long File::find(const ByteVector &pattern, long fromOffset, const ByteVector &before)
{
if(!d->stream || pattern.size() > d->bufferSize)
return -1;
// The position in the file that the current buffer starts at.
long bufferOffset = fromOffset;
ByteVector buffer;
// These variables are used to keep track of a partial match that happens at
// the end of a buffer.
int previousPartialMatch = -1;
int beforePreviousPartialMatch = -1;
// Save the location of the current read pointer. We will restore the
// position using seek() before all returns.
long originalPosition = tell();
// Start the search at the offset.
seek(fromOffset);
// This loop is the crux of the find method. There are three cases that we
// want to account for:
//
// (1) The previously searched buffer contained a partial match of the search
// pattern and we want to see if the next one starts with the remainder of
// that pattern.
//
// (2) The search pattern is wholly contained within the current buffer.
//
// (3) The current buffer ends with a partial match of the pattern. We will
// note this for use in the next itteration, where we will check for the rest
// of the pattern.
//
// All three of these are done in two steps. First we check for the pattern
// and do things appropriately if a match (or partial match) is found. We
// then check for "before". The order is important because it gives priority
// to "real" matches.
for(buffer = readBlock(d->bufferSize); buffer.size() > 0; buffer = readBlock(d->bufferSize)) {
// (1) previous partial match
if(previousPartialMatch >= 0 && int(d->bufferSize) > previousPartialMatch) {
const int patternOffset = (d->bufferSize - previousPartialMatch);
if(buffer.containsAt(pattern, 0, patternOffset)) {
seek(originalPosition);
return bufferOffset - d->bufferSize + previousPartialMatch;
}
}
if(!before.isNull() && beforePreviousPartialMatch >= 0 && int(d->bufferSize) > beforePreviousPartialMatch) {
const int beforeOffset = (d->bufferSize - beforePreviousPartialMatch);
if(buffer.containsAt(before, 0, beforeOffset)) {
seek(originalPosition);
return -1;
}
}
// (2) pattern contained in current buffer
long location = buffer.find(pattern);
if(location >= 0) {
seek(originalPosition);
return bufferOffset + location;
}
if(!before.isNull() && buffer.find(before) >= 0) {
seek(originalPosition);
return -1;
}
// (3) partial match
previousPartialMatch = buffer.endsWithPartialMatch(pattern);
if(!before.isNull())
beforePreviousPartialMatch = buffer.endsWithPartialMatch(before);
bufferOffset += d->bufferSize;
}
// Since we hit the end of the file, reset the status before continuing.
clear();
seek(originalPosition);
return -1;
}
long MPEG::File::findID3v2()
{
// This method is based on the contents of TagLib::File::find(), but because
// of some subtlteies -- specifically the need to look for the bit pattern of
// an MPEG sync, it has been modified for use here.
if(isValid() && ID3v2::Header::fileIdentifier().size() <= bufferSize()) {
// The position in the file that the current buffer starts at.
long bufferOffset = 0;
ByteVector buffer;
// These variables are used to keep track of a partial match that happens at
// the end of a buffer.
int previousPartialMatch = -1;
bool previousPartialSynchMatch = false;
// Save the location of the current read pointer. We will restore the
// position using seek() before all returns.
long originalPosition = tell();
// Start the search at the beginning of the file.
seek(0);
// This loop is the crux of the find method. There are three cases that we
// want to account for:
// (1) The previously searched buffer contained a partial match of the search
// pattern and we want to see if the next one starts with the remainder of
// that pattern.
//
// (2) The search pattern is wholly contained within the current buffer.
//
// (3) The current buffer ends with a partial match of the pattern. We will
// note this for use in the next itteration, where we will check for the rest
// of the pattern.
for(buffer = readBlock(bufferSize()); buffer.size() > 0; buffer = readBlock(bufferSize())) {
// (1) previous partial match
if(previousPartialSynchMatch && secondSynchByte(buffer[0]))
return -1;
if(previousPartialMatch >= 0 && int(bufferSize()) > previousPartialMatch) {
const int patternOffset = (bufferSize() - previousPartialMatch);
if(buffer.containsAt(ID3v2::Header::fileIdentifier(), 0, patternOffset)) {
seek(originalPosition);
return bufferOffset - bufferSize() + previousPartialMatch;
}
}
// (2) pattern contained in current buffer
long location = buffer.find(ID3v2::Header::fileIdentifier());
if(location >= 0) {
seek(originalPosition);
return bufferOffset + location;
}
int firstSynchByte = buffer.find(char(uchar(255)));
// Here we have to loop because there could be several of the first
// (11111111) byte, and we want to check all such instances until we find
// a full match (11111111 111) or hit the end of the buffer.
while(firstSynchByte >= 0) {
// if this *is not* at the end of the buffer
if(firstSynchByte < int(buffer.size()) - 1) {
if(secondSynchByte(buffer[firstSynchByte + 1])) {
// We've found the frame synch pattern.
seek(originalPosition);
return -1;
}
else {
// We found 11111111 at the end of the current buffer indicating a
// partial match of the synch pattern. The find() below should
// return -1 and break out of the loop.
previousPartialSynchMatch = true;
}
}
// Check in the rest of the buffer.
firstSynchByte = buffer.find(char(uchar(255)), firstSynchByte + 1);
}
// (3) partial match
previousPartialMatch = buffer.endsWithPartialMatch(ID3v2::Header::fileIdentifier());
bufferOffset += bufferSize();
}
// Since we hit the end of the file, reset the status before continuing.
clear();
seek(originalPosition);
}
return -1;
}
