basic_regex_creator.hpp

C++的一个好库。。。现在很流行

         s1 = this->m_traits.transform(a1, (a1[1] ? a1+2 : a1+1));
         s2 = this->m_traits.transform(a2, (a2[1] ? a2+2 : a2+1));
#endif
         if(s1.size() == 0)
            s1 = string_type(1, charT(0));
         if(s2.size() == 0)
            s2 = string_type(1, charT(0));
      }
      else
      {
         if(c1.second)
         {
            s1.insert(s1.end(), c1.first);
            s1.insert(s1.end(), c1.second);
         }
         else
            s1 = string_type(1, c1.first);
         if(c2.second)
         {
            s2.insert(s2.end(), c2.first);
            s2.insert(s2.end(), c2.second);
         }
         else
            s2.insert(s2.end(), c2.first);
      }
      if(s1 > s2)
      {
         // Oops error:
         return 0;
      }
      charT* p = static_cast<charT*>(this->m_pdata->m_data.extend(sizeof(charT) * (s1.size() + s2.size() + 2) ) );
      re_detail::copy(s1.begin(), s1.end(), p);
      p[s1.size()] = charT(0);
      p += s1.size() + 1;
      re_detail::copy(s2.begin(), s2.end(), p);
      p[s2.size()] = charT(0);
   }
   //
   // now process the equivalence classes:
   //
   first = char_set.equivalents_begin();
   last = char_set.equivalents_end();
   while(first != last)
   {
      string_type s;
      if(first->second)
      {
#if BOOST_WORKAROUND(__GNUC__, < 3)
         string_type in(3, charT(0));
         in[0] = first->first;
         in[1] = first->second;
         s = m_traits.transform_primary(in.c_str(), in.c_str()+2);
#else
         charT cs[3] = { first->first, first->second, charT(0), };
         s = m_traits.transform_primary(cs, cs+2);
#endif
      }
      else
         s = m_traits.transform_primary(&first->first, &first->first+1);
      if(s.empty())
         return 0;  // invalid or unsupported equivalence class
      charT* p = static_cast<charT*>(this->m_pdata->m_data.extend(sizeof(charT) * (s.size()+1) ) );
      re_detail::copy(s.begin(), s.end(), p);
      p[s.size()] = charT(0);
      ++first;
   }
   //
   // finally reset the address of our last state:
   //
   m_last_state = result = static_cast<re_set_long<mask_type>*>(getaddress(offset));
   return result;
}

namespace{

template<class T>
inline bool char_less(T t1, T t2)
{
   return t1 < t2;
}
template<>
inline bool char_less<char>(char t1, char t2)
{
   return static_cast<unsigned char>(t1) < static_cast<unsigned char>(t2);
}
template<>
inline bool char_less<signed char>(signed char t1, signed char t2)
{
   return static_cast<unsigned char>(t1) < static_cast<unsigned char>(t2);
}
}

template <class charT, class traits>
re_syntax_base* basic_regex_creator<charT, traits>::append_set(
   const basic_char_set<charT, traits>& char_set, mpl::true_*)
{
   typedef typename traits::string_type string_type;
   typedef typename basic_char_set<charT, traits>::list_iterator item_iterator;
   
   re_set* result = static_cast<re_set*>(append_state(syntax_element_set, sizeof(re_set)));
   bool negate = char_set.is_negated();
   std::memset(result->_map, 0, sizeof(result->_map));
   //
   // handle singles first:
   //
   item_iterator first, last;
   first = char_set.singles_begin();
   last = char_set.singles_end();
   while(first != last)
   {
      for(unsigned int i = 0; i < (1 << CHAR_BIT); ++i)
      {
         if(this->m_traits.translate(static_cast<charT>(i), this->m_icase)
            == this->m_traits.translate(first->first, this->m_icase))
            result->_map[i] = true;
      }
      ++first;
   }
   //
   // OK now handle ranges:
   //
   first = char_set.ranges_begin();
   last = char_set.ranges_end();
   while(first != last)
   {
      // first grab the endpoints of the range:
      charT c1 = this->m_traits.translate(first->first, this->m_icase);
      ++first;
      charT c2 = this->m_traits.translate(first->first, this->m_icase);
      ++first;
      // different actions now depending upon whether collation is turned on:
      if(flags() & regex_constants::collate)
      {
         // we need to transform our range into sort keys:
         charT c3[2] = { c1, charT(0), };
         string_type s1 = this->m_traits.transform(c3, c3+1);
         c3[0] = c2;
         string_type s2 = this->m_traits.transform(c3, c3+1);
         if(s1 > s2)
         {
            // Oops error:
            return 0;
         }
         for(unsigned i = 0; i < (1u << CHAR_BIT); ++i)
         {
            charT c3[2] = { static_cast<charT>(i), charT(0), };
            string_type s3 = this->m_traits.transform(c3, c3 +1);
            if((s1 <= s3) && (s3 <= s2))
               result->_map[i] = true;
         }
      }
      else
      {
         if(char_less<charT>(c2, c1))
         {
            // Oops error:
            return 0;
         }
         // everything in range matches:
         std::memset(result->_map + static_cast<unsigned char>(c1), true, 1 + static_cast<unsigned char>(c2) - static_cast<unsigned char>(c1));
      }
   }
   //
   // and now the classes:
   //
   typedef typename traits::char_class_type mask_type;
   mask_type m = char_set.classes();
   if(flags() & regbase::icase)
   {
      // adjust m as needed:
      if(((m & m_lower_mask) == m_lower_mask) || ((m & m_upper_mask) == m_upper_mask))
         m |= m_alpha_mask;
   }
   if(m != 0)
   {
      for(unsigned i = 0; i < (1u << CHAR_BIT); ++i)
      {
         if(this->m_traits.isctype(static_cast<charT>(i), m))
            result->_map[i] = true;
      }
   }
   //
   // and now the negated classes:
   //
   m = char_set.negated_classes();
   if(flags() & regbase::icase)
   {
      // adjust m as needed:
      if(((m & m_lower_mask) == m_lower_mask) || ((m & m_upper_mask) == m_upper_mask))
         m |= m_alpha_mask;
   }
   if(m != 0)
   {
      for(unsigned i = 0; i < (1u << CHAR_BIT); ++i)
      {
         if(0 == this->m_traits.isctype(static_cast<charT>(i), m))
            result->_map[i] = true;
      }
   }
   //
   // now process the equivalence classes:
   //
   first = char_set.equivalents_begin();
   last = char_set.equivalents_end();
   while(first != last)
   {
      string_type s;
      BOOST_ASSERT(static_cast<charT>(0) == first->second);
      s = m_traits.transform_primary(&first->first, &first->first+1);
      if(s.empty())
         return 0;  // invalid or unsupported equivalence class
      for(unsigned i = 0; i < (1u << CHAR_BIT); ++i)
      {
         charT c[2] = { (static_cast<charT>(i)), charT(0), };
         string_type s2 = this->m_traits.transform_primary(c, c+1);
         if(s == s2)
            result->_map[i] = true;
      }
      ++first;
   }
   if(negate)
   {
      for(unsigned i = 0; i < (1u << CHAR_BIT); ++i)
      {
         result->_map[i] = !(result->_map[i]);
      }
   }
   return result;
}

template <class charT, class traits>
void basic_regex_creator<charT, traits>::finalize(const charT* p1, const charT* p2)
{
   // we've added all the states we need, now finish things off.
   // start by adding a terminating state:
   append_state(syntax_element_match);
   // extend storage to store original expression:
   std::ptrdiff_t len = p2 - p1;
   m_pdata->m_expression_len = len;
   charT* ps = static_cast<charT*>(m_pdata->m_data.extend(sizeof(charT) * (1 + (p2 - p1))));
   m_pdata->m_expression = ps;
   re_detail::copy(p1, p2, ps);
   ps[p2 - p1] = 0;
   // fill in our other data...
   // successful parsing implies a zero status:
   m_pdata->m_status = 0;
   // get the first state of the machine:
   m_pdata->m_first_state = static_cast<re_syntax_base*>(m_pdata->m_data.data());
   // fixup pointers in the machine:
   fixup_pointers(m_pdata->m_first_state);
   // create nested startmaps:
   create_startmaps(m_pdata->m_first_state);
   // create main startmap:
   std::memset(m_pdata->m_startmap, 0, sizeof(m_pdata->m_startmap));
   m_pdata->m_can_be_null = 0;

   m_bad_repeats = 0;
   create_startmap(m_pdata->m_first_state, m_pdata->m_startmap, &(m_pdata->m_can_be_null), mask_all);
   // get the restart type:
   m_pdata->m_restart_type = get_restart_type(m_pdata->m_first_state);
   // optimise a leading repeat if there is one:
   probe_leading_repeat(m_pdata->m_first_state);
}

template <class charT, class traits>
void basic_regex_creator<charT, traits>::fixup_pointers(re_syntax_base* state)
{
   while(state)
   {
      switch(state->type)
      {
      case syntax_element_rep:
      case syntax_element_dot_rep:
      case syntax_element_char_rep:
      case syntax_element_short_set_rep:
      case syntax_element_long_set_rep:
         // set the id of this repeat:
         static_cast<re_repeat*>(state)->id = m_repeater_id++;
         // fall through:
      case syntax_element_alt:
         std::memset(static_cast<re_alt*>(state)->_map, 0, sizeof(static_cast<re_alt*>(state)->_map));
         static_cast<re_alt*>(state)->can_be_null = 0;
         // fall through:
      case syntax_element_jump:
         static_cast<re_jump*>(state)->alt.p = getaddress(static_cast<re_jump*>(state)->alt.i, state);
         // fall through again:
      default:
         if(state->next.i)
            state->next.p = getaddress(state->next.i, state);
         else
            state->next.p = 0;
      }
      state = state->next.p;
   }
}

template <class charT, class traits>
void basic_regex_creator<charT, traits>::create_startmaps(re_syntax_base* state)
{
   // non-recursive implementation:
   // create the last map in the machine first, so that earlier maps
   // can make use of the result...
   //
   // This was originally a recursive implementation, but that caused stack
   // overflows with complex expressions on small stacks (think COM+).

   // start by saving the case setting:
   bool l_icase = m_icase;
   std::vector<std::pair<bool, re_syntax_base*> > v;

   while(state)
   {
      switch(state->type)
      {
      case syntax_element_toggle_case:
         // we need to track case changes here:
         m_icase = static_cast<re_case*>(state)->icase;
         state = state->next.p;
         continue;
      case syntax_element_alt:
      case syntax_element_rep:
      case syntax_element_dot_rep:
      case syntax_element_char_rep:
      case syntax_element_short_set_rep:
      case syntax_element_long_set_rep:
         // just push the state onto our stack for now:
         v.push_back(std::pair<bool, re_syntax_base*>(m_icase, state));
         state = state->next.p;
         break;
      case syntax_element_backstep:
         // we need to calculate how big the backstep is:
         static_cast<re_brace*>(state)->index
            = this->calculate_backstep(state->next.p);
         if(static_cast<re_brace*>(state)->index < 0)
         {
            // Oops error:
            if(0 == this->m_pdata->m_status) // update the error code if not already set
               this->m_pdata->m_status = boost::regex_constants::error_bad_pattern;
            //
            // clear the expression, we should be empty:
            //
            this->m_pdata->m_expression = 0;
            this->m_pdata->m_expression_len = 0;
            //
            // and throw if required:
            //
            if(0 == (this->flags() & regex_constants::no_except))
            {
               std::string message = this->m_pdata->m_ptraits->error_string(boost::regex_constants::error_bad_pattern);
               boost::regex_error e(message, boost::regex_constants::error_bad_pattern, 0);
               e.raise();
            }
         }
         // fall through:
      default:
         state = state->next.p;
      }
   }
   // now work through our list, building all the maps as we go:
   while(v.size())
   {
      const std::pair<bool, re_syntax_base*>& p = v.back();
      m_icase = p.first;
      state = p.second;
      v.pop_back();

      // Build maps:
      create_startmap(state->next.p, static_cast<re_alt*>(state)->_map, &static_cast<re_alt*>(state)->can_be_null, mask_take);
      m_bad_repeats = 0;
      create_startmap(static_cast<re_alt*>(state)->alt.p, static_cast<re_alt*>(state)->_map, &static_cast<re_alt*>(state)->can_be_null, mask_skip);
      // adjust the type of the state to allow for faster matching:
      state->type = this->get_repeat_type(state);
   }
   // restore case sensitivity:
   m_icase = l_icase;
}

template <class charT, class traits>
int basic_regex_creator<charT, traits>::calculate_backstep(re_syntax_base* state)
{
   typedef typename traits::char_class_type mask_type;
   int result = 0;
   while(state)
   {
      switch(state->type)
      {
      case syntax_element_startmark:
         if((static_cast<re_brace*>(state)->index == -1)
            || (static_cast<re_brace*>(state)->index == -2))
         {
            state = static_cast<re_jump*>(state->next.p)->alt.p->next.p;
            continue;
         }
         else if(static_cast<re_brace*>(state)->index == -3)
         {
            state = state->next.p->next.p;
            continue;
         }
         break;
      case syntax_element_endmark:
         if((static_cast<re_brace*>(state)->index == -1)
            || (static_cast<re_brace*>(state)->index == -2))
            return result;
         break;
      case syntax_element_literal:
         result += static_cast<re_literal*>(state)->length;
         break;
      case syntax_element_wild:
      case syntax_element_set:
         result += 1;
         break;
      case syntax_element_dot_rep:
      case syntax_element_char_rep:
      case syntax_element_short_set_rep:
      case syntax_element_backref:
      case syntax_element_rep:
      case syntax_element_combining:
      case syntax_element_long_set_rep:
      case syntax_element_backstep:
         {
            re_repeat* rep = static_cast<re_repeat *>(state);
            // adjust the type of the state to allow for faster matching:
            state->type = this->get_repeat_type(state);
            if((state->type == syntax_element_dot_rep) 
               || (state->type == syntax_element_char_rep)
               || (state->type == syntax_element_short_set_rep))
            {
               if(rep->max != rep->min)
                  return -1;
               result += static_cast<int>(rep->min);