daw_json_parse_std_string.h

Go to the documentation of this file.

// Copyright (c) Darrell Wright
//
// Distributed under the Boost Software License, Version 1.0. (See accompanying
// file LICENSE or copy at http://www.boost.org/LICENSE_1_0.txt)
//
// Official repository: https://github.com/beached/daw_json_link
//
 
#pragma once
 
#include "daw/json/impl/version.h"
 
#include "daw/json/daw_json_switches.h"
#include "daw/json/impl/daw_json_assert.h"
#include "daw/json/impl/daw_json_parse_common.h"
#include "daw/json/impl/daw_not_const_ex_functions.h"
 
#include <daw/algorithms/daw_algorithm_copy.h>
#include <daw/algorithms/daw_algorithm_copy_n.h>
#include <daw/daw_data_end.h>
#include <daw/daw_likely.h>
#include <daw/daw_not_null.h>
#include <daw/daw_span.h>
 
#include <cstddef>
#include <daw/stdinc/data_access.h>
#include <daw/stdinc/range_access.h>
#include <type_traits>
 
namespace daw::json {
  inline namespace DAW_JSON_VER {
    namespace json_details {
      [[nodiscard]] static constexpr UInt8 to_nibble( unsigned char chr ) {
        int const b = static_cast<int>( chr );
        int const maskLetter = ( ( '9' - b ) >> 31 );
        int const maskSmall = ( ( 'Z' - b ) >> 31 );
        int const offset = '0' + ( maskLetter & int( 'A' - '0' - 10 ) ) +
                           ( maskSmall & int( 'a' - 'A' ) );
        auto const result = static_cast<unsigned>( b - offset );
        return to_uint8( result );
      }
 
      template<bool is_unchecked_input>
      [[nodiscard]] static constexpr UInt16
      byte_from_nibbles( daw::not_null<char const *> &first ) {
        auto const n0 = to_nibble( static_cast<unsigned char>( *first++ ) );
        auto const n1 = to_nibble( static_cast<unsigned char>( *first++ ) );
        if constexpr( not is_unchecked_input ) {
          daw_json_ensure( n0 < 16 and n1 < 16, ErrorReason::InvalidUTFEscape );
        }
        return to_uint16( ( n0 << 4U ) | n1 );
      }
 
      static constexpr char u32toC( UInt32 value ) {
        return static_cast<char>( static_cast<unsigned char>( value ) );
      }
 
      template<typename ParseState>
      [[nodiscard]] static constexpr daw::not_null<char *>
      decode_utf16( ParseState &parse_state, daw::not_null<char *> it ) {
        daw_json_assert_weak( parse_state.size( ) >= 5,
                              ErrorReason::UnexpectedEndOfData,
                              parse_state );
        auto first = daw::not_null<char const *>( parse_state.first );
        ++first;
        UInt32 cp =
          to_uint32(
            byte_from_nibbles<ParseState::is_unchecked_input>( first ) )
          << 8U;
        cp |= byte_from_nibbles<ParseState::is_unchecked_input>( first );
        if( cp <= 0x7FU ) {
          *it++ = static_cast<char>( static_cast<unsigned char>( cp ) );
          parse_state.first = input_pointer( parse_state.first, first.get( ) );
          return it;
        }
 
        //******************************
        daw_json_assert_weak( cp < 0xDC00U or cp > 0xDFFFU,
                              ErrorReason::InvalidUTFEscape,
                              parse_state ); // Lone/unpaired low surrogate
        if( 0xD800U <= cp and cp <= 0xDBFFU ) {
          cp = ( cp - 0xD800U ) * 0x400U;
          daw_json_assert_weak(
            *first == '\\',
            ErrorReason::InvalidUTFEscape,
            parse_state ); // High surrogate must be followed by \\u
          ++first;
          daw_json_assert_weak(
            ( parse_state.last - first >= 5 ) and *first == 'u',
            ErrorReason::InvalidUTFEscape,
            parse_state ); // Expected parse_state to start with a \\u
          ++first;
          auto trailing =
            to_uint32(
              byte_from_nibbles<ParseState::is_unchecked_input>( first ) )
            << 8U;
          trailing |=
            byte_from_nibbles<ParseState::is_unchecked_input>( first );
          daw_json_assert_weak(
            0xDC00U <= trailing and trailing <= 0xDFFFU,
            ErrorReason::InvalidUTFEscape,
            parse_state ); // Expected a low surrogate to complete the pair
          trailing -= 0xDC00U;
          cp += trailing;
          cp += 0x10000;
        }
        // UTF32-> UTF8
        if( cp >= 0x10000U ) {
          // 4 bytes
          char const enc3 = u32toC( ( cp & 0b0011'1111U ) | 0b1000'0000U );
          char const enc2 =
            u32toC( ( ( cp >> 6U ) & 0b0011'1111U ) | 0b1000'0000U );
          char const enc1 =
            u32toC( ( ( cp >> 12U ) & 0b0011'1111U ) | 0b1000'0000U );
          char const enc0 = u32toC( ( cp >> 18U ) | 0b1111'0000U );
          *it++ = enc0;
          *it++ = enc1;
          *it++ = enc2;
          *it++ = enc3;
          parse_state.first = input_pointer( parse_state.first, first.get( ) );
          return it;
        }
        //******************************
        if( cp >= 0x800U ) {
          // 3 bytes
          char const enc2 = u32toC( ( cp & 0b0011'1111U ) | 0b1000'0000U );
          char const enc1 =
            u32toC( ( ( cp >> 6U ) & 0b0011'1111U ) | 0b1000'0000U );
          char const enc0 = u32toC( ( cp >> 12U ) | 0b1110'0000U );
          *it++ = enc0;
          *it++ = enc1;
          *it++ = enc2;
          parse_state.first = input_pointer( parse_state.first, first.get( ) );
          return it;
        }
        //******************************
        // cp >= 0x80U
        // 2 bytes
        char const enc1 = u32toC( ( cp & 0b0011'1111U ) | 0b1000'0000U );
        char const enc0 = u32toC( ( cp >> 6U ) | 0b1100'0000U );
        *it++ = enc0;
        *it++ = enc1;
        parse_state.first = input_pointer( parse_state.first, first.get( ) );
        return it;
      }
 
      template<typename ParseState, typename Appender>
      static constexpr void decode_utf16( ParseState &parse_state,
                                          Appender &app ) {
        auto first = daw::not_null<char const *>( parse_state.first );
        ++first;
        UInt32 cp =
          to_uint32(
            byte_from_nibbles<ParseState::is_unchecked_input>( first ) )
          << 8U;
        cp |= byte_from_nibbles<ParseState::is_unchecked_input>( first );
        if( cp <= 0x7FU ) {
          app( u32toC( cp ) );
          parse_state.first = input_pointer( parse_state.first, first.get( ) );
          return;
        }
        daw_json_assert_weak( cp < 0xDC00U or cp > 0xDFFFU,
                              ErrorReason::InvalidUTFEscape,
                              parse_state ); // Lone/unpaired low surrogate
        if( 0xD800U <= cp and cp <= 0xDBFFU ) {
          cp = ( cp - 0xD800U ) * 0x400U;
          daw_json_assert_weak(
            *first == '\\',
            ErrorReason::InvalidUTFEscape,
            parse_state ); // High surrogate must be followed by \\u
          ++first;
          daw_json_assert_weak(
            *first == 'u', ErrorReason::InvalidUTFEscape, parse_state );
          ++first;
          auto trailing =
            to_uint32(
              byte_from_nibbles<ParseState::is_unchecked_input>( first ) )
            << 8U;
          trailing |=
            byte_from_nibbles<ParseState::is_unchecked_input>( first );
          daw_json_assert_weak(
            0xDC00U <= trailing and trailing <= 0xDFFFU,
            ErrorReason::InvalidUTFEscape,
            parse_state ); // Expected a low surrogate to complete the pair
          trailing -= 0xDC00U;
          cp += trailing;
          cp += 0x10000;
        }
        // UTF32-> UTF8
        if( cp >= 0x10000U ) {
          // 4 bytes
          char const enc3 = u32toC( ( cp & 0b0011'1111U ) | 0b1000'0000U );
          char const enc2 =
            u32toC( ( ( cp >> 6U ) & 0b0011'1111U ) | 0b1000'0000U );
          char const enc1 =
            u32toC( ( ( cp >> 12U ) & 0b0011'1111U ) | 0b1000'0000U );
          char const enc0 = u32toC( ( cp >> 18U ) | 0b1111'0000U );
          app( enc0 );
          app( enc1 );
          app( enc2 );
          app( enc3 );
          parse_state.first = input_pointer( parse_state.first, first.get( ) );
          return;
        }
        if( cp >= 0x800U ) {
          // 3 bytes
          char const enc2 = u32toC( ( cp & 0b0011'1111U ) | 0b1000'0000U );
          char const enc1 =
            u32toC( ( ( cp >> 6U ) & 0b0011'1111U ) | 0b1000'0000U );
          char const enc0 = u32toC( ( cp >> 12U ) | 0b1110'0000U );
          app( enc0 );
          app( enc1 );
          app( enc2 );
          parse_state.first = input_pointer( parse_state.first, first.get( ) );
          return;
        }
        // cp >= 0x80U
        // 2 bytes
        char const enc1 = u32toC( ( cp & 0b0011'1111U ) | 0b1000'0000U );
        char const enc0 = u32toC( ( cp >> 6U ) | 0b1100'0000U );
        app( enc0 );
        app( enc1 );
        parse_state.first = input_pointer( parse_state.first, first.get( ) );
      }
 
      namespace parse_tokens {
        inline constexpr char escape_quotes[] = "\\\"";
      }
 
      // Fast path for parsing escaped strings to a std::string with the default
      // appender
      template<bool AllowHighEight, typename JsonMember, bool KnownBounds,
               typename ParseState>
      [[nodiscard]] constexpr json_result_t<JsonMember>
      parse_string_known_stdstring( ParseState &parse_state ) {
        DAW_CPP23_STATIC_LOCAL constexpr bool is_insitu =
          JsonMember::expected_type == JsonParseTypes::StringInsitu;
 
        static_assert(
          not is_insitu or ParseState::allow_string_mutation,
          "In-situ strings can only be parsed from writable input documents" );
        static_assert( not is_insitu or
                         std::is_same_v<typename ParseState::iterator, char *>,
                       "In-situ strings require a mutable character buffer" );
 
        bool const has_quote = parse_state.front( ) == '"';
        if( has_quote ) {
          parse_state.remove_prefix( );
        }
 
        using string_type = json_base_type_t<JsonMember>;
        using result_t =
          std::conditional_t<is_insitu, daw::span<char>, string_type>;
        result_t result = [&] {
          if constexpr( is_insitu ) {
            return daw::span( std::data( parse_state ),
                              std::size( parse_state ) );
          } else {
            return string_type(
              std::size( parse_state ) + 1U +
                static_cast<unsigned>( has_quote ),
              '\0',
              parse_state.template get_allocator_for<char>( ) );
          }
        }( );
        daw::not_null<char *> it = std::data( result );
 
        if( auto const first_slash =
              static_cast<std::ptrdiff_t>( parse_state.counter ) - 1;
            first_slash > 1 ) {
          if constexpr( is_insitu ) {
            it += first_slash;
          } else {
            it =
              daw::algorithm::copy_n( parse_state.first,
                                      it.get( ),
                                      static_cast<std::size_t>( first_slash ) )
                .output;
          }
          parse_state.first += first_slash;
        }
 
        DAW_CPP23_STATIC_LOCAL constexpr auto in_json_string =
          []( auto const &r ) DAW_JSON_CPP23_STATIC_CALL_OP -> bool {
          if constexpr( KnownBounds or not ParseState::is_unchecked_input ) {
            if( not DAW_LIKELY( r.has_more( ) ) ) {
              return false;
            }
          }
          return DAW_LIKELY( r.front( ) != '"' );
        };
 
        while( in_json_string( parse_state ) ) {
          {
            daw::not_null<char const *> first = parse_state.first;
            daw::not_null<char const *> const last = parse_state.last;
 
            if( not json_details::use_constexpr_exec_mode<
                  typename ParseState::exec_tag_t>( ) ) {
              first = mem_move_to_next_of<
                ( not KnownBounds and
                  ( ParseState::is_unchecked_input or
                    ParseState::is_zero_terminated_string ) ),
                typename ParseState::exec_tag_t,
                '"',
                '\\'>( first, last );
            } else {
              while( first < last and *first != '"' and *first != '\\' ) {
                ++first;
              }
              daw_json_assert_weak( KnownBounds or first < last,
                                    ErrorReason::UnexpectedEndOfData,
                                    parse_state );
            }
 
            auto const run_size = first.get( ) - parse_state.first;
            daw_json_assert_weak(
              static_cast<std::ptrdiff_t>( result.size( ) ) -
                  ( it.get( ) - result.data( ) ) >=
                run_size,
              ErrorReason::UnexpectedEndOfData );
 
            if( it.get( ) == parse_state.first ) {
              it += run_size;
            } else {
              it = daw::algorithm::copy(
                parse_state.first, first.get( ), it.get( ) );
            }
            parse_state.first =
              input_pointer( parse_state.first, first.get( ) );
          }
          if( parse_state.front( ) == '\\' ) {
            parse_state.remove_prefix( );
            daw_json_assert_weak( not parse_state.is_space_unchecked( ),
                                  ErrorReason::InvalidUTFCodepoint,
                                  parse_state );
            switch( parse_state.front( ) ) {
            case 'b':
              *it++ = '\b';
              parse_state.remove_prefix( );
              break;
            case 'f':
              *it++ = '\f';
              parse_state.remove_prefix( );
              break;
            case 'n':
              *it++ = '\n';
              parse_state.remove_prefix( );
              break;
            case 'r':
              *it++ = '\r';
              parse_state.remove_prefix( );
              break;
            case 't':
              *it++ = '\t';
              parse_state.remove_prefix( );
              break;
            case 'u':
              it = decode_utf16( parse_state, it );
              break;
            case '/':
            case '\\':
            case '"':
              *it++ = parse_state.front( );
              parse_state.remove_prefix( );
              break;
            default:
              // Every legal escape character (", \, /, b, f, n, r, t, u) is
              // handled by an explicit case above; anything else following a
              // backslash is not a valid JSON escape.
              daw_json_assert_weak( parse_state.front( ) != '"',
                                    ErrorReason::InvalidString,
                                    parse_state );
              *it++ = parse_state.front( );
              parse_state.remove_prefix( );
            }
          } else {
            daw_json_assert_weak( not has_quote or
                                    parse_state.is_quotes_checked( ),
                                  ErrorReason::InvalidString,
                                  parse_state );
          }
          daw_json_assert_weak( not has_quote or parse_state.has_more( ),
                                ErrorReason::UnexpectedEndOfData,
                                parse_state );
        }
        auto const sz = static_cast<std::size_t>(
          std::distance( std::data( result ), it.get( ) ) );
        daw_json_assert_weak(
          std::size( result ) >= sz, ErrorReason::InvalidString, parse_state );
 
        if constexpr( is_insitu ) {
          *it = '"';
          ++it;
          while( it != ( parse_state.last + 1 ) ) {
            *it = ' ';
            ++it;
          }
          result = result.subspan( 0, sz );
        } else {
          result.resize( sz );
        }
 
        if constexpr( not is_insitu and
                      std::is_convertible_v<result_t,
                                            json_result_t<JsonMember>> ) {
          return result;
        } else {
          using constructor_t = json_constructor_t<JsonMember>;
          return construct_value<json_result_t<JsonMember>, constructor_t>(
            parse_state, std::data( result ), std::size( result ) );
        }
      }
    } // namespace json_details
  } // namespace DAW_JSON_VER
} // namespace daw::json