xmpscanner.cpp

flash xmp sdk,flash官方SDK

	switch ( ths->fPosition ) {
	
		case 0 :	// The name should haved ended at the '=', nulls already skipped.

			if ( ths->fBufferPtr >= ths->fBufferLimit ) return eTriMaybe;
			if ( *ths->fBufferPtr != '=' ) return eTriNo;
			ths->fBufferPtr += bytesPerChar;	
			ths->fPosition = 1;
			// fall through OK because MatchOpenQuote will check the buffer limit and nulls ...
		
		case 1 :	// Look for the open quote.

			result = MatchOpenQuote ( ths, NULL );
			if ( result != eTriYes ) return result;
			ths->fPosition = 2;
			// fall through OK because the buffer limit and nulls are checked below ...
		
		default :	// Look for the close quote, capturing the value along the way.
			
			assert ( ths->fPosition == 2 );
			
			const char quoteChar = ths->fQuoteChar;

			while ( ths->fBufferPtr < ths->fBufferLimit ) {
				currChar = *ths->fBufferPtr;
				if ( currChar == quoteChar ) break;
				#if UseStringPushBack
					ths->fAttrValue.push_back ( currChar );
				#else
					ths->fAttrValue.insert ( ths->fAttrValue.end(), currChar );
				#endif
				ths->fBufferPtr += bytesPerChar;
			}
			
			if ( ths->fBufferPtr >= ths->fBufferLimit ) return eTriMaybe;
			assert ( currChar == quoteChar );
			ths->fBufferPtr += bytesPerChar;	// Advance past the closing quote.
			return eTriYes;
	
	}

}	// CaptureAttrValue


// =================================================================================================
// RecordStart
// ===========
//
// Note that this routine looks at bytes, not logical characters.  It has to figure out how many
// bytes per character there are so that the other recognizers can skip intervening nulls.

XMPScanner::PacketMachine::TriState
XMPScanner::PacketMachine::RecordStart ( PacketMachine * ths, const char * /* unused */ )
{

	while ( true ) {

		if ( ths->fBufferPtr >= ths->fBufferLimit ) return eTriMaybe;
		
		const char currByte = *ths->fBufferPtr;
	
		switch ( ths->fPosition ) {
		
			case 0 :	// Record the length.
				assert ( ths->fCharForm == eChar8Bit );
				assert ( ths->fBytesPerChar == 1 );
				ths->fPacketStart = ths->fBufferOffset + ((ths->fBufferPtr - 1) - ths->fBufferOrigin);
				ths->fPacketLength = 0;
				ths->fPosition = 1;
				// ! OK to fall through here, we didn't consume a byte in this step.

			case 1 :	// Look for the first null byte.
				if ( currByte != 0 ) return eTriYes;	// No nulls found.
				ths->fCharForm = eChar16BitBig;			// Assume 16 bit big endian for now.
				ths->fBytesPerChar = 2;
				ths->fBufferPtr++;
				ths->fPosition = 2;
				break;	// ! Don't fall through, have to check for the end of the buffer between each byte.

			case 2 :	// One null was found, look for a second.
				if ( currByte != 0 ) return eTriYes;	// Just one null found.
				ths->fBufferPtr++;
				ths->fPosition = 3;
				break;

			case 3 :	// Two nulls were found, look for a third.
				if ( currByte != 0 ) return eTriNo;	// Just two nulls is not valid.
				ths->fCharForm = eChar32BitBig;		// Assume 32 bit big endian for now.
				ths->fBytesPerChar = 4;
				ths->fBufferPtr++;
				return eTriYes;
				break;

		}
	
	}
	
}	// RecordStart


// =================================================================================================
// RecognizeBOM
// ============
//
// Recognizing the byte order marker is a surprisingly messy thing to do.  It can't be done by the
// normal string matcher, there are no intervening nulls.  There are 4 transitions after the opening
// quote, the closing quote or one of the three encodings.  For the actual BOM there are then 1 or 2
// following bytes that depend on which of the encodings we're in.  Not to mention that the buffer
// might end at any point.
//
// The intervening null count done earlier determined 8, 16, or 32 bits per character, but not the
// big or little endian nature for the 16/32 bit cases.  The BOM must be present for the 16 and 32
// bit cases in order to determine the endian mode.  There are six possible byte sequences for the
// quoted BOM string, ignoring the differences for quoting with ''' versus '"'.
//
// Keep in mind that for the 16 and 32 bit cases there will be nulls for the quote.  In the table
// below the symbol <quote> means just the one byte containing the ''' or '"'.  The nulls for the
// quote character are explicitly shown.
//
//	<quote> <quote>					- 1: No BOM, this must be an 8 bit case.
//	<quote> \xEF \xBB \xBF <quote>	- 1.12-13: The 8 bit form.
//
//	<quote> \xFE \xFF \x00 <quote>	- 1.22-23: The 16 bit, big endian form
//	<quote> \x00 \xFF \xFE <quote>	- 1.32-33: The 16 bit, little endian form.
//
//	<quote> \x00 \x00 \xFE \xFF \x00 \x00 \x00 <quote>	- 1.32.43-45.56-57: The 32 bit, big endian form.
//	<quote> \x00 \x00 \x00 \xFF \xFE \x00 \x00 <quote>	- 1.32.43.54-57: The 32 bit, little endian form.

enum {
	eBOM_8_1		= 0xEF,
	eBOM_8_2		= 0xBB,
	eBOM_8_3		= 0xBF,
	eBOM_Big_1		= 0xFE,
	eBOM_Big_2		= 0xFF,
	eBOM_Little_1	= eBOM_Big_2,
	eBOM_Little_2	= eBOM_Big_1
};

XMPScanner::PacketMachine::TriState
XMPScanner::PacketMachine::RecognizeBOM ( PacketMachine * ths, const char * /* unused */ )
{
	const int	bytesPerChar	= ths->fBytesPerChar;

	while ( true ) {	// Handle one character at a time, the micro-state (fPosition) changes for each.
	
		if ( ths->fBufferPtr >= ths->fBufferLimit ) return eTriMaybe;

		const unsigned char currChar = *ths->fBufferPtr;	// ! The BOM bytes look like integers bigger than 127.
		
		switch ( ths->fPosition ) {
		
			case  0 :	// Look for the opening quote.
				if ( (currChar != '\'') && (currChar != '"') ) return eTriNo;
				ths->fQuoteChar = currChar;
				ths->fBufferPtr++;
				ths->fPosition = 1;
				break;	// ! Don't fall through, have to check for the end of the buffer between each byte.
			
			case 1 :	// Look at the byte immediately following the opening quote.
				if ( currChar == ths->fQuoteChar ) {	// Closing quote, no BOM character, must be 8 bit.
					if ( ths->fCharForm != eChar8Bit ) return eTriNo;
					ths->fBufferPtr += bytesPerChar;	// Skip the nulls after the closing quote.
					return eTriYes;
				} else if ( currChar == eBOM_8_1 ) {	// Start of the 8 bit form.
					if ( ths->fCharForm != eChar8Bit ) return eTriNo;
					ths->fBufferPtr++;
					ths->fPosition = 12;
				} else if ( currChar == eBOM_Big_1 ) {	// Start of the 16 bit big endian form.
					if ( ths->fCharForm != eChar16BitBig ) return eTriNo;
					ths->fBufferPtr++;
					ths->fPosition = 22;
				} else if ( currChar == 0 ) {	// Start of the 16 bit little endian or either 32 bit form.
					if ( ths->fCharForm == eChar8Bit ) return eTriNo;
					ths->fBufferPtr++;
					ths->fPosition = 32;
				} else {
					return eTriNo;
				}
				break;
			
			case 12 :	// Look for the second byte of the 8 bit form.
				if ( currChar != eBOM_8_2 ) return eTriNo;
				ths->fPosition = 13;
				ths->fBufferPtr++;
				break;
			
			case 13 :	// Look for the third byte of the 8 bit form.
				if ( currChar != eBOM_8_3 ) return eTriNo;
				ths->fPosition = 99;
				ths->fBufferPtr++;
				break;
			
			case 22 :	// Look for the second byte of the 16 bit big endian form.
				if ( currChar != eBOM_Big_2 ) return eTriNo;
				ths->fPosition = 23;
				ths->fBufferPtr++;
				break;
			
			case 23 :	// Look for the null before the closing quote of the 16 bit big endian form.
				if ( currChar != 0 ) return eTriNo;
				ths->fBufferPtr++;
				ths->fPosition = 99;
				break;
			
			case 32 :	// Look at the second byte of the 16 bit little endian or either 32 bit form.
				if ( currChar == eBOM_Little_1 ) {
					ths->fPosition = 33;
				} else if ( currChar == 0 ) {
					ths->fPosition = 43;
				} else {
					return eTriNo;
				}
				ths->fBufferPtr++;
				break;
			
			case 33 :	// Look for the third byte of the 16 bit little endian form.
				if ( ths->fCharForm != eChar16BitBig ) return eTriNo;	// Null count before assumed big endian.
				if ( currChar != eBOM_Little_2 ) return eTriNo;
				ths->fCharForm = eChar16BitLittle;
				ths->fPosition = 99;
				ths->fBufferPtr++;
				break;
			
			case 43 :	// Look at the third byte of either 32 bit form.
				if ( ths->fCharForm != eChar32BitBig ) return eTriNo;	// Null count before assumed big endian.
				if ( currChar == eBOM_Big_1 ) {
					ths->fPosition = 44;
				} else if ( currChar == 0 ) {
					ths->fPosition = 54;
				} else {
					return eTriNo;
				}
				ths->fBufferPtr++;
				break;
			
			case 44 :	// Look for the fourth byte of the 32 bit big endian form.
				if ( currChar != eBOM_Big_2 ) return eTriNo;
				ths->fPosition = 45;
				ths->fBufferPtr++;
				break;
			
			case 45 :	// Look for the first null before the closing quote of the 32 bit big endian form.
				if ( currChar != 0 ) return eTriNo;
				ths->fPosition = 56;
				ths->fBufferPtr++;
				break;
			
			case 54 :	// Look for the fourth byte of the 32 bit little endian form.
				ths->fCharForm = eChar32BitLittle;
				if ( currChar != eBOM_Little_1 ) return eTriNo;
				ths->fPosition = 55;
				ths->fBufferPtr++;
				break;
			
			case 55 :	// Look for the fifth byte of the 32 bit little endian form.
				if ( currChar != eBOM_Little_2 ) return eTriNo;
				ths->fPosition = 56;
				ths->fBufferPtr++;
				break;
			
			case 56 :	// Look for the next to last null before the closing quote of the 32 bit forms.
				if ( currChar != 0 ) return eTriNo;
				ths->fPosition = 57;
				ths->fBufferPtr++;
				break;
			
			case 57 :	// Look for the last null before the closing quote of the 32 bit forms.
				if ( currChar != 0 ) return eTriNo;
				ths->fPosition = 99;
				ths->fBufferPtr++;
				break;
			
			default :	// Look for the closing quote.
				assert ( ths->fPosition == 99 );
				if ( currChar != ths->fQuoteChar ) return eTriNo;
				ths->fBufferPtr += bytesPerChar;	// Skip the nulls after the closing quote.
				return eTriYes;
				break;

		}
	
	}

}	// RecognizeBOM


// =================================================================================================
// RecordHeadAttr
// ==============

XMPScanner::PacketMachine::TriState
XMPScanner::PacketMachine::RecordHeadAttr ( PacketMachine * ths, const char * /* unused */ )
{
	
	if ( ths->fAttrName == "encoding" ) {
	
		assert ( ths->fEncodingAttr.empty() );
		ths->fEncodingAttr = ths->fAttrValue;
	
	} else if ( ths->fAttrName == "bytes" ) {
			
		long	value	= 0;
		int		count	= ths->fAttrValue.size();
		int		i;
		
		assert ( ths->fBytesAttr == -1 );
		
		if ( count > 0 ) {	// Allow bytes='' to be the same as no bytes attribute.

			for ( i = 0; i < count; i++ ) {
				const char	currChar	= ths->fAttrValue[i];
				if ( ('0' <= currChar) && (currChar <= '9') ) {
					value = (value * 10) + (currChar - '0');
				} else {
					ths->fBogusPacket = true;
					value = -1;
					break;
				}
			}
			ths->fBytesAttr = value;
			
			if ( CharFormIs16Bit ( ths->fCharForm ) ) {
				if ( (ths->fBytesAttr & 1) != 0 ) ths->fBogusPacket = true;
			} else if ( CharFormIs32Bit ( ths->fCharForm ) ) {
				if ( (ths->fBytesAttr & 3) != 0 ) ths->fBogusPacket = true;
			}

		}

	}
	
	ths->fAttrName.erase ( ths->fAttrName.begin(), ths->fAttrName.end() );
	ths->fAttrValue.erase ( ths->fAttrValue.begin(), ths->fAttrValue.end() );

	return eTriYes;

}	// RecordHeadAttr


// =================================================================================================
// CaptureAccess
// =============

XMPScanner::PacketMachine::TriState
XMPScanner::PacketMachine::CaptureAccess ( PacketMachine * ths, const char * /* unused */ )
{
	const int	bytesPerChar	= ths->fBytesPerChar;

	while ( true ) {

		if ( ths->fBufferPtr >= ths->fBufferLimit ) return eTriMaybe;
		
		const char currChar = *ths->fBufferPtr;
		
		switch ( ths->fPosition ) {
		
			case  0 :	// Look for the opening quote.
				if ( (currChar != '\'') && (currChar != '"') ) return eTriNo;
				ths->fQuoteChar = currChar;
				ths->fBufferPtr += bytesPerChar;
				ths->fPosition = 1;
				break;	// ! Don't fall through, have to check for the end of the buffer between each byte.
		
			case  1 :	// Look for the 'r' or 'w'.
				if ( (currChar != 'r') && (currChar != 'w') ) return eTriNo;
				ths->fAccess = currChar;
				ths->fBufferPtr += bytesPerChar;
				ths->fPosition = 2;
				break;