regex_compile.hpp

boost库提供标准的C++ API 配合dev c++使用,功能更加强大

      singles.pop();
   }
   while(ranges.empty() == false)
   {
      traits_string_type c1, c2;
      if(nocollate_state)
         c1 = ranges.peek();
      else
         traits_inst.transform(c1, ranges.peek());
      ranges.pop();
      if(nocollate_state)
         c2 = ranges.peek();
      else
         traits_inst.transform(c2, ranges.peek());
      ranges.pop();
      if(c1 < c2)
      {
         // for some reason bc5 crashes when throwing exceptions
         // from here - probably an EH-compiler bug, but hard to
         // be sure...
         // delay throw to later:
         #ifdef __BORLANDC__
         boost::uint_fast32_t f = _flags;
         _flags &= ~regex_constants::use_except;
         #endif
         fail(REG_ERANGE);
         #ifdef __BORLANDC__
         _flags = f;
         #endif
         return 0;
      }
      ++cranges;
      std::size_t len = (re_detail::re_strlen(c1.c_str()) + 1) * sizeof(charT);
      std::memcpy(data.extend(len), c1.c_str(), len);
      len = (re_detail::re_strlen(c2.c_str()) + 1) * sizeof(charT);
      std::memcpy(data.extend(len), c2.c_str(), len);
   }
   while(classes.empty() == false)
   {
      cclasses |= classes.peek();
      classes.pop();
   }
   while(equivalents.empty() == false)
   {
      ++cequivalents;
      const traits_string_type& s = equivalents.peek();
      std::size_t len = (re_detail::re_strlen(s.c_str()) + 1) * sizeof(charT);
      std::memcpy(reinterpret_cast<charT*>(data.extend(len)), s.c_str(), len);
      equivalents.pop();
   }

   re_detail::re_set_long* dat = reinterpret_cast<re_detail::re_set_long*>(reinterpret_cast<unsigned char*>(data.data()) + base);
   dat->type = re_detail::syntax_element_long_set;
   dat->csingles = csingles;
   dat->cranges = cranges;
   dat->cclasses = cclasses;
   dat->cequivalents = cequivalents;
   dat->isnot = isnot;
   dat->next.i = 0;
   dat->singleton = isnot ? true : singleton;
   return dat;
}

template <class charT, class traits, class Allocator>
re_detail::re_syntax_base* BOOST_REGEX_CALL reg_expression<charT, traits, Allocator>::compile_set_aux(re_detail::jstack<traits_string_type, Allocator>& singles, re_detail::jstack<traits_string_type, Allocator>& ranges, re_detail::jstack<boost::uint_fast32_t, Allocator>& classes, re_detail::jstack<traits_string_type, Allocator>& equivalents, bool isnot, const re_detail::_narrow_type&)
{
   re_detail::re_set* dat = reinterpret_cast<re_detail::re_set*>(data.extend(sizeof(re_detail::re_set)));
   std::memset(dat, 0, sizeof(re_detail::re_set));

   while(singles.empty() == false)
   {
      dat->_map[(traits_size_type)(traits_uchar_type)*(singles.peek().c_str())] = re_detail::mask_all;
      singles.pop();
   }
   while(ranges.empty() == false)
   {
      traits_string_type c1, c2, c3, c4;

      if((flags() & regex_constants::collate) == 0)
         c1 = ranges.peek();
      else
         traits_inst.transform(c1, ranges.peek());
      ranges.pop();
      if((flags() & regex_constants::collate) == 0)
         c2 = ranges.peek();
      else
         traits_inst.transform(c2, ranges.peek());
      ranges.pop();

      if(c1 < c2)
      {
         // for some reason bc5 crashes when throwing exceptions
         // from here - probably an EH-compiler bug, but hard to
         // be sure...
         // delay throw to later:
         #ifdef __BORLANDC__
         boost::uint_fast32_t f = _flags;
         _flags &= ~regex_constants::use_except;
         #endif
         fail(REG_ERANGE);
         #ifdef __BORLANDC__
         _flags = f;
         #endif
         return 0;
      }
      for(unsigned int i = 0; i < 256; ++i)
      {
         c4 = (charT)i;
         if((flags() & regex_constants::collate) == 0)
            c3 = c4;
         else
            traits_inst.transform(c3, c4);
         if((c3 <= c1) && (c3 >= c2))
            dat->_map[i] = re_detail::mask_all;
      }
   }
   while(equivalents.empty() == false)
   {
      traits_string_type c1, c2;
      for(unsigned int i = 0; i < 256; ++i)
      {
         c2 = (charT)i;
         traits_inst.transform_primary(c1, c2);
         if(c1 == equivalents.peek())
            dat->_map[i] = re_detail::mask_all;
      }
      equivalents.pop();
   }

   boost::uint_fast32_t l_flags = 0;
   while(classes.empty() == false)
   {
      l_flags |= classes.peek();
      classes.pop();
   }
   if(l_flags)
   {
      for(unsigned int i = 0; i < 256; ++i)
      {
         if(traits_inst.is_class(charT(i), l_flags))
            dat->_map[(traits_uchar_type)traits_inst.translate((charT)i, (_flags & regex_constants::icase))] = re_detail::mask_all;
      }
   }

   if(isnot)
   {
      for(unsigned int i = 0; i < 256; ++i)
      {
         dat->_map[i] = !dat->_map[i];
      }
   }

   dat->type = re_detail::syntax_element_set;
   dat->next.i = 0;
   return dat;
}

#ifndef __CODEGUARD__
// this must not be inline when Borland's codeguard support is turned
// on, otherwise we _will_ get surious codeguard errors...
inline
#endif
 re_detail::re_syntax_base* add_offset(void* base, std::ptrdiff_t off)
{
   return reinterpret_cast<re_detail::re_syntax_base*>(reinterpret_cast<char*>(base) + off);
}

template <class charT, class traits, class Allocator>
void BOOST_REGEX_CALL reg_expression<charT, traits, Allocator>::fixup_apply(re_detail::re_syntax_base* b, unsigned cbraces)
{
   typedef typename boost::detail::rebind_allocator<bool, Allocator>::type b_alloc;
   
   register unsigned char* base = reinterpret_cast<unsigned char*>(b);
   register re_detail::re_syntax_base* ptr = b;
   bool* pb = 0;
   b_alloc a(data.allocator());
#ifndef BOOST_NO_EXCEPTIONS
   try
   {
#endif
      pb = a.allocate(cbraces);
      BOOST_REGEX_NOEH_ASSERT(pb)
      for(unsigned i = 0; i < cbraces; ++i)
         pb[i] = false;

      repeats = 0;

      while(ptr->next.i)
      {
         switch(ptr->type)
         {
         case re_detail::syntax_element_rep:
            jm_assert(data.size() > static_cast<re_detail::re_jump*>(ptr)->alt.i);
            static_cast<re_detail::re_jump*>(ptr)->alt.p = add_offset(base, static_cast<re_detail::re_jump*>(ptr)->alt.i);
#ifdef BOOST_REGEX_DEBUG
            if((re_detail::padding_mask & reinterpret_cast<int>(static_cast<re_detail::re_jump*>(ptr)->alt.p)) && (static_cast<re_detail::re_jump*>(ptr)->alt.p != b))
            {
               jm_trace("padding mis-aligment in repeat jump to object type: " << static_cast<re_detail::re_jump*>(ptr)->alt.p->type)
               //jm_assert(0 == (padding_mask & (int)((re_detail::re_jump*)ptr)->alt.p));
            }
#endif
            static_cast<re_detail::re_repeat*>(ptr)->id = repeats;
            ++repeats;
            goto rebase;
         case re_detail::syntax_element_jump:
         case re_detail::syntax_element_alt:
            jm_assert(data.size() > static_cast<re_detail::re_jump*>(ptr)->alt.i);
            static_cast<re_detail::re_jump*>(ptr)->alt.p = add_offset(base, static_cast<re_detail::re_jump*>(ptr)->alt.i);
#ifdef BOOST_REGEX_DEBUG
            if((re_detail::padding_mask & reinterpret_cast<int>(static_cast<re_detail::re_jump*>(ptr)->alt.p) && (static_cast<re_detail::re_jump*>(ptr)->alt.p != b)))
            {
               jm_trace("padding mis-aligment in alternation jump to object type: " << static_cast<re_detail::re_jump*>(ptr)->alt.p->type)
               //jm_assert(0 == (padding_mask & (int)((re_detail::re_jump*)ptr)->alt.p));
            }
#endif
            goto rebase;
         case re_detail::syntax_element_backref:
            if((static_cast<re_detail::re_brace*>(ptr)->index >= (int)cbraces) || (pb[static_cast<re_detail::re_brace*>(ptr)->index] == false) )
            {
               fail(REG_ESUBREG);
               a.deallocate(pb, cbraces);
               return;
            }
            goto rebase;
         case re_detail::syntax_element_endmark:
            if(static_cast<re_detail::re_brace*>(ptr)->index > 0)
               pb[static_cast<re_detail::re_brace*>(ptr)->index] = true;
            goto rebase;
         default:
            rebase:
            jm_assert(data.size() > ptr->next.i);
            ptr->next.p = add_offset(base, ptr->next.i);
#ifdef BOOST_REGEX_DEBUG
            if((re_detail::padding_mask & (int)(ptr->next.p)) && (static_cast<re_detail::re_jump*>(ptr)->alt.p != b))
            {
               jm_trace("padding mis-alignment in next record of type " << ptr->next.p->type)
               jm_assert(0 == (re_detail::padding_mask & (int)(ptr->next.p)));
            }
#endif
            ptr = ptr->next.p;
         }
      }
      a.deallocate(pb, cbraces);
      pb = 0;
#ifndef BOOST_NO_EXCEPTIONS
   }
   catch(...)
   {
      if(pb)
         a.deallocate(pb, cbraces);
      throw;
   }
#endif
}


template <class charT, class traits, class Allocator>
unsigned int BOOST_REGEX_CALL reg_expression<charT, traits, Allocator>::set_expression(const charT* arg_first, const charT* arg_last, flag_type f)
{
# ifdef BOOST_MSVC
#  pragma warning(push)
#  pragma warning(disable: 4127)
#endif
#ifdef __OpenBSD__ 
   // strxfrm not working on OpenBSD?? 
   f &= ~regex_constants::collate; 
#endif 

   if(arg_first == expression())
   {
      traits_string_type s(arg_first, arg_last);
      return set_expression(s.c_str(), s.c_str() + s.size(), f);
   }
   typedef typename traits_type::sentry sentry_t;
   sentry_t sent(traits_inst);
   if(sent){

   const charT* base = arg_first;
   data.clear();
   _flags = f;
   fail(REG_NOERROR);  // clear any error

   if(arg_first >= arg_last)
   {
      fail(REG_EMPTY);
      return error_code();
   }

   const charT* ptr = arg_first;
   marks = 0;
   re_detail::jstack<std::size_t, Allocator> mark(64, data.allocator());
   re_detail::jstack<int, Allocator> markid(64, data.allocator());
   std::size_t last_mark_popped = 0;
   register traits_size_type c;
   register re_detail::re_syntax_base* dat;

   unsigned rep_min = 0;
   unsigned rep_max = 0;

   //
   // set up header:
   //
   ++marks;
   dat = 0;

   if(_flags & regex_constants::literal)
   {
      while(ptr != arg_last)
      {
         dat = add_literal(dat, traits_inst.translate(*ptr, (_flags & regex_constants::icase)));
         ++ptr;
      }
   }

   while (ptr < arg_last)
   {
      c = (traits_size_type)(traits_uchar_type)*ptr;
      // this is only used for the switch(), but cannot be folded in
      // due to a bug in Comeau 4.2.44beta3
      traits_size_type syntax = traits_inst.syntax_type(c);
      switch(syntax)
      {
      case traits_type::syntax_open_bracket:
         if(_flags & bk_parens)
         {
            dat = add_literal(dat, (charT)c);
            ++ptr;
            continue;
         }
         open_bracked_jump:
         // extend:
         dat = add_simple(dat, re_detail::syntax_element_startmark, sizeof(re_detail::re_brace));
         if(_flags & nosubs)
         {
            markid.push(0);
            static_cast<re_detail::re_brace*>(dat)->index = 0;
         }
         else
         {
            markid.push(marks);
            static_cast<re_detail::re_brace*>(dat)->index = marks++;
         }
         mark.push(data.index(dat));
         ++ptr;
         //
         // check for perl like (?...) extention syntax
         c = (traits_size_type)(traits_uchar_type)*ptr;
         if(((_flags & (bk_parens|perlex)) == perlex) && (traits_type::syntax_question == traits_inst.syntax_type(c)))
         {
            ++ptr;
            c = (traits_size_type)(traits_uchar_type)*ptr;
            // this is only used for the switch(), but cannot be folded in
            // due to a bug in Comeau 4.2.44beta3
            traits_size_type syntax2 = traits_inst.syntax_type(c);
            switch(syntax2)
            {
            case traits_type::syntax_colon: