xmpscanner.cpp

flash xmp sdk,flash官方SDK

			default :	// Look for the closing quote.
				assert ( ths->fPosition == 2 );
				if ( currChar != ths->fQuoteChar ) return eTriNo;
				ths->fBufferPtr += bytesPerChar;
				return eTriYes;
				break;

		}

	}

}	// CaptureAccess


// =================================================================================================
// RecordTailAttr
// ==============

XMPScanner::PacketMachine::TriState
XMPScanner::PacketMachine::RecordTailAttr ( PacketMachine * ths, const char * /* unused */ )
{
	
	// There are no known "general" attributes for the packet trailer.
	
	ths->fAttrName.erase ( ths->fAttrName.begin(), ths->fAttrName.end() );
	ths->fAttrValue.erase ( ths->fAttrValue.begin(), ths->fAttrValue.end() );

	return eTriYes;


}	// RecordTailAttr


// =================================================================================================
// CheckPacketEnd
// ==============
//
// Check for trailing padding and record the packet length.  We have trailing padding if the bytes
// attribute is present and has a value greater than the current length.

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

	if ( ths->fPosition == 0 ) {	// First call, decide if there is trailing padding.
	
		const XMP_Int64 currLen64 = (ths->fBufferOffset + (ths->fBufferPtr - ths->fBufferOrigin)) - ths->fPacketStart;
		if ( currLen64 > 0x7FFFFFFF ) throw std::runtime_error ( "Packet length exceeds 2GB-1" );
		const XMP_Int32 currLength = (XMP_Int32)currLen64;
		
		if ( (ths->fBytesAttr != -1) && (ths->fBytesAttr != currLength) ) {
			if ( ths->fBytesAttr < currLength ) {
				ths->fBogusPacket = true;	// The bytes attribute value is too small.
			} else {
				ths->fPosition = ths->fBytesAttr - currLength;
				if ( (ths->fPosition % ths->fBytesPerChar) != 0 ) {
					ths->fBogusPacket = true;	// The padding is not a multiple of the character size.
					ths->fPosition = (ths->fPosition / ths->fBytesPerChar) * ths->fBytesPerChar;
				}
			}
		}
		
	}
	
	while ( ths->fPosition > 0 ) {

		if ( ths->fBufferPtr >= ths->fBufferLimit ) return eTriMaybe;

		const char currChar = *ths->fBufferPtr;

		if ( (currChar != ' ') && (currChar != '\t') && (currChar != '\n') && (currChar != '\r') ) {
			ths->fBogusPacket = true;	// The padding is not whitespace.
			break;						// Stop the packet here.
		}

		ths->fPosition -= bytesPerChar;
		ths->fBufferPtr += bytesPerChar;

	}
	
	const XMP_Int64 currLen64 = (ths->fBufferOffset + (ths->fBufferPtr - ths->fBufferOrigin)) - ths->fPacketStart;
	if ( currLen64 > 0x7FFFFFFF ) throw std::runtime_error ( "Packet length exceeds 2GB-1" );
	ths->fPacketLength = (XMP_Int32)currLen64;
	return eTriYes;

}	// CheckPacketEnd


// =================================================================================================
// CheckFinalNulls
// ===============
//
// Do some special case processing for little endian characters.  We have to make sure the presumed
// nulls after the last character actually exist, i.e. that the stream does not end too soon.  Note
// that the prior character scanning has moved the buffer pointer to the address following the last
// byte of the last character.  I.e. we're already past the presumed nulls, so we can't check their
// content.  All we can do is verify that the stream does not end too soon.
//
// Doing this check is simple yet subtle.  If we're still in the current buffer then the trailing
// bytes obviously exist.  If we're exactly at the end of the buffer then the bytes also exist.
// The only question is when we're actually past this buffer, partly into the next buffer.  This is
// when "ths->fBufferPtr > ths->fBufferLimit" on entry.  For that case we have to wait until we've
// actually seen enough extra bytes of input.
//
// Since the normal buffer processing is already adjusting for this partial character overrun, all
// that needs to be done here is wait until "ths->fBufferPtr <= ths->fBufferLimit" on entry.  In
// other words, if we're presently too far, ths->fBufferPtr will be adjusted by the amount of the
// overflow the next time XMPScanner::Scan is called.  This might still be too far, so just keep
// waiting for enough data to pass by.
//
// Note that there is a corresponding special case for big endian characters, we must decrement the
// starting offset by the number of leading nulls.  But we don't do that here, we leave it to the
// outer code.  This is because the leading nulls might have been at the exact end of a previous
// buffer, in which case we have to also decrement the length of that raw data snip.

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

	if ( (ths->fCharForm != eChar8Bit) && CharFormIsLittleEndian ( ths->fCharForm ) ) {
		if ( ths->fBufferPtr > ths->fBufferLimit ) return eTriMaybe;
	}
	
	return eTriYes;

}	// CheckFinalNulls


// =================================================================================================
// SetNextRecognizer
// =================

void
XMPScanner::PacketMachine::SetNextRecognizer ( RecognizerKind nextRecognizer )
{

	fRecognizer = nextRecognizer;
	fPosition = 0;
	
}	// SetNextRecognizer


// =================================================================================================
// FindNextPacket
// ==============

// *** When we start validating intervening nulls for 2 and 4 bytes characters, throw an exception
// *** for errors.  Don't return eTriNo, that might skip at an optional point.

XMPScanner::PacketMachine::TriState
XMPScanner::PacketMachine::FindNextPacket ()
{

	TriState	status;
		
	#define kPacketHead		"?xpacket begin="
	#define kPacketID		"W5M0MpCehiHzreSzNTczkc9d"
	#define kPacketTail		"?xpacket end="
	
	static const RecognizerInfo	recognizerTable [eRecognizerCount]	= {		// ! Would be safer to assign these explicitly.

		// proc				successNext					failureNext					literal

		{ NULL,				eFailureRecognizer,			eFailureRecognizer,			NULL},			// eFailureRecognizer
		{ NULL,				eSuccessRecognizer,			eSuccessRecognizer,			NULL},			// eSuccessRecognizer

		{ FindLessThan,		eHeadStartRecorder,			eFailureRecognizer,			"H" },			// eLeadInRecognizer
		{ RecordStart,	 	eHeadStartRecognizer,		eLeadInRecognizer,			NULL },			// eHeadStartRecorder
		{ MatchString, 		eBOMRecognizer,				eLeadInRecognizer,			kPacketHead },	// eHeadStartRecognizer

		{ RecognizeBOM, 	eIDTagRecognizer,			eLeadInRecognizer,			NULL },			// eBOMRecognizer

		{ MatchString, 		eIDOpenRecognizer,			eLeadInRecognizer,			" id=" },		// eIDTagRecognizer
		{ MatchOpenQuote,	eIDValueRecognizer,			eLeadInRecognizer,			NULL },			// eIDOpenRecognizer
		{ MatchString, 		eIDCloseRecognizer,			eLeadInRecognizer,			kPacketID },	// eIDValueRecognizer
		{ MatchCloseQuote,	eAttrSpaceRecognizer_1,		eLeadInRecognizer,			NULL },			// eIDCloseRecognizer

		{ MatchChar, 		eAttrNameRecognizer_1,		eHeadEndRecognizer,			" " },			// eAttrSpaceRecognizer_1
		{ CaptureAttrName,	eAttrValueRecognizer_1,		eLeadInRecognizer,			NULL },			// eAttrNameRecognizer_1
		{ CaptureAttrValue,	eAttrValueRecorder_1,		eLeadInRecognizer,			NULL },			// eAttrValueRecognizer_1
		{ RecordHeadAttr,	eAttrSpaceRecognizer_1,		eLeadInRecognizer,			NULL },			// eAttrValueRecorder_1

		{ MatchString, 		eBodyRecognizer,			eLeadInRecognizer,			"?>" },			// eHeadEndRecognizer
		
		{ FindLessThan,		eTailStartRecognizer,		eBodyRecognizer,			"T"},			// eBodyRecognizer

		{ MatchString, 		eAccessValueRecognizer,		eBodyRecognizer,			kPacketTail },	// eTailStartRecognizer
		{ CaptureAccess,	eAttrSpaceRecognizer_2,		eBodyRecognizer,			NULL },			// eAccessValueRecognizer

		{ MatchChar, 		eAttrNameRecognizer_2,		eTailEndRecognizer,			" " },			// eAttrSpaceRecognizer_2
		{ CaptureAttrName,	eAttrValueRecognizer_2,		eBodyRecognizer,			NULL },			// eAttrNameRecognizer_2
		{ CaptureAttrValue,	eAttrValueRecorder_2,		eBodyRecognizer,			NULL },			// eAttrValueRecognizer_2
		{ RecordTailAttr,	eAttrSpaceRecognizer_2,		eBodyRecognizer,			NULL },			// eAttrValueRecorder_2

		{ MatchString, 		ePacketEndRecognizer,		eBodyRecognizer,			"?>" },			// eTailEndRecognizer
		{ CheckPacketEnd,	eCloseOutRecognizer,		eBodyRecognizer,			"" },			// ePacketEndRecognizer
		{ CheckFinalNulls,	eSuccessRecognizer,			eBodyRecognizer,			"" }			// eCloseOutRecognizer

	};
	
	while ( true ) {
	
		switch ( fRecognizer ) {
		
			case eFailureRecognizer :
				return eTriNo;
			
			case eSuccessRecognizer :
				return eTriYes;
			
			default :
			
				// -------------------------------------------------------------------
				// For everything else, the normal cases, use the state machine table.
				
				const RecognizerInfo *	thisState	= &recognizerTable [fRecognizer];
				
				status = thisState->proc ( this, thisState->literal );
				
				switch ( status ) {
				
					case eTriNo :
						SetNextRecognizer ( thisState->failureNext );
						continue;
					
					case eTriYes :
						SetNextRecognizer ( thisState->successNext );
						continue;
					
					case eTriMaybe :
						fBufferOverrun = fBufferPtr - fBufferLimit;
						return eTriMaybe;	// Keep this recognizer intact, to be resumed later.
				
				}
		
		}	// switch ( fRecognizer ) { ...
		
	}	// while ( true ) { ...

}	// FindNextPacket


// =================================================================================================
// =================================================================================================
// class InternalSnip
// ==================


// =================================================================================================
// InternalSnip
// ============

XMPScanner::InternalSnip::InternalSnip ( XMP_Int64 offset, XMP_Int64 length )
{

	fInfo.fOffset = offset;
	fInfo.fLength = length;
	
}	// InternalSnip


// =================================================================================================
// InternalSnip
// ============

XMPScanner::InternalSnip::InternalSnip ( const InternalSnip & rhs ) :
	fInfo ( rhs.fInfo ),
	fMachine ( NULL )
{

	assert ( rhs.fMachine.get() == NULL );	// Don't copy a snip with a machine.
	assert ( (rhs.fInfo.fEncodingAttr == 0) || (*rhs.fInfo.fEncodingAttr == 0) ); // Don't copy a snip with an encoding.

}	// InternalSnip


// =================================================================================================
// ~InternalSnip
// =============

XMPScanner::InternalSnip::~InternalSnip ()
{
}	// ~InternalSnip



// =================================================================================================
// =================================================================================================
// class XMPScanner
// ================


// =================================================================================================
// DumpSnipList
// ============

#if DEBUG

static const char *	snipStateName [6] = { "not-seen", "pending", "raw-data", "good-packet", "partial", "bad-packet" };

void
XMPScanner::DumpSnipList ( const char * title )
{
	InternalSnipIterator currPos = fInternalSnips.begin();
	InternalSnipIterator endPos  = fInternalSnips.end();
	
	cout << endl << title << " snip list: " << fInternalSnips.size() << endl;
	
	for ( ; currPos != endPos; ++currPos ) {
		SnipInfo * currSnip = &currPos->fInfo;
		cout << '\t' << currSnip << ' ' << snipStateName[currSnip->fState] << ' '
		     << currSnip->fOffset << ".." << (currSnip->fOffset + currSnip->fLength - 1)
			 << ' ' << currSnip->fLength << ' ' << endl;
	}
}	// DumpSnipList

#endif


// =================================================================================================
// PrevSnip and NextSnip
// =====================

XMPScanner::InternalSnipIterator
XMPScanner::PrevSnip ( InternalSnipIterator snipPos )
{

	InternalSnipIterator prev = snipPos;
	return --prev;

}	// PrevSnip

XMPScanner::InternalSnipIterator
XMPScanner::NextSnip ( InternalSnipIterator snipPos )
{

	InternalSnipIterator next = snipPos;
	return ++next;

}	// NextSnip


// =================================================================================================
// XMPScanner
// ==========
//
// Initialize the scanner object with one "not seen" snip covering the whole stream.

XMPScanner::XMPScanner ( XMP_Int64 streamLength ) :

	fStreamLength ( streamLength )
	
{
	InternalSnip	rootSnip ( 0, streamLength );
	
	if ( streamLength > 0 ) fInternalSnips.push_front ( rootSnip );		// Be nice for empty files.
	// DumpSnipList ( "New XMPScanner" );
	
}	// XMPScanner


// =================================================================================================
// ~XMPScanner
// ===========

XMPScanner::~XMPScanner()
{
	
}	// ~XMPScanner