         ostr << "backref\n";
00476         break;
00477       case _S_token_bracket_begin:
00478         ostr << "bracket-begin\n";
00479         break;
00480       case _S_token_bracket_neg_begin:
00481         ostr << "bracket-neg-begin\n";
00482         break;
00483       case _S_token_bracket_end:
00484         ostr << "bracket-end\n";
00485         break;
00486       case _S_token_char_class_name:
00487         ostr << "char-class-name \"" << _M_value << "\"\n";
00488         break;
00489       case _S_token_closure0:
00490         ostr << "closure0\n";
00491         break;
00492       case _S_token_closure1:
00493         ostr << "closure1\n";
00494         break;
00495       case _S_token_collsymbol:
00496         ostr << "collsymbol \"" << _M_value << "\"\n";
00497         break;
00498       case _S_token_comma:
00499         ostr << "comma\n";
00500         break;
00501       case _S_token_dup_count:
00502         ostr << "dup count: " << _M_value << "\n";
00503         break;
00504       case _S_token_eof:
00505         ostr << "EOF\n";
00506         break;
00507       case _S_token_equiv_class_name:
00508         ostr << "equiv-class-name \"" << _M_value << "\"\n";
00509         break;
00510       case _S_token_interval_begin:
00511         ostr << "interval begin\n";
00512         break;
00513       case _S_token_interval_end:
00514         ostr << "interval end\n";
00515         break;
00516       case _S_token_line_begin:
00517         ostr << "line begin\n";
00518         break;
00519       case _S_token_line_end:
00520         ostr << "line end\n";
00521         break;
00522       case _S_token_opt:
00523         ostr << "opt\n";
00524         break;
00525       case _S_token_or:
00526         ostr << "or\n";
00527         break;
00528       case _S_token_ord_char:
00529         ostr << "ordinary character: \"" << _M_value << "\"\n";
00530         break;
00531       case _S_token_subexpr_begin:
00532         ostr << "subexpr begin\n";
00533         break;
00534       case _S_token_subexpr_no_group_begin:
00535         ostr << "no grouping subexpr begin\n";
00536         break;
00537       case _S_token_subexpr_lookahead_begin:
00538         ostr << "lookahead subexpr begin\n";
00539         break;
00540       case _S_token_subexpr_end:
00541         ostr << "subexpr end\n";
00542         break;
00543       case _S_token_unknown:
00544         ostr << "-- unknown token --\n";
00545         break;
00546       case _S_token_oct_num:
00547         ostr << "oct number " << _M_value << "\n";
00548         break;
00549       case _S_token_hex_num:
00550         ostr << "hex number " << _M_value << "\n";
00551         break;
00552       case _S_token_quoted_class:
00553         ostr << "quoted class " << "\\" << _M_value << "\n";
00554         break;
00555       default:
00556         _GLIBCXX_DEBUG_ASSERT(false);
00557       }
00558       return ostr;
00559     }
00560 #endif
00561 
00562 _GLIBCXX_END_NAMESPACE_VERSION
00563 } // namespace __detail
00564 } // namespace