// Copyright (c) 2014-2018 Dr. Colin Hirsch and Daniel Frey
// Please see LICENSE for license or visit https://github.com/taocpp/PEGTL/

#ifndef TAO_PEGTL_CONTRIB_UNESCAPE_HPP
#define TAO_PEGTL_CONTRIB_UNESCAPE_HPP

#include <cassert>
#include <string>

#include "../ascii.hpp"
#include "../config.hpp"
#include "../parse_error.hpp"

namespace tao
{
   namespace TAO_PEGTL_NAMESPACE
   {
      namespace unescape
      {
         struct state
         {
            std::string unescaped;
         };

         // Utility functions for the unescape actions.

         inline bool utf8_append_utf32( std::string& string, const unsigned utf32 )
         {
            if( utf32 <= 0x7f ) {
               string += char( utf32 & 0xff );
               return true;
            }
            if( utf32 <= 0x7ff ) {
               char tmp[] = { char( ( ( utf32 & 0x7c0 ) >> 6 ) | 0xc0 ),
                              char( ( ( utf32 & 0x03f ) ) | 0x80 ) };
               string.append( tmp, sizeof( tmp ) );
               return true;
            }
            if( utf32 <= 0xffff ) {
               if( utf32 >= 0xd800 && utf32 <= 0xdfff ) {
                  // nope, this is a UTF-16 surrogate
                  return false;
               }
               char tmp[] = { char( ( ( utf32 & 0xf000 ) >> 12 ) | 0xe0 ),
                              char( ( ( utf32 & 0x0fc0 ) >> 6 ) | 0x80 ),
                              char( ( ( utf32 & 0x003f ) ) | 0x80 ) };
               string.append( tmp, sizeof( tmp ) );
               return true;
            }
            if( utf32 <= 0x10ffff ) {
               char tmp[] = { char( ( ( utf32 & 0x1c0000 ) >> 18 ) | 0xf0 ),
                              char( ( ( utf32 & 0x03f000 ) >> 12 ) | 0x80 ),
                              char( ( ( utf32 & 0x000fc0 ) >> 6 ) | 0x80 ),
                              char( ( ( utf32 & 0x00003f ) ) | 0x80 ) };
               string.append( tmp, sizeof( tmp ) );
               return true;
            }
            return false;
         }

         // This function MUST only be called for characters matching tao::TAO_PEGTL_NAMESPACE::ascii::xdigit!
         template< typename I >
         I unhex_char( const char c )
         {
            switch( c ) {
               case '0':
               case '1':
               case '2':
               case '3':
               case '4':
               case '5':
               case '6':
               case '7':
               case '8':
               case '9':
                  return I( c - '0' );
               case 'a':
               case 'b':
               case 'c':
               case 'd':
               case 'e':
               case 'f':
                  return I( c - 'a' + 10 );
               case 'A':
               case 'B':
               case 'C':
               case 'D':
               case 'E':
               case 'F':
                  return I( c - 'A' + 10 );
            }
            throw std::runtime_error( "invalid character in unhex" );  // NOLINT, LCOV_EXCL_LINE
         }

         template< typename I >
         I unhex_string( const char* begin, const char* end )
         {
            I r = 0;
            while( begin != end ) {
               r <<= 4;
               r += unhex_char< I >( *begin++ );
            }
            return r;
         }

         // Actions for common unescape situations.

         struct append_all
         {
            template< typename Input, typename State >
            static void apply( const Input& in, State& st )
            {
               st.unescaped.append( in.begin(), in.size() );
            }
         };

         // This action MUST be called for a character matching T which MUST be tao::TAO_PEGTL_NAMESPACE::one< ... >.
         template< typename T, char... Rs >
         struct unescape_c
         {
            template< typename Input, typename State >
            static void apply( const Input& in, State& st )
            {
               assert( in.size() == 1 );
               st.unescaped += apply_one( *in.begin(), static_cast< const T* >( nullptr ) );
            }

            template< char... Qs >
            static char apply_one( const char c, const one< Qs... >* /*unused*/ )
            {
               static_assert( sizeof...( Qs ) == sizeof...( Rs ), "size mismatch between escaped characters and their mappings" );
               return apply_two( c, { Qs... }, { Rs... } );
            }

            static char apply_two( const char c, const std::initializer_list< char >& q, const std::initializer_list< char >& r )
            {
               for( std::size_t i = 0; i < q.size(); ++i ) {
                  if( *( q.begin() + i ) == c ) {
                     return *( r.begin() + i );
                  }
               }
               throw std::runtime_error( "invalid character in unescape" );  // NOLINT, LCOV_EXCL_LINE
            }
         };

         // See src/example/pegtl/unescape.cpp for why the following two actions
         // skip the first input character. They also MUST be called
         // with non-empty matched inputs!

         struct unescape_u
         {
            template< typename Input, typename State >
            static void apply( const Input& in, State& st )
            {
               assert( !in.empty() );  // First character MUST be present, usually 'u' or 'U'.
               if( !utf8_append_utf32( st.unescaped, unhex_string< unsigned >( in.begin() + 1, in.end() ) ) ) {
                  throw parse_error( "invalid escaped unicode code point", in );
               }
            }
         };

         struct unescape_x
         {
            template< typename Input, typename State >
            static void apply( const Input& in, State& st )
            {
               assert( !in.empty() );  // First character MUST be present, usually 'x'.
               st.unescaped += unhex_string< char >( in.begin() + 1, in.end() );
            }
         };

         // The unescape_j action is similar to unescape_u, however unlike
         // unescape_u it
         // (a) assumes exactly 4 hexdigits per escape sequence,
         // (b) accepts multiple consecutive escaped 16-bit values.
         // When applied to more than one escape sequence, unescape_j
         // translates UTF-16 surrogate pairs in the input into a single
         // UTF-8 sequence in st.unescaped, as required for JSON by RFC 8259.

         struct unescape_j
         {
            template< typename Input, typename State >
            static void apply( const Input& in, State& st )
            {
               assert( ( ( in.size() + 1 ) % 6 ) == 0 );  // Expects multiple "\\u1234", starting with the first "u".
               for( const char* b = in.begin() + 1; b < in.end(); b += 6 ) {
                  const auto c = unhex_string< unsigned >( b, b + 4 );
                  if( ( 0xd800 <= c ) && ( c <= 0xdbff ) && ( b + 6 < in.end() ) ) {
                     const auto d = unhex_string< unsigned >( b + 6, b + 10 );
                     if( ( 0xdc00 <= d ) && ( d <= 0xdfff ) ) {
                        b += 6;
                        // note: no need to check the result code, as we are always >= 0x10000 and < 0x110000.
                        utf8_append_utf32( st.unescaped, ( ( ( c & 0x03ff ) << 10 ) | ( d & 0x03ff ) ) + 0x10000 );
                        continue;
                     }
                  }
                  if( !utf8_append_utf32( st.unescaped, c ) ) {
                     throw parse_error( "invalid escaped unicode code point", in );
                  }
               }
            }
         };

      }  // namespace unescape

   }  // namespace TAO_PEGTL_NAMESPACE

}  // namespace tao

#endif