regex_match.hpp

来自「CGAL is a collaborative effort of severa」· HPP 代码 · 共 1,914 行 · 第 1/5 页

            {
               // either we don't care what we match or we've matched
               // the whole string and can't match anything longer.
               while(matches.empty() == false)
                  matches.pop();
               while(prev_pos.empty() == false)
                  prev_pos.pop();
               while(prev_record.empty() == false)
                  prev_record.pop();
               while(prev_acc.empty() == false)
                  prev_acc.pop();
               return true;
            }
         }
         goto failure;
      case syntax_element_startmark:
         start_mark_jump:
         if(static_cast<const re_brace*>(ptr)->index > 0)
         {
            temp_match.set_first(first, static_cast<const re_brace*>(ptr)->index);
         }
         else if(
            (static_cast<const re_brace*>(ptr)->index == -1)
            || (static_cast<const re_brace*>(ptr)->index == -2)
         )
         {
           matches.push(temp_match);
            for(k = 0; k <= cur_acc; ++k)
               prev_pos.push(start_loop[k]);
            prev_pos.push(first);
            prev_record.push(ptr);
            for(k = 0; k <= cur_acc; ++k)
               prev_acc.push(accumulators[k]);
            prev_acc.push(cur_acc);
            prev_acc.push(match_found);
            match_found = false;
            // skip next jump and fall through:
            ptr = ptr->next.p;
         }
         ptr = ptr->next.p;
         break;
      case syntax_element_endmark:
         end_mark_jump:
         if(static_cast<const re_brace*>(ptr)->index > 0)
         {
            temp_match.set_second(first, static_cast<const re_brace*>(ptr)->index);
         }
         else if(
            (static_cast<const re_brace*>(ptr)->index == -1)
            || (static_cast<const re_brace*>(ptr)->index == -2)
         )
         {
            match_found = true;
            unwind_stack = true;
            goto failure;
         }
         ptr = ptr->next.p;
         break;
      case syntax_element_literal:
      {
         unsigned int len = static_cast<const re_literal*>(ptr)->length;
         const charT* what = reinterpret_cast<const charT*>(static_cast<const re_literal*>(ptr) + 1);
         //
         // compare string with what we stored in
         // our records:
         for(unsigned int i = 0; i < len; ++i, ++first)
         {
            if((first == last) || (traits_inst.translate(*first, icase) != what[i]))
               goto failure;
         }
         ptr = ptr->next.p;
         break;
      }
      case syntax_element_start_line:
         outer_line_check:
         if(first == temp_match[0].first)
         {
            // we're at the start of the buffer
            if(flags & match_prev_avail)
            {
               inner_line_check:
               // check the previous value even though its before
               // the start of our "buffer".
               iterator t(first);
               --t;
               if(traits::is_separator(*t) && !((*t == '\r') && (*first == '\n')) )
               {
                  ptr = ptr->next.p;
                  continue;
               }
               goto failure;
            }
            if((flags & match_not_bol) == 0)
            {
               ptr = ptr->next.p;
               continue;
            }
            goto failure;
         }
         // we're in the middle of the string
         goto inner_line_check;
      case syntax_element_end_line:
         // we're not yet at the end so *first is always valid:
         if(traits::is_separator(*first))
         {
            if((first != base) || (flags & match_prev_avail))
            {
               // check that we're not in the middle of \r\n sequence
               iterator t(first);
               --t;
               if((*t == '\r') && (*first == '\n'))
               {
                  goto failure;
               }
            }
            ptr = ptr->next.p;
            continue;
         }
         goto failure;
      case syntax_element_wild:
         // anything except possibly NULL or \n:
         if(traits::is_separator(*first))
         {
            if(flags & match_not_dot_newline)
               goto failure;
            ptr = ptr->next.p;
            ++first;
            continue;
         }
         if(*first == charT(0))
         {
            if(flags & match_not_dot_null)
               goto failure;
            ptr = ptr->next.p;
            ++first;
            continue;
         }
         ptr = ptr->next.p;
         ++first;
         break;
      case syntax_element_word_boundary:
      {
         // prev and this character must be opposites:
#if defined(BOOST_REGEX_USE_C_LOCALE) && defined(__GNUC__) && (__GNUC__ == 2) && (__GNUC_MINOR__ < 95)
         bool b = traits::is_class(*first, traits::char_class_word);
#else
         bool b = traits_inst.is_class(*first, traits::char_class_word);
#endif
         if((first == temp_match[0].first)  && ((flags & match_prev_avail) == 0))
         {
            if(flags & match_not_bow)
               b ^= true;
            else
               b ^= false;
         }
         else
         {
            --first;
            b ^= traits_inst.is_class(*first, traits::char_class_word);
            ++first;
         }
         if(b)
         {
            ptr = ptr->next.p;
            continue;
         }
         goto failure;
      }
      case syntax_element_within_word:
         // both prev and this character must be traits::char_class_word:
         if(traits_inst.is_class(*first, traits::char_class_word))
         {
            bool b;
            if((first == temp_match[0].first) && ((flags & match_prev_avail) == 0))
               b = false;
            else
            {
               --first;
               b = traits_inst.is_class(*first, traits::char_class_word);
               ++first;
            }
            if(b)
            {
               ptr = ptr->next.p;
               continue;
            }
         }
         goto failure;
      case syntax_element_word_start:
         if((first == temp_match[0].first) && ((flags & match_prev_avail) == 0))
         {
            // start of buffer:
            if(flags & match_not_bow)
               goto failure;
            if(traits_inst.is_class(*first, traits::char_class_word))
            {
               ptr = ptr->next.p;
               continue;
            }
            goto failure;
         }
         // otherwise inside buffer:
         if(traits_inst.is_class(*first, traits::char_class_word))
         {
            iterator t(first);
            --t;
            if(traits_inst.is_class(*t, traits::char_class_word) == false)
            {
               ptr = ptr->next.p;
               continue;
            }
         }
         goto failure;      // if we fall through to here then we've failed
      case syntax_element_word_end:
         if((first == temp_match[0].first) && ((flags & match_prev_avail) == 0))
            goto failure;  // start of buffer can't be end of word

         // otherwise inside buffer:
         if(traits_inst.is_class(*first, traits::char_class_word) == false)
         {
            iterator t(first);
            --t;
            if(traits_inst.is_class(*t, traits::char_class_word))
            {
               ptr = ptr->next.p;
               continue;
            }
         }
         goto failure;      // if we fall through to here then we've failed
      case syntax_element_buffer_start:
         if((first != temp_match[0].first) || (flags & match_not_bob))
            goto failure;
         // OK match:
         ptr = ptr->next.p;
         break;
      case syntax_element_buffer_end:
         if((first != last) || (flags & match_not_eob))
            goto failure;
         // OK match:
         ptr = ptr->next.p;
         break;
      case syntax_element_backref:
      {
         // compare with what we previously matched:
         iterator i = temp_match[static_cast<const re_brace*>(ptr)->index].first;
         iterator j = temp_match[static_cast<const re_brace*>(ptr)->index].second;
         while(i != j)
         {
            if((first == last) || (traits_inst.translate(*first, icase) != traits_inst.translate(*i, icase)))
               goto failure;
            ++i;
            ++first;
         }
         ptr = ptr->next.p;
         break;
      }
      case syntax_element_long_set:
      {
         // let the traits class do the work:
         iterator t = re_is_set_member(first, last, static_cast<const re_set_long*>(ptr), e);
         if(t != first)
         {
            ptr = ptr->next.p;
            first = t;
            continue;
         }
         goto failure;
      }
      case syntax_element_set:
         // lookup character in table:
         if(static_cast<const re_set*>(ptr)->_map[(traits_uchar_type)traits_inst.translate(*first, icase)])
         {
            ptr = ptr->next.p;
            ++first;
            continue;
         }
         goto failure;
      case syntax_element_jump:
         ptr = static_cast<const re_jump*>(ptr)->alt.p;
         continue;
      case syntax_element_alt:
      {
         // alt_jump:
         if(access::can_start(*first, static_cast<const re_jump*>(ptr)->_map, (unsigned char)mask_take))
         {
            // we can take the first alternative,
            // see if we need to push next alternative:
            if(access::can_start(*first, static_cast<const re_jump*>(ptr)->_map, mask_skip))
            {
               if(need_push_match)
                  matches.push(temp_match);
               for(k = 0; k <= cur_acc; ++k)
                  prev_pos.push(start_loop[k]);
               prev_pos.push(first);
               prev_record.push(ptr);
               for(k = 0; k <= cur_acc; ++k)
                  prev_acc.push(accumulators[k]);
               prev_acc.push(cur_acc);
            }
            ptr = ptr->next.p;
            continue;
         }
         if(access::can_start(*first, static_cast<const re_jump*>(ptr)->_map, mask_skip))
         {
            ptr = static_cast<const re_jump*>(ptr)->alt.p;
            continue;
         }
         goto failure;  // neither option is possible
      }
      case syntax_element_rep:
      {
         // repeater_jump:
         // if we're moving to a higher id (nested repeats etc)
         // zero out our accumualtors:
         if(cur_acc < static_cast<const re_repeat*>(ptr)->id)
         {
            cur_acc = static_cast<const re_repeat*>(ptr)->id;
            accumulators[cur_acc] = 0;
            start_loop[cur_acc] = first;
         }

         cur_acc = static_cast<const re_repeat*>(ptr)->id;

         if(static_cast<const re_repeat*>(ptr)->leading)
            *restart = first;

         //charT c = traits_inst.translate(*first);

         // first of all test for special case where this is last element,
         // if that is the case then repeat as many times as possible,
         // as long as the repeat is greedy:

         if((static_cast<const re_repeat*>(ptr)->alt.p->type == syntax_element_match)
            && (static_cast<const re_repeat*>(ptr)->greedy == true))
         {
            // see if we can take the repeat:
            if(((unsigned int)accumulators[cur_acc] < static_cast<const re_repeat*>(ptr)->max)
                  && access::can_start(*first, static_cast<const re_repeat*>(ptr)->_map, mask_take))
            {
               // push terminating match as fallback:
               if((unsigned int)accumulators[cur_acc] >= static_cast<const re_repeat*>(ptr)->min)
               {
                  if((prev_record.empty() == false) && (prev_record.peek() == static_cast<const re_repeat*>(ptr)->alt.p))
                  {
                     // we already have the required fallback
                     // don't add any more, just update this one:
                     if(need_push_match)
                        matches.peek() = temp_match;
                     prev_pos.peek() = first;
                  }
                  else
                  {
                     if(need_push_match)
                        matches.push(temp_match);
                     prev_pos.push(first);
                     prev_record.push(static_cast<const re_repeat*>(ptr)->alt.p);
                  }
               }
               // move to next item in list:
               if((first != start_loop[cur_acc]) || !accumulators[cur_acc])
               {
                  ++accumulators[cur_acc];
                  ptr = ptr->next.p;
                  start_loop[cur_acc] = first;
                  continue;
               }
               else if((unsigned int)accumulators[cur_acc] < static_cast<const re_repeat*>(ptr)->min)
               {
                  // the repeat was null, and we haven't gone round min times yet,
                  // since all subsequent repeats will be null as well, just update
                  // our repeat count and skip out.
                  accumulators[cur_acc] = static_cast<const re_repeat*>(ptr)->min;
                  ptr = static_cast<const re_repeat*>(ptr)->alt.p;
                  continue;
               }
               goto failure;
            }
            // see if we can skip the repeat:
            if(((unsigned int)accumulators[cur_acc] >= static_cast<const re_repeat*>(ptr)->min)
               && access::can_start(*first, static_cast<const re_repeat*>(ptr)->_map, mask_skip))
            {
               ptr = static_cast<const re_repeat*>(ptr)->alt.p;
               continue;