ian@0: // Copyright 2007-2011 Baptiste Lepilleur
ian@0: // Distributed under MIT license, or public domain if desired and
ian@0: // recognized in your jurisdiction.
ian@0: // See file LICENSE for detail or copy at http://jsoncpp.sourceforge.net/LICENSE
ian@0: 
ian@0: #if !defined(JSON_IS_AMALGAMATION)
ian@0: # include <json/assertions.h>
ian@0: # include <json/reader.h>
ian@0: # include <json/value.h>
ian@0: # include "json_tool.h"
ian@0: #endif // if !defined(JSON_IS_AMALGAMATION)
ian@0: #include <utility>
ian@0: #include <cstdio>
ian@0: #include <cassert>
ian@0: #include <cstring>
ian@0: #include <iostream>
ian@0: #include <stdexcept>
ian@0: 
ian@0: #if _MSC_VER >= 1400 // VC++ 8.0
ian@0: #pragma warning( disable : 4996 )   // disable warning about strdup being deprecated.
ian@0: #endif
ian@0: 
ian@0: namespace Json {
ian@0: 
ian@0: // Implementation of class Features
ian@0: // ////////////////////////////////
ian@0: 
ian@0: Features::Features()
ian@0:    : allowComments_( true )
ian@0:    , strictRoot_( false )
ian@0: {
ian@0: }
ian@0: 
ian@0: 
ian@0: Features 
ian@0: Features::all()
ian@0: {
ian@0:    return Features();
ian@0: }
ian@0: 
ian@0: 
ian@0: Features 
ian@0: Features::strictMode()
ian@0: {
ian@0:    Features features;
ian@0:    features.allowComments_ = false;
ian@0:    features.strictRoot_ = true;
ian@0:    return features;
ian@0: }
ian@0: 
ian@0: // Implementation of class Reader
ian@0: // ////////////////////////////////
ian@0: 
ian@0: 
ian@0: static inline bool 
ian@0: in( Reader::Char c, Reader::Char c1, Reader::Char c2, Reader::Char c3, Reader::Char c4 )
ian@0: {
ian@0:    return c == c1  ||  c == c2  ||  c == c3  ||  c == c4;
ian@0: }
ian@0: 
ian@0: static inline bool 
ian@0: in( Reader::Char c, Reader::Char c1, Reader::Char c2, Reader::Char c3, Reader::Char c4, Reader::Char c5 )
ian@0: {
ian@0:    return c == c1  ||  c == c2  ||  c == c3  ||  c == c4  ||  c == c5;
ian@0: }
ian@0: 
ian@0: 
ian@0: static bool 
ian@0: containsNewLine( Reader::Location begin, 
ian@0:                  Reader::Location end )
ian@0: {
ian@0:    for ( ;begin < end; ++begin )
ian@0:       if ( *begin == '\n'  ||  *begin == '\r' )
ian@0:          return true;
ian@0:    return false;
ian@0: }
ian@0: 
ian@0: 
ian@0: // Class Reader
ian@0: // //////////////////////////////////////////////////////////////////
ian@0: 
ian@0: Reader::Reader()
ian@0:     : errors_(),
ian@0:       document_(),
ian@0:       begin_(),
ian@0:       end_(),
ian@0:       current_(),
ian@0:       lastValueEnd_(),
ian@0:       lastValue_(),
ian@0:       commentsBefore_(),
ian@0:       features_( Features::all() ),
ian@0:       collectComments_()
ian@0: {
ian@0: }
ian@0: 
ian@0: 
ian@0: Reader::Reader( const Features &features )
ian@0:     : errors_(),
ian@0:       document_(),
ian@0:       begin_(),
ian@0:       end_(),
ian@0:       current_(),
ian@0:       lastValueEnd_(),
ian@0:       lastValue_(),
ian@0:       commentsBefore_(),
ian@0:       features_( features ),
ian@0:       collectComments_()
ian@0: {
ian@0: }
ian@0: 
ian@0: 
ian@0: bool
ian@0: Reader::parse( const std::string &document, 
ian@0:                Value &root,
ian@0:                bool collectComments )
ian@0: {
ian@0:    document_ = document;
ian@0:    const char *begin = document_.c_str();
ian@0:    const char *end = begin + document_.length();
ian@0:    return parse( begin, end, root, collectComments );
ian@0: }
ian@0: 
ian@0: 
ian@0: bool
ian@0: Reader::parse( std::istream& sin,
ian@0:                Value &root,
ian@0:                bool collectComments )
ian@0: {
ian@0:    //std::istream_iterator<char> begin(sin);
ian@0:    //std::istream_iterator<char> end;
ian@0:    // Those would allow streamed input from a file, if parse() were a
ian@0:    // template function.
ian@0: 
ian@0:    // Since std::string is reference-counted, this at least does not
ian@0:    // create an extra copy.
ian@0:    std::string doc;
ian@0:    std::getline(sin, doc, (char)EOF);
ian@0:    return parse( doc, root, collectComments );
ian@0: }
ian@0: 
ian@0: bool 
ian@0: Reader::parse( const char *beginDoc, const char *endDoc, 
ian@0:                Value &root,
ian@0:                bool collectComments )
ian@0: {
ian@0:    if ( !features_.allowComments_ )
ian@0:    {
ian@0:       collectComments = false;
ian@0:    }
ian@0: 
ian@0:    begin_ = beginDoc;
ian@0:    end_ = endDoc;
ian@0:    collectComments_ = collectComments;
ian@0:    current_ = begin_;
ian@0:    lastValueEnd_ = 0;
ian@0:    lastValue_ = 0;
ian@0:    commentsBefore_ = "";
ian@0:    errors_.clear();
ian@0:    while ( !nodes_.empty() )
ian@0:       nodes_.pop();
ian@0:    nodes_.push( &root );
ian@0:    
ian@0:    bool successful = readValue();
ian@0:    Token token;
ian@0:    skipCommentTokens( token );
ian@0:    if ( collectComments_  &&  !commentsBefore_.empty() )
ian@0:       root.setComment( commentsBefore_, commentAfter );
ian@0:    if ( features_.strictRoot_ )
ian@0:    {
ian@0:       if ( !root.isArray()  &&  !root.isObject() )
ian@0:       {
ian@0:          // Set error location to start of doc, ideally should be first token found in doc
ian@0:          token.type_ = tokenError;
ian@0:          token.start_ = beginDoc;
ian@0:          token.end_ = endDoc;
ian@0:          addError( "A valid JSON document must be either an array or an object value.",
ian@0:                    token );
ian@0:          return false;
ian@0:       }
ian@0:    }
ian@0:    return successful;
ian@0: }
ian@0: 
ian@0: 
ian@0: bool
ian@0: Reader::readValue()
ian@0: {
ian@0:    Token token;
ian@0:    skipCommentTokens( token );
ian@0:    bool successful = true;
ian@0: 
ian@0:    if ( collectComments_  &&  !commentsBefore_.empty() )
ian@0:    {
ian@0:       currentValue().setComment( commentsBefore_, commentBefore );
ian@0:       commentsBefore_ = "";
ian@0:    }
ian@0: 
ian@0: 
ian@0:    switch ( token.type_ )
ian@0:    {
ian@0:    case tokenObjectBegin:
ian@0:       successful = readObject( token );
ian@0:       break;
ian@0:    case tokenArrayBegin:
ian@0:       successful = readArray( token );
ian@0:       break;
ian@0:    case tokenNumber:
ian@0:       successful = decodeNumber( token );
ian@0:       break;
ian@0:    case tokenString:
ian@0:       successful = decodeString( token );
ian@0:       break;
ian@0:    case tokenTrue:
ian@0:       currentValue() = true;
ian@0:       break;
ian@0:    case tokenFalse:
ian@0:       currentValue() = false;
ian@0:       break;
ian@0:    case tokenNull:
ian@0:       currentValue() = Value();
ian@0:       break;
ian@0:    default:
ian@0:       return addError( "Syntax error: value, object or array expected.", token );
ian@0:    }
ian@0: 
ian@0:    if ( collectComments_ )
ian@0:    {
ian@0:       lastValueEnd_ = current_;
ian@0:       lastValue_ = &currentValue();
ian@0:    }
ian@0: 
ian@0:    return successful;
ian@0: }
ian@0: 
ian@0: 
ian@0: void 
ian@0: Reader::skipCommentTokens( Token &token )
ian@0: {
ian@0:    if ( features_.allowComments_ )
ian@0:    {
ian@0:       do
ian@0:       {
ian@0:          readToken( token );
ian@0:       }
ian@0:       while ( token.type_ == tokenComment );
ian@0:    }
ian@0:    else
ian@0:    {
ian@0:       readToken( token );
ian@0:    }
ian@0: }
ian@0: 
ian@0: 
ian@0: bool 
ian@0: Reader::expectToken( TokenType type, Token &token, const char *message )
ian@0: {
ian@0:    readToken( token );
ian@0:    if ( token.type_ != type )
ian@0:       return addError( message, token );
ian@0:    return true;
ian@0: }
ian@0: 
ian@0: 
ian@0: bool 
ian@0: Reader::readToken( Token &token )
ian@0: {
ian@0:    skipSpaces();
ian@0:    token.start_ = current_;
ian@0:    Char c = getNextChar();
ian@0:    bool ok = true;
ian@0:    switch ( c )
ian@0:    {
ian@0:    case '{':
ian@0:       token.type_ = tokenObjectBegin;
ian@0:       break;
ian@0:    case '}':
ian@0:       token.type_ = tokenObjectEnd;
ian@0:       break;
ian@0:    case '[':
ian@0:       token.type_ = tokenArrayBegin;
ian@0:       break;
ian@0:    case ']':
ian@0:       token.type_ = tokenArrayEnd;
ian@0:       break;
ian@0:    case '"':
ian@0:       token.type_ = tokenString;
ian@0:       ok = readString();
ian@0:       break;
ian@0:    case '/':
ian@0:       token.type_ = tokenComment;
ian@0:       ok = readComment();
ian@0:       break;
ian@0:    case '0':
ian@0:    case '1':
ian@0:    case '2':
ian@0:    case '3':
ian@0:    case '4':
ian@0:    case '5':
ian@0:    case '6':
ian@0:    case '7':
ian@0:    case '8':
ian@0:    case '9':
ian@0:    case '-':
ian@0:       token.type_ = tokenNumber;
ian@0:       readNumber();
ian@0:       break;
ian@0:    case 't':
ian@0:       token.type_ = tokenTrue;
ian@0:       ok = match( "rue", 3 );
ian@0:       break;
ian@0:    case 'f':
ian@0:       token.type_ = tokenFalse;
ian@0:       ok = match( "alse", 4 );
ian@0:       break;
ian@0:    case 'n':
ian@0:       token.type_ = tokenNull;
ian@0:       ok = match( "ull", 3 );
ian@0:       break;
ian@0:    case ',':
ian@0:       token.type_ = tokenArraySeparator;
ian@0:       break;
ian@0:    case ':':
ian@0:       token.type_ = tokenMemberSeparator;
ian@0:       break;
ian@0:    case 0:
ian@0:       token.type_ = tokenEndOfStream;
ian@0:       break;
ian@0:    default:
ian@0:       ok = false;
ian@0:       break;
ian@0:    }
ian@0:    if ( !ok )
ian@0:       token.type_ = tokenError;
ian@0:    token.end_ = current_;
ian@0:    return true;
ian@0: }
ian@0: 
ian@0: 
ian@0: void 
ian@0: Reader::skipSpaces()
ian@0: {
ian@0:    while ( current_ != end_ )
ian@0:    {
ian@0:       Char c = *current_;
ian@0:       if ( c == ' '  ||  c == '\t'  ||  c == '\r'  ||  c == '\n' )
ian@0:          ++current_;
ian@0:       else
ian@0:          break;
ian@0:    }
ian@0: }
ian@0: 
ian@0: 
ian@0: bool 
ian@0: Reader::match( Location pattern, 
ian@0:                int patternLength )
ian@0: {
ian@0:    if ( end_ - current_ < patternLength )
ian@0:       return false;
ian@0:    int index = patternLength;
ian@0:    while ( index-- )
ian@0:       if ( current_[index] != pattern[index] )
ian@0:          return false;
ian@0:    current_ += patternLength;
ian@0:    return true;
ian@0: }
ian@0: 
ian@0: 
ian@0: bool
ian@0: Reader::readComment()
ian@0: {
ian@0:    Location commentBegin = current_ - 1;
ian@0:    Char c = getNextChar();
ian@0:    bool successful = false;
ian@0:    if ( c == '*' )
ian@0:       successful = readCStyleComment();
ian@0:    else if ( c == '/' )
ian@0:       successful = readCppStyleComment();
ian@0:    if ( !successful )
ian@0:       return false;
ian@0: 
ian@0:    if ( collectComments_ )
ian@0:    {
ian@0:       CommentPlacement placement = commentBefore;
ian@0:       if ( lastValueEnd_  &&  !containsNewLine( lastValueEnd_, commentBegin ) )
ian@0:       {
ian@0:          if ( c != '*'  ||  !containsNewLine( commentBegin, current_ ) )
ian@0:             placement = commentAfterOnSameLine;
ian@0:       }
ian@0: 
ian@0:       addComment( commentBegin, current_, placement );
ian@0:    }
ian@0:    return true;
ian@0: }
ian@0: 
ian@0: 
ian@0: void 
ian@0: Reader::addComment( Location begin, 
ian@0:                     Location end, 
ian@0:                     CommentPlacement placement )
ian@0: {
ian@0:    assert( collectComments_ );
ian@0:    if ( placement == commentAfterOnSameLine )
ian@0:    {
ian@0:       assert( lastValue_ != 0 );
ian@0:       lastValue_->setComment( std::string( begin, end ), placement );
ian@0:    }
ian@0:    else
ian@0:    {
ian@0:       if ( !commentsBefore_.empty() )
ian@0:          commentsBefore_ += "\n";
ian@0:       commentsBefore_ += std::string( begin, end );
ian@0:    }
ian@0: }
ian@0: 
ian@0: 
ian@0: bool 
ian@0: Reader::readCStyleComment()
ian@0: {
ian@0:    while ( current_ != end_ )
ian@0:    {
ian@0:       Char c = getNextChar();
ian@0:       if ( c == '*'  &&  *current_ == '/' )
ian@0:          break;
ian@0:    }
ian@0:    return getNextChar() == '/';
ian@0: }
ian@0: 
ian@0: 
ian@0: bool 
ian@0: Reader::readCppStyleComment()
ian@0: {
ian@0:    while ( current_ != end_ )
ian@0:    {
ian@0:       Char c = getNextChar();
ian@0:       if (  c == '\r'  ||  c == '\n' )
ian@0:          break;
ian@0:    }
ian@0:    return true;
ian@0: }
ian@0: 
ian@0: 
ian@0: void 
ian@0: Reader::readNumber()
ian@0: {
ian@0:    while ( current_ != end_ )
ian@0:    {
ian@0:       if ( !(*current_ >= '0'  &&  *current_ <= '9')  &&
ian@0:            !in( *current_, '.', 'e', 'E', '+', '-' ) )
ian@0:          break;
ian@0:       ++current_;
ian@0:    }
ian@0: }
ian@0: 
ian@0: bool
ian@0: Reader::readString()
ian@0: {
ian@0:    Char c = 0;
ian@0:    while ( current_ != end_ )
ian@0:    {
ian@0:       c = getNextChar();
ian@0:       if ( c == '\\' )
ian@0:          getNextChar();
ian@0:       else if ( c == '"' )
ian@0:          break;
ian@0:    }
ian@0:    return c == '"';
ian@0: }
ian@0: 
ian@0: 
ian@0: bool 
ian@0: Reader::readObject( Token &/*tokenStart*/ )
ian@0: {
ian@0:    Token tokenName;
ian@0:    std::string name;
ian@0:    currentValue() = Value( objectValue );
ian@0:    while ( readToken( tokenName ) )
ian@0:    {
ian@0:       bool initialTokenOk = true;
ian@0:       while ( tokenName.type_ == tokenComment  &&  initialTokenOk )
ian@0:          initialTokenOk = readToken( tokenName );
ian@0:       if  ( !initialTokenOk )
ian@0:          break;
ian@0:       if ( tokenName.type_ == tokenObjectEnd  &&  name.empty() )  // empty object
ian@0:          return true;
ian@0:       if ( tokenName.type_ != tokenString )
ian@0:          break;
ian@0:       
ian@0:       name = "";
ian@0:       if ( !decodeString( tokenName, name ) )
ian@0:          return recoverFromError( tokenObjectEnd );
ian@0: 
ian@0:       Token colon;
ian@0:       if ( !readToken( colon ) ||  colon.type_ != tokenMemberSeparator )
ian@0:       {
ian@0:          return addErrorAndRecover( "Missing ':' after object member name", 
ian@0:                                     colon, 
ian@0:                                     tokenObjectEnd );
ian@0:       }
ian@0:       Value &value = currentValue()[ name ];
ian@0:       nodes_.push( &value );
ian@0:       bool ok = readValue();
ian@0:       nodes_.pop();
ian@0:       if ( !ok ) // error already set
ian@0:          return recoverFromError( tokenObjectEnd );
ian@0: 
ian@0:       Token comma;
ian@0:       if ( !readToken( comma )
ian@0:             ||  ( comma.type_ != tokenObjectEnd  &&  
ian@0:                   comma.type_ != tokenArraySeparator &&
ian@0:                   comma.type_ != tokenComment ) )
ian@0:       {
ian@0:          return addErrorAndRecover( "Missing ',' or '}' in object declaration", 
ian@0:                                     comma, 
ian@0:                                     tokenObjectEnd );
ian@0:       }
ian@0:       bool finalizeTokenOk = true;
ian@0:       while ( comma.type_ == tokenComment &&
ian@0:               finalizeTokenOk )
ian@0:          finalizeTokenOk = readToken( comma );
ian@0:       if ( comma.type_ == tokenObjectEnd )
ian@0:          return true;
ian@0:    }
ian@0:    return addErrorAndRecover( "Missing '}' or object member name", 
ian@0:                               tokenName, 
ian@0:                               tokenObjectEnd );
ian@0: }
ian@0: 
ian@0: 
ian@0: bool 
ian@0: Reader::readArray( Token &/*tokenStart*/ )
ian@0: {
ian@0:    currentValue() = Value( arrayValue );
ian@0:    skipSpaces();
ian@0:    if ( *current_ == ']' ) // empty array
ian@0:    {
ian@0:       Token endArray;
ian@0:       readToken( endArray );
ian@0:       return true;
ian@0:    }
ian@0:    int index = 0;
ian@0:    for (;;)
ian@0:    {
ian@0:       Value &value = currentValue()[ index++ ];
ian@0:       nodes_.push( &value );
ian@0:       bool ok = readValue();
ian@0:       nodes_.pop();
ian@0:       if ( !ok ) // error already set
ian@0:          return recoverFromError( tokenArrayEnd );
ian@0: 
ian@0:       Token token;
ian@0:       // Accept Comment after last item in the array.
ian@0:       ok = readToken( token );
ian@0:       while ( token.type_ == tokenComment  &&  ok )
ian@0:       {
ian@0:          ok = readToken( token );
ian@0:       }
ian@0:       bool badTokenType = ( token.type_ != tokenArraySeparator  &&
ian@0:                             token.type_ != tokenArrayEnd );
ian@0:       if ( !ok  ||  badTokenType )
ian@0:       {
ian@0:          return addErrorAndRecover( "Missing ',' or ']' in array declaration", 
ian@0:                                     token, 
ian@0:                                     tokenArrayEnd );
ian@0:       }
ian@0:       if ( token.type_ == tokenArrayEnd )
ian@0:          break;
ian@0:    }
ian@0:    return true;
ian@0: }
ian@0: 
ian@0: 
ian@0: bool 
ian@0: Reader::decodeNumber( Token &token )
ian@0: {
ian@0:    bool isDouble = false;
ian@0:    for ( Location inspect = token.start_; inspect != token.end_; ++inspect )
ian@0:    {
ian@0:       isDouble = isDouble  
ian@0:                  ||  in( *inspect, '.', 'e', 'E', '+' )  
ian@0:                  ||  ( *inspect == '-'  &&  inspect != token.start_ );
ian@0:    }
ian@0:    if ( isDouble )
ian@0:       return decodeDouble( token );
ian@0:    // Attempts to parse the number as an integer. If the number is
ian@0:    // larger than the maximum supported value of an integer then
ian@0:    // we decode the number as a double.
ian@0:    Location current = token.start_;
ian@0:    bool isNegative = *current == '-';
ian@0:    if ( isNegative )
ian@0:       ++current;
ian@0:    Value::LargestUInt maxIntegerValue = isNegative ? Value::LargestUInt(-Value::minLargestInt) 
ian@0:                                                    : Value::maxLargestUInt;
ian@0:    Value::LargestUInt threshold = maxIntegerValue / 10;
ian@0:    Value::LargestUInt value = 0;
ian@0:    while ( current < token.end_ )
ian@0:    {
ian@0:       Char c = *current++;
ian@0:       if ( c < '0'  ||  c > '9' )
ian@0:          return addError( "'" + std::string( token.start_, token.end_ ) + "' is not a number.", token );
ian@0:       Value::UInt digit(c - '0');
ian@0:       if ( value >= threshold )
ian@0:       {
ian@0:          // We've hit or exceeded the max value divided by 10 (rounded down). If
ian@0:          // a) we've only just touched the limit, b) this is the last digit, and
ian@0:          // c) it's small enough to fit in that rounding delta, we're okay.
ian@0:          // Otherwise treat this number as a double to avoid overflow.
ian@0:          if (value > threshold ||
ian@0:              current != token.end_ ||
ian@0:              digit > maxIntegerValue % 10)
ian@0:          {
ian@0:             return decodeDouble( token );
ian@0:          }
ian@0:       }
ian@0:       value = value * 10 + digit;
ian@0:    }
ian@0:    if ( isNegative )
ian@0:       currentValue() = -Value::LargestInt( value );
ian@0:    else if ( value <= Value::LargestUInt(Value::maxInt) )
ian@0:       currentValue() = Value::LargestInt( value );
ian@0:    else
ian@0:       currentValue() = value;
ian@0:    return true;
ian@0: }
ian@0: 
ian@0: 
ian@0: bool 
ian@0: Reader::decodeDouble( Token &token )
ian@0: {
ian@0:    double value = 0;
ian@0:    const int bufferSize = 32;
ian@0:    int count;
ian@0:    int length = int(token.end_ - token.start_);
ian@0: 
ian@0:    // Sanity check to avoid buffer overflow exploits.
ian@0:    if (length < 0) {
ian@0:       return addError( "Unable to parse token length", token );
ian@0:    }
ian@0: 
ian@0:    // Avoid using a string constant for the format control string given to
ian@0:    // sscanf, as this can cause hard to debug crashes on OS X. See here for more
ian@0:    // info:
ian@0:    //
ian@0:    //     http://developer.apple.com/library/mac/#DOCUMENTATION/DeveloperTools/gcc-4.0.1/gcc/Incompatibilities.html
ian@0:    char format[] = "%lf";
ian@0: 
ian@0:    if ( length <= bufferSize )
ian@0:    {
ian@0:       Char buffer[bufferSize+1];
ian@0:       memcpy( buffer, token.start_, length );
ian@0:       buffer[length] = 0;
ian@0:       count = sscanf( buffer, format, &value );
ian@0:    }
ian@0:    else
ian@0:    {
ian@0:       std::string buffer( token.start_, token.end_ );
ian@0:       count = sscanf( buffer.c_str(), format, &value );
ian@0:    }
ian@0: 
ian@0:    if ( count != 1 )
ian@0:       return addError( "'" + std::string( token.start_, token.end_ ) + "' is not a number.", token );
ian@0:    currentValue() = value;
ian@0:    return true;
ian@0: }
ian@0: 
ian@0: 
ian@0: bool 
ian@0: Reader::decodeString( Token &token )
ian@0: {
ian@0:    std::string decoded;
ian@0:    if ( !decodeString( token, decoded ) )
ian@0:       return false;
ian@0:    currentValue() = decoded;
ian@0:    return true;
ian@0: }
ian@0: 
ian@0: 
ian@0: bool 
ian@0: Reader::decodeString( Token &token, std::string &decoded )
ian@0: {
ian@0:    decoded.reserve( token.end_ - token.start_ - 2 );
ian@0:    Location current = token.start_ + 1; // skip '"'
ian@0:    Location end = token.end_ - 1;      // do not include '"'
ian@0:    while ( current != end )
ian@0:    {
ian@0:       Char c = *current++;
ian@0:       if ( c == '"' )
ian@0:          break;
ian@0:       else if ( c == '\\' )
ian@0:       {
ian@0:          if ( current == end )
ian@0:             return addError( "Empty escape sequence in string", token, current );
ian@0:          Char escape = *current++;
ian@0:          switch ( escape )
ian@0:          {
ian@0:          case '"': decoded += '"'; break;
ian@0:          case '/': decoded += '/'; break;
ian@0:          case '\\': decoded += '\\'; break;
ian@0:          case 'b': decoded += '\b'; break;
ian@0:          case 'f': decoded += '\f'; break;
ian@0:          case 'n': decoded += '\n'; break;
ian@0:          case 'r': decoded += '\r'; break;
ian@0:          case 't': decoded += '\t'; break;
ian@0:          case 'u':
ian@0:             {
ian@0:                unsigned int unicode;
ian@0:                if ( !decodeUnicodeCodePoint( token, current, end, unicode ) )
ian@0:                   return false;
ian@0:                decoded += codePointToUTF8(unicode);
ian@0:             }
ian@0:             break;
ian@0:          default:
ian@0:             return addError( "Bad escape sequence in string", token, current );
ian@0:          }
ian@0:       }
ian@0:       else
ian@0:       {
ian@0:          decoded += c;
ian@0:       }
ian@0:    }
ian@0:    return true;
ian@0: }
ian@0: 
ian@0: bool
ian@0: Reader::decodeUnicodeCodePoint( Token &token, 
ian@0:                                      Location &current, 
ian@0:                                      Location end, 
ian@0:                                      unsigned int &unicode )
ian@0: {
ian@0: 
ian@0:    if ( !decodeUnicodeEscapeSequence( token, current, end, unicode ) )
ian@0:       return false;
ian@0:    if (unicode >= 0xD800 && unicode <= 0xDBFF)
ian@0:    {
ian@0:       // surrogate pairs
ian@0:       if (end - current < 6)
ian@0:          return addError( "additional six characters expected to parse unicode surrogate pair.", token, current );
ian@0:       unsigned int surrogatePair;
ian@0:       if (*(current++) == '\\' && *(current++)== 'u')
ian@0:       {
ian@0:          if (decodeUnicodeEscapeSequence( token, current, end, surrogatePair ))
ian@0:          {
ian@0:             unicode = 0x10000 + ((unicode & 0x3FF) << 10) + (surrogatePair & 0x3FF);
ian@0:          } 
ian@0:          else
ian@0:             return false;
ian@0:       } 
ian@0:       else
ian@0:          return addError( "expecting another \\u token to begin the second half of a unicode surrogate pair", token, current );
ian@0:    }
ian@0:    return true;
ian@0: }
ian@0: 
ian@0: bool 
ian@0: Reader::decodeUnicodeEscapeSequence( Token &token, 
ian@0:                                      Location &current, 
ian@0:                                      Location end, 
ian@0:                                      unsigned int &unicode )
ian@0: {
ian@0:    if ( end - current < 4 )
ian@0:       return addError( "Bad unicode escape sequence in string: four digits expected.", token, current );
ian@0:    unicode = 0;
ian@0:    for ( int index =0; index < 4; ++index )
ian@0:    {
ian@0:       Char c = *current++;
ian@0:       unicode *= 16;
ian@0:       if ( c >= '0'  &&  c <= '9' )
ian@0:          unicode += c - '0';
ian@0:       else if ( c >= 'a'  &&  c <= 'f' )
ian@0:          unicode += c - 'a' + 10;
ian@0:       else if ( c >= 'A'  &&  c <= 'F' )
ian@0:          unicode += c - 'A' + 10;
ian@0:       else
ian@0:          return addError( "Bad unicode escape sequence in string: hexadecimal digit expected.", token, current );
ian@0:    }
ian@0:    return true;
ian@0: }
ian@0: 
ian@0: 
ian@0: bool 
ian@0: Reader::addError( const std::string &message, 
ian@0:                   Token &token,
ian@0:                   Location extra )
ian@0: {
ian@0:    ErrorInfo info;
ian@0:    info.token_ = token;
ian@0:    info.message_ = message;
ian@0:    info.extra_ = extra;
ian@0:    errors_.push_back( info );
ian@0:    return false;
ian@0: }
ian@0: 
ian@0: 
ian@0: bool 
ian@0: Reader::recoverFromError( TokenType skipUntilToken )
ian@0: {
ian@0:    int errorCount = int(errors_.size());
ian@0:    Token skip;
ian@0:    for (;;)
ian@0:    {
ian@0:       if ( !readToken(skip) )
ian@0:          errors_.resize( errorCount ); // discard errors caused by recovery
ian@0:       if ( skip.type_ == skipUntilToken  ||  skip.type_ == tokenEndOfStream )
ian@0:          break;
ian@0:    }
ian@0:    errors_.resize( errorCount );
ian@0:    return false;
ian@0: }
ian@0: 
ian@0: 
ian@0: bool 
ian@0: Reader::addErrorAndRecover( const std::string &message, 
ian@0:                             Token &token,
ian@0:                             TokenType skipUntilToken )
ian@0: {
ian@0:    addError( message, token );
ian@0:    return recoverFromError( skipUntilToken );
ian@0: }
ian@0: 
ian@0: 
ian@0: Value &
ian@0: Reader::currentValue()
ian@0: {
ian@0:    return *(nodes_.top());
ian@0: }
ian@0: 
ian@0: 
ian@0: Reader::Char 
ian@0: Reader::getNextChar()
ian@0: {
ian@0:    if ( current_ == end_ )
ian@0:       return 0;
ian@0:    return *current_++;
ian@0: }
ian@0: 
ian@0: 
ian@0: void 
ian@0: Reader::getLocationLineAndColumn( Location location,
ian@0:                                   int &line,
ian@0:                                   int &column ) const
ian@0: {
ian@0:    Location current = begin_;
ian@0:    Location lastLineStart = current;
ian@0:    line = 0;
ian@0:    while ( current < location  &&  current != end_ )
ian@0:    {
ian@0:       Char c = *current++;
ian@0:       if ( c == '\r' )
ian@0:       {
ian@0:          if ( *current == '\n' )
ian@0:             ++current;
ian@0:          lastLineStart = current;
ian@0:          ++line;
ian@0:       }
ian@0:       else if ( c == '\n' )
ian@0:       {
ian@0:          lastLineStart = current;
ian@0:          ++line;
ian@0:       }
ian@0:    }
ian@0:    // column & line start at 1
ian@0:    column = int(location - lastLineStart) + 1;
ian@0:    ++line;
ian@0: }
ian@0: 
ian@0: 
ian@0: std::string
ian@0: Reader::getLocationLineAndColumn( Location location ) const
ian@0: {
ian@0:    int line, column;
ian@0:    getLocationLineAndColumn( location, line, column );
ian@0:    char buffer[18+16+16+1];
ian@0:    sprintf( buffer, "Line %d, Column %d", line, column );
ian@0:    return buffer;
ian@0: }
ian@0: 
ian@0: 
ian@0: // Deprecated. Preserved for backward compatibility
ian@0: std::string 
ian@0: Reader::getFormatedErrorMessages() const
ian@0: {
ian@0:     return getFormattedErrorMessages();
ian@0: }
ian@0: 
ian@0: 
ian@0: std::string 
ian@0: Reader::getFormattedErrorMessages() const
ian@0: {
ian@0:    std::string formattedMessage;
ian@0:    for ( Errors::const_iterator itError = errors_.begin();
ian@0:          itError != errors_.end();
ian@0:          ++itError )
ian@0:    {
ian@0:       const ErrorInfo &error = *itError;
ian@0:       formattedMessage += "* " + getLocationLineAndColumn( error.token_.start_ ) + "\n";
ian@0:       formattedMessage += "  " + error.message_ + "\n";
ian@0:       if ( error.extra_ )
ian@0:          formattedMessage += "See " + getLocationLineAndColumn( error.extra_ ) + " for detail.\n";
ian@0:    }
ian@0:    return formattedMessage;
ian@0: }
ian@0: 
ian@0: 
ian@0: std::istream& operator>>( std::istream &sin, Value &root )
ian@0: {
ian@0:     Json::Reader reader;
ian@0:     bool ok = reader.parse(sin, root, true);
ian@0:     if (!ok) JSON_FAIL_MESSAGE(reader.getFormattedErrorMessages());
ian@0:     return sin;
ian@0: }
ian@0: 
ian@0: 
ian@0: } // namespace Json