The R Project SVN R-packages

#include "JSONWorker.h"

bool used_ascii_one = false;  //used to know whether or not to check for intermediates when writing, once flipped, can't be unflipped
inline json_char ascii_one(void) json_nothrow {
    used_ascii_one = true;
    return JSON_TEXT('\1');
}

#ifdef JSON_READ_PRIORITY

JSONNode JSONWorker::parse(const json_string & json) json_throws(std::invalid_argument) {
    json_auto<json_char> s;
    size_t len;
    s.set(RemoveWhiteSpace(json, len, true));
    return _parse_unformatted(s.ptr, s.ptr + len);
}

JSONNode JSONWorker::parse_unformatted(const json_string & json) json_throws(std::invalid_argument) {
    #if defined JSON_DEBUG || defined JSON_SAFE
       #ifndef JSON_NO_EXCEPTIONS
          JSON_ASSERT_SAFE((json[0] == JSON_TEXT('{')) || (json[0] == JSON_TEXT('[')), JSON_TEXT("Not JSON!"), throw std::invalid_argument(json_global(EMPTY_STD_STRING)););
       #else
          JSON_ASSERT_SAFE((json[0] == JSON_TEXT('{')) || (json[0] == JSON_TEXT('[')), JSON_TEXT("Not JSON!"), return JSONNode(JSON_NULL););
       #endif
    #endif
    return _parse_unformatted(json.data(), json.data() + json.length());
}

JSONNode JSONWorker::_parse_unformatted(const json_char * json, const json_char * const end) json_throws(std::invalid_argument) {
    #ifdef JSON_COMMENTS
       json_char firstchar = *json;
       json_string _comment;
       json_char * runner = (json_char*)json;
       if (json_unlikely(firstchar == JSON_TEMP_COMMENT_IDENTIFIER)){  //multiple comments will be consolidated into one
          newcomment:
          while(*(++runner) != JSON_TEMP_COMMENT_IDENTIFIER){
             JSON_ASSERT(runner != end, JSON_TEXT("Removing white space failed"));
             _comment += *runner;
          }
          firstchar = *(++runner); //step past the trailing tag
          if (json_unlikely(firstchar == JSON_TEMP_COMMENT_IDENTIFIER)){
             _comment += JSON_TEXT('\n');
             goto newcomment;
          }
       }
    #else
       const json_char firstchar = *json;
    #endif

    switch (firstchar){
        case JSON_TEXT('{'):
        case JSON_TEXT('['):
          #if defined JSON_DEBUG || defined JSON_SAFE
             if (firstchar == JSON_TEXT('[')){
                if (json_unlikely(*(end - 1) != JSON_TEXT(']'))){
                    JSON_FAIL(JSON_TEXT("Missing final ]"));
                    break;
                }
             } else {
                if (json_unlikely(*(end - 1) != JSON_TEXT('}'))){
                    JSON_FAIL(JSON_TEXT("Missing final }"));
                    break;
                }
             }
          #endif
          #ifdef JSON_COMMENTS
             JSONNode foo(json_string(runner, end - runner));
             foo.set_comment(_comment);
             return JSONNode(true, foo);  //forces it to simply return the original interal, even with ref counting off
          #else
             return JSONNode(json_string(json, end - json));
          #endif
    }

    JSON_FAIL(JSON_TEXT("Not JSON!"));
    #ifndef JSON_NO_EXCEPTIONS
       throw std::invalid_argument(json_global(EMPTY_STD_STRING));
    #else
       return JSONNode(JSON_NULL);
    #endif
}
#endif

#define QUOTECASE()\
    case JSON_TEXT('\"'):\
       while (*(++p) != JSON_TEXT('\"')){\
          JSON_ASSERT_SAFE(*p, JSON_TEXT("Null terminator inside of a quotation"), return json_string::npos;);\
       }\
       break;

#if defined(JSON_DEBUG) || defined(JSON_SAFE)
    #define NULLCASE(error)\
       case JSON_TEXT('\0'):\
          JSON_FAIL_SAFE(error, return json_string::npos;);\
          break;
#else
    #define NULLCASE(error)
#endif

#define BRACKET(left, right)\
    case left: {\
       size_t brac = 1;\
       while (brac){\
          switch (*(++p)){\
             case right:\
                --brac;\
                break;\
             case left:\
                ++brac;\
                break;\
             QUOTECASE()\
             NULLCASE(JSON_TEXT("Null terminator inside of a bracket"))\
          }\
       }\
       break;}\
    case right:\
       return json_string::npos;



#if defined(JSON_READ_PRIORITY) || defined(JSON_STREAM)
    #if (JSON_READ_PRIORITY == HIGH) && (!(defined(JSON_LESS_MEMORY)))
        #define FIND_NEXT_RELEVANT(ch, vt, po) JSONWorker::FindNextRelevant<ch>(vt, po)
        template<json_char ch>
        size_t JSONWorker::FindNextRelevant(const json_string & value_t, const size_t pos) json_nothrow {
    #else
        #define FIND_NEXT_RELEVANT(ch, vt, po) JSONWorker::FindNextRelevant(ch, vt, po)
        size_t JSONWorker::FindNextRelevant(json_char ch, const json_string & value_t, const size_t pos) json_nothrow {
    #endif
        json_string::const_iterator start = value_t.begin();
        json_string::const_iterator e = value_t.end();
       for (json_string::const_iterator p = value_t.begin() + pos; p != e; ++p){
          if (json_unlikely(*p == ch)) return p - start;
          switch (*p){
                BRACKET(JSON_TEXT('['), JSON_TEXT(']'))
                BRACKET(JSON_TEXT('{'), JSON_TEXT('}'))
                QUOTECASE()
          }
       };
       return json_string::npos;
    }
#endif

#ifdef JSON_COMMENTS
    #define COMMENT_DELIMITER() *runner++ = JSON_TEMP_COMMENT_IDENTIFIER
    #define AND_RUNNER ,runner
    inline void SingleLineComment(const json_char * & p, const json_char * const end, json_char * & runner) json_nothrow {
       //It is okay to add two '\5' characters here because at minimun the # and '\n' are replaced, so it's at most the same size
       COMMENT_DELIMITER();
       while((++p != end) && (*p != JSON_TEXT('\n'))){
          *runner++ = *p;
       }
       COMMENT_DELIMITER();
    }
#else
    #define COMMENT_DELIMITER() (void)0
    #define AND_RUNNER
#endif

#ifndef JSON_STRICT
inline void SingleLineComment(const json_char * & p, const json_char * const end) json_nothrow {
    while((++p != end) && (*p != JSON_TEXT('\n')));
}
#endif

#if defined(JSON_LESS_MEMORY) && defined(JSON_READ_PRIORITY)
    #define PRIVATE_REMOVEWHITESPACE(T, value_t, escapeQuotes, len) private_RemoveWhiteSpace(T, value_t, escapeQuotes, len)
    json_char * private_RemoveWhiteSpace(bool T, const json_string & value_t, bool escapeQuotes, size_t & len) json_nothrow {
#else
    #define PRIVATE_REMOVEWHITESPACE(T, value_t, escapeQuotes, len) private_RemoveWhiteSpace<T>(value_t, escapeQuotes, len)
    template<bool T>
    json_char * private_RemoveWhiteSpace(const json_string & value_t, bool escapeQuotes, size_t & len) json_nothrow {
#endif
    json_char * result;
    json_char * runner = result = json_malloc<json_char>(value_t.length() + 1);  //dealing with raw memory is faster than adding to a json_string
    JSON_ASSERT(result != 0, json_global(ERROR_OUT_OF_MEMORY));
    const json_char * const end = value_t.data() + value_t.length();
    for(const json_char * p = value_t.data(); p != end; ++p){
      switch(*p){
         case JSON_TEXT(' '):   //defined as white space
         case JSON_TEXT('\t'):  //defined as white space
         case JSON_TEXT('\n'):  //defined as white space
         case JSON_TEXT('\r'):  //defined as white space
            break;
        #ifndef JSON_STRICT
            case JSON_TEXT('/'):  //a C comment
                if (*(++p) == JSON_TEXT('*')){  //a multiline comment
                   if (T) COMMENT_DELIMITER();
                   while ((*(++p) != JSON_TEXT('*')) || (*(p + 1) != JSON_TEXT('/'))){
                      if(p == end){
                            COMMENT_DELIMITER(); 
                            goto endofrunner;
                        }
                      if (T) *runner++ = *p;
                   }
                   ++p;
                   if (T) COMMENT_DELIMITER();
                   break;
                }
                //Should be a single line C comment, so let it fall through to use the bash comment stripper
                JSON_ASSERT_SAFE(*p == JSON_TEXT('/'), JSON_TEXT("stray / character, not quoted, or a comment"), goto endofrunner;);
            case JSON_TEXT('#'):  //a bash comment
                if (T){
                    SingleLineComment(p, end AND_RUNNER);
                } else {
                    SingleLineComment(p, end);
                }
                break;
         #endif
         case JSON_TEXT('\"'):  //a quote
            *runner++ = JSON_TEXT('\"');
            while(*(++p) != JSON_TEXT('\"')){  //find the end of the quotation, as white space is preserved within it
                if(p == end) goto endofrunner;
                switch(*p){
                   case JSON_TEXT('\\'):
                      *runner++ = JSON_TEXT('\\');
                      if (escapeQuotes){
                            *runner++ = (*++p == JSON_TEXT('\"')) ? ascii_one() : *p;  //an escaped quote will reak havoc will all of my searching functions, so change it into an illegal character in JSON for convertion later on
                      } else {
                            *runner++ = *++p;
                      }
                      break;
                   default:
                      *runner++ = *p;
                      break;
                }
            }
            //no break, let it fall through so that the trailing quote gets added
         default:
            JSON_ASSERT_SAFE((json_uchar)*p >= 32, JSON_TEXT("Invalid JSON character detected (lo)"), goto endofrunner;);
            JSON_ASSERT_SAFE((json_uchar)*p <= 126, JSON_TEXT("Invalid JSON character detected (hi)"), goto endofrunner;);
            *runner++ = *p;
            break;
      }
    }
    endofrunner:
    len = runner - result;
    return result;
}

#ifdef JSON_READ_PRIORITY
    json_char * JSONWorker::RemoveWhiteSpace(const json_string & value_t, size_t & len, bool escapeQuotes) json_nothrow  {
        json_char * result = PRIVATE_REMOVEWHITESPACE(true, value_t, escapeQuotes, len); 
        result[len] = JSON_TEXT('\0');
        return result;
    }
#endif

json_char * JSONWorker::RemoveWhiteSpaceAndCommentsC(const json_string & value_t, bool escapeQuotes) json_nothrow {
    size_t len;
    json_char * result = PRIVATE_REMOVEWHITESPACE(false, value_t, escapeQuotes, len);
    result[len] = JSON_TEXT('\0');
    return result;
}

json_string JSONWorker::RemoveWhiteSpaceAndComments(const json_string & value_t, bool escapeQuotes) json_nothrow {
    json_auto<json_char> s;
    size_t len;
    s.set(PRIVATE_REMOVEWHITESPACE(false, value_t, escapeQuotes, len)); 
    return json_string(s.ptr, len);
}

#ifdef JSON_READ_PRIORITY
/*
 These three functions analyze json_string literals and convert them into std::strings
 This includes dealing with special characters and utf characters
 */
#ifdef JSON_UNICODE
    inline json_uchar SurrogatePair(const json_uchar hi, const json_uchar lo) json_pure;
    inline json_uchar SurrogatePair(const json_uchar hi, const json_uchar lo) json_nothrow {
       JSON_ASSERT(sizeof(unsigned int) == 4, JSON_TEXT("size of unsigned int is not 32-bit"));
       JSON_ASSERT(sizeof(json_uchar) == 4, JSON_TEXT("size of json_char is not 32-bit"));
       return (((hi << 10) & 0x1FFC00) + 0x10000) | lo & 0x3FF;
    }

    void JSONWorker::UTF(const json_char * & pos, json_string & result, const json_char * const end) json_nothrow {
    JSON_ASSERT_SAFE((end > pos + 4), JSON_TEXT("UTF will go out of bounds"), return;);
       json_uchar first = UTF8(pos, end);
       if (json_unlikely((first > 0xD800) && (first < 0xDBFF) &&
          (*(pos + 1) == '\\') && (*(pos + 2) == 'u'))){
             const json_char * original_pos = pos;  //if the 2nd character is not correct I need to roll back the iterator
             pos += 2;
             json_uchar second = UTF8(pos, end);
             //surrogate pair, not two characters
             if (json_unlikely((second > 0xDC00) && (second < 0xDFFF))){
                result += SurrogatePair(first, second);
             } else {
                pos = original_pos;
             }
       } else {
          result += first;
       }
    }
#endif

json_uchar JSONWorker::UTF8(const json_char * & pos, const json_char * const end) json_nothrow {
    JSON_ASSERT_SAFE((end > pos + 4), JSON_TEXT("UTF will go out of bounds"), return JSON_TEXT('\0'););
    #ifdef JSON_UNICODE
       ++pos;
       json_uchar temp = Hex(pos) << 8;
       ++pos;
       return temp | Hex(pos);
    #else
       JSON_ASSERT(*(pos + 1) == JSON_TEXT('0'), JSON_TEXT("wide utf character (hihi)"));
       JSON_ASSERT(*(pos + 2) == JSON_TEXT('0'), JSON_TEXT("wide utf character (hilo)"));
       pos += 3;
       return Hex(pos);
    #endif
}


json_char JSONWorker::Hex(const json_char * & pos) json_nothrow {
    /*
    takes the numeric value of the next two characters and convert them
    \u0058 becomes 0x58

    In case of \u, it's SpecialChar's responsibility to move past the first two chars
    as this method is also used for \x
    */
    //First character
    json_uchar hi = *pos++ - 48;
    if (hi > 48){  //A-F don't immediately follow 0-9, so have to pull them down a little
       hi -= 39;
    } else if (hi > 9){  //neither do a-f
       hi -= 7;
    }
    //second character
    json_uchar lo = *pos - 48;
    if (lo > 48){  //A-F don't immediately follow 0-9, so have to pull them down a little
       lo -= 39;
    } else if (lo > 9){  //neither do a-f
       lo -= 7;
    }
    //combine them
    return (json_char)((hi << 4) | lo);
}

#ifndef JSON_STRICT
    inline json_char FromOctal(const json_char * & str, const json_char * const end) json_nothrow {
    JSON_ASSERT_SAFE((end > str + 3), JSON_TEXT("Octal will go out of bounds"), return JSON_TEXT('\0'););
       str += 2;
       return (json_char)(((((json_uchar)(*(str - 2) - 48))) << 6) | (((json_uchar)(*(str - 1) - 48)) << 3) | ((json_uchar)(*str - 48)));
    }
#endif

void JSONWorker::SpecialChar(const json_char * & pos, const json_char * const end, json_string & res) json_nothrow {
    JSON_ASSERT_SAFE(pos != end, JSON_TEXT("Special char termantion"), return;);
    /*
       Since JSON uses forward slash escaping for special characters within strings, I have to
       convert these escaped characters into C characters
    */
    switch(*pos){
       case JSON_TEXT('\1'):  //quote character (altered by RemoveWhiteSpace)
          res += JSON_TEXT('\"');
          break;
       case JSON_TEXT('t'): //tab character
          res += JSON_TEXT('\t');
          break;
       case JSON_TEXT('n'): //newline character
          res += JSON_TEXT('\n');
          break;
       case JSON_TEXT('r'): //return character
          res += JSON_TEXT('\r');
          break;
       case JSON_TEXT('\\'):    //backslash
          res += JSON_TEXT('\\');
          break;
       case JSON_TEXT('/'): //forward slash
          res += JSON_TEXT('/');
          break;
       case JSON_TEXT('b'): //backspace
          res += JSON_TEXT('\b');
          break;
       case JSON_TEXT('f'): //formfeed
          res += JSON_TEXT('\f');
          break;
       case JSON_TEXT('v'): //vertical tab
          res += JSON_TEXT('\v');
          break;
       case JSON_TEXT('u'): //utf character
          #ifdef JSON_UNICODE
             UTF(pos, res, end);
          #else
             res += UTF8(pos, end);
          #endif
          break;
       #ifndef JSON_STRICT
          case JSON_TEXT('x'):   //hexidecimal ascii code
              JSON_ASSERT_SAFE((end > pos + 3), JSON_TEXT("Hex will go out of bounds"), res += JSON_TEXT('\0'); return;);
             res += Hex(++pos);
             break;

          //#ifdef __GNUC__
        //   case JSON_TEXT('0') ... JSON_TEXT('7'):
          //#else
             //octal encoding
             case JSON_TEXT('0'):
             case JSON_TEXT('1'):
             case JSON_TEXT('2'):
             case JSON_TEXT('3'):
             case JSON_TEXT('4'):
             case JSON_TEXT('5'):
             case JSON_TEXT('6'):
             case JSON_TEXT('7'):
          //#endif
             res += FromOctal(pos, end);
             break;
          default:
             res += *pos;
             break;
       #elif defined(JSON_DEBUG)
          default:
             JSON_FAIL(JSON_TEXT("Unsupported escaped character"));
             break;
       #endif
    }
}

#ifdef JSON_LESS_MEMORY
    inline void doflag(const internalJSONNode * flag, bool which, bool x) json_nothrow {
       if (json_likely(which)){
          flag -> _name_encoded = x;
       } else {
          flag -> _string_encoded = x;
       }
    }

    json_string JSONWorker::FixString(const json_string & value_t, const internalJSONNode * flag, bool which) json_nothrow {
    #define setflag(x) doflag(flag, which, x)
#else
    json_string JSONWorker::FixString(const json_string & value_t, bool & flag) json_nothrow {
    #define setflag(x) flag = x
#endif

    //Do things like unescaping
    setflag(false);
    json_string res;
    res.reserve(value_t.length());   //since it goes one character at a time, want to reserve it first so that it doens't have to reallocating
    const json_char * const end = value_t.data() + value_t.length();
    for(const json_char * p = value_t.data(); p != end; ++p){
       switch (*p){
          case JSON_TEXT('\\'):
             setflag(true);
             SpecialChar(++p, end, res);
             break;
          default:
             res += *p;
             break;
       }
    }
    shrinkString(res);  //because this is actually setting something to be stored, shrink it it need be
    return res;
}
#endif

#ifdef JSON_UNICODE
    #ifdef JSON_ESCAPE_WRITES
       json_string JSONWorker::toSurrogatePair(json_uchar C) json_nothrow {
          JSON_ASSERT(sizeof(unsigned int) == 4, JSON_TEXT("size of unsigned int is not 32-bit"));
          JSON_ASSERT(sizeof(unsigned short) == 2, JSON_TEXT("size of unsigned short is not 16-bit"));
          JSON_ASSERT(sizeof(json_uchar) == 4, JSON_TEXT("json_char is not 32-bit"));

          //Compute the high surrogate
          unsigned short HiSurrogate = 0xD800 | (((unsigned short)((unsigned int)((C >> 16) & 31)) - 1) << 6) | ((unsigned short)C) >> 10;

          //compute the low surrogate
          unsigned short LoSurrogate = (unsigned short) (0xDC00 | ((unsigned short)C & 1023));

          json_string res;
          res += toUTF8(HiSurrogate);
          res += toUTF8(LoSurrogate);
          return res;
       }
    #endif
#endif

#ifdef JSON_ESCAPE_WRITES
    json_string JSONWorker::toUTF8(json_uchar p) json_nothrow {
       #ifdef JSON_UNICODE
          if (json_unlikely(p > 0xFFFF)) return toSurrogatePair(p);
       #endif
       json_string res(JSON_TEXT("\\u"));
       #ifdef JSON_UNICODE
          START_MEM_SCOPE
             json_uchar hihi = ((p & 0xF000) >> 12) + 48;
             if (hihi > 57) hihi += 7; //A-F don't immediately follow 0-9, so have to further adjust those
             json_uchar hilo = ((p & 0x0F00) >> 8) + 48;
             if (hilo > 57) hilo += 7; //A-F don't immediately follow 0-9, so have to further adjust those
             res += hihi;
             res += hilo;
          END_MEM_SCOPE
          json_uchar hi = ((p & 0x00F0) >> 4) + 48;
       #else
          res += JSON_TEXT("00");
          json_uchar hi = (p >> 4) + 48;
       #endif
       //convert the character to be escaped into two digits between 0 and 15
       if (hi > 57) hi += 7; //A-F don't immediately follow 0-9, so have to further adjust those
       json_uchar lo = (p & 0x000F) + 48;
       if (lo > 57) lo += 7; //A-F don't immediately follow 0-9, so have to further adjust those
       res += hi;
       res += lo;
       return res;
    }
#endif

void JSONWorker::UnfixString(const json_string & value_t, bool flag, json_string & res) json_nothrow {
    if (!flag){
        res += value_t;
        return;
    }
    //Re-escapes a json_string so that it can be written out into a JSON file
    const json_char * const end = value_t.data() + value_t.length();
    for(const json_char * p = value_t.data(); p != end; ++p){
       switch(*p){
          case JSON_TEXT('\"'):  //quote character
             res += JSON_TEXT("\\\"");
             break;
          case JSON_TEXT('\\'): //backslash
             res += JSON_TEXT("\\\\");
             break;
          #ifdef JSON_ESCAPE_WRITES
             case JSON_TEXT('\t'):  //tab character
                res += JSON_TEXT("\\t");
                break;
             case JSON_TEXT('\n'):  //newline character
                res += JSON_TEXT("\\n");
                break;
             case JSON_TEXT('\r'):  //return character
                res += JSON_TEXT("\\r");
                break;
             case JSON_TEXT('/'):   //forward slash
                res += JSON_TEXT("\\/");
                break;
             case JSON_TEXT('\b'):  //backspace
                res += JSON_TEXT("\\b");
                break;
             case JSON_TEXT('\f'):  //formfeed
                res += JSON_TEXT("\\f");
                break;
             default:
             {
                if (json_unlikely(((json_uchar)(*p) < 32) || ((json_uchar)(*p) > 126))){
                    res += toUTF8((json_uchar)(*p));
                } else {
                    res += *p;
                }
            }
                break;
          #else
             default:
                res += *p;
                break;
          #endif
       }
    }
}

#ifdef JSON_READ_PRIORITY
//Create a childnode
#ifdef JSON_COMMENTS
    #define ARRAY_PARAM bool array  //Just to supress warnings
#else
    #define ARRAY_PARAM bool
#endif
inline void JSONWorker::NewNode(const internalJSONNode * parent, const json_string & name, const json_string & value, ARRAY_PARAM) json_nothrow {
    #ifdef JSON_COMMENTS
       JSONNode * child;
       START_MEM_SCOPE
          json_string _comment;
          START_MEM_SCOPE
             const json_char * runner = ((array) ? value.data() : name.data());
             #ifdef JSON_DEBUG
                const json_char * const end = runner + value.length();
            #endif
             if (json_unlikely(*runner == JSON_TEMP_COMMENT_IDENTIFIER)){  //multiple comments will be consolidated into one
                size_t count;
                const json_char * start;
                newcomment:
                count = 0;
                start = runner + 1;
                while(*(++runner) != JSON_TEMP_COMMENT_IDENTIFIER){
                    JSON_ASSERT(runner != end, JSON_TEXT("Removing white space failed"));
                    ++count;
                }
                if (count) _comment += json_string(start, count);
                if (json_unlikely(*(++runner) == JSON_TEMP_COMMENT_IDENTIFIER)){ //step past the trailing tag
                    _comment += JSON_TEXT('\n');
                    goto newcomment;
                }
             }
             internalJSONNode * myinternal;
             if (array){
                myinternal = internalJSONNode::newInternal(name, runner);
             } else {
                myinternal = internalJSONNode::newInternal(++runner, value);
             }
             child = JSONNode::newJSONNode(myinternal);
          END_MEM_SCOPE
          child -> set_comment(_comment);
       END_MEM_SCOPE
       const_cast<internalJSONNode*>(parent) -> CHILDREN -> push_back(child);   //attach it to the parent node
    #else
    if (name.empty()){
        const_cast<internalJSONNode*>(parent) -> CHILDREN -> push_back(JSONNode::newJSONNode(internalJSONNode::newInternal(name, value)));      //attach it to the parent node
    } else {
        const_cast<internalJSONNode*>(parent) -> CHILDREN -> push_back(JSONNode::newJSONNode(internalJSONNode::newInternal(json_string(name.begin() + 1, name.end()), value)));     //attach it to the parent node
    }
    #endif
}

//Create a subarray
void JSONWorker::DoArray(const internalJSONNode * parent, const json_string & value_t) json_nothrow {
    //This takes an array and creates nodes out of them
    JSON_ASSERT(!value_t.empty(), JSON_TEXT("DoArray is empty"));
    JSON_ASSERT_SAFE(value_t[0] == JSON_TEXT('['), JSON_TEXT("DoArray is not an array"), parent -> Nullify(); return;);
    if (json_unlikely(value_t.length() <= 2)) return;  // just a [] (blank array)
    
    #ifdef JSON_SAFE
        json_string newValue;  //share this so it has a reserved buffer
    #endif
    size_t starting = 1;  //ignore the [
    
    //Not sure what's in the array, so we have to use commas
    for(size_t ending = FIND_NEXT_RELEVANT(JSON_TEXT(','), value_t, 1);
        ending != json_string::npos;
        ending = FIND_NEXT_RELEVANT(JSON_TEXT(','), value_t, starting)){
        
        #ifdef JSON_SAFE
            newValue.assign(value_t.begin() + starting, value_t.begin() + ending);
            JSON_ASSERT_SAFE(FIND_NEXT_RELEVANT(JSON_TEXT(':'), newValue, 0) == json_string::npos, JSON_TEXT("Key/Value pairs are not allowed in arrays"), parent -> Nullify(); return;);
            NewNode(parent, json_global(EMPTY_JSON_STRING), newValue, true);
        #else
            NewNode(parent, json_global(EMPTY_JSON_STRING), json_string(value_t.begin() + starting, value_t.begin() + ending), true);
        #endif
        starting = ending + 1;
    }
    //since the last one will not find the comma, we have to add it here, but ignore the final ]
    
    #ifdef JSON_SAFE
        newValue.assign(value_t.begin() + starting, value_t.end() - 1);
        JSON_ASSERT_SAFE(FIND_NEXT_RELEVANT(JSON_TEXT(':'), newValue, 0) == json_string::npos, JSON_TEXT("Key/Value pairs are not allowed in arrays"), parent -> Nullify(); return;);
        NewNode(parent, json_global(EMPTY_JSON_STRING), newValue, true);
    #else
        NewNode(parent, json_global(EMPTY_JSON_STRING), json_string(value_t.begin() + starting, value_t.end() - 1), true);
    #endif
}


//Create all child nodes
void JSONWorker::DoNode(const internalJSONNode * parent, const json_string & value_t) json_nothrow {    
    //This take a node and creates its members and such
    JSON_ASSERT(!value_t.empty(), JSON_TEXT("DoNode is empty"));
    JSON_ASSERT_SAFE(value_t[0] == JSON_TEXT('{'), JSON_TEXT("DoNode is not an node"), parent -> Nullify(); return;);
    if (json_unlikely(value_t.length() <= 2)) return;  // just a {} (blank node)
    
    size_t name_ending = FIND_NEXT_RELEVANT(JSON_TEXT(':'), value_t, 1);  //find where the name ends
    JSON_ASSERT_SAFE(name_ending != json_string::npos, JSON_TEXT("Missing :"), parent -> Nullify(); return;);
    json_string name(value_t.begin() + 1, value_t.begin() + name_ending - 1);     //pull the name out
    for (size_t value_ending = FIND_NEXT_RELEVANT(JSON_TEXT(','), value_t, name_ending),  //find the end of the value
         name_starting = 1;  //ignore the {
         value_ending != json_string::npos;
         value_ending = FIND_NEXT_RELEVANT(JSON_TEXT(','), value_t, name_ending)){
        
        NewNode(parent, name, json_string(value_t.begin() + name_ending + 1, value_t.begin() + value_ending), false);
        name_starting = value_ending + 1;
        name_ending = FIND_NEXT_RELEVANT(JSON_TEXT(':'), value_t, name_starting);
        JSON_ASSERT_SAFE(name_ending != json_string::npos, JSON_TEXT("Missing :"), parent -> Nullify(); return;);
        name.assign(value_t.begin() + name_starting, value_t.begin() + name_ending - 1);
    }
    //since the last one will not find the comma, we have to add it here
    NewNode(parent, name, json_string(value_t.begin() + name_ending + 1, value_t.end() - 1), false);
}
#endif