// Copyright (c) 2010 The WebM project authors. All Rights Reserved.
//
// Use of this source code is governed by a BSD-style license
// that can be found in the LICENSE file in the root of the source
// tree. An additional intellectual property rights grant can be found
// in the file PATENTS. All contributing project authors may
// be found in the AUTHORS file in the root of the source tree.
#include "mkvparser.hpp"
#include <cassert>
#include <cstring>
#include <new>
#include <climits>
mkvparser::IMkvReader::~IMkvReader()
{
}
void mkvparser::GetVersion(int& major, int& minor, int& build, int& revision)
{
major = 1;
minor = 0;
build = 0;
revision = 24;
}
long long mkvparser::ReadUInt(IMkvReader* pReader, long long pos, long& len)
{
assert(pReader);
assert(pos >= 0);
int status;
//#ifdef _DEBUG
// long long total, available;
// status = pReader->Length(&total, &available);
// assert(status >= 0);
// assert((total < 0) || (available <= total));
// assert(pos < available);
// assert((available - pos) >= 1); //assume here max u-int len is 8
//#endif
len = 1;
unsigned char b;
status = pReader->Read(pos, 1, &b);
if (status < 0) //error or underflow
return status;
if (status > 0) //interpreted as "underflow"
return E_BUFFER_NOT_FULL;
if (b == 0) //we can't handle u-int values larger than 8 bytes
return E_FILE_FORMAT_INVALID;
unsigned char m = 0x80;
while (!(b & m))
{
m >>= 1;
++len;
}
//#ifdef _DEBUG
// assert((available - pos) >= len);
//#endif
long long result = b & (~m);
++pos;
for (int i = 1; i < len; ++i)
{
status = pReader->Read(pos, 1, &b);
if (status < 0)
{
len = 1;
return status;
}
if (status > 0)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
result <<= 8;
result |= b;
++pos;
}
return result;
}
long long mkvparser::GetUIntLength(
IMkvReader* pReader,
long long pos,
long& len)
{
assert(pReader);
assert(pos >= 0);
long long total, available;
int status = pReader->Length(&total, &available);
assert(status >= 0);
assert((total < 0) || (available <= total));
len = 1;
if (pos >= available)
return pos; //too few bytes available
unsigned char b;
status = pReader->Read(pos, 1, &b);
if (status < 0)
return status;
assert(status == 0);
if (b == 0) //we can't handle u-int values larger than 8 bytes
return E_FILE_FORMAT_INVALID;
unsigned char m = 0x80;
while (!(b & m))
{
m >>= 1;
++len;
}
return 0; //success
}
long long mkvparser::UnserializeUInt(
IMkvReader* pReader,
long long pos,
long long size)
{
assert(pReader);
assert(pos >= 0);
if ((size <= 0) || (size > 8))
return E_FILE_FORMAT_INVALID;
long long result = 0;
for (long long i = 0; i < size; ++i)
{
unsigned char b;
const long status = pReader->Read(pos, 1, &b);
if (status < 0)
return status;
result <<= 8;
result |= b;
++pos;
}
return result;
}
long mkvparser::UnserializeFloat(
IMkvReader* pReader,
long long pos,
long long size_,
double& result)
{
assert(pReader);
assert(pos >= 0);
if ((size_ != 4) && (size_ != 8))
return E_FILE_FORMAT_INVALID;
const long size = static_cast<long>(size_);
unsigned char buf[8];
const int status = pReader->Read(pos, size, buf);
if (status < 0) //error
return status;
if (size == 4)
{
union
{
float f;
unsigned long ff;
};
ff = 0;
for (int i = 0;;)
{
ff |= buf[i];
if (++i >= 4)
break;
ff <<= 8;
}
result = f;
}
else
{
assert(size == 8);
union
{
double d;
unsigned long long dd;
};
dd = 0;
for (int i = 0;;)
{
dd |= buf[i];
if (++i >= 8)
break;
dd <<= 8;
}
result = d;
}
return 0;
}
long mkvparser::UnserializeInt(
IMkvReader* pReader,
long long pos,
long size,
long long& result)
{
assert(pReader);
assert(pos >= 0);
assert(size > 0);
assert(size <= 8);
{
signed char b;
const long status = pReader->Read(pos, 1, (unsigned char*)&b);
if (status < 0)
return status;
result = b;
++pos;
}
for (long i = 1; i < size; ++i)
{
unsigned char b;
const long status = pReader->Read(pos, 1, &b);
if (status < 0)
return status;
result <<= 8;
result |= b;
++pos;
}
return 0; //success
}
long mkvparser::UnserializeString(
IMkvReader* pReader,
long long pos,
long long size_,
char*& str)
{
delete[] str;
str = NULL;
if (size_ >= LONG_MAX) //we need (size+1) chars
return E_FILE_FORMAT_INVALID;
const long size = static_cast<long>(size_);
str = new (std::nothrow) char[size+1];
if (str == NULL)
return -1;
unsigned char* const buf = reinterpret_cast<unsigned char*>(str);
const long status = pReader->Read(pos, size, buf);
if (status)
{
delete[] str;
str = NULL;
return status;
}
str[size] = '\0';
return 0; //success
}
long mkvparser::ParseElementHeader(
IMkvReader* pReader,
long long& pos,
long long stop,
long long& id,
long long& size)
{
if ((stop >= 0) && (pos >= stop))
return E_FILE_FORMAT_INVALID;
long len;
id = ReadUInt(pReader, pos, len);
if (id <= 0)
return E_FILE_FORMAT_INVALID;
pos += len; //consume id
if ((stop >= 0) && (pos >= stop))
return E_FILE_FORMAT_INVALID;
size = ReadUInt(pReader, pos, len);
if (size < 0)
return E_FILE_FORMAT_INVALID;
pos += len; //consume length of size
//pos now designates payload
if ((stop >= 0) && ((pos + size) > stop))
return E_FILE_FORMAT_INVALID;
return 0; //success
}
bool mkvparser::Match(
IMkvReader* pReader,
long long& pos,
unsigned long id_,
long long& val)
{
assert(pReader);
assert(pos >= 0);
long long total, available;
const long status = pReader->Length(&total, &available);
assert(status >= 0);
assert((total < 0) || (available <= total));
long len;
const long long id = ReadUInt(pReader, pos, len);
assert(id >= 0);
assert(len > 0);
assert(len <= 8);
assert((pos + len) <= available);
if ((unsigned long)id != id_)
return false;
pos += len; //consume id
const long long size = ReadUInt(pReader, pos, len);
assert(size >= 0);
assert(size <= 8);
assert(len > 0);
assert(len <= 8);
assert((pos + len) <= available);
pos += len; //consume length of size of payload
val = UnserializeUInt(pReader, pos, size);
assert(val >= 0);
pos += size; //consume size of payload
return true;
}
bool mkvparser::Match(
IMkvReader* pReader,
long long& pos,
unsigned long id_,
unsigned char*& buf,
size_t& buflen)
{
assert(pReader);
assert(pos >= 0);
long long total, available;
long status = pReader->Length(&total, &available);
assert(status >= 0);
assert((total < 0) || (available <= total));
long len;
const long long id = ReadUInt(pReader, pos, len);
assert(id >= 0);
assert(len > 0);
assert(len <= 8);
assert((pos + len) <= available);
if ((unsigned long)id != id_)
return false;
pos += len; //consume id
const long long size_ = ReadUInt(pReader, pos, len);
assert(size_ >= 0);
assert(len > 0);
assert(len <= 8);
assert((pos + len) <= available);
pos += len; //consume length of size of payload
assert((pos + size_) <= available);
const long buflen_ = static_cast<long>(size_);
buf = new (std::nothrow) unsigned char[buflen_];
assert(buf); //TODO
status = pReader->Read(pos, buflen_, buf);
assert(status == 0); //TODO
buflen = buflen_;
pos += size_; //consume size of payload
return true;
}
namespace mkvparser
{
EBMLHeader::EBMLHeader() :
m_docType(NULL)
{
Init();
}
EBMLHeader::~EBMLHeader()
{
delete[] m_docType;
}
void EBMLHeader::Init()
{
m_version = 1;
m_readVersion = 1;
m_maxIdLength = 4;
m_maxSizeLength = 8;
if (m_docType)
{
delete[] m_docType;
m_docType = NULL;
}
m_docTypeVersion = 1;
m_docTypeReadVersion = 1;
}
long long EBMLHeader::Parse(
IMkvReader* pReader,
long long& pos)
{
assert(pReader);
long long total, available;
long status = pReader->Length(&total, &available);
if (status < 0) //error
return status;
pos = 0;
long long end = (available >= 1024) ? 1024 : available;
for (;;)
{
unsigned char b = 0;
while (pos < end)
{
status = pReader->Read(pos, 1, &b);
if (status < 0) //error
return status;
if (b == 0x1A)
break;
++pos;
}
if (b != 0x1A)
{
if (pos >= 1024)
return E_FILE_FORMAT_INVALID; //don't bother looking anymore
if ((total >= 0) && ((total - available) < 5))
return E_FILE_FORMAT_INVALID;
return available + 5; //5 = 4-byte ID + 1st byte of size
}
if ((total >= 0) && ((total - pos) < 5))
return E_FILE_FORMAT_INVALID;
if ((available - pos) < 5)
return pos + 5; //try again later
long len;
const long long result = ReadUInt(pReader, pos, len);
if (result < 0) //error
return result;
if (result == 0x0A45DFA3) //EBML Header ID
{
pos += len; //consume ID
break;
}
++pos; //throw away just the 0x1A byte, and try again
}
//pos designates start of size field
//get length of size field
long len;
long long result = GetUIntLength(pReader, pos, len);
if (result < 0) //error
return result;
if (result > 0) //need more data
return result;
assert(len > 0);
assert(len <= 8);
if ((total >= 0) && ((total - pos) < len))
return E_FILE_FORMAT_INVALID;
if ((available - pos) < len)
return pos + len; //try again later
//get the EBML header size
result = ReadUInt(pReader, pos, len);
if (result < 0) //error
return result;
pos += len; //consume size field
//pos now designates start of payload
if ((total >= 0) && ((total - pos) < result))
return E_FILE_FORMAT_INVALID;
if ((available - pos) < result)
return pos + result;
end = pos + result;
Init();
while (pos < end)
{
long long id, size;
status = ParseElementHeader(
pReader,
pos,
end,
id,
size);
if (status < 0) //error
return status;
if (size == 0) //weird
return E_FILE_FORMAT_INVALID;
if (id == 0x0286) //version
{
m_version = UnserializeUInt(pReader, pos, size);
if (m_version <= 0)
return E_FILE_FORMAT_INVALID;
}
else if (id == 0x02F7) //read version
{
m_readVersion = UnserializeUInt(pReader, pos, size);
if (m_readVersion <= 0)
return E_FILE_FORMAT_INVALID;
}
else if (id == 0x02F2) //max id length
{
m_maxIdLength = UnserializeUInt(pReader, pos, size);
if (m_maxIdLength <= 0)
return E_FILE_FORMAT_INVALID;
}
else if (id == 0x02F3) //max size length
{
m_maxSizeLength = UnserializeUInt(pReader, pos, size);
if (m_maxSizeLength <= 0)
return E_FILE_FORMAT_INVALID;
}
else if (id == 0x0282) //doctype
{
if (m_docType)
return E_FILE_FORMAT_INVALID;
status = UnserializeString(pReader, pos, size, m_docType);
if (status) //error
return status;
}
else if (id == 0x0287) //doctype version
{
m_docTypeVersion = UnserializeUInt(pReader, pos, size);
if (m_docTypeVersion <= 0)
return E_FILE_FORMAT_INVALID;
}
else if (id == 0x0285) //doctype read version
{
m_docTypeReadVersion = UnserializeUInt(pReader, pos, size);
if (m_docTypeReadVersion <= 0)
return E_FILE_FORMAT_INVALID;
}
pos += size;
}
assert(pos == end);
return 0;
}
Segment::Segment(
IMkvReader* pReader,
long long elem_start,
//long long elem_size,
long long start,
long long size) :
m_pReader(pReader),
m_element_start(elem_start),
//m_element_size(elem_size),
m_start(start),
m_size(size),
m_pos(start),
m_pUnknownSize(0),
m_pSeekHead(NULL),
m_pInfo(NULL),
m_pTracks(NULL),
m_pCues(NULL),
m_clusters(NULL),
m_clusterCount(0),
m_clusterPreloadCount(0),
m_clusterSize(0)
{
}
Segment::~Segment()
{
const long count = m_clusterCount + m_clusterPreloadCount;
Cluster** i = m_clusters;
Cluster** j = m_clusters + count;
while (i != j)
{
Cluster* const p = *i++;
assert(p);
delete p;
}
delete[] m_clusters;
delete m_pTracks;
delete m_pInfo;
delete m_pCues;
delete m_pSeekHead;
}
long long Segment::CreateInstance(
IMkvReader* pReader,
long long pos,
Segment*& pSegment)
{
assert(pReader);
assert(pos >= 0);
pSegment = NULL;
long long total, available;
const long status = pReader->Length(&total, &available);
if (status < 0) //error
return status;
if (available < 0)
return -1;
if ((total >= 0) && (available > total))
return -1;
const long long end = (total >= 0) ? total : available;
//TODO: this might need to be liberalized
//I would assume that in practice this loop would execute
//exactly once, but we allow for other elements (e.g. Void)
//to immediately follow the EBML header. This is fine for
//the source filter case (since the entire file is available),
//but in the splitter case over a network we should probably
//just give up early. We could for example decide only to
//execute this loop a maximum of, say, 10 times.
//TODO:
//There is an implied "give up early" by only parsing up
//to the available limit. We do do that, but only if the
//total file size is unknown. We could decide to always
//use what's available as our limit (irrespective of whether
//we happen to know the total file length). This would have
//as its sense "parse this much of the file before giving up",
//which a slightly different sense from "try to parse up to
//10 EMBL elements before giving up".
while (pos < end)
{
//Read ID
long len;
long long result = GetUIntLength(pReader, pos, len);
if (result) //error, or too few available bytes
return result;
if ((pos + len) > end)
return E_FILE_FORMAT_INVALID;
if ((pos + len) > available)
return pos + len;
const long long idpos = pos;
const long long id = ReadUInt(pReader, pos, len);
if (id < 0) //error
return id;
pos += len; //consume ID
//Read Size
result = GetUIntLength(pReader, pos, len);
if (result) //error, or too few available bytes
return result;
if ((pos + len) > end)
return E_FILE_FORMAT_INVALID;
if ((pos + len) > available)
return pos + len;
long long size = ReadUInt(pReader, pos, len);
if (size < 0) //error
return size;
pos += len; //consume length of size of element
//Pos now points to start of payload
//Handle "unknown size" for live streaming of webm files.
const long long unknown_size = (1LL << (7 * len)) - 1;
if (id == 0x08538067) //Segment ID
{
if (size == unknown_size)
size = -1;
else if (total < 0)
size = -1;
else if ((pos + size) > total)
size = -1;
pSegment = new (std::nothrow) Segment(
pReader,
idpos,
//elem_size
pos,
size);
if (pSegment == 0)
return -1; //generic error
return 0; //success
}
if (size == unknown_size)
return E_FILE_FORMAT_INVALID;
if ((pos + size) > end)
return E_FILE_FORMAT_INVALID;
pos += size; //consume payload
}
return E_FILE_FORMAT_INVALID; //there is no segment
//TODO: this might need to be liberalized. See comments above.
}
long long Segment::ParseHeaders()
{
//Outermost (level 0) segment object has been constructed,
//and pos designates start of payload. We need to find the
//inner (level 1) elements.
long long total, available;
const int status = m_pReader->Length(&total, &available);
if (status < 0) //error
return status;
assert((total < 0) || (available <= total));
const long long segment_stop = (m_size < 0) ? -1 : m_start + m_size;
assert((segment_stop < 0) || (total < 0) || (segment_stop <= total));
assert((segment_stop < 0) || (m_pos <= segment_stop));
for (;;)
{
if ((total >= 0) && (m_pos >= total))
break;
if ((segment_stop >= 0) && (m_pos >= segment_stop))
break;
long long pos = m_pos;
const long long element_start = pos;
if ((pos + 1) > available)
return (pos + 1);
long len;
long long result = GetUIntLength(m_pReader, pos, len);
if (result < 0) //error
return result;
if (result > 0) //underflow (weird)
return (pos + 1);
if ((segment_stop >= 0) && ((pos + len) > segment_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > available)
return pos + len;
const long long idpos = pos;
const long long id = ReadUInt(m_pReader, idpos, len);
if (id < 0) //error
return id;
if (id == 0x0F43B675) //Cluster ID
break;
pos += len; //consume ID
if ((pos + 1) > available)
return (pos + 1);
//Read Size
result = GetUIntLength(m_pReader, pos, len);
if (result < 0) //error
return result;
if (result > 0) //underflow (weird)
return (pos + 1);
if ((segment_stop >= 0) && ((pos + len) > segment_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > available)
return pos + len;
const long long size = ReadUInt(m_pReader, pos, len);
if (size < 0) //error
return size;
pos += len; //consume length of size of element
const long long element_size = size + pos - element_start;
//Pos now points to start of payload
if ((segment_stop >= 0) && ((pos + size) > segment_stop))
return E_FILE_FORMAT_INVALID;
//We read EBML elements either in total or nothing at all.
if ((pos + size) > available)
return pos + size;
if (id == 0x0549A966) //Segment Info ID
{
if (m_pInfo)
return E_FILE_FORMAT_INVALID;
m_pInfo = new (std::nothrow) SegmentInfo(
this,
pos,
size,
element_start,
element_size);
if (m_pInfo == NULL)
return -1;
const long status = m_pInfo->Parse();
if (status)
return status;
}
else if (id == 0x0654AE6B) //Tracks ID
{
if (m_pTracks)
return E_FILE_FORMAT_INVALID;
m_pTracks = new (std::nothrow) Tracks(this,
pos,
size,
element_start,
element_size);
if (m_pTracks == NULL)
return -1;
const long status = m_pTracks->Parse();
if (status)
return status;
}
else if (id == 0x0C53BB6B) //Cues ID
{
if (m_pCues == NULL)
{
m_pCues = new (std::nothrow) Cues(
this,
pos,
size,
element_start,
element_size);
if (m_pCues == NULL)
return -1;
}
}
else if (id == 0x014D9B74) //SeekHead ID
{
if (m_pSeekHead == NULL)
{
m_pSeekHead = new (std::nothrow) SeekHead(
this,
pos,
size,
element_start,
element_size);
if (m_pSeekHead == NULL)
return -1;
const long status = m_pSeekHead->Parse();
if (status)
return status;
}
}
m_pos = pos + size; //consume payload
}
assert((segment_stop < 0) || (m_pos <= segment_stop));
if (m_pInfo == NULL) //TODO: liberalize this behavior
return E_FILE_FORMAT_INVALID;
if (m_pTracks == NULL)
return E_FILE_FORMAT_INVALID;
return 0; //success
}
long Segment::LoadCluster(
long long& pos,
long& len)
{
for (;;)
{
const long result = DoLoadCluster(pos, len);
if (result <= 1)
return result;
}
}
long Segment::DoLoadCluster(
long long& pos,
long& len)
{
if (m_pos < 0)
return DoLoadClusterUnknownSize(pos, len);
long long total, avail;
long status = m_pReader->Length(&total, &avail);
if (status < 0) //error
return status;
assert((total < 0) || (avail <= total));
const long long segment_stop = (m_size < 0) ? -1 : m_start + m_size;
long long cluster_off = -1; //offset relative to start of segment
long long cluster_size = -1; //size of cluster payload
for (;;)
{
if ((total >= 0) && (m_pos >= total))
return 1; //no more clusters
if ((segment_stop >= 0) && (m_pos >= segment_stop))
return 1; //no more clusters
pos = m_pos;
//Read ID
if ((pos + 1) > avail)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
long long result = GetUIntLength(m_pReader, pos, len);
if (result < 0) //error
return static_cast<long>(result);
if (result > 0) //weird
return E_BUFFER_NOT_FULL;
if ((segment_stop >= 0) && ((pos + len) > segment_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
return E_BUFFER_NOT_FULL;
const long long idpos = pos;
const long long id = ReadUInt(m_pReader, idpos, len);
if (id < 0) //error (or underflow)
return static_cast<long>(id);
pos += len; //consume ID
//Read Size
if ((pos + 1) > avail)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
result = GetUIntLength(m_pReader, pos, len);
if (result < 0) //error
return static_cast<long>(result);
if (result > 0) //weird
return E_BUFFER_NOT_FULL;
if ((segment_stop >= 0) && ((pos + len) > segment_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
return E_BUFFER_NOT_FULL;
const long long size = ReadUInt(m_pReader, pos, len);
if (size < 0) //error
return static_cast<long>(size);
pos += len; //consume length of size of element
//pos now points to start of payload
if (size == 0) //weird
{
m_pos = pos;
continue;
}
const long long unknown_size = (1LL << (7 * len)) - 1;
#if 0 //we must handle this to support live webm
if (size == unknown_size)
return E_FILE_FORMAT_INVALID; //TODO: allow this
#endif
if ((segment_stop >= 0) &&
(size != unknown_size) &&
((pos + size) > segment_stop))
{
return E_FILE_FORMAT_INVALID;
}
#if 0 //commented-out, to support incremental cluster parsing
len = static_cast<long>(size);
if ((pos + size) > avail)
return E_BUFFER_NOT_FULL;
#endif
if (id == 0x0C53BB6B) //Cues ID
{
if (size == unknown_size)
return E_FILE_FORMAT_INVALID; //TODO: liberalize
if (m_pCues == NULL)
{
const long long element_size = (pos - idpos) + size;
m_pCues = new Cues(this,
pos,
size,
idpos,
element_size);
assert(m_pCues); //TODO
}
m_pos = pos + size; //consume payload
continue;
}
if (id != 0x0F43B675) //Cluster ID
{
if (size == unknown_size)
return E_FILE_FORMAT_INVALID; //TODO: liberalize
m_pos = pos + size; //consume payload
continue;
}
//We have a cluster.
cluster_off = idpos - m_start; //relative pos
if (size != unknown_size)
cluster_size = size;
break;
}
assert(cluster_off >= 0); //have cluster
long long pos_;
long len_;
status = Cluster::HasBlockEntries(this, cluster_off, pos_, len_);
if (status < 0) //error, or underflow
{
pos = pos_;
len = len_;
return status;
}
//status == 0 means "no block entries found"
//status > 0 means "found at least one block entry"
//TODO:
//The issue here is that the segment increments its own
//pos ptr past the most recent cluster parsed, and then
//starts from there to parse the next cluster. If we
//don't know the size of the current cluster, then we
//must either parse its payload (as we do below), looking
//for the cluster (or cues) ID to terminate the parse.
//This isn't really what we want: rather, we really need
//a way to create the curr cluster object immediately.
//The pity is that cluster::parse can determine its own
//boundary, and we largely duplicate that same logic here.
//
//Maybe we need to get rid of our look-ahead preloading
//in source::parse???
//
//As we're parsing the blocks in the curr cluster
//(in cluster::parse), we should have some way to signal
//to the segment that we have determined the boundary,
//so it can adjust its own segment::m_pos member.
//
//The problem is that we're asserting in asyncreadinit,
//because we adjust the pos down to the curr seek pos,
//and the resulting adjusted len is > 2GB. I'm suspicious
//that this is even correct, but even if it is, we can't
//be loading that much data in the cache anyway.
const long idx = m_clusterCount;
if (m_clusterPreloadCount > 0)
{
assert(idx < m_clusterSize);
Cluster* const pCluster = m_clusters[idx];
assert(pCluster);
assert(pCluster->m_index < 0);
const long long off = pCluster->GetPosition();
assert(off >= 0);
if (off == cluster_off) //preloaded already
{
if (status == 0) //no entries found
return E_FILE_FORMAT_INVALID;
if (cluster_size >= 0)
pos += cluster_size;
else
{
const long long element_size = pCluster->GetElementSize();
if (element_size <= 0)
return E_FILE_FORMAT_INVALID; //TODO: handle this case
pos = pCluster->m_element_start + element_size;
}
pCluster->m_index = idx; //move from preloaded to loaded
++m_clusterCount;
--m_clusterPreloadCount;
m_pos = pos; //consume payload
assert((segment_stop < 0) || (m_pos <= segment_stop));
return 0; //success
}
}
if (status == 0) //no entries found
{
if (cluster_size < 0)
return E_FILE_FORMAT_INVALID; //TODO: handle this
pos += cluster_size;
if ((total >= 0) && (pos >= total))
{
m_pos = total;
return 1; //no more clusters
}
if ((segment_stop >= 0) && (pos >= segment_stop))
{
m_pos = segment_stop;
return 1; //no more clusters
}
m_pos = pos;
return 2; //try again
}
//status > 0 means we have an entry
Cluster* const pCluster = Cluster::Create(this,
idx,
cluster_off);
//element_size);
assert(pCluster);
AppendCluster(pCluster);
assert(m_clusters);
assert(idx < m_clusterSize);
assert(m_clusters[idx] == pCluster);
if (cluster_size >= 0)
{
pos += cluster_size;
m_pos = pos;
assert((segment_stop < 0) || (m_pos <= segment_stop));
return 0;
}
m_pUnknownSize = pCluster;
m_pos = -pos;
return 0; //partial success, since we have a new cluster
//status == 0 means "no block entries found"
//pos designates start of payload
//m_pos has NOT been adjusted yet (in case we need to come back here)
#if 0
if (cluster_size < 0) //unknown size
{
const long long payload_pos = pos; //absolute pos of cluster payload
for (;;) //determine cluster size
{
if ((total >= 0) && (pos >= total))
break;
if ((segment_stop >= 0) && (pos >= segment_stop))
break; //no more clusters
//Read ID
if ((pos + 1) > avail)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
long long result = GetUIntLength(m_pReader, pos, len);
if (result < 0) //error
return static_cast<long>(result);
if (result > 0) //weird
return E_BUFFER_NOT_FULL;
if ((segment_stop >= 0) && ((pos + len) > segment_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
return E_BUFFER_NOT_FULL;
const long long idpos = pos;
const long long id = ReadUInt(m_pReader, idpos, len);
if (id < 0) //error (or underflow)
return static_cast<long>(id);
//This is the distinguished set of ID's we use to determine
//that we have exhausted the sub-element's inside the cluster
//whose ID we parsed earlier.
if (id == 0x0F43B675) //Cluster ID
break;
if (id == 0x0C53BB6B) //Cues ID
break;
switch (id)
{
case 0x20: //BlockGroup
case 0x23: //Simple Block
case 0x67: //TimeCode
case 0x2B: //PrevSize
break;
default:
assert(false);
break;
}
pos += len; //consume ID (of sub-element)
//Read Size
if ((pos + 1) > avail)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
result = GetUIntLength(m_pReader, pos, len);
if (result < 0) //error
return static_cast<long>(result);
if (result > 0) //weird
return E_BUFFER_NOT_FULL;
if ((segment_stop >= 0) && ((pos + len) > segment_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
return E_BUFFER_NOT_FULL;
const long long size = ReadUInt(m_pReader, pos, len);
if (size < 0) //error
return static_cast<long>(size);
pos += len; //consume size field of element
//pos now points to start of sub-element's payload
if (size == 0) //weird
continue;
const long long unknown_size = (1LL << (7 * len)) - 1;
if (size == unknown_size)
return E_FILE_FORMAT_INVALID; //not allowed for sub-elements
if ((segment_stop >= 0) && ((pos + size) > segment_stop)) //weird
return E_FILE_FORMAT_INVALID;
pos += size; //consume payload of sub-element
assert((segment_stop < 0) || (pos <= segment_stop));
} //determine cluster size
cluster_size = pos - payload_pos;
assert(cluster_size >= 0);
pos = payload_pos; //reset and re-parse original cluster
}
if (m_clusterPreloadCount > 0)
{
assert(idx < m_clusterSize);
Cluster* const pCluster = m_clusters[idx];
assert(pCluster);
assert(pCluster->m_index < 0);
const long long off = pCluster->GetPosition();
assert(off >= 0);
if (off == cluster_off) //preloaded already
return E_FILE_FORMAT_INVALID; //subtle
}
m_pos = pos + cluster_size; //consume payload
assert((segment_stop < 0) || (m_pos <= segment_stop));
return 2; //try to find another cluster
#endif
}
long Segment::DoLoadClusterUnknownSize(
long long& pos,
long& len)
{
assert(m_pos < 0);
assert(m_pUnknownSize);
#if 0
assert(m_pUnknownSize->GetElementSize() < 0); //TODO: verify this
const long long element_start = m_pUnknownSize->m_element_start;
pos = -m_pos;
assert(pos > element_start);
//We have already consumed the (cluster) ID and size fields.
//We just need to consume the blocks and other sub-elements
//of this cluster, until we discover the boundary.
long long total, avail;
long status = m_pReader->Length(&total, &avail);
if (status < 0) //error
return status;
assert((total < 0) || (avail <= total));
const long long segment_stop = (m_size < 0) ? -1 : m_start + m_size;
long long element_size = -1;
for (;;) //determine cluster size
{
if ((total >= 0) && (pos >= total))
{
element_size = total - element_start;
assert(element_size > 0);
break;
}
if ((segment_stop >= 0) && (pos >= segment_stop))
{
element_size = segment_stop - element_start;
assert(element_size > 0);
break;
}
//Read ID
if ((pos + 1) > avail)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
long long result = GetUIntLength(m_pReader, pos, len);
if (result < 0) //error
return static_cast<long>(result);
if (result > 0) //weird
return E_BUFFER_NOT_FULL;
if ((segment_stop >= 0) && ((pos + len) > segment_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
return E_BUFFER_NOT_FULL;
const long long idpos = pos;
const long long id = ReadUInt(m_pReader, idpos, len);
if (id < 0) //error (or underflow)
return static_cast<long>(id);
//This is the distinguished set of ID's we use to determine
//that we have exhausted the sub-element's inside the cluster
//whose ID we parsed earlier.
if ((id == 0x0F43B675) || (id == 0x0C53BB6B)) //Cluster ID or Cues ID
{
element_size = pos - element_start;
assert(element_size > 0);
break;
}
#ifdef _DEBUG
switch (id)
{
case 0x20: //BlockGroup
case 0x23: //Simple Block
case 0x67: //TimeCode
case 0x2B: //PrevSize
break;
default:
assert(false);
break;
}
#endif
pos += len; //consume ID (of sub-element)
//Read Size
if ((pos + 1) > avail)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
result = GetUIntLength(m_pReader, pos, len);
if (result < 0) //error
return static_cast<long>(result);
if (result > 0) //weird
return E_BUFFER_NOT_FULL;
if ((segment_stop >= 0) && ((pos + len) > segment_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
return E_BUFFER_NOT_FULL;
const long long size = ReadUInt(m_pReader, pos, len);
if (size < 0) //error
return static_cast<long>(size);
pos += len; //consume size field of element
//pos now points to start of sub-element's payload
if (size == 0) //weird
continue;
const long long unknown_size = (1LL << (7 * len)) - 1;
if (size == unknown_size)
return E_FILE_FORMAT_INVALID; //not allowed for sub-elements
if ((segment_stop >= 0) && ((pos + size) > segment_stop)) //weird
return E_FILE_FORMAT_INVALID;
pos += size; //consume payload of sub-element
assert((segment_stop < 0) || (pos <= segment_stop));
} //determine cluster size
assert(element_size >= 0);
m_pos = element_start + element_size;
m_pUnknownSize = 0;
return 2; //continue parsing
#else
const long status = m_pUnknownSize->Parse(pos, len);
if (status < 0) //error or underflow
return status;
if (status == 0) //parsed a block
return 2; //continue parsing
assert(status > 0); //nothing left to parse of this cluster
const long long start = m_pUnknownSize->m_element_start;
const long long size = m_pUnknownSize->GetElementSize();
assert(size >= 0);
pos = start + size;
m_pos = pos;
m_pUnknownSize = 0;
return 2; //continue parsing
#endif
}
void Segment::AppendCluster(Cluster* pCluster)
{
assert(pCluster);
assert(pCluster->m_index >= 0);
const long count = m_clusterCount + m_clusterPreloadCount;
long& size = m_clusterSize;
assert(size >= count);
const long idx = pCluster->m_index;
assert(idx == m_clusterCount);
if (count >= size)
{
const long n = (size <= 0) ? 2048 : 2*size;
Cluster** const qq = new Cluster*[n];
Cluster** q = qq;
Cluster** p = m_clusters;
Cluster** const pp = p + count;
while (p != pp)
*q++ = *p++;
delete[] m_clusters;
m_clusters = qq;
size = n;
}
if (m_clusterPreloadCount > 0)
{
assert(m_clusters);
Cluster** const p = m_clusters + m_clusterCount;
assert(*p);
assert((*p)->m_index < 0);
Cluster** q = p + m_clusterPreloadCount;
assert(q < (m_clusters + size));
for (;;)
{
Cluster** const qq = q - 1;
assert((*qq)->m_index < 0);
*q = *qq;
q = qq;
if (q == p)
break;
}
}
m_clusters[idx] = pCluster;
++m_clusterCount;
}
void Segment::PreloadCluster(Cluster* pCluster, ptrdiff_t idx)
{
assert(pCluster);
assert(pCluster->m_index < 0);
assert(idx >= m_clusterCount);
const long count = m_clusterCount + m_clusterPreloadCount;
long& size = m_clusterSize;
assert(size >= count);
if (count >= size)
{
const long n = (size <= 0) ? 2048 : 2*size;
Cluster** const qq = new Cluster*[n];
Cluster** q = qq;
Cluster** p = m_clusters;
Cluster** const pp = p + count;
while (p != pp)
*q++ = *p++;
delete[] m_clusters;
m_clusters = qq;
size = n;
}
assert(m_clusters);
Cluster** const p = m_clusters + idx;
Cluster** q = m_clusters + count;
assert(q >= p);
assert(q < (m_clusters + size));
while (q > p)
{
Cluster** const qq = q - 1;
assert((*qq)->m_index < 0);
*q = *qq;
q = qq;
}
m_clusters[idx] = pCluster;
++m_clusterPreloadCount;
}
long Segment::Load()
{
assert(m_clusters == NULL);
assert(m_clusterSize == 0);
assert(m_clusterCount == 0);
//assert(m_size >= 0);
//Outermost (level 0) segment object has been constructed,
//and pos designates start of payload. We need to find the
//inner (level 1) elements.
const long long header_status = ParseHeaders();
if (header_status < 0) //error
return static_cast<long>(header_status);
if (header_status > 0) //underflow
return E_BUFFER_NOT_FULL;
assert(m_pInfo);
assert(m_pTracks);
for (;;)
{
const int status = LoadCluster();
if (status < 0) //error
return status;
if (status >= 1) //no more clusters
return 0;
}
}
SeekHead::SeekHead(
Segment* pSegment,
long long start,
long long size_,
long long element_start,
long long element_size) :
m_pSegment(pSegment),
m_start(start),
m_size(size_),
m_element_start(element_start),
m_element_size(element_size),
m_entries(0),
m_entry_count(0),
m_void_elements(0),
m_void_element_count(0)
{
}
SeekHead::~SeekHead()
{
delete[] m_entries;
delete[] m_void_elements;
}
long SeekHead::Parse()
{
IMkvReader* const pReader = m_pSegment->m_pReader;
long long pos = m_start;
const long long stop = m_start + m_size;
//first count the seek head entries
int entry_count = 0;
int void_element_count = 0;
while (pos < stop)
{
long long id, size;
const long status = ParseElementHeader(
pReader,
pos,
stop,
id,
size);
if (status < 0) //error
return status;
if (id == 0x0DBB) //SeekEntry ID
++entry_count;
else if (id == 0x6C) //Void ID
++void_element_count;
pos += size; //consume payload
assert(pos <= stop);
}
assert(pos == stop);
m_entries = new (std::nothrow) Entry[entry_count];
if (m_entries == NULL)
return -1;
m_void_elements = new (std::nothrow) VoidElement[void_element_count];
if (m_void_elements == NULL)
return -1;
//now parse the entries and void elements
Entry* pEntry = m_entries;
VoidElement* pVoidElement = m_void_elements;
pos = m_start;
while (pos < stop)
{
const long long idpos = pos;
long long id, size;
const long status = ParseElementHeader(
pReader,
pos,
stop,
id,
size);
if (status < 0) //error
return status;
if (id == 0x0DBB) //SeekEntry ID
{
if (ParseEntry(pReader, pos, size, pEntry))
{
Entry& e = *pEntry++;
e.element_start = idpos;
e.element_size = (pos + size) - idpos;
}
}
else if (id == 0x6C) //Void ID
{
VoidElement& e = *pVoidElement++;
e.element_start = idpos;
e.element_size = (pos + size) - idpos;
}
pos += size; //consume payload
assert(pos <= stop);
}
assert(pos == stop);
ptrdiff_t count_ = ptrdiff_t(pEntry - m_entries);
assert(count_ >= 0);
assert(count_ <= entry_count);
m_entry_count = static_cast<int>(count_);
count_ = ptrdiff_t(pVoidElement - m_void_elements);
assert(count_ >= 0);
assert(count_ <= void_element_count);
m_void_element_count = static_cast<int>(count_);
return 0;
}
int SeekHead::GetCount() const
{
return m_entry_count;
}
const SeekHead::Entry* SeekHead::GetEntry(int idx) const
{
if (idx < 0)
return 0;
if (idx >= m_entry_count)
return 0;
return m_entries + idx;
}
int SeekHead::GetVoidElementCount() const
{
return m_void_element_count;
}
const SeekHead::VoidElement* SeekHead::GetVoidElement(int idx) const
{
if (idx < 0)
return 0;
if (idx >= m_void_element_count)
return 0;
return m_void_elements + idx;
}
#if 0
void Segment::ParseCues(long long off)
{
if (m_pCues)
return;
//odbgstream os;
//os << "Segment::ParseCues (begin)" << endl;
long long pos = m_start + off;
const long long element_start = pos;
const long long stop = m_start + m_size;
long len;
long long result = GetUIntLength(m_pReader, pos, len);
assert(result == 0);
assert((pos + len) <= stop);
const long long idpos = pos;
const long long id = ReadUInt(m_pReader, idpos, len);
assert(id == 0x0C53BB6B); //Cues ID
pos += len; //consume ID
assert(pos < stop);
//Read Size
result = GetUIntLength(m_pReader, pos, len);
assert(result == 0);
assert((pos + len) <= stop);
const long long size = ReadUInt(m_pReader, pos, len);
assert(size >= 0);
pos += len; //consume length of size of element
assert((pos + size) <= stop);
const long long element_size = size + pos - element_start;
//Pos now points to start of payload
m_pCues = new Cues(this, pos, size, element_start, element_size);
assert(m_pCues); //TODO
//os << "Segment::ParseCues (end)" << endl;
}
#else
long Segment::ParseCues(
long long off,
long long& pos,
long& len)
{
if (m_pCues)
return 0; //success
if (off < 0)
return -1;
long long total, avail;
const int status = m_pReader->Length(&total, &avail);
if (status < 0) //error
return status;
assert((total < 0) || (avail <= total));
pos = m_start + off;
if ((total < 0) || (pos >= total))
return 1; //don't bother parsing cues
const long long element_start = pos;
const long long segment_stop = (m_size < 0) ? -1 : m_start + m_size;
if ((pos + 1) > avail)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
long long result = GetUIntLength(m_pReader, pos, len);
if (result < 0) //error
return static_cast<long>(result);
if (result > 0) //underflow (weird)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
if ((segment_stop >= 0) && ((pos + len) > segment_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
return E_BUFFER_NOT_FULL;
const long long idpos = pos;
const long long id = ReadUInt(m_pReader, idpos, len);
if (id != 0x0C53BB6B) //Cues ID
return E_FILE_FORMAT_INVALID;
pos += len; //consume ID
assert((segment_stop < 0) || (pos <= segment_stop));
//Read Size
if ((pos + 1) > avail)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
result = GetUIntLength(m_pReader, pos, len);
if (result < 0) //error
return static_cast<long>(result);
if (result > 0) //underflow (weird)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
if ((segment_stop >= 0) && ((pos + len) > segment_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
return E_BUFFER_NOT_FULL;
const long long size = ReadUInt(m_pReader, pos, len);
if (size < 0) //error
return static_cast<long>(size);
if (size == 0) //weird, although technically not illegal
return 1; //done
pos += len; //consume length of size of element
assert((segment_stop < 0) || (pos <= segment_stop));
//Pos now points to start of payload
const long long element_stop = pos + size;
if ((segment_stop >= 0) && (element_stop > segment_stop))
return E_FILE_FORMAT_INVALID;
if ((total >= 0) && (element_stop > total))
return 1; //don't bother parsing anymore
len = static_cast<long>(size);
if (element_stop > avail)
return E_BUFFER_NOT_FULL;
const long long element_size = element_stop - element_start;
m_pCues = new (std::nothrow) Cues(
this,
pos,
size,
element_start,
element_size);
assert(m_pCues); //TODO
return 0; //success
}
#endif
#if 0
void Segment::ParseSeekEntry(
long long start,
long long size_)
{
long long pos = start;
const long long stop = start + size_;
long len;
const long long seekIdId = ReadUInt(m_pReader, pos, len);
//seekIdId;
assert(seekIdId == 0x13AB); //SeekID ID
assert((pos + len) <= stop);
pos += len; //consume id
const long long seekIdSize = ReadUInt(m_pReader, pos, len);
assert(seekIdSize >= 0);
assert((pos + len) <= stop);
pos += len; //consume size
const long long seekId = ReadUInt(m_pReader, pos, len); //payload
assert(seekId >= 0);
assert(len == seekIdSize);
assert((pos + len) <= stop);
pos += seekIdSize; //consume payload
const long long seekPosId = ReadUInt(m_pReader, pos, len);
//seekPosId;
assert(seekPosId == 0x13AC); //SeekPos ID
assert((pos + len) <= stop);
pos += len; //consume id
const long long seekPosSize = ReadUInt(m_pReader, pos, len);
assert(seekPosSize >= 0);
assert((pos + len) <= stop);
pos += len; //consume size
assert((pos + seekPosSize) <= stop);
const long long seekOff = UnserializeUInt(m_pReader, pos, seekPosSize);
assert(seekOff >= 0);
assert(seekOff < m_size);
pos += seekPosSize; //consume payload
assert(pos == stop);
const long long seekPos = m_start + seekOff;
assert(seekPos < (m_start + m_size));
if (seekId == 0x0C53BB6B) //Cues ID
ParseCues(seekOff);
}
#else
bool SeekHead::ParseEntry(
IMkvReader* pReader,
long long start,
long long size_,
Entry* pEntry)
{
if (size_ <= 0)
return false;
long long pos = start;
const long long stop = start + size_;
long len;
//parse the container for the level-1 element ID
const long long seekIdId = ReadUInt(pReader, pos, len);
//seekIdId;
if (seekIdId != 0x13AB) //SeekID ID
return false;
if ((pos + len) > stop)
return false;
pos += len; //consume SeekID id
const long long seekIdSize = ReadUInt(pReader, pos, len);
if (seekIdSize <= 0)
return false;
if ((pos + len) > stop)
return false;
pos += len; //consume size of field
if ((pos + seekIdSize) > stop)
return false;
//Note that the SeekId payload really is serialized
//as a "Matroska integer", not as a plain binary value.
//In fact, Matroska requires that ID values in the
//stream exactly match the binary representation as listed
//in the Matroska specification.
//
//This parser is more liberal, and permits IDs to have
//any width. (This could make the representation in the stream
//different from what's in the spec, but it doesn't matter here,
//since we always normalize "Matroska integer" values.)
pEntry->id = ReadUInt(pReader, pos, len); //payload
if (pEntry->id <= 0)
return false;
if (len != seekIdSize)
return false;
pos += seekIdSize; //consume SeekID payload
const long long seekPosId = ReadUInt(pReader, pos, len);
if (seekPosId != 0x13AC) //SeekPos ID
return false;
if ((pos + len) > stop)
return false;
pos += len; //consume id
const long long seekPosSize = ReadUInt(pReader, pos, len);
if (seekPosSize <= 0)
return false;
if ((pos + len) > stop)
return false;
pos += len; //consume size
if ((pos + seekPosSize) > stop)
return false;
pEntry->pos = UnserializeUInt(pReader, pos, seekPosSize);
if (pEntry->pos < 0)
return false;
pos += seekPosSize; //consume payload
if (pos != stop)
return false;
return true;
}
#endif
Cues::Cues(
Segment* pSegment,
long long start_,
long long size_,
long long element_start,
long long element_size) :
m_pSegment(pSegment),
m_start(start_),
m_size(size_),
m_element_start(element_start),
m_element_size(element_size),
m_cue_points(NULL),
m_count(0),
m_preload_count(0),
m_pos(start_)
{
}
Cues::~Cues()
{
const long n = m_count + m_preload_count;
CuePoint** p = m_cue_points;
CuePoint** const q = p + n;
while (p != q)
{
CuePoint* const pCP = *p++;
assert(pCP);
delete pCP;
}
delete[] m_cue_points;
}
long Cues::GetCount() const
{
if (m_cue_points == NULL)
return -1;
return m_count; //TODO: really ignore preload count?
}
bool Cues::DoneParsing() const
{
const long long stop = m_start + m_size;
return (m_pos >= stop);
}
void Cues::Init() const
{
if (m_cue_points)
return;
assert(m_count == 0);
assert(m_preload_count == 0);
IMkvReader* const pReader = m_pSegment->m_pReader;
const long long stop = m_start + m_size;
long long pos = m_start;
long cue_points_size = 0;
while (pos < stop)
{
const long long idpos = pos;
long len;
const long long id = ReadUInt(pReader, pos, len);
assert(id >= 0); //TODO
assert((pos + len) <= stop);
pos += len; //consume ID
const long long size = ReadUInt(pReader, pos, len);
assert(size >= 0);
assert((pos + len) <= stop);
pos += len; //consume Size field
assert((pos + size) <= stop);
if (id == 0x3B) //CuePoint ID
PreloadCuePoint(cue_points_size, idpos);
pos += size; //consume payload
assert(pos <= stop);
}
}
void Cues::PreloadCuePoint(
long& cue_points_size,
long long pos) const
{
assert(m_count == 0);
if (m_preload_count >= cue_points_size)
{
const long n = (cue_points_size <= 0) ? 2048 : 2*cue_points_size;
CuePoint** const qq = new CuePoint*[n];
CuePoint** q = qq; //beginning of target
CuePoint** p = m_cue_points; //beginning of source
CuePoint** const pp = p + m_preload_count; //end of source
while (p != pp)
*q++ = *p++;
delete[] m_cue_points;
m_cue_points = qq;
cue_points_size = n;
}
CuePoint* const pCP = new CuePoint(m_preload_count, pos);
m_cue_points[m_preload_count++] = pCP;
}
bool Cues::LoadCuePoint() const
{
//odbgstream os;
//os << "Cues::LoadCuePoint" << endl;
const long long stop = m_start + m_size;
if (m_pos >= stop)
return false; //nothing else to do
Init();
IMkvReader* const pReader = m_pSegment->m_pReader;
while (m_pos < stop)
{
const long long idpos = m_pos;
long len;
const long long id = ReadUInt(pReader, m_pos, len);
assert(id >= 0); //TODO
assert((m_pos + len) <= stop);
m_pos += len; //consume ID
const long long size = ReadUInt(pReader, m_pos, len);
assert(size >= 0);
assert((m_pos + len) <= stop);
m_pos += len; //consume Size field
assert((m_pos + size) <= stop);
if (id != 0x3B) //CuePoint ID
{
m_pos += size; //consume payload
assert(m_pos <= stop);
continue;
}
assert(m_preload_count > 0);
CuePoint* const pCP = m_cue_points[m_count];
assert(pCP);
assert((pCP->GetTimeCode() >= 0) || (-pCP->GetTimeCode() == idpos));
pCP->Load(pReader);
++m_count;
--m_preload_count;
m_pos += size; //consume payload
assert(m_pos <= stop);
return true; //yes, we loaded a cue point
}
//return (m_pos < stop);
return false; //no, we did not load a cue point
}
bool Cues::Find(
long long time_ns,
const Track* pTrack,
const CuePoint*& pCP,
const CuePoint::TrackPosition*& pTP) const
{
assert(time_ns >= 0);
assert(pTrack);
#if 0
LoadCuePoint(); //establish invariant
assert(m_cue_points);
assert(m_count > 0);
CuePoint** const ii = m_cue_points;
CuePoint** i = ii;
CuePoint** const jj = ii + m_count + m_preload_count;
CuePoint** j = jj;
pCP = *i;
assert(pCP);
if (time_ns <= pCP->GetTime(m_pSegment))
{
pTP = pCP->Find(pTrack);
return (pTP != NULL);
}
IMkvReader* const pReader = m_pSegment->m_pReader;
while (i < j)
{
//INVARIANT:
//[ii, i) <= time_ns
//[i, j) ?
//[j, jj) > time_ns
CuePoint** const k = i + (j - i) / 2;
assert(k < jj);
CuePoint* const pCP = *k;
assert(pCP);
pCP->Load(pReader);
const long long t = pCP->GetTime(m_pSegment);
if (t <= time_ns)
i = k + 1;
else
j = k;
assert(i <= j);
}
assert(i == j);
assert(i <= jj);
assert(i > ii);
pCP = *--i;
assert(pCP);
assert(pCP->GetTime(m_pSegment) <= time_ns);
#else
if (m_cue_points == NULL)
return false;
if (m_count == 0)
return false;
CuePoint** const ii = m_cue_points;
CuePoint** i = ii;
CuePoint** const jj = ii + m_count;
CuePoint** j = jj;
pCP = *i;
assert(pCP);
if (time_ns <= pCP->GetTime(m_pSegment))
{
pTP = pCP->Find(pTrack);
return (pTP != NULL);
}
while (i < j)
{
//INVARIANT:
//[ii, i) <= time_ns
//[i, j) ?
//[j, jj) > time_ns
CuePoint** const k = i + (j - i) / 2;
assert(k < jj);
CuePoint* const pCP = *k;
assert(pCP);
const long long t = pCP->GetTime(m_pSegment);
if (t <= time_ns)
i = k + 1;
else
j = k;
assert(i <= j);
}
assert(i == j);
assert(i <= jj);
assert(i > ii);
pCP = *--i;
assert(pCP);
assert(pCP->GetTime(m_pSegment) <= time_ns);
#endif
//It's probably not correct to search for the cue point with this
//time, and then search for a matching track. In principle, the
//matching track could be on some earlier cue point. So we recheck
//the earlier cue point if the track doesn't match.
long index = pCP->GetIndex();
while (index >= 0) {
pTP = pCP->Find(pTrack);
if (pTP == NULL)
pCP = m_cue_points[index--];
else
break;
}
return (pTP != NULL);
}
#if 0
bool Cues::FindNext(
long long time_ns,
const Track* pTrack,
const CuePoint*& pCP,
const CuePoint::TrackPosition*& pTP) const
{
pCP = 0;
pTP = 0;
if (m_count == 0)
return false;
assert(m_cue_points);
const CuePoint* const* const ii = m_cue_points;
const CuePoint* const* i = ii;
const CuePoint* const* const jj = ii + m_count;
const CuePoint* const* j = jj;
while (i < j)
{
//INVARIANT:
//[ii, i) <= time_ns
//[i, j) ?
//[j, jj) > time_ns
const CuePoint* const* const k = i + (j - i) / 2;
assert(k < jj);
pCP = *k;
assert(pCP);
const long long t = pCP->GetTime(m_pSegment);
if (t <= time_ns)
i = k + 1;
else
j = k;
assert(i <= j);
}
assert(i == j);
assert(i <= jj);
if (i >= jj) //time_ns is greater than max cue point
return false;
pCP = *i;
assert(pCP);
assert(pCP->GetTime(m_pSegment) > time_ns);
pTP = pCP->Find(pTrack);
return (pTP != NULL);
}
#endif
const CuePoint* Cues::GetFirst() const
{
if (m_cue_points == NULL)
return NULL;
if (m_count == 0)
return NULL;
#if 0
LoadCuePoint(); //init cues
const size_t count = m_count + m_preload_count;
if (count == 0) //weird
return NULL;
#endif
CuePoint* const* const pp = m_cue_points;
assert(pp);
CuePoint* const pCP = pp[0];
assert(pCP);
assert(pCP->GetTimeCode() >= 0);
return pCP;
}
const CuePoint* Cues::GetLast() const
{
if (m_cue_points == NULL)
return NULL;
if (m_count <= 0)
return NULL;
#if 0
LoadCuePoint(); //init cues
const size_t count = m_count + m_preload_count;
if (count == 0) //weird
return NULL;
const size_t index = count - 1;
CuePoint* const* const pp = m_cue_points;
assert(pp);
CuePoint* const pCP = pp[index];
assert(pCP);
pCP->Load(m_pSegment->m_pReader);
assert(pCP->GetTimeCode() >= 0);
#else
const long index = m_count - 1;
CuePoint* const* const pp = m_cue_points;
assert(pp);
CuePoint* const pCP = pp[index];
assert(pCP);
assert(pCP->GetTimeCode() >= 0);
#endif
return pCP;
}
const CuePoint* Cues::GetNext(const CuePoint* pCurr) const
{
if (pCurr == NULL)
return NULL;
assert(pCurr->GetTimeCode() >= 0);
assert(m_cue_points);
assert(m_count >= 1);
#if 0
const size_t count = m_count + m_preload_count;
size_t index = pCurr->m_index;
assert(index < count);
CuePoint* const* const pp = m_cue_points;
assert(pp);
assert(pp[index] == pCurr);
++index;
if (index >= count)
return NULL;
CuePoint* const pNext = pp[index];
assert(pNext);
pNext->Load(m_pSegment->m_pReader);
#else
long index = pCurr->m_index;
assert(index < m_count);
CuePoint* const* const pp = m_cue_points;
assert(pp);
assert(pp[index] == pCurr);
++index;
if (index >= m_count)
return NULL;
CuePoint* const pNext = pp[index];
assert(pNext);
assert(pNext->GetTimeCode() >= 0);
#endif
return pNext;
}
const BlockEntry* Cues::GetBlock(
const CuePoint* pCP,
const CuePoint::TrackPosition* pTP) const
{
if (pCP == NULL)
return NULL;
if (pTP == NULL)
return NULL;
return m_pSegment->GetBlock(*pCP, *pTP);
}
const BlockEntry* Segment::GetBlock(
const CuePoint& cp,
const CuePoint::TrackPosition& tp)
{
Cluster** const ii = m_clusters;
Cluster** i = ii;
const long count = m_clusterCount + m_clusterPreloadCount;
Cluster** const jj = ii + count;
Cluster** j = jj;
while (i < j)
{
//INVARIANT:
//[ii, i) < pTP->m_pos
//[i, j) ?
//[j, jj) > pTP->m_pos
Cluster** const k = i + (j - i) / 2;
assert(k < jj);
Cluster* const pCluster = *k;
assert(pCluster);
//const long long pos_ = pCluster->m_pos;
//assert(pos_);
//const long long pos = pos_ * ((pos_ < 0) ? -1 : 1);
const long long pos = pCluster->GetPosition();
assert(pos >= 0);
if (pos < tp.m_pos)
i = k + 1;
else if (pos > tp.m_pos)
j = k;
else
return pCluster->GetEntry(cp, tp);
}
assert(i == j);
//assert(Cluster::HasBlockEntries(this, tp.m_pos));
Cluster* const pCluster = Cluster::Create(this, -1, tp.m_pos); //, -1);
assert(pCluster);
const ptrdiff_t idx = i - m_clusters;
PreloadCluster(pCluster, idx);
assert(m_clusters);
assert(m_clusterPreloadCount > 0);
assert(m_clusters[idx] == pCluster);
return pCluster->GetEntry(cp, tp);
}
const Cluster* Segment::FindOrPreloadCluster(long long requested_pos)
{
if (requested_pos < 0)
return 0;
Cluster** const ii = m_clusters;
Cluster** i = ii;
const long count = m_clusterCount + m_clusterPreloadCount;
Cluster** const jj = ii + count;
Cluster** j = jj;
while (i < j)
{
//INVARIANT:
//[ii, i) < pTP->m_pos
//[i, j) ?
//[j, jj) > pTP->m_pos
Cluster** const k = i + (j - i) / 2;
assert(k < jj);
Cluster* const pCluster = *k;
assert(pCluster);
//const long long pos_ = pCluster->m_pos;
//assert(pos_);
//const long long pos = pos_ * ((pos_ < 0) ? -1 : 1);
const long long pos = pCluster->GetPosition();
assert(pos >= 0);
if (pos < requested_pos)
i = k + 1;
else if (pos > requested_pos)
j = k;
else
return pCluster;
}
assert(i == j);
//assert(Cluster::HasBlockEntries(this, tp.m_pos));
Cluster* const pCluster = Cluster::Create(
this,
-1,
requested_pos);
//-1);
assert(pCluster);
const ptrdiff_t idx = i - m_clusters;
PreloadCluster(pCluster, idx);
assert(m_clusters);
assert(m_clusterPreloadCount > 0);
assert(m_clusters[idx] == pCluster);
return pCluster;
}
CuePoint::CuePoint(long idx, long long pos) :
m_element_start(0),
m_element_size(0),
m_index(idx),
m_timecode(-1 * pos),
m_track_positions(NULL),
m_track_positions_count(0)
{
assert(pos > 0);
}
CuePoint::~CuePoint()
{
delete[] m_track_positions;
}
void CuePoint::Load(IMkvReader* pReader)
{
//odbgstream os;
//os << "CuePoint::Load(begin): timecode=" << m_timecode << endl;
if (m_timecode >= 0) //already loaded
return;
assert(m_track_positions == NULL);
assert(m_track_positions_count == 0);
long long pos_ = -m_timecode;
const long long element_start = pos_;
long long stop;
{
long len;
const long long id = ReadUInt(pReader, pos_, len);
assert(id == 0x3B); //CuePoint ID
//assert((pos + len) <= stop);
pos_ += len; //consume ID
const long long size = ReadUInt(pReader, pos_, len);
assert(size >= 0);
//assert((pos + len) <= stop);
pos_ += len; //consume Size field
//assert((pos + size) <= stop);
//pos_ now points to start of payload
stop = pos_ + size;
}
const long long element_size = stop - element_start;
long long pos = pos_;
//First count number of track positions
while (pos < stop)
{
long len;
const long long id = ReadUInt(pReader, pos, len);
assert(id >= 0); //TODO
assert((pos + len) <= stop);
pos += len; //consume ID
const long long size = ReadUInt(pReader, pos, len);
assert(size >= 0);
assert((pos + len) <= stop);
pos += len; //consume Size field
assert((pos + size) <= stop);
if (id == 0x33) //CueTime ID
m_timecode = UnserializeUInt(pReader, pos, size);
else if (id == 0x37) //CueTrackPosition(s) ID
++m_track_positions_count;
pos += size; //consume payload
assert(pos <= stop);
}
assert(m_timecode >= 0);
assert(m_track_positions_count > 0);
//os << "CuePoint::Load(cont'd): idpos=" << idpos
// << " timecode=" << m_timecode
// << endl;
m_track_positions = new TrackPosition[m_track_positions_count];
//Now parse track positions
TrackPosition* p = m_track_positions;
pos = pos_;
while (pos < stop)
{
long len;
const long long id = ReadUInt(pReader, pos, len);
assert(id >= 0); //TODO
assert((pos + len) <= stop);
pos += len; //consume ID
const long long size = ReadUInt(pReader, pos, len);
assert(size >= 0);
assert((pos + len) <= stop);
pos += len; //consume Size field
assert((pos + size) <= stop);
if (id == 0x37) //CueTrackPosition(s) ID
{
TrackPosition& tp = *p++;
tp.Parse(pReader, pos, size);
}
pos += size; //consume payload
assert(pos <= stop);
}
assert(size_t(p - m_track_positions) == m_track_positions_count);
m_element_start = element_start;
m_element_size = element_size;
}
void CuePoint::TrackPosition::Parse(
IMkvReader* pReader,
long long start_,
long long size_)
{
const long long stop = start_ + size_;
long long pos = start_;
m_track = -1;
m_pos = -1;
m_block = 1; //default
while (pos < stop)
{
long len;
const long long id = ReadUInt(pReader, pos, len);
assert(id >= 0); //TODO
assert((pos + len) <= stop);
pos += len; //consume ID
const long long size = ReadUInt(pReader, pos, len);
assert(size >= 0);
assert((pos + len) <= stop);
pos += len; //consume Size field
assert((pos + size) <= stop);
if (id == 0x77) //CueTrack ID
m_track = UnserializeUInt(pReader, pos, size);
else if (id == 0x71) //CueClusterPos ID
m_pos = UnserializeUInt(pReader, pos, size);
else if (id == 0x1378) //CueBlockNumber
m_block = UnserializeUInt(pReader, pos, size);
pos += size; //consume payload
assert(pos <= stop);
}
assert(m_pos >= 0);
assert(m_track > 0);
//assert(m_block > 0);
}
const CuePoint::TrackPosition* CuePoint::Find(const Track* pTrack) const
{
assert(pTrack);
const long long n = pTrack->GetNumber();
const TrackPosition* i = m_track_positions;
const TrackPosition* const j = i + m_track_positions_count;
while (i != j)
{
const TrackPosition& p = *i++;
if (p.m_track == n)
return &p;
}
return NULL; //no matching track number found
}
long long CuePoint::GetTimeCode() const
{
return m_timecode;
}
long long CuePoint::GetTime(const Segment* pSegment) const
{
assert(pSegment);
assert(m_timecode >= 0);
const SegmentInfo* const pInfo = pSegment->GetInfo();
assert(pInfo);
const long long scale = pInfo->GetTimeCodeScale();
assert(scale >= 1);
const long long time = scale * m_timecode;
return time;
}
long CuePoint::GetIndex() const
{
return m_index;
}
#if 0
long long Segment::Unparsed() const
{
if (m_size < 0)
return LLONG_MAX;
const long long stop = m_start + m_size;
const long long result = stop - m_pos;
assert(result >= 0);
return result;
}
#else
bool Segment::DoneParsing() const
{
if (m_size < 0)
{
long long total, avail;
const int status = m_pReader->Length(&total, &avail);
if (status < 0) //error
return true; //must assume done
if (total < 0)
return false; //assume live stream
return (m_pos >= total);
}
const long long stop = m_start + m_size;
return (m_pos >= stop);
}
#endif
const Cluster* Segment::GetFirst() const
{
if ((m_clusters == NULL) || (m_clusterCount <= 0))
return &m_eos;
Cluster* const pCluster = m_clusters[0];
assert(pCluster);
return pCluster;
}
const Cluster* Segment::GetLast() const
{
if ((m_clusters == NULL) || (m_clusterCount <= 0))
return &m_eos;
const long idx = m_clusterCount - 1;
Cluster* const pCluster = m_clusters[idx];
assert(pCluster);
return pCluster;
}
unsigned long Segment::GetCount() const
{
return m_clusterCount;
}
const Cluster* Segment::GetNext(const Cluster* pCurr)
{
assert(pCurr);
assert(pCurr != &m_eos);
assert(m_clusters);
long idx = pCurr->m_index;
if (idx >= 0)
{
assert(m_clusterCount > 0);
assert(idx < m_clusterCount);
assert(pCurr == m_clusters[idx]);
++idx;
if (idx >= m_clusterCount)
return &m_eos; //caller will LoadCluster as desired
Cluster* const pNext = m_clusters[idx];
assert(pNext);
assert(pNext->m_index >= 0);
assert(pNext->m_index == idx);
return pNext;
}
assert(m_clusterPreloadCount > 0);
//const long long off_ = pCurr->m_pos;
//const long long off = off_ * ((off_ < 0) ? -1 : 1);
//long long pos = m_start + off;
long long pos = pCurr->m_element_start;
assert(m_size >= 0); //TODO
const long long stop = m_start + m_size; //end of segment
{
long len;
long long result = GetUIntLength(m_pReader, pos, len);
assert(result == 0); //TODO
assert((pos + len) <= stop); //TODO
const long long id = ReadUInt(m_pReader, pos, len);
assert(id == 0x0F43B675); //Cluster ID //TODO
pos += len; //consume ID
//Read Size
result = GetUIntLength(m_pReader, pos, len);
assert(result == 0); //TODO
assert((pos + len) <= stop); //TODO
const long long size = ReadUInt(m_pReader, pos, len);
assert(size > 0); //TODO
//assert((pCurr->m_size <= 0) || (pCurr->m_size == size));
pos += len; //consume length of size of element
assert((pos + size) <= stop); //TODO
//Pos now points to start of payload
pos += size; //consume payload
}
long long off_next = 0;
//long long element_start_next = 0;
long long element_size_next = 0;
while (pos < stop)
{
long len;
long long result = GetUIntLength(m_pReader, pos, len);
assert(result == 0); //TODO
assert((pos + len) <= stop); //TODO
const long long idpos = pos; //pos of next (potential) cluster
const long long id = ReadUInt(m_pReader, idpos, len);
assert(id > 0); //TODO
pos += len; //consume ID
//Read Size
result = GetUIntLength(m_pReader, pos, len);
assert(result == 0); //TODO
assert((pos + len) <= stop); //TODO
const long long size = ReadUInt(m_pReader, pos, len);
assert(size >= 0); //TODO
pos += len; //consume length of size of element
assert((pos + size) <= stop); //TODO
const long long element_size = size + pos - idpos;
//Pos now points to start of payload
if (size == 0) //weird
continue;
if (id == 0x0F43B675) //Cluster ID
{
const long long off_next_ = idpos - m_start;
long long pos_;
long len_;
const long status = Cluster::HasBlockEntries(
this,
off_next_,
pos_,
len_);
assert(status >= 0);
if (status > 0)
{
off_next = off_next_;
//element_start_next = idpos;
element_size_next = element_size;
break;
}
}
pos += size; //consume payload
}
if (off_next <= 0)
return 0;
Cluster** const ii = m_clusters + m_clusterCount;
Cluster** i = ii;
Cluster** const jj = ii + m_clusterPreloadCount;
Cluster** j = jj;
while (i < j)
{
//INVARIANT:
//[0, i) < pos_next
//[i, j) ?
//[j, jj) > pos_next
Cluster** const k = i + (j - i) / 2;
assert(k < jj);
Cluster* const pNext = *k;
assert(pNext);
assert(pNext->m_index < 0);
//const long long pos_ = pNext->m_pos;
//assert(pos_);
//pos = pos_ * ((pos_ < 0) ? -1 : 1);
pos = pNext->GetPosition();
if (pos < off_next)
i = k + 1;
else if (pos > off_next)
j = k;
else
return pNext;
}
assert(i == j);
Cluster* const pNext = Cluster::Create(this,
-1,
off_next);
//element_size_next);
assert(pNext);
const ptrdiff_t idx_next = i - m_clusters; //insertion position
PreloadCluster(pNext, idx_next);
assert(m_clusters);
assert(idx_next < m_clusterSize);
assert(m_clusters[idx_next] == pNext);
return pNext;
}
long Segment::ParseNext(
const Cluster* pCurr,
const Cluster*& pResult,
long long& pos,
long& len)
{
assert(pCurr);
assert(!pCurr->EOS());
assert(m_clusters);
pResult = 0;
if (pCurr->m_index >= 0) //loaded (not merely preloaded)
{
assert(m_clusters[pCurr->m_index] == pCurr);
const long next_idx = pCurr->m_index + 1;
if (next_idx < m_clusterCount)
{
pResult = m_clusters[next_idx];
return 0; //success
}
//curr cluster is last among loaded
const long result = LoadCluster(pos, len);
if (result < 0) //error or underflow
return result;
if (result > 0) //no more clusters
{
//pResult = &m_eos;
return 1;
}
pResult = GetLast();
return 0; //success
}
assert(m_pos > 0);
long long total, avail;
long status = m_pReader->Length(&total, &avail);
if (status < 0) //error
return status;
assert((total < 0) || (avail <= total));
const long long segment_stop = (m_size < 0) ? -1 : m_start + m_size;
//interrogate curr cluster
pos = pCurr->m_element_start;
if (pCurr->m_element_size >= 0)
pos += pCurr->m_element_size;
else
{
if ((pos + 1) > avail)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
long long result = GetUIntLength(m_pReader, pos, len);
if (result < 0) //error
return static_cast<long>(result);
if (result > 0) //weird
return E_BUFFER_NOT_FULL;
if ((segment_stop >= 0) && ((pos + len) > segment_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
return E_BUFFER_NOT_FULL;
const long long id = ReadUInt(m_pReader, pos, len);
if (id != 0x0F43B675) //weird: not Cluster ID
return -1;
pos += len; //consume ID
//Read Size
if ((pos + 1) > avail)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
result = GetUIntLength(m_pReader, pos, len);
if (result < 0) //error
return static_cast<long>(result);
if (result > 0) //weird
return E_BUFFER_NOT_FULL;
if ((segment_stop >= 0) && ((pos + len) > segment_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
return E_BUFFER_NOT_FULL;
const long long size = ReadUInt(m_pReader, pos, len);
if (size < 0) //error
return static_cast<long>(size);
pos += len; //consume size field
const long long unknown_size = (1LL << (7 * len)) - 1;
if (size == unknown_size) //TODO: should never happen
return E_FILE_FORMAT_INVALID; //TODO: resolve this
//assert((pCurr->m_size <= 0) || (pCurr->m_size == size));
if ((segment_stop >= 0) && ((pos + size) > segment_stop))
return E_FILE_FORMAT_INVALID;
//Pos now points to start of payload
pos += size; //consume payload (that is, the current cluster)
assert((segment_stop < 0) || (pos <= segment_stop));
//By consuming the payload, we are assuming that the curr
//cluster isn't interesting. That is, we don't bother checking
//whether the payload of the curr cluster is less than what
//happens to be available (obtained via IMkvReader::Length).
//Presumably the caller has already dispensed with the current
//cluster, and really does want the next cluster.
}
//pos now points to just beyond the last fully-loaded cluster
for (;;)
{
const long status = DoParseNext(pResult, pos, len);
if (status <= 1)
return status;
}
}
long Segment::DoParseNext(
const Cluster*& pResult,
long long& pos,
long& len)
{
long long total, avail;
long status = m_pReader->Length(&total, &avail);
if (status < 0) //error
return status;
assert((total < 0) || (avail <= total));
const long long segment_stop = (m_size < 0) ? -1 : m_start + m_size;
//Parse next cluster. This is strictly a parsing activity.
//Creation of a new cluster object happens later, after the
//parsing is done.
long long off_next = 0;
long long cluster_size = -1;
for (;;)
{
if ((total >= 0) && (pos >= total))
return 1; //EOF
if ((segment_stop >= 0) && (pos >= segment_stop))
return 1; //EOF
if ((pos + 1) > avail)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
long long result = GetUIntLength(m_pReader, pos, len);
if (result < 0) //error
return static_cast<long>(result);
if (result > 0) //weird
return E_BUFFER_NOT_FULL;
if ((segment_stop >= 0) && ((pos + len) > segment_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
return E_BUFFER_NOT_FULL;
const long long idpos = pos; //absolute
const long long idoff = pos - m_start; //relative
const long long id = ReadUInt(m_pReader, idpos, len); //absolute
if (id < 0) //error
return static_cast<long>(id);
if (id == 0) //weird
return -1; //generic error
pos += len; //consume ID
//Read Size
if ((pos + 1) > avail)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
result = GetUIntLength(m_pReader, pos, len);
if (result < 0) //error
return static_cast<long>(result);
if (result > 0) //weird
return E_BUFFER_NOT_FULL;
if ((segment_stop >= 0) && ((pos + len) > segment_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
return E_BUFFER_NOT_FULL;
const long long size = ReadUInt(m_pReader, pos, len);
if (size < 0) //error
return static_cast<long>(size);
pos += len; //consume length of size of element
//Pos now points to start of payload
if (size == 0) //weird
continue;
const long long unknown_size = (1LL << (7 * len)) - 1;
if ((segment_stop >= 0) &&
(size != unknown_size) &&
((pos + size) > segment_stop))
{
return E_FILE_FORMAT_INVALID;
}
if (id == 0x0C53BB6B) //Cues ID
{
if (size == unknown_size)
return E_FILE_FORMAT_INVALID;
const long long element_stop = pos + size;
if ((segment_stop >= 0) && (element_stop > segment_stop))
return E_FILE_FORMAT_INVALID;
const long long element_start = idpos;
const long long element_size = element_stop - element_start;
if (m_pCues == NULL)
{
m_pCues = new Cues(this,
pos,
size,
element_start,
element_size);
assert(m_pCues); //TODO
}
pos += size; //consume payload
assert((segment_stop < 0) || (pos <= segment_stop));
continue;
}
if (id != 0x0F43B675) //not a Cluster ID
{
if (size == unknown_size)
return E_FILE_FORMAT_INVALID;
pos += size; //consume payload
assert((segment_stop < 0) || (pos <= segment_stop));
continue;
}
#if 0 //this is commented-out to support incremental cluster parsing
len = static_cast<long>(size);
if (element_stop > avail)
return E_BUFFER_NOT_FULL;
#endif
//We have a cluster.
off_next = idoff;
if (size != unknown_size)
cluster_size = size;
break;
}
assert(off_next > 0); //have cluster
//We have parsed the next cluster.
//We have not created a cluster object yet. What we need
//to do now is determine whether it has already be preloaded
//(in which case, an object for this cluster has already been
//created), and if not, create a new cluster object.
Cluster** const ii = m_clusters + m_clusterCount;
Cluster** i = ii;
Cluster** const jj = ii + m_clusterPreloadCount;
Cluster** j = jj;
while (i < j)
{
//INVARIANT:
//[0, i) < pos_next
//[i, j) ?
//[j, jj) > pos_next
Cluster** const k = i + (j - i) / 2;
assert(k < jj);
const Cluster* const pNext = *k;
assert(pNext);
assert(pNext->m_index < 0);
pos = pNext->GetPosition();
assert(pos >= 0);
if (pos < off_next)
i = k + 1;
else if (pos > off_next)
j = k;
else
{
pResult = pNext;
return 0; //success
}
}
assert(i == j);
long long pos_;
long len_;
status = Cluster::HasBlockEntries(this, off_next, pos_, len_);
if (status < 0) //error or underflow
{
pos = pos_;
len = len_;
return status;
}
if (status > 0) //means "found at least one block entry"
{
Cluster* const pNext = Cluster::Create(this,
-1, //preloaded
off_next);
//element_size);
assert(pNext);
const ptrdiff_t idx_next = i - m_clusters; //insertion position
PreloadCluster(pNext, idx_next);
assert(m_clusters);
assert(idx_next < m_clusterSize);
assert(m_clusters[idx_next] == pNext);
pResult = pNext;
return 0; //success
}
//status == 0 means "no block entries found"
if (cluster_size < 0) //unknown size
{
const long long payload_pos = pos; //absolute pos of cluster payload
for (;;) //determine cluster size
{
if ((total >= 0) && (pos >= total))
break;
if ((segment_stop >= 0) && (pos >= segment_stop))
break; //no more clusters
//Read ID
if ((pos + 1) > avail)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
long long result = GetUIntLength(m_pReader, pos, len);
if (result < 0) //error
return static_cast<long>(result);
if (result > 0) //weird
return E_BUFFER_NOT_FULL;
if ((segment_stop >= 0) && ((pos + len) > segment_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
return E_BUFFER_NOT_FULL;
const long long idpos = pos;
const long long id = ReadUInt(m_pReader, idpos, len);
if (id < 0) //error (or underflow)
return static_cast<long>(id);
//This is the distinguished set of ID's we use to determine
//that we have exhausted the sub-element's inside the cluster
//whose ID we parsed earlier.
if (id == 0x0F43B675) //Cluster ID
break;
if (id == 0x0C53BB6B) //Cues ID
break;
pos += len; //consume ID (of sub-element)
//Read Size
if ((pos + 1) > avail)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
result = GetUIntLength(m_pReader, pos, len);
if (result < 0) //error
return static_cast<long>(result);
if (result > 0) //weird
return E_BUFFER_NOT_FULL;
if ((segment_stop >= 0) && ((pos + len) > segment_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
return E_BUFFER_NOT_FULL;
const long long size = ReadUInt(m_pReader, pos, len);
if (size < 0) //error
return static_cast<long>(size);
pos += len; //consume size field of element
//pos now points to start of sub-element's payload
if (size == 0) //weird
continue;
const long long unknown_size = (1LL << (7 * len)) - 1;
if (size == unknown_size)
return E_FILE_FORMAT_INVALID; //not allowed for sub-elements
if ((segment_stop >= 0) && ((pos + size) > segment_stop)) //weird
return E_FILE_FORMAT_INVALID;
pos += size; //consume payload of sub-element
assert((segment_stop < 0) || (pos <= segment_stop));
} //determine cluster size
cluster_size = pos - payload_pos;
assert(cluster_size >= 0); //TODO: handle cluster_size = 0
pos = payload_pos; //reset and re-parse original cluster
}
pos += cluster_size; //consume payload
assert((segment_stop < 0) || (pos <= segment_stop));
return 2; //try to find a cluster that follows next
}
const Cluster* Segment::FindCluster(long long time_ns) const
{
if ((m_clusters == NULL) || (m_clusterCount <= 0))
return &m_eos;
{
Cluster* const pCluster = m_clusters[0];
assert(pCluster);
assert(pCluster->m_index == 0);
if (time_ns <= pCluster->GetTime())
return pCluster;
}
//Binary search of cluster array
long i = 0;
long j = m_clusterCount;
while (i < j)
{
//INVARIANT:
//[0, i) <= time_ns
//[i, j) ?
//[j, m_clusterCount) > time_ns
const long k = i + (j - i) / 2;
assert(k < m_clusterCount);
Cluster* const pCluster = m_clusters[k];
assert(pCluster);
assert(pCluster->m_index == k);
const long long t = pCluster->GetTime();
if (t <= time_ns)
i = k + 1;
else
j = k;
assert(i <= j);
}
assert(i == j);
assert(i > 0);
assert(i <= m_clusterCount);
const long k = i - 1;
Cluster* const pCluster = m_clusters[k];
assert(pCluster);
assert(pCluster->m_index == k);
assert(pCluster->GetTime() <= time_ns);
return pCluster;
}
#if 0
const BlockEntry* Segment::Seek(
long long time_ns,
const Track* pTrack) const
{
assert(pTrack);
if ((m_clusters == NULL) || (m_clusterCount <= 0))
return pTrack->GetEOS();
Cluster** const i = m_clusters;
assert(i);
{
Cluster* const pCluster = *i;
assert(pCluster);
assert(pCluster->m_index == 0); //m_clusterCount > 0
assert(pCluster->m_pSegment == this);
if (time_ns <= pCluster->GetTime())
return pCluster->GetEntry(pTrack);
}
Cluster** const j = i + m_clusterCount;
if (pTrack->GetType() == 2) //audio
{
//TODO: we could decide to use cues for this, as we do for video.
//But we only use it for video because looking around for a keyframe
//can get expensive. Audio doesn't require anything special so a
//straight cluster search is good enough (we assume).
Cluster** lo = i;
Cluster** hi = j;
while (lo < hi)
{
//INVARIANT:
//[i, lo) <= time_ns
//[lo, hi) ?
//[hi, j) > time_ns
Cluster** const mid = lo + (hi - lo) / 2;
assert(mid < hi);
Cluster* const pCluster = *mid;
assert(pCluster);
assert(pCluster->m_index == long(mid - m_clusters));
assert(pCluster->m_pSegment == this);
const long long t = pCluster->GetTime();
if (t <= time_ns)
lo = mid + 1;
else
hi = mid;
assert(lo <= hi);
}
assert(lo == hi);
assert(lo > i);
assert(lo <= j);
while (lo > i)
{
Cluster* const pCluster = *--lo;
assert(pCluster);
assert(pCluster->GetTime() <= time_ns);
const BlockEntry* const pBE = pCluster->GetEntry(pTrack);
if ((pBE != 0) && !pBE->EOS())
return pBE;
//landed on empty cluster (no entries)
}
return pTrack->GetEOS(); //weird
}
assert(pTrack->GetType() == 1); //video
Cluster** lo = i;
Cluster** hi = j;
while (lo < hi)
{
//INVARIANT:
//[i, lo) <= time_ns
//[lo, hi) ?
//[hi, j) > time_ns
Cluster** const mid = lo + (hi - lo) / 2;
assert(mid < hi);
Cluster* const pCluster = *mid;
assert(pCluster);
const long long t = pCluster->GetTime();
if (t <= time_ns)
lo = mid + 1;
else
hi = mid;
assert(lo <= hi);
}
assert(lo == hi);
assert(lo > i);
assert(lo <= j);
Cluster* pCluster = *--lo;
assert(pCluster);
assert(pCluster->GetTime() <= time_ns);
{
const BlockEntry* const pBE = pCluster->GetEntry(pTrack, time_ns);
if ((pBE != 0) && !pBE->EOS()) //found a keyframe
return pBE;
}
const VideoTrack* const pVideo = static_cast<const VideoTrack*>(pTrack);
while (lo != i)
{
pCluster = *--lo;
assert(pCluster);
assert(pCluster->GetTime() <= time_ns);
const BlockEntry* const pBlockEntry = pCluster->GetMaxKey(pVideo);
if ((pBlockEntry != 0) && !pBlockEntry->EOS())
return pBlockEntry;
}
//weird: we're on the first cluster, but no keyframe found
//should never happen but we must return something anyway
return pTrack->GetEOS();
}
#endif
#if 0
bool Segment::SearchCues(
long long time_ns,
Track* pTrack,
Cluster*& pCluster,
const BlockEntry*& pBlockEntry,
const CuePoint*& pCP,
const CuePoint::TrackPosition*& pTP)
{
if (pTrack->GetType() != 1) //not video
return false; //TODO: for now, just handle video stream
if (m_pCues == NULL)
return false;
if (!m_pCues->Find(time_ns, pTrack, pCP, pTP))
return false; //weird
assert(pCP);
assert(pTP);
assert(pTP->m_track == pTrack->GetNumber());
//We have the cue point and track position we want,
//so we now need to search for the cluster having
//the indicated position.
return GetCluster(pCP, pTP, pCluster, pBlockEntry);
}
#endif
const Tracks* Segment::GetTracks() const
{
return m_pTracks;
}
const SegmentInfo* Segment::GetInfo() const
{
return m_pInfo;
}
const Cues* Segment::GetCues() const
{
return m_pCues;
}
const SeekHead* Segment::GetSeekHead() const
{
return m_pSeekHead;
}
long long Segment::GetDuration() const
{
assert(m_pInfo);
return m_pInfo->GetDuration();
}
SegmentInfo::SegmentInfo(
Segment* pSegment,
long long start,
long long size_,
long long element_start,
long long element_size) :
m_pSegment(pSegment),
m_start(start),
m_size(size_),
m_element_start(element_start),
m_element_size(element_size),
m_pMuxingAppAsUTF8(NULL),
m_pWritingAppAsUTF8(NULL),
m_pTitleAsUTF8(NULL)
{
}
SegmentInfo::~SegmentInfo()
{
delete[] m_pMuxingAppAsUTF8;
m_pMuxingAppAsUTF8 = NULL;
delete[] m_pWritingAppAsUTF8;
m_pWritingAppAsUTF8 = NULL;
delete[] m_pTitleAsUTF8;
m_pTitleAsUTF8 = NULL;
}
long SegmentInfo::Parse()
{
assert(m_pMuxingAppAsUTF8 == NULL);
assert(m_pWritingAppAsUTF8 == NULL);
assert(m_pTitleAsUTF8 == NULL);
IMkvReader* const pReader = m_pSegment->m_pReader;
long long pos = m_start;
const long long stop = m_start + m_size;
m_timecodeScale = 1000000;
m_duration = -1;
while (pos < stop)
{
long long id, size;
const long status = ParseElementHeader(
pReader,
pos,
stop,
id,
size);
if (status < 0) //error
return status;
if (id == 0x0AD7B1) //Timecode Scale
{
m_timecodeScale = UnserializeUInt(pReader, pos, size);
if (m_timecodeScale <= 0)
return E_FILE_FORMAT_INVALID;
}
else if (id == 0x0489) //Segment duration
{
const long status = UnserializeFloat(
pReader,
pos,
size,
m_duration);
if (status < 0)
return status;
if (m_duration < 0)
return E_FILE_FORMAT_INVALID;
}
else if (id == 0x0D80) //MuxingApp
{
const long status = UnserializeString(
pReader,
pos,
size,
m_pMuxingAppAsUTF8);
if (status)
return status;
}
else if (id == 0x1741) //WritingApp
{
const long status = UnserializeString(
pReader,
pos,
size,
m_pWritingAppAsUTF8);
if (status)
return status;
}
else if (id == 0x3BA9) //Title
{
const long status = UnserializeString(
pReader,
pos,
size,
m_pTitleAsUTF8);
if (status)
return status;
}
pos += size;
assert(pos <= stop);
}
assert(pos == stop);
return 0;
}
long long SegmentInfo::GetTimeCodeScale() const
{
return m_timecodeScale;
}
long long SegmentInfo::GetDuration() const
{
if (m_duration < 0)
return -1;
assert(m_timecodeScale >= 1);
const double dd = double(m_duration) * double(m_timecodeScale);
const long long d = static_cast<long long>(dd);
return d;
}
const char* SegmentInfo::GetMuxingAppAsUTF8() const
{
return m_pMuxingAppAsUTF8;
}
const char* SegmentInfo::GetWritingAppAsUTF8() const
{
return m_pWritingAppAsUTF8;
}
const char* SegmentInfo::GetTitleAsUTF8() const
{
return m_pTitleAsUTF8;
}
///////////////////////////////////////////////////////////////
// ContentEncoding element
ContentEncoding::ContentCompression::ContentCompression()
: algo(0),
settings(NULL) {
}
ContentEncoding::ContentCompression::~ContentCompression() {
delete [] settings;
}
ContentEncoding::ContentEncryption::ContentEncryption()
: algo(0),
key_id(NULL),
key_id_len(0),
signature(NULL),
signature_len(0),
sig_key_id(NULL),
sig_key_id_len(0),
sig_algo(0),
sig_hash_algo(0) {
}
ContentEncoding::ContentEncryption::~ContentEncryption() {
delete [] key_id;
delete [] signature;
delete [] sig_key_id;
}
ContentEncoding::ContentEncoding()
: compression_entries_(NULL),
compression_entries_end_(NULL),
encryption_entries_(NULL),
encryption_entries_end_(NULL),
encoding_order_(0),
encoding_scope_(1),
encoding_type_(0) {
}
ContentEncoding::~ContentEncoding() {
ContentCompression** comp_i = compression_entries_;
ContentCompression** const comp_j = compression_entries_end_;
while (comp_i != comp_j) {
ContentCompression* const comp = *comp_i++;
delete comp;
}
delete [] compression_entries_;
ContentEncryption** enc_i = encryption_entries_;
ContentEncryption** const enc_j = encryption_entries_end_;
while (enc_i != enc_j) {
ContentEncryption* const enc = *enc_i++;
delete enc;
}
delete [] encryption_entries_;
}
const ContentEncoding::ContentCompression*
ContentEncoding::GetCompressionByIndex(unsigned long idx) const {
const ptrdiff_t count = compression_entries_end_ - compression_entries_;
assert(count >= 0);
if (idx >= static_cast<unsigned long>(count))
return NULL;
return compression_entries_[idx];
}
unsigned long ContentEncoding::GetCompressionCount() const {
const ptrdiff_t count = compression_entries_end_ - compression_entries_;
assert(count >= 0);
return static_cast<unsigned long>(count);
}
const ContentEncoding::ContentEncryption*
ContentEncoding::GetEncryptionByIndex(unsigned long idx) const {
const ptrdiff_t count = encryption_entries_end_ - encryption_entries_;
assert(count >= 0);
if (idx >= static_cast<unsigned long>(count))
return NULL;
return encryption_entries_[idx];
}
unsigned long ContentEncoding::GetEncryptionCount() const {
const ptrdiff_t count = encryption_entries_end_ - encryption_entries_;
assert(count >= 0);
return static_cast<unsigned long>(count);
}
void ContentEncoding::ParseEncryptionEntry(
long long start,
long long size,
IMkvReader* const pReader,
ContentEncryption* const encryption) {
assert(pReader);
assert(encryption);
long long pos = start;
const long long stop = start + size;
while (pos < stop) {
#ifdef _DEBUG
long len;
const long long id = ReadUInt(pReader, pos, len);
assert(id >= 0); //TODO: handle error case
assert((pos + len) <= stop);
#endif
long long value;
unsigned char* buf;
size_t buf_len;
if (Match(pReader, pos, 0x7E1, value)) {
// ContentEncAlgo
encryption->algo = value;
} else if (Match(pReader, pos, 0x7E2, buf, buf_len)) {
// ContentEncKeyID
encryption->key_id = buf;
encryption->key_id_len = buf_len;
} else if (Match(pReader, pos, 0x7E3, buf, buf_len)) {
// ContentSignature
encryption->signature = buf;
encryption->signature_len = buf_len;
} else if (Match(pReader, pos, 0x7E4, buf, buf_len)) {
// ContentSigKeyID
encryption->sig_key_id = buf;
encryption->sig_key_id_len = buf_len;
} else if (Match(pReader, pos, 0x7E5, value)) {
// ContentSigAlgo
encoding_type_ = value;
} else if (Match(pReader, pos, 0x7E6, value)) {
// ContentSigHashAlgo
encoding_type_ = value;
} else {
long len;
const long long id = ReadUInt(pReader, pos, len);
assert(id >= 0); //TODO: handle error case
assert((pos + len) <= stop);
pos += len; //consume id
const long long size = ReadUInt(pReader, pos, len);
assert(size >= 0); //TODO: handle error case
assert((pos + len) <= stop);
pos += len; //consume length of size
assert((pos + size) <= stop);
pos += size; //consume payload
assert(pos <= stop);
}
}
}
bool ContentEncoding::ParseContentEncodingEntry(long long start,
long long size,
IMkvReader* const pReader) {
assert(pReader);
long long pos = start;
const long long stop = start + size;
// Count ContentCompression and ContentEncryption elements.
long long pos1 = start;
int compression_count = 0;
int encryption_count = 0;
while (pos1 < stop) {
long len;
const long long id = ReadUInt(pReader, pos1, len);
assert(id >= 0);
assert((pos1 + len) <= stop);
pos1 += len; //consume id
const long long size = ReadUInt(pReader, pos1, len);
assert(size >= 0);
assert((pos1 + len) <= stop);
pos1 += len; //consume length of size
//pos now designates start of element
if (id == 0x1034) // ContentCompression ID
++compression_count;
if (id == 0x1035) // ContentEncryption ID
++encryption_count;
pos1 += size; //consume payload
assert(pos1 <= stop);
}
if (compression_count <= 0 && encryption_count <= 0)
return false;
if (compression_count > 0) {
compression_entries_ = new ContentCompression*[compression_count];
compression_entries_end_ = compression_entries_;
}
if (encryption_count > 0) {
encryption_entries_ = new ContentEncryption*[encryption_count];
encryption_entries_end_ = encryption_entries_;
}
while (pos < stop) {
#ifdef _DEBUG
long len;
const long long id = ReadUInt(pReader, pos, len);
assert(id >= 0); //TODO: handle error case
assert((pos + len) <= stop);
#endif
long long value;
if (Match(pReader, pos, 0x1031, value)) {
// ContentEncodingOrder
encoding_order_ = value;
} else if (Match(pReader, pos, 0x1032, value)) {
// ContentEncodingScope
encoding_scope_ = value;
assert(encoding_scope_ > 0);
} else if (Match(pReader, pos, 0x1033, value)) {
// ContentEncodingType
encoding_type_ = value;
} else {
long len;
const long long id = ReadUInt(pReader, pos, len);
assert(id >= 0); //TODO: handle error case
assert((pos + len) <= stop);
pos += len; //consume id
const long long size = ReadUInt(pReader, pos, len);
assert(size >= 0); //TODO: handle error case
assert((pos + len) <= stop);
pos += len; //consume length of size
assert((pos + size) <= stop);
//pos now designates start of payload
if (id == 0x1034) {
// ContentCompression ID
// TODO(fgaligan): Add code to parse ContentCompression elements.
} else if (id == 0x1035) {
// ContentEncryption ID
ContentEncryption* const encryption = new ContentEncryption();
ParseEncryptionEntry(pos, size, pReader, encryption);
*encryption_entries_end_ = encryption;
++encryption_entries_end_;
}
pos += size; //consume payload
assert(pos <= stop);
}
}
assert(pos == stop);
return true;
}
Track::Track(
Segment* pSegment,
long long element_start,
long long element_size) :
m_pSegment(pSegment),
m_element_start(element_start),
m_element_size(element_size),
content_encoding_entries_(NULL),
content_encoding_entries_end_(NULL)
{
}
Track::~Track()
{
Info& info = const_cast<Info&>(m_info);
info.Clear();
ContentEncoding** i = content_encoding_entries_;
ContentEncoding** const j = content_encoding_entries_end_;
while (i != j) {
ContentEncoding* const encoding = *i++;
delete encoding;
}
delete [] content_encoding_entries_;
}
Track::Info::Info():
nameAsUTF8(NULL),
codecId(NULL),
codecNameAsUTF8(NULL),
codecPrivate(NULL),
codecPrivateSize(0)
{
}
Track::Info::~Info()
{
Clear();
}
void Track::Info::Clear()
{
delete[] nameAsUTF8;
nameAsUTF8 = NULL;
delete[] codecId;
codecId = NULL;
delete[] codecPrivate;
codecPrivate = NULL;
codecPrivateSize = 0;
delete[] codecNameAsUTF8;
codecNameAsUTF8 = NULL;
}
int Track::Info::CopyStr(char* Info::*str, Info& dst_) const
{
if (str == static_cast<char* Info::*>(NULL))
return -1;
char*& dst = dst_.*str;
if (dst) //should be NULL already
return -1;
const char* const src = this->*str;
if (src == NULL)
return 0;
const size_t len = strlen(src);
dst = new (std::nothrow) char[len+1];
if (dst == NULL)
return -1;
strcpy(dst, src);
return 0;
}
int Track::Info::Copy(Info& dst) const
{
if (&dst == this)
return 0;
dst.type = type;
dst.number = number;
dst.uid = uid;
dst.lacing = lacing;
dst.settings = settings;
//We now copy the string member variables from src to dst.
//This involves memory allocation so in principle the operation
//can fail (indeed, that's why we have Info::Copy), so we must
//report this to the caller. An error return from this function
//therefore implies that the copy was only partially successful.
if (int status = CopyStr(&Info::nameAsUTF8, dst))
return status;
if (int status = CopyStr(&Info::codecId, dst))
return status;
if (int status = CopyStr(&Info::codecNameAsUTF8, dst))
return status;
if (codecPrivateSize > 0)
{
if (codecPrivate == NULL)
return -1;
if (dst.codecPrivate)
return -1;
if (dst.codecPrivateSize != 0)
return -1;
dst.codecPrivate = new (std::nothrow) unsigned char[codecPrivateSize];
if (dst.codecPrivate == NULL)
return -1;
memcpy(dst.codecPrivate, codecPrivate, codecPrivateSize);
dst.codecPrivateSize = codecPrivateSize;
}
return 0;
}
const BlockEntry* Track::GetEOS() const
{
return &m_eos;
}
long Track::GetType() const
{
return m_info.type;
}
long Track::GetNumber() const
{
return m_info.number;
}
unsigned long long Track::GetUid() const
{
return m_info.uid;
}
const char* Track::GetNameAsUTF8() const
{
return m_info.nameAsUTF8;
}
const char* Track::GetCodecNameAsUTF8() const
{
return m_info.codecNameAsUTF8;
}
const char* Track::GetCodecId() const
{
return m_info.codecId;
}
const unsigned char* Track::GetCodecPrivate(size_t& size) const
{
size = m_info.codecPrivateSize;
return m_info.codecPrivate;
}
bool Track::GetLacing() const
{
return m_info.lacing;
}
long Track::GetFirst(const BlockEntry*& pBlockEntry) const
{
const Cluster* pCluster = m_pSegment->GetFirst();
for (int i = 0; ; )
{
if (pCluster == NULL)
{
pBlockEntry = GetEOS();
return 1;
}
if (pCluster->EOS())
{
#if 0
if (m_pSegment->Unparsed() <= 0) //all clusters have been loaded
{
pBlockEntry = GetEOS();
return 1;
}
#else
if (m_pSegment->DoneParsing())
{
pBlockEntry = GetEOS();
return 1;
}
#endif
pBlockEntry = 0;
return E_BUFFER_NOT_FULL;
}
long status = pCluster->GetFirst(pBlockEntry);
if (status < 0) //error
return status;
if (pBlockEntry == 0) //empty cluster
{
pCluster = m_pSegment->GetNext(pCluster);
continue;
}
for (;;)
{
const Block* const pBlock = pBlockEntry->GetBlock();
assert(pBlock);
const long long tn = pBlock->GetTrackNumber();
if ((tn == m_info.number) && VetEntry(pBlockEntry))
return 0;
const BlockEntry* pNextEntry;
status = pCluster->GetNext(pBlockEntry, pNextEntry);
if (status < 0) //error
return status;
if (pNextEntry == 0)
break;
pBlockEntry = pNextEntry;
}
++i;
if (i >= 100)
break;
pCluster = m_pSegment->GetNext(pCluster);
}
//NOTE: if we get here, it means that we didn't find a block with
//a matching track number. We interpret that as an error (which
//might be too conservative).
pBlockEntry = GetEOS(); //so we can return a non-NULL value
return 1;
}
long Track::GetNext(
const BlockEntry* pCurrEntry,
const BlockEntry*& pNextEntry) const
{
assert(pCurrEntry);
assert(!pCurrEntry->EOS()); //?
const Block* const pCurrBlock = pCurrEntry->GetBlock();
assert(pCurrBlock->GetTrackNumber() == m_info.number);
const Cluster* pCluster = pCurrEntry->GetCluster();
assert(pCluster);
assert(!pCluster->EOS());
long status = pCluster->GetNext(pCurrEntry, pNextEntry);
if (status < 0) //error
return status;
for (int i = 0; ; )
{
while (pNextEntry)
{
const Block* const pNextBlock = pNextEntry->GetBlock();
assert(pNextBlock);
if (pNextBlock->GetTrackNumber() == m_info.number)
return 0;
pCurrEntry = pNextEntry;
status = pCluster->GetNext(pCurrEntry, pNextEntry);
if (status < 0) //error
return status;
}
pCluster = m_pSegment->GetNext(pCluster);
if (pCluster == NULL)
{
pNextEntry = GetEOS();
return 1;
}
if (pCluster->EOS())
{
#if 0
if (m_pSegment->Unparsed() <= 0) //all clusters have been loaded
{
pNextEntry = GetEOS();
return 1;
}
#else
if (m_pSegment->DoneParsing())
{
pNextEntry = GetEOS();
return 1;
}
#endif
//TODO: there is a potential O(n^2) problem here: we tell the
//caller to (pre)load another cluster, which he does, but then he
//calls GetNext again, which repeats the same search. This is
//a pathological case, since the only way it can happen is if
//there exists a long sequence of clusters none of which contain a
// block from this track. One way around this problem is for the
//caller to be smarter when he loads another cluster: don't call
//us back until you have a cluster that contains a block from this
//track. (Of course, that's not cheap either, since our caller
//would have to scan the each cluster as it's loaded, so that
//would just push back the problem.)
pNextEntry = NULL;
return E_BUFFER_NOT_FULL;
}
status = pCluster->GetFirst(pNextEntry);
if (status < 0) //error
return status;
if (pNextEntry == NULL) //empty cluster
continue;
++i;
if (i >= 100)
break;
}
//NOTE: if we get here, it means that we didn't find a block with
//a matching track number after lots of searching, so we give
//up trying.
pNextEntry = GetEOS(); //so we can return a non-NULL value
return 1;
}
const ContentEncoding*
Track::GetContentEncodingByIndex(unsigned long idx) const {
const ptrdiff_t count =
content_encoding_entries_end_ - content_encoding_entries_;
assert(count >= 0);
if (idx >= static_cast<unsigned long>(count))
return NULL;
return content_encoding_entries_[idx];
}
unsigned long Track::GetContentEncodingCount() const {
const ptrdiff_t count =
content_encoding_entries_end_ - content_encoding_entries_;
assert(count >= 0);
return static_cast<unsigned long>(count);
}
void Track::ParseContentEncodingsEntry(long long start, long long size) {
IMkvReader* const pReader = m_pSegment->m_pReader;
assert(pReader);
long long pos = start;
const long long stop = start + size;
// Count ContentEncoding elements.
long long pos1 = start;
int count = 0;
while (pos1 < stop) {
long len;
const long long id = ReadUInt(pReader, pos1, len);
assert(id >= 0);
assert((pos1 + len) <= stop);
pos1 += len; //consume id
const long long size = ReadUInt(pReader, pos1, len);
assert(size >= 0);
assert((pos1 + len) <= stop);
pos1 += len; //consume length of size
//pos now designates start of element
if (id == 0x2240) // ContentEncoding ID
++count;
pos1 += size; //consume payload
assert(pos1 <= stop);
}
if (count <= 0)
return;
content_encoding_entries_ = new ContentEncoding*[count];
content_encoding_entries_end_ = content_encoding_entries_;
while (pos < stop) {
long len;
const long long id = ReadUInt(pReader, pos, len);
assert(id >= 0);
assert((pos + len) <= stop);
pos += len; //consume id
const long long size1 = ReadUInt(pReader, pos, len);
assert(size1 >= 0);
assert((pos + len) <= stop);
pos += len; //consume length of size
//pos now designates start of element
if (id == 0x2240) { // ContentEncoding ID
ContentEncoding* const content_encoding = new ContentEncoding();
if (!content_encoding->ParseContentEncodingEntry(pos, size1, pReader)) {
delete content_encoding;
} else {
*content_encoding_entries_end_ = content_encoding;
++content_encoding_entries_end_;
}
}
pos += size1; //consume payload
assert(pos <= stop);
}
assert(pos == stop);
return;
}
Track::EOSBlock::EOSBlock() :
BlockEntry(NULL, LONG_MIN)
{
}
BlockEntry::Kind Track::EOSBlock::GetKind() const
{
return kBlockEOS;
}
const Block* Track::EOSBlock::GetBlock() const
{
return NULL;
}
VideoTrack::VideoTrack(
Segment* pSegment,
long long element_start,
long long element_size) :
Track(pSegment, element_start, element_size)
{
}
long VideoTrack::Parse(
Segment* pSegment,
const Info& i,
long long elem_st,
long long elem_sz,
VideoTrack*& pTrack)
{
if (pTrack)
return -1;
if (i.type != Track::kVideo)
return -1;
long long width = 0;
long long height = 0;
double rate = 0.0;
IMkvReader* const pReader = pSegment->m_pReader;
const Settings& s = i.settings;
assert(s.start >= 0);
assert(s.size >= 0);
long long pos = s.start;
assert(pos >= 0);
const long long stop = pos + s.size;
while (pos < stop)
{
long long id, size;
const long status = ParseElementHeader(
pReader,
pos,
stop,
id,
size);
if (status < 0) //error
return status;
if (id == 0x30) //pixel width
{
width = UnserializeUInt(pReader, pos, size);
if (width <= 0)
return E_FILE_FORMAT_INVALID;
}
else if (id == 0x3A) //pixel height
{
height = UnserializeUInt(pReader, pos, size);
if (height <= 0)
return E_FILE_FORMAT_INVALID;
}
else if (id == 0x0383E3) //frame rate
{
const long status = UnserializeFloat(
pReader,
pos,
size,
rate);
if (status < 0)
return status;
if (rate <= 0)
return E_FILE_FORMAT_INVALID;
}
pos += size; //consume payload
assert(pos <= stop);
}
assert(pos == stop);
pTrack = new (std::nothrow) VideoTrack(pSegment, elem_st, elem_sz);
if (pTrack == NULL)
return -1; //generic error
const int status = i.Copy(pTrack->m_info);
if (status)
return status;
pTrack->m_width = width;
pTrack->m_height = height;
pTrack->m_rate = rate;
return 0; //success
}
bool VideoTrack::VetEntry(const BlockEntry* pBlockEntry) const
{
assert(pBlockEntry);
const Block* const pBlock = pBlockEntry->GetBlock();
assert(pBlock);
assert(pBlock->GetTrackNumber() == m_info.number);
return pBlock->IsKey();
}
long VideoTrack::Seek(
long long time_ns,
const BlockEntry*& pResult) const
{
const long status = GetFirst(pResult);
if (status < 0) //buffer underflow, etc
return status;
assert(pResult);
if (pResult->EOS())
return 0;
const Cluster* pCluster = pResult->GetCluster();
assert(pCluster);
assert(pCluster->GetIndex() >= 0);
if (time_ns <= pResult->GetBlock()->GetTime(pCluster))
return 0;
Cluster** const clusters = m_pSegment->m_clusters;
assert(clusters);
const long count = m_pSegment->GetCount(); //loaded only, not pre-loaded
assert(count > 0);
Cluster** const i = clusters + pCluster->GetIndex();
assert(i);
assert(*i == pCluster);
assert(pCluster->GetTime() <= time_ns);
Cluster** const j = clusters + count;
Cluster** lo = i;
Cluster** hi = j;
while (lo < hi)
{
//INVARIANT:
//[i, lo) <= time_ns
//[lo, hi) ?
//[hi, j) > time_ns
Cluster** const mid = lo + (hi - lo) / 2;
assert(mid < hi);
pCluster = *mid;
assert(pCluster);
assert(pCluster->GetIndex() >= 0);
assert(pCluster->GetIndex() == long(mid - m_pSegment->m_clusters));
const long long t = pCluster->GetTime();
if (t <= time_ns)
lo = mid + 1;
else
hi = mid;
assert(lo <= hi);
}
assert(lo == hi);
assert(lo > i);
assert(lo <= j);
pCluster = *--lo;
assert(pCluster);
assert(pCluster->GetTime() <= time_ns);
pResult = pCluster->GetEntry(this, time_ns);
if ((pResult != 0) && !pResult->EOS()) //found a keyframe
return 0;
while (lo != i)
{
pCluster = *--lo;
assert(pCluster);
assert(pCluster->GetTime() <= time_ns);
#if 0
//TODO:
//We need to handle the case when a cluster
//contains multiple keyframes. Simply returning
//the largest keyframe on the cluster isn't
//good enough.
pResult = pCluster->GetMaxKey(this);
#else
pResult = pCluster->GetEntry(this, time_ns);
#endif
if ((pResult != 0) && !pResult->EOS())
return 0;
}
//weird: we're on the first cluster, but no keyframe found
//should never happen but we must return something anyway
pResult = GetEOS();
return 0;
}
long long VideoTrack::GetWidth() const
{
return m_width;
}
long long VideoTrack::GetHeight() const
{
return m_height;
}
double VideoTrack::GetFrameRate() const
{
return m_rate;
}
AudioTrack::AudioTrack(
Segment* pSegment,
long long element_start,
long long element_size) :
Track(pSegment, element_start, element_size)
{
}
long AudioTrack::Parse(
Segment* pSegment,
const Info& i,
long long elem_st,
long long elem_sz,
AudioTrack*& pTrack)
{
if (pTrack)
return -1;
if (i.type != Track::kAudio)
return -1;
IMkvReader* const pReader = pSegment->m_pReader;
const Settings& s = i.settings;
assert(s.start >= 0);
assert(s.size >= 0);
long long pos = s.start;
assert(pos >= 0);
const long long stop = pos + s.size;
double rate = 8000.0;
long long channels = 1;
long long bit_depth = 0;
while (pos < stop)
{
long long id, size;
long status = ParseElementHeader(
pReader,
pos,
stop,
id,
size);
if (status < 0) //error
return status;
if (id == 0x35) //Sample Rate
{
status = UnserializeFloat(pReader, pos, size, rate);
if (status < 0)
return status;
if (rate <= 0)
return E_FILE_FORMAT_INVALID;
}
else if (id == 0x1F) //Channel Count
{
channels = UnserializeUInt(pReader, pos, size);
if (channels <= 0)
return E_FILE_FORMAT_INVALID;
}
else if (id == 0x2264) //Bit Depth
{
bit_depth = UnserializeUInt(pReader, pos, size);
if (bit_depth <= 0)
return E_FILE_FORMAT_INVALID;
}
pos += size; //consume payload
assert(pos <= stop);
}
assert(pos == stop);
pTrack = new (std::nothrow) AudioTrack(pSegment, elem_st, elem_sz);
if (pTrack == NULL)
return -1; //generic error
const int status = i.Copy(pTrack->m_info);
if (status)
return status;
pTrack->m_rate = rate;
pTrack->m_channels = channels;
pTrack->m_bitDepth = bit_depth;
return 0; //success
}
bool AudioTrack::VetEntry(const BlockEntry* pBlockEntry) const
{
assert(pBlockEntry);
const Block* const pBlock = pBlockEntry->GetBlock();
assert(pBlock);
assert(pBlock->GetTrackNumber() == m_info.number);
return true;
}
long AudioTrack::Seek(
long long time_ns,
const BlockEntry*& pResult) const
{
const long status = GetFirst(pResult);
if (status < 0) //buffer underflow, etc
return status;
assert(pResult);
if (pResult->EOS())
return 0;
const Cluster* pCluster = pResult->GetCluster();
assert(pCluster);
assert(pCluster->GetIndex() >= 0);
if (time_ns <= pResult->GetBlock()->GetTime(pCluster))
return 0;
Cluster** const clusters = m_pSegment->m_clusters;
assert(clusters);
const long count = m_pSegment->GetCount(); //loaded only, not preloaded
assert(count > 0);
Cluster** const i = clusters + pCluster->GetIndex();
assert(i);
assert(*i == pCluster);
assert(pCluster->GetTime() <= time_ns);
Cluster** const j = clusters + count;
Cluster** lo = i;
Cluster** hi = j;
while (lo < hi)
{
//INVARIANT:
//[i, lo) <= time_ns
//[lo, hi) ?
//[hi, j) > time_ns
Cluster** const mid = lo + (hi - lo) / 2;
assert(mid < hi);
pCluster = *mid;
assert(pCluster);
assert(pCluster->GetIndex() >= 0);
assert(pCluster->GetIndex() == long(mid - m_pSegment->m_clusters));
const long long t = pCluster->GetTime();
if (t <= time_ns)
lo = mid + 1;
else
hi = mid;
assert(lo <= hi);
}
assert(lo == hi);
assert(lo > i);
assert(lo <= j);
while (lo > i)
{
pCluster = *--lo;
assert(pCluster);
assert(pCluster->GetTime() <= time_ns);
pResult = pCluster->GetEntry(this);
if ((pResult != 0) && !pResult->EOS())
return 0;
//landed on empty cluster (no entries)
}
pResult = GetEOS(); //weird
return 0;
}
double AudioTrack::GetSamplingRate() const
{
return m_rate;
}
long long AudioTrack::GetChannels() const
{
return m_channels;
}
long long AudioTrack::GetBitDepth() const
{
return m_bitDepth;
}
Tracks::Tracks(
Segment* pSegment,
long long start,
long long size_,
long long element_start,
long long element_size) :
m_pSegment(pSegment),
m_start(start),
m_size(size_),
m_element_start(element_start),
m_element_size(element_size),
m_trackEntries(NULL),
m_trackEntriesEnd(NULL)
{
}
long Tracks::Parse()
{
assert(m_trackEntries == NULL);
assert(m_trackEntriesEnd == NULL);
const long long stop = m_start + m_size;
IMkvReader* const pReader = m_pSegment->m_pReader;
int count = 0;
long long pos = m_start;
while (pos < stop)
{
long long id, size;
const long status = ParseElementHeader(
pReader,
pos,
stop,
id,
size);
if (status < 0) //error
return status;
if (size == 0) //weird
continue;
if (id == 0x2E) //TrackEntry ID
++count;
pos += size; //consume payload
assert(pos <= stop);
}
assert(pos == stop);
if (count <= 0)
return 0; //success
m_trackEntries = new (std::nothrow) Track*[count];
if (m_trackEntries == NULL)
return -1;
m_trackEntriesEnd = m_trackEntries;
pos = m_start;
while (pos < stop)
{
const long long element_start = pos;
long long id, payload_size;
const long status = ParseElementHeader(
pReader,
pos,
stop,
id,
payload_size);
if (status < 0) //error
return status;
if (payload_size == 0) //weird
continue;
const long long payload_stop = pos + payload_size;
assert(payload_stop <= stop); //checked in ParseElement
const long long element_size = payload_stop - element_start;
if (id == 0x2E) //TrackEntry ID
{
Track*& pTrack = *m_trackEntriesEnd;
pTrack = NULL;
const long status = ParseTrackEntry(
pos,
payload_size,
element_start,
element_size,
pTrack);
if (status)
return status;
if (pTrack)
++m_trackEntriesEnd;
}
pos = payload_stop;
assert(pos <= stop);
}
assert(pos == stop);
return 0; //success
}
unsigned long Tracks::GetTracksCount() const
{
const ptrdiff_t result = m_trackEntriesEnd - m_trackEntries;
assert(result >= 0);
return static_cast<unsigned long>(result);
}
long Tracks::ParseTrackEntry(
long long track_start,
long long track_size,
long long elem_st,
long long elem_sz,
Track*& pTrack) const
{
if (pTrack)
return -1;
IMkvReader* const pReader = m_pSegment->m_pReader;
long long pos = track_start;
const long long track_stop = track_start + track_size;
Track::Info i;
i.type = 0;
i.number = 0;
i.uid = 0;
Track::Settings v;
v.start = -1;
v.size = -1;
Track::Settings a;
a.start = -1;
a.size = -1;
Track::Settings e; //content_encodings_settings;
e.start = -1;
e.size = -1;
long long lacing = 1; //default is true
while (pos < track_stop)
{
long long id, size;
const long status = ParseElementHeader(
pReader,
pos,
track_stop,
id,
size);
if (status < 0) //error
return status;
const long long start = pos;
if (id == 0x60) // VideoSettings ID
{
if (size < 0)
return E_FILE_FORMAT_INVALID;
v.start = start;
v.size = size;
}
else if (id == 0x61) // AudioSettings ID
{
if (size < 0)
return E_FILE_FORMAT_INVALID;
a.start = start;
a.size = size;
}
else if (id == 0x2D80) // ContentEncodings ID
{
if (size < 0)
return E_FILE_FORMAT_INVALID;
e.start = start;
e.size = size;
}
else if (id == 0x33C5) //Track UID
{
if ((size <= 0) || (size > 8))
return E_FILE_FORMAT_INVALID;
i.uid = 0;
long long pos_ = start;
const long long pos_end = start + size;
while (pos_ != pos_end)
{
unsigned char b;
const int status = pReader->Read(pos_, 1, &b);
if (status)
return status;
i.uid <<= 8;
i.uid |= b;
++pos_;
}
}
else if (id == 0x57) //Track Number
{
const long long num = UnserializeUInt(pReader, pos, size);
if ((num <= 0) || (num > 127))
return E_FILE_FORMAT_INVALID;
i.number = static_cast<long>(num);
}
else if (id == 0x03) //Track Type
{
const long long type = UnserializeUInt(pReader, pos, size);
if ((type <= 0) || (type > 254))
return E_FILE_FORMAT_INVALID;
i.type = static_cast<long>(type);
}
else if (id == 0x136E) //Track Name
{
const long status = UnserializeString(
pReader,
pos,
size,
i.nameAsUTF8);
if (status)
return status;
}
else if (id == 0x06) //CodecID
{
const long status = UnserializeString(
pReader,
pos,
size,
i.codecId);
if (status)
return status;
}
else if (id == 0x1C) //lacing
{
lacing = UnserializeUInt(pReader, pos, size);
if ((lacing < 0) || (lacing > 1))
return E_FILE_FORMAT_INVALID;
}
else if (id == 0x23A2) //Codec Private
{
delete[] i.codecPrivate;
i.codecPrivate = NULL;
i.codecPrivateSize = 0;
if (size <= 0)
return E_FILE_FORMAT_INVALID;
const size_t buflen = static_cast<size_t>(size);
typedef unsigned char* buf_t;
const buf_t buf = new (std::nothrow) unsigned char[buflen];
if (buf == NULL)
return -1;
const int status = pReader->Read(pos, buflen, buf);
if (status)
{
delete[] buf;
return status;
}
i.codecPrivate = buf;
i.codecPrivateSize = buflen;
}
else if (id == 0x058688) //Codec Name
{
const long status = UnserializeString(
pReader,
pos,
size,
i.codecNameAsUTF8);
if (status)
return status;
}
pos += size; //consume payload
assert(pos <= track_stop);
}
assert(pos == track_stop);
if (i.number <= 0) //not specified
return E_FILE_FORMAT_INVALID;
if (GetTrackByNumber(i.number))
return E_FILE_FORMAT_INVALID;
if (i.type <= 0) //not specified
return E_FILE_FORMAT_INVALID;
if ((i.type != Track::kVideo) && (i.type != Track::kAudio))
{
//TODO(matthewjheaney): go ahead and create a "generic" track
//object, so that GetTrackByXXX always returns something, even
//if the object it returns has a type that is not kVideo or kAudio.
return 0; //no error
}
i.lacing = (lacing > 0) ? true : false;
long status;
if (i.type == Track::kVideo)
{
if (v.start < 0)
return E_FILE_FORMAT_INVALID;
if (a.start >= 0)
return E_FILE_FORMAT_INVALID;
i.settings = v;
VideoTrack* p = NULL;
status = VideoTrack::Parse(m_pSegment, i, elem_st, elem_sz, p);
pTrack = p;
}
else
{
assert(i.type == Track::kAudio);
if (a.start < 0)
return E_FILE_FORMAT_INVALID;
if (v.start >= 0)
return E_FILE_FORMAT_INVALID;
i.settings = a;
AudioTrack* p = NULL;
status = AudioTrack::Parse(m_pSegment, i, elem_st, elem_sz, p);
pTrack = p;
}
if (status)
return status;
assert(pTrack);
if (e.start >= 0)
pTrack->ParseContentEncodingsEntry(e.start, e.size);
return 0; //success
}
Tracks::~Tracks()
{
Track** i = m_trackEntries;
Track** const j = m_trackEntriesEnd;
while (i != j)
{
Track* const pTrack = *i++;
delete pTrack;
}
delete[] m_trackEntries;
}
const Track* Tracks::GetTrackByNumber(long tn) const
{
if (tn < 0)
return NULL;
Track** i = m_trackEntries;
Track** const j = m_trackEntriesEnd;
while (i != j)
{
Track* const pTrack = *i++;
if (pTrack == NULL)
continue;
if (tn == pTrack->GetNumber())
return pTrack;
}
return NULL; //not found
}
const Track* Tracks::GetTrackByIndex(unsigned long idx) const
{
const ptrdiff_t count = m_trackEntriesEnd - m_trackEntries;
if (idx >= static_cast<unsigned long>(count))
return NULL;
return m_trackEntries[idx];
}
#if 0
long long Cluster::Unparsed() const
{
if (m_timecode < 0) //not even partially loaded
return LLONG_MAX;
assert(m_pos >= m_element_start);
//assert(m_element_size > m_size);
const long long element_stop = m_element_start + m_element_size;
assert(m_pos <= element_stop);
const long long result = element_stop - m_pos;
assert(result >= 0);
return result;
}
#endif
long Cluster::Load(long long& pos, long& len) const
{
assert(m_pSegment);
assert(m_pos >= m_element_start);
if (m_timecode >= 0) //at least partially loaded
return 0;
assert(m_pos == m_element_start);
assert(m_element_size < 0);
IMkvReader* const pReader = m_pSegment->m_pReader;
long long total, avail;
const int status = pReader->Length(&total, &avail);
if (status < 0) //error
return status;
assert((total < 0) || (avail <= total));
assert((total < 0) || (m_pos <= total)); //TODO: verify this
pos = m_pos;
long long cluster_size = -1;
{
if ((pos + 1) > avail)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
long long result = GetUIntLength(pReader, pos, len);
if (result < 0) //error or underflow
return static_cast<long>(result);
if (result > 0) //underflow (weird)
return E_BUFFER_NOT_FULL;
//if ((pos + len) > segment_stop)
// return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
return E_BUFFER_NOT_FULL;
const long long id_ = ReadUInt(pReader, pos, len);
if (id_ < 0) //error
return static_cast<long>(id_);
if (id_ != 0x0F43B675) //Cluster ID
return E_FILE_FORMAT_INVALID;
pos += len; //consume id
//read cluster size
if ((pos + 1) > avail)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
result = GetUIntLength(pReader, pos, len);
if (result < 0) //error
return static_cast<long>(result);
if (result > 0) //weird
return E_BUFFER_NOT_FULL;
//if ((pos + len) > segment_stop)
// return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
return E_BUFFER_NOT_FULL;
const long long size = ReadUInt(pReader, pos, len);
if (size < 0) //error
return static_cast<long>(cluster_size);
if (size == 0)
return E_FILE_FORMAT_INVALID; //TODO: verify this
pos += len; //consume length of size of element
const long long unknown_size = (1LL << (7 * len)) - 1;
if (size != unknown_size)
cluster_size = size;
}
//pos points to start of payload
#if 0
len = static_cast<long>(size_);
if (cluster_stop > avail)
return E_BUFFER_NOT_FULL;
#endif
long long timecode = -1;
long long new_pos = -1;
bool bBlock = false;
long long cluster_stop = (cluster_size < 0) ? -1 : pos + cluster_size;
for (;;)
{
if ((cluster_stop >= 0) && (pos >= cluster_stop))
break;
//Parse ID
if ((pos + 1) > avail)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
long long result = GetUIntLength(pReader, pos, len);
if (result < 0) //error
return static_cast<long>(result);
if (result > 0) //weird
return E_BUFFER_NOT_FULL;
if ((cluster_stop >= 0) && ((pos + len) > cluster_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
return E_BUFFER_NOT_FULL;
const long long id = ReadUInt(pReader, pos, len);
if (id < 0) //error
return static_cast<long>(id);
if (id == 0)
return E_FILE_FORMAT_INVALID;
//This is the distinguished set of ID's we use to determine
//that we have exhausted the sub-element's inside the cluster
//whose ID we parsed earlier.
if (id == 0x0F43B675) //Cluster ID
break;
if (id == 0x0C53BB6B) //Cues ID
break;
pos += len; //consume ID field
//Parse Size
if ((pos + 1) > avail)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
result = GetUIntLength(pReader, pos, len);
if (result < 0) //error
return static_cast<long>(result);
if (result > 0) //weird
return E_BUFFER_NOT_FULL;
if ((cluster_stop >= 0) && ((pos + len) > cluster_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
return E_BUFFER_NOT_FULL;
const long long size = ReadUInt(pReader, pos, len);
if (size < 0) //error
return static_cast<long>(size);
const long long unknown_size = (1LL << (7 * len)) - 1;
if (size == unknown_size)
return E_FILE_FORMAT_INVALID;
pos += len; //consume size field
if ((cluster_stop >= 0) && (pos > cluster_stop))
return E_FILE_FORMAT_INVALID;
//pos now points to start of payload
if (size == 0) //weird
continue;
if ((cluster_stop >= 0) && ((pos + size) > cluster_stop))
return E_FILE_FORMAT_INVALID;
if (id == 0x67) //TimeCode ID
{
len = static_cast<long>(size);
if ((pos + size) > avail)
return E_BUFFER_NOT_FULL;
timecode = UnserializeUInt(pReader, pos, size);
if (timecode < 0) //error (or underflow)
return static_cast<long>(timecode);
new_pos = pos + size;
if (bBlock)
break;
}
else if (id == 0x20) //BlockGroup ID
{
bBlock = true;
break;
}
else if (id == 0x23) //SimpleBlock ID
{
bBlock = true;
break;
}
pos += size; //consume payload
assert((cluster_stop < 0) || (pos <= cluster_stop));
}
assert((cluster_stop < 0) || (pos <= cluster_stop));
if (timecode < 0) //no timecode found
return E_FILE_FORMAT_INVALID;
if (!bBlock)
return E_FILE_FORMAT_INVALID;
m_pos = new_pos; //designates position just beyond timecode payload
m_timecode = timecode; // m_timecode >= 0 means we're partially loaded
if (cluster_size >= 0)
m_element_size = cluster_stop - m_element_start;
return 0;
}
long Cluster::Parse(long long& pos, long& len) const
{
long status = Load(pos, len);
if (status < 0)
return status;
assert(m_pos >= m_element_start);
assert(m_timecode >= 0);
//assert(m_size > 0);
//assert(m_element_size > m_size);
const long long cluster_stop =
(m_element_size < 0) ? -1 : m_element_start + m_element_size;
if ((cluster_stop >= 0) && (m_pos >= cluster_stop))
return 1; //nothing else to do
IMkvReader* const pReader = m_pSegment->m_pReader;
long long total, avail;
status = pReader->Length(&total, &avail);
if (status < 0) //error
return status;
assert((total < 0) || (avail <= total));
pos = m_pos;
for (;;)
{
if ((cluster_stop >= 0) && (pos >= cluster_stop))
break;
if ((total >= 0) && (pos >= total))
{
if (m_element_size < 0)
m_element_size = pos - m_element_start;
break;
}
//Parse ID
if ((pos + 1) > avail)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
long long result = GetUIntLength(pReader, pos, len);
if (result < 0) //error
return static_cast<long>(result);
if (result > 0) //weird
return E_BUFFER_NOT_FULL;
if ((cluster_stop >= 0) && ((pos + len) > cluster_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
return E_BUFFER_NOT_FULL;
const long long id = ReadUInt(pReader, pos, len);
if (id < 0) //error
return static_cast<long>(id);
if (id == 0) //weird
return E_FILE_FORMAT_INVALID;
//This is the distinguished set of ID's we use to determine
//that we have exhausted the sub-element's inside the cluster
//whose ID we parsed earlier.
if ((id == 0x0F43B675) || (id == 0x0C53BB6B)) //Cluster or Cues ID
{
if (m_element_size < 0)
m_element_size = pos - m_element_start;
break;
}
pos += len; //consume ID field
//Parse Size
if ((pos + 1) > avail)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
result = GetUIntLength(pReader, pos, len);
if (result < 0) //error
return static_cast<long>(result);
if (result > 0) //weird
return E_BUFFER_NOT_FULL;
if ((cluster_stop >= 0) && ((pos + len) > cluster_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
return E_BUFFER_NOT_FULL;
const long long size = ReadUInt(pReader, pos, len);
if (size < 0) //error
return static_cast<long>(size);
const long long unknown_size = (1LL << (7 * len)) - 1;
if (size == unknown_size)
return E_FILE_FORMAT_INVALID;
pos += len; //consume size field
if ((cluster_stop >= 0) && (pos > cluster_stop))
return E_FILE_FORMAT_INVALID;
//pos now points to start of payload
if (size == 0) //weird
continue;
//const long long block_start = pos;
const long long block_stop = pos + size;
if (cluster_stop >= 0)
{
if (block_stop > cluster_stop)
return E_FILE_FORMAT_INVALID;
}
else if ((total >= 0) && (block_stop > total))
{
m_element_size = total - m_element_start;
pos = total;
break;
}
else if (block_stop > avail)
{
len = static_cast<long>(size);
return E_BUFFER_NOT_FULL;
}
Cluster* const this_ = const_cast<Cluster*>(this);
if (id == 0x20) //BlockGroup
return this_->ParseBlockGroup(size, pos, len);
if (id == 0x23) //SimpleBlock
return this_->ParseSimpleBlock(size, pos, len);
pos += size; //consume payload
assert((cluster_stop < 0) || (pos <= cluster_stop));
}
assert(m_element_size > 0);
m_pos = pos;
assert((cluster_stop < 0) || (m_pos <= cluster_stop));
if (m_entries_count > 0)
{
const long idx = m_entries_count - 1;
const BlockEntry* const pLast = m_entries[idx];
assert(pLast);
const Block* const pBlock = pLast->GetBlock();
assert(pBlock);
const long long start = pBlock->m_start;
if ((total >= 0) && (start > total))
return -1; //defend against trucated stream
const long long size = pBlock->m_size;
const long long stop = start + size;
assert((cluster_stop < 0) || (stop <= cluster_stop));
if ((total >= 0) && (stop > total))
return -1; //defend against trucated stream
}
return 1; //no more entries
}
long Cluster::ParseSimpleBlock(
long long block_size,
long long& pos,
long& len)
{
const long long block_start = pos;
const long long block_stop = pos + block_size;
IMkvReader* const pReader = m_pSegment->m_pReader;
long long total, avail;
long status = pReader->Length(&total, &avail);
if (status < 0) //error
return status;
assert((total < 0) || (avail <= total));
//parse track number
if ((pos + 1) > avail)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
long long result = GetUIntLength(pReader, pos, len);
if (result < 0) //error
return static_cast<long>(result);
if (result > 0) //weird
return E_BUFFER_NOT_FULL;
if ((pos + len) > block_stop)
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
return E_BUFFER_NOT_FULL;
const long long track = ReadUInt(pReader, pos, len);
if (track < 0) //error
return static_cast<long>(track);
if (track == 0)
return E_FILE_FORMAT_INVALID;
#if 0
//TODO(matthewjheaney)
//This turned out to be too conservative. The problem is that
//if we see a track header in the tracks element with an unsupported
//track type, we throw that track header away, so it is not present
//in the track map. But even though we don't understand the track
//header, there are still blocks in the cluster with that track
//number. It was our decision to ignore that track header, so it's
//up to us to deal with blocks associated with that track -- we
//cannot simply report an error since technically there's nothing
//wrong with the file.
//
//For now we go ahead and finish the parse, creating a block entry
//for this block. This is somewhat wasteful, because without a
//track header there's nothing you can do with the block. What
//we really need here is a special return value that indicates to
//the caller that he should ignore this particular block, and
//continue parsing.
const Tracks* const pTracks = m_pSegment->GetTracks();
assert(pTracks);
const long tn = static_cast<long>(track);
const Track* const pTrack = pTracks->GetTrackByNumber(tn);
if (pTrack == NULL)
return E_FILE_FORMAT_INVALID;
#endif
pos += len; //consume track number
if ((pos + 2) > block_stop)
return E_FILE_FORMAT_INVALID;
if ((pos + 2) > avail)
{
len = 2;
return E_BUFFER_NOT_FULL;
}
pos += 2; //consume timecode
if ((pos + 1) > block_stop)
return E_FILE_FORMAT_INVALID;
if ((pos + 1) > avail)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
unsigned char flags;
status = pReader->Read(pos, 1, &flags);
if (status < 0) //error or underflow
{
len = 1;
return status;
}
++pos; //consume flags byte
assert(pos <= avail);
if (pos >= block_stop)
return E_FILE_FORMAT_INVALID;
const int lacing = int(flags & 0x06) >> 1;
if ((lacing != 0) && (block_stop > avail))
{
len = static_cast<long>(block_stop - pos);
return E_BUFFER_NOT_FULL;
}
status = CreateBlock(0x23, block_start, block_size); //simple block id
if (status != 0)
return status;
m_pos = block_stop;
return 0; //success
}
long Cluster::ParseBlockGroup(
long long payload_size,
long long& pos,
long& len)
{
const long long payload_start = pos;
const long long payload_stop = pos + payload_size;
IMkvReader* const pReader = m_pSegment->m_pReader;
long long total, avail;
long status = pReader->Length(&total, &avail);
if (status < 0) //error
return status;
assert((total < 0) || (avail <= total));
if ((total >= 0) && (payload_stop > total))
return E_FILE_FORMAT_INVALID;
if (payload_stop > avail)
{
len = static_cast<long>(payload_size);
return E_BUFFER_NOT_FULL;
}
while (pos < payload_stop)
{
//parse sub-block element ID
if ((pos + 1) > avail)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
long long result = GetUIntLength(pReader, pos, len);
if (result < 0) //error
return static_cast<long>(result);
if (result > 0) //weird
return E_BUFFER_NOT_FULL;
if ((pos + len) > payload_stop)
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
return E_BUFFER_NOT_FULL;
const long long id = ReadUInt(pReader, pos, len);
if (id < 0) //error
return static_cast<long>(id);
if (id == 0) //not a value ID
return E_FILE_FORMAT_INVALID;
pos += len; //consume ID field
//Parse Size
if ((pos + 1) > avail)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
result = GetUIntLength(pReader, pos, len);
if (result < 0) //error
return static_cast<long>(result);
if (result > 0) //weird
return E_BUFFER_NOT_FULL;
if ((pos + len) > payload_stop)
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
return E_BUFFER_NOT_FULL;
const long long size = ReadUInt(pReader, pos, len);
if (size < 0) //error
return static_cast<long>(size);
pos += len; //consume size field
//pos now points to start of sub-block group payload
if (pos > payload_stop)
return E_FILE_FORMAT_INVALID;
if (size == 0) //weird
continue;
const long long unknown_size = (1LL << (7 * len)) - 1;
if (size == unknown_size)
return E_FILE_FORMAT_INVALID;
if (id != 0x21) //sub-part of BlockGroup is not a Block
{
pos += size; //consume sub-part of block group
if (pos > payload_stop)
return E_FILE_FORMAT_INVALID;
continue;
}
const long long block_stop = pos + size;
if (block_stop > payload_stop)
return E_FILE_FORMAT_INVALID;
//parse track number
if ((pos + 1) > avail)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
result = GetUIntLength(pReader, pos, len);
if (result < 0) //error
return static_cast<long>(result);
if (result > 0) //weird
return E_BUFFER_NOT_FULL;
if ((pos + len) > block_stop)
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
return E_BUFFER_NOT_FULL;
const long long track = ReadUInt(pReader, pos, len);
if (track < 0) //error
return static_cast<long>(track);
if (track == 0)
return E_FILE_FORMAT_INVALID;
#if 0
//TODO(matthewjheaney)
//This turned out to be too conservative. The problem is that
//if we see a track header in the tracks element with an unsupported
//track type, we throw that track header away, so it is not present
//in the track map. But even though we don't understand the track
//header, there are still blocks in the cluster with that track
//number. It was our decision to ignore that track header, so it's
//up to us to deal with blocks associated with that track -- we
//cannot simply report an error since technically there's nothing
//wrong with the file.
//
//For now we go ahead and finish the parse, creating a block entry
//for this block. This is somewhat wasteful, because without a
//track header there's nothing you can do with the block. What
//we really need here is a special return value that indicates to
//the caller that he should ignore this particular block, and
//continue parsing.
const Tracks* const pTracks = m_pSegment->GetTracks();
assert(pTracks);
const long tn = static_cast<long>(track);
const Track* const pTrack = pTracks->GetTrackByNumber(tn);
if (pTrack == NULL)
return E_FILE_FORMAT_INVALID;
#endif
pos += len; //consume track number
if ((pos + 2) > block_stop)
return E_FILE_FORMAT_INVALID;
if ((pos + 2) > avail)
{
len = 2;
return E_BUFFER_NOT_FULL;
}
pos += 2; //consume timecode
if ((pos + 1) > block_stop)
return E_FILE_FORMAT_INVALID;
if ((pos + 1) > avail)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
unsigned char flags;
status = pReader->Read(pos, 1, &flags);
if (status < 0) //error or underflow
{
len = 1;
return status;
}
++pos; //consume flags byte
assert(pos <= avail);
if (pos >= block_stop)
return E_FILE_FORMAT_INVALID;
const int lacing = int(flags & 0x06) >> 1;
if ((lacing != 0) && (block_stop > avail))
{
len = static_cast<long>(block_stop - pos);
return E_BUFFER_NOT_FULL;
}
pos = block_stop; //consume block-part of block group
assert(pos <= payload_stop);
}
assert(pos == payload_stop);
status = CreateBlock(0x20, payload_start, payload_size); //BlockGroup ID
if (status != 0)
return status;
m_pos = payload_stop;
return 0; //success
}
long Cluster::GetEntry(long index, const mkvparser::BlockEntry*& pEntry) const
{
assert(m_pos >= m_element_start);
pEntry = NULL;
if (index < 0)
return -1; //generic error
if (m_entries_count < 0)
return E_BUFFER_NOT_FULL;
assert(m_entries);
assert(m_entries_size > 0);
assert(m_entries_count <= m_entries_size);
if (index < m_entries_count)
{
pEntry = m_entries[index];
assert(pEntry);
return 1; //found entry
}
if (m_element_size < 0) //we don't know cluster end yet
return E_BUFFER_NOT_FULL; //underflow
const long long element_stop = m_element_start + m_element_size;
if (m_pos >= element_stop)
return 0; //nothing left to parse
return E_BUFFER_NOT_FULL; //underflow, since more remains to be parsed
}
Cluster* Cluster::Create(
Segment* pSegment,
long idx,
long long off)
//long long element_size)
{
assert(pSegment);
assert(off >= 0);
const long long element_start = pSegment->m_start + off;
Cluster* const pCluster = new Cluster(pSegment,
idx,
element_start);
//element_size);
assert(pCluster);
return pCluster;
}
Cluster::Cluster() :
m_pSegment(NULL),
m_element_start(0),
m_index(0),
m_pos(0),
m_element_size(0),
m_timecode(0),
m_entries(NULL),
m_entries_size(0),
m_entries_count(0) //means "no entries"
{
}
Cluster::Cluster(
Segment* pSegment,
long idx,
long long element_start
/* long long element_size */ ) :
m_pSegment(pSegment),
m_element_start(element_start),
m_index(idx),
m_pos(element_start),
m_element_size(-1 /* element_size */ ),
m_timecode(-1),
m_entries(NULL),
m_entries_size(0),
m_entries_count(-1) //means "has not been parsed yet"
{
}
Cluster::~Cluster()
{
if (m_entries_count <= 0)
return;
BlockEntry** i = m_entries;
BlockEntry** const j = m_entries + m_entries_count;
while (i != j)
{
BlockEntry* p = *i++;
assert(p);
delete p;
}
delete[] m_entries;
}
bool Cluster::EOS() const
{
return (m_pSegment == NULL);
}
long Cluster::GetIndex() const
{
return m_index;
}
long long Cluster::GetPosition() const
{
const long long pos = m_element_start - m_pSegment->m_start;
assert(pos >= 0);
return pos;
}
long long Cluster::GetElementSize() const
{
return m_element_size;
}
#if 0
bool Cluster::HasBlockEntries(
const Segment* pSegment,
long long off) //relative to start of segment payload
{
assert(pSegment);
assert(off >= 0); //relative to segment
IMkvReader* const pReader = pSegment->m_pReader;
long long pos = pSegment->m_start + off; //absolute
long long size;
{
long len;
const long long id = ReadUInt(pReader, pos, len);
(void)id;
assert(id >= 0);
assert(id == 0x0F43B675); //Cluster ID
pos += len; //consume id
size = ReadUInt(pReader, pos, len);
assert(size > 0);
pos += len; //consume size
//pos now points to start of payload
}
const long long stop = pos + size;
while (pos < stop)
{
long len;
const long long id = ReadUInt(pReader, pos, len);
assert(id >= 0); //TODO
assert((pos + len) <= stop);
pos += len; //consume id
const long long size = ReadUInt(pReader, pos, len);
assert(size >= 0); //TODO
assert((pos + len) <= stop);
pos += len; //consume size
if (id == 0x20) //BlockGroup ID
return true;
if (id == 0x23) //SimpleBlock ID
return true;
pos += size; //consume payload
assert(pos <= stop);
}
return false;
}
#endif
long Cluster::HasBlockEntries(
const Segment* pSegment,
long long off, //relative to start of segment payload
long long& pos,
long& len)
{
assert(pSegment);
assert(off >= 0); //relative to segment
IMkvReader* const pReader = pSegment->m_pReader;
long long total, avail;
long status = pReader->Length(&total, &avail);
if (status < 0) //error
return status;
assert((total < 0) || (avail <= total));
pos = pSegment->m_start + off; //absolute
if ((total >= 0) && (pos >= total))
return 0; //we don't even have a complete cluster
const long long segment_stop =
(pSegment->m_size < 0) ? -1 : pSegment->m_start + pSegment->m_size;
long long cluster_stop = -1; //interpreted later to mean "unknown size"
{
if ((pos + 1) > avail)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
long long result = GetUIntLength(pReader, pos, len);
if (result < 0) //error
return static_cast<long>(result);
if (result > 0) //need more data
return E_BUFFER_NOT_FULL;
if ((segment_stop >= 0) && ((pos + len) > segment_stop))
return E_FILE_FORMAT_INVALID;
if ((total >= 0) && ((pos + len) > total))
return 0;
if ((pos + len) > avail)
return E_BUFFER_NOT_FULL;
const long long id = ReadUInt(pReader, pos, len);
if (id < 0) //error
return static_cast<long>(id);
if (id != 0x0F43B675) //weird: not cluster ID
return -1; //generic error
pos += len; //consume Cluster ID field
//read size field
if ((pos + 1) > avail)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
result = GetUIntLength(pReader, pos, len);
if (result < 0) //error
return static_cast<long>(result);
if (result > 0) //weird
return E_BUFFER_NOT_FULL;
if ((segment_stop >= 0) && ((pos + len) > segment_stop))
return E_FILE_FORMAT_INVALID;
if ((total >= 0) && ((pos + len) > total))
return 0;
if ((pos + len) > avail)
return E_BUFFER_NOT_FULL;
const long long size = ReadUInt(pReader, pos, len);
if (size < 0) //error
return static_cast<long>(size);
if (size == 0)
return 0; //cluster does not have entries
pos += len; //consume size field
//pos now points to start of payload
const long long unknown_size = (1LL << (7 * len)) - 1;
if (size != unknown_size)
{
cluster_stop = pos + size;
assert(cluster_stop >= 0);
if ((segment_stop >= 0) && (cluster_stop > segment_stop))
return E_FILE_FORMAT_INVALID;
if ((total >= 0) && (cluster_stop > total))
//return E_FILE_FORMAT_INVALID; //too conservative
return 0; //cluster does not have any entries
}
}
for (;;)
{
if ((cluster_stop >= 0) && (pos >= cluster_stop))
return 0; //no entries detected
if ((pos + 1) > avail)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
long long result = GetUIntLength(pReader, pos, len);
if (result < 0) //error
return static_cast<long>(result);
if (result > 0) //need more data
return E_BUFFER_NOT_FULL;
if ((cluster_stop >= 0) && ((pos + len) > cluster_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
return E_BUFFER_NOT_FULL;
const long long id = ReadUInt(pReader, pos, len);
if (id < 0) //error
return static_cast<long>(id);
//This is the distinguished set of ID's we use to determine
//that we have exhausted the sub-element's inside the cluster
//whose ID we parsed earlier.
if (id == 0x0F43B675) //Cluster ID
return 0; //no entries found
if (id == 0x0C53BB6B) //Cues ID
return 0; //no entries found
pos += len; //consume id field
if ((cluster_stop >= 0) && (pos >= cluster_stop))
return E_FILE_FORMAT_INVALID;
//read size field
if ((pos + 1) > avail)
{
len = 1;
return E_BUFFER_NOT_FULL;
}
result = GetUIntLength(pReader, pos, len);
if (result < 0) //error
return static_cast<long>(result);
if (result > 0) //underflow
return E_BUFFER_NOT_FULL;
if ((cluster_stop >= 0) && ((pos + len) > cluster_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
return E_BUFFER_NOT_FULL;
const long long size = ReadUInt(pReader, pos, len);
if (size < 0) //error
return static_cast<long>(size);
pos += len; //consume size field
//pos now points to start of payload
if ((cluster_stop >= 0) && (pos > cluster_stop))
return E_FILE_FORMAT_INVALID;
if (size == 0) //weird
continue;
const long long unknown_size = (1LL << (7 * len)) - 1;
if (size == unknown_size)
return E_FILE_FORMAT_INVALID; //not supported inside cluster
if ((cluster_stop >= 0) && ((pos + size) > cluster_stop))
return E_FILE_FORMAT_INVALID;
if (id == 0x20) //BlockGroup ID
return 1; //have at least one entry
if (id == 0x23) //SimpleBlock ID
return 1; //have at least one entry
pos += size; //consume payload
assert((cluster_stop < 0) || (pos <= cluster_stop));
}
}
long long Cluster::GetTimeCode() const
{
long long pos;
long len;
const long status = Load(pos, len);
if (status < 0) //error
return status;
return m_timecode;
}
long long Cluster::GetTime() const
{
const long long tc = GetTimeCode();
if (tc < 0)
return tc;
const SegmentInfo* const pInfo = m_pSegment->GetInfo();
assert(pInfo);
const long long scale = pInfo->GetTimeCodeScale();
assert(scale >= 1);
const long long t = m_timecode * scale;
return t;
}
long long Cluster::GetFirstTime() const
{
const BlockEntry* pEntry;
const long status = GetFirst(pEntry);
if (status < 0) //error
return status;
if (pEntry == NULL) //empty cluster
return GetTime();
const Block* const pBlock = pEntry->GetBlock();
assert(pBlock);
return pBlock->GetTime(this);
}
long long Cluster::GetLastTime() const
{
const BlockEntry* pEntry;
const long status = GetLast(pEntry);
if (status < 0) //error
return status;
if (pEntry == NULL) //empty cluster
return GetTime();
const Block* const pBlock = pEntry->GetBlock();
assert(pBlock);
return pBlock->GetTime(this);
}
long Cluster::CreateBlock(
long long id,
long long pos, //absolute pos of payload
long long size)
{
assert((id == 0x20) || (id == 0x23)); //BlockGroup or SimpleBlock
if (m_entries_count < 0) //haven't parsed anything yet
{
assert(m_entries == NULL);
assert(m_entries_size == 0);
m_entries_size = 1024;
m_entries = new BlockEntry*[m_entries_size];
m_entries_count = 0;
}
else
{
assert(m_entries);
assert(m_entries_size > 0);
assert(m_entries_count <= m_entries_size);
if (m_entries_count >= m_entries_size)
{
const long entries_size = 2 * m_entries_size;
BlockEntry** const entries = new BlockEntry*[entries_size];
assert(entries);
BlockEntry** src = m_entries;
BlockEntry** const src_end = src + m_entries_count;
BlockEntry** dst = entries;
while (src != src_end)
*dst++ = *src++;
delete[] m_entries;
m_entries = entries;
m_entries_size = entries_size;
}
}
if (id == 0x20) //BlockGroup ID
return CreateBlockGroup(pos, size);
else //SimpleBlock ID
return CreateSimpleBlock(pos, size);
}
long Cluster::CreateBlockGroup(
long long st,
long long sz)
{
assert(m_entries);
assert(m_entries_size > 0);
assert(m_entries_count >= 0);
assert(m_entries_count < m_entries_size);
IMkvReader* const pReader = m_pSegment->m_pReader;
long long pos = st;
const long long stop = st + sz;
//For WebM files, there is a bias towards previous reference times
//(in order to support alt-ref frames, which refer back to the previous
//keyframe). Normally a 0 value is not possible, but here we tenatively
//allow 0 as the value of a reference frame, with the interpretation
//that this is a "previous" reference time.
long long prev = 1; //nonce
long long next = 0; //nonce
long long duration = -1; //really, this is unsigned
long long bpos = -1;
long long bsize = -1;
while (pos < stop)
{
long len;
const long long id = ReadUInt(pReader, pos, len);
assert(id >= 0); //TODO
assert((pos + len) <= stop);
pos += len; //consume ID
const long long size = ReadUInt(pReader, pos, len);
assert(size >= 0); //TODO
assert((pos + len) <= stop);
pos += len; //consume size
if (id == 0x21) //Block ID
{
if (bpos < 0) //Block ID
{
bpos = pos;
bsize = size;
}
}
else if (id == 0x1B) //Duration ID
{
assert(size <= 8);
duration = UnserializeUInt(pReader, pos, size);
assert(duration >= 0); //TODO
}
else if (id == 0x7B) //ReferenceBlock
{
assert(size <= 8);
const long size_ = static_cast<long>(size);
long long time;
long status = UnserializeInt(pReader, pos, size_, time);
assert(status == 0); //TODO
if (time <= 0) //see note above
prev = time;
else //weird
next = time;
}
pos += size; //consume payload
assert(pos <= stop);
}
assert(pos == stop);
assert(bpos >= 0);
assert(bsize >= 0);
const long idx = m_entries_count;
BlockEntry** const ppEntry = m_entries + idx;
BlockEntry*& pEntry = *ppEntry;
pEntry = new (std::nothrow) BlockGroup(
this,
idx,
bpos,
bsize,
prev,
next,
duration);
if (pEntry == NULL)
return -1; //generic error
BlockGroup* const p = static_cast<BlockGroup*>(pEntry);
const long status = p->Parse();
if (status == 0) //success
{
++m_entries_count;
return 0;
}
delete pEntry;
pEntry = 0;
return status;
}
long Cluster::CreateSimpleBlock(
long long st,
long long sz)
{
assert(m_entries);
assert(m_entries_size > 0);
assert(m_entries_count >= 0);
assert(m_entries_count < m_entries_size);
const long idx = m_entries_count;
BlockEntry** const ppEntry = m_entries + idx;
BlockEntry*& pEntry = *ppEntry;
pEntry = new (std::nothrow) SimpleBlock(this, idx, st, sz);
if (pEntry == NULL)
return -1; //generic error
SimpleBlock* const p = static_cast<SimpleBlock*>(pEntry);
const long status = p->Parse();
if (status == 0)
{
++m_entries_count;
return 0;
}
delete pEntry;
pEntry = 0;
return status;
}
long Cluster::GetFirst(const BlockEntry*& pFirst) const
{
if (m_entries_count <= 0)
{
long long pos;
long len;
const long status = Parse(pos, len);
if (status < 0) //error
{
pFirst = NULL;
return status;
}
if (m_entries_count <= 0) //empty cluster
{
pFirst = NULL;
return 0;
}
}
assert(m_entries);
pFirst = m_entries[0];
assert(pFirst);
return 0; //success
}
long Cluster::GetLast(const BlockEntry*& pLast) const
{
for (;;)
{
long long pos;
long len;
const long status = Parse(pos, len);
if (status < 0) //error
{
pLast = NULL;
return status;
}
if (status > 0) //no new block
break;
}
if (m_entries_count <= 0)
{
pLast = NULL;
return 0;
}
assert(m_entries);
const long idx = m_entries_count - 1;
pLast = m_entries[idx];
assert(pLast);
return 0;
}
long Cluster::GetNext(
const BlockEntry* pCurr,
const BlockEntry*& pNext) const
{
assert(pCurr);
assert(m_entries);
assert(m_entries_count > 0);
size_t idx = pCurr->GetIndex();
assert(idx < size_t(m_entries_count));
assert(m_entries[idx] == pCurr);
++idx;
if (idx >= size_t(m_entries_count))
{
long long pos;
long len;
const long status = Parse(pos, len);
if (status < 0) //error
{
pNext = NULL;
return status;
}
if (status > 0)
{
pNext = NULL;
return 0;
}
assert(m_entries);
assert(m_entries_count > 0);
assert(idx < size_t(m_entries_count));
}
pNext = m_entries[idx];
assert(pNext);
return 0;
}
long Cluster::GetEntryCount() const
{
return m_entries_count;
}
const BlockEntry* Cluster::GetEntry(
const Track* pTrack,
long long time_ns) const
{
assert(pTrack);
if (m_pSegment == NULL) //this is the special EOS cluster
return pTrack->GetEOS();
#if 0
LoadBlockEntries();
if ((m_entries == NULL) || (m_entries_count <= 0))
return NULL; //return EOS here?
const BlockEntry* pResult = pTrack->GetEOS();
BlockEntry** i = m_entries;
assert(i);
BlockEntry** const j = i + m_entries_count;
while (i != j)
{
const BlockEntry* const pEntry = *i++;
assert(pEntry);
assert(!pEntry->EOS());
const Block* const pBlock = pEntry->GetBlock();
assert(pBlock);
if (pBlock->GetTrackNumber() != pTrack->GetNumber())
continue;
if (pTrack->VetEntry(pEntry))
{
if (time_ns < 0) //just want first candidate block
return pEntry;
const long long ns = pBlock->GetTime(this);
if (ns > time_ns)
break;
pResult = pEntry;
}
else if (time_ns >= 0)
{
const long long ns = pBlock->GetTime(this);
if (ns > time_ns)
break;
}
}
return pResult;
#else
const BlockEntry* pResult = pTrack->GetEOS();
long index = 0;
for (;;)
{
if (index >= m_entries_count)
{
long long pos;
long len;
const long status = Parse(pos, len);
assert(status >= 0);
if (status > 0) //completely parsed, and no more entries
return pResult;
if (status < 0) //should never happen
return 0;
assert(m_entries);
assert(index < m_entries_count);
}
const BlockEntry* const pEntry = m_entries[index];
assert(pEntry);
assert(!pEntry->EOS());
const Block* const pBlock = pEntry->GetBlock();
assert(pBlock);
if (pBlock->GetTrackNumber() != pTrack->GetNumber())
{
++index;
continue;
}
if (pTrack->VetEntry(pEntry))
{
if (time_ns < 0) //just want first candidate block
return pEntry;
const long long ns = pBlock->GetTime(this);
if (ns > time_ns)
return pResult;
pResult = pEntry; //have a candidate
}
else if (time_ns >= 0)
{
const long long ns = pBlock->GetTime(this);
if (ns > time_ns)
return pResult;
}
++index;
}
#endif
}
const BlockEntry*
Cluster::GetEntry(
const CuePoint& cp,
const CuePoint::TrackPosition& tp) const
{
assert(m_pSegment);
#if 0
LoadBlockEntries();
if (m_entries == NULL)
return NULL;
const long long count = m_entries_count;
if (count <= 0)
return NULL;
const long long tc = cp.GetTimeCode();
if ((tp.m_block > 0) && (tp.m_block <= count))
{
const size_t block = static_cast<size_t>(tp.m_block);
const size_t index = block - 1;
const BlockEntry* const pEntry = m_entries[index];
assert(pEntry);
assert(!pEntry->EOS());
const Block* const pBlock = pEntry->GetBlock();
assert(pBlock);
if ((pBlock->GetTrackNumber() == tp.m_track) &&
(pBlock->GetTimeCode(this) == tc))
{
return pEntry;
}
}
const BlockEntry* const* i = m_entries;
const BlockEntry* const* const j = i + count;
while (i != j)
{
#ifdef _DEBUG
const ptrdiff_t idx = i - m_entries;
idx;
#endif
const BlockEntry* const pEntry = *i++;
assert(pEntry);
assert(!pEntry->EOS());
const Block* const pBlock = pEntry->GetBlock();
assert(pBlock);
if (pBlock->GetTrackNumber() != tp.m_track)
continue;
const long long tc_ = pBlock->GetTimeCode(this);
assert(tc_ >= 0);
if (tc_ < tc)
continue;
if (tc_ > tc)
return NULL;
const Tracks* const pTracks = m_pSegment->GetTracks();
assert(pTracks);
const long tn = static_cast<long>(tp.m_track);
const Track* const pTrack = pTracks->GetTrackByNumber(tn);
if (pTrack == NULL)
return NULL;
const long long type = pTrack->GetType();
if (type == 2) //audio
return pEntry;
if (type != 1) //not video
return NULL;
if (!pBlock->IsKey())
return NULL;
return pEntry;
}
return NULL;
#else
const long long tc = cp.GetTimeCode();
if (tp.m_block > 0)
{
const long block = static_cast<long>(tp.m_block);
const long index = block - 1;
while (index >= m_entries_count)
{
long long pos;
long len;
const long status = Parse(pos, len);
if (status < 0) //TODO: can this happen?
return NULL;
if (status > 0) //nothing remains to be parsed
return NULL;
}
const BlockEntry* const pEntry = m_entries[index];
assert(pEntry);
assert(!pEntry->EOS());
const Block* const pBlock = pEntry->GetBlock();
assert(pBlock);
if ((pBlock->GetTrackNumber() == tp.m_track) &&
(pBlock->GetTimeCode(this) == tc))
{
return pEntry;
}
}
long index = 0;
for (;;)
{
if (index >= m_entries_count)
{
long long pos;
long len;
const long status = Parse(pos, len);
if (status < 0) //TODO: can this happen?
return NULL;
if (status > 0) //nothing remains to be parsed
return NULL;
assert(m_entries);
assert(index < m_entries_count);
}
const BlockEntry* const pEntry = m_entries[index];
assert(pEntry);
assert(!pEntry->EOS());
const Block* const pBlock = pEntry->GetBlock();
assert(pBlock);
if (pBlock->GetTrackNumber() != tp.m_track)
{
++index;
continue;
}
const long long tc_ = pBlock->GetTimeCode(this);
assert(tc_ >= 0);
if (tc_ < tc)
{
++index;
continue;
}
if (tc_ > tc)
return NULL;
const Tracks* const pTracks = m_pSegment->GetTracks();
assert(pTracks);
const long tn = static_cast<long>(tp.m_track);
const Track* const pTrack = pTracks->GetTrackByNumber(tn);
if (pTrack == NULL)
return NULL;
const long long type = pTrack->GetType();
if (type == 2) //audio
return pEntry;
if (type != 1) //not video
return NULL;
if (!pBlock->IsKey())
return NULL;
return pEntry;
}
#endif
}
#if 0
const BlockEntry* Cluster::GetMaxKey(const VideoTrack* pTrack) const
{
assert(pTrack);
if (m_pSegment == NULL) //EOS
return pTrack->GetEOS();
LoadBlockEntries();
if ((m_entries == NULL) || (m_entries_count <= 0))
return pTrack->GetEOS();
BlockEntry** i = m_entries + m_entries_count;
BlockEntry** const j = m_entries;
while (i != j)
{
const BlockEntry* const pEntry = *--i;
assert(pEntry);
assert(!pEntry->EOS());
const Block* const pBlock = pEntry->GetBlock();
assert(pBlock);
if (pBlock->GetTrackNumber() != pTrack->GetNumber())
continue;
if (pBlock->IsKey())
return pEntry;
}
return pTrack->GetEOS(); //no satisfactory block found
}
#endif
BlockEntry::BlockEntry(Cluster* p, long idx) :
m_pCluster(p),
m_index(idx)
{
}
BlockEntry::~BlockEntry()
{
}
bool BlockEntry::EOS() const
{
return (GetKind() == kBlockEOS);
}
const Cluster* BlockEntry::GetCluster() const
{
return m_pCluster;
}
long BlockEntry::GetIndex() const
{
return m_index;
}
SimpleBlock::SimpleBlock(
Cluster* pCluster,
long idx,
long long start,
long long size) :
BlockEntry(pCluster, idx),
m_block(start, size)
{
}
long SimpleBlock::Parse()
{
return m_block.Parse(m_pCluster->m_pSegment->m_pReader);
}
BlockEntry::Kind SimpleBlock::GetKind() const
{
return kBlockSimple;
}
const Block* SimpleBlock::GetBlock() const
{
return &m_block;
}
BlockGroup::BlockGroup(
Cluster* pCluster,
long idx,
long long block_start,
long long block_size,
long long prev,
long long next,
long long duration) :
BlockEntry(pCluster, idx),
m_block(block_start, block_size),
m_prev(prev),
m_next(next),
m_duration(duration)
{
}
long BlockGroup::Parse()
{
const long status = m_block.Parse(m_pCluster->m_pSegment->m_pReader);
if (status)
return status;
m_block.SetKey((m_prev > 0) && (m_next <= 0));
return 0;
}
#if 0
void BlockGroup::ParseBlock(long long start, long long size)
{
IMkvReader* const pReader = m_pCluster->m_pSegment->m_pReader;
Block* const pBlock = new Block(start, size, pReader);
assert(pBlock); //TODO
//TODO: the Matroska spec says you have multiple blocks within the
//same block group, with blocks ranked by priority (the flag bits).
assert(m_pBlock == NULL);
m_pBlock = pBlock;
}
#endif
BlockEntry::Kind BlockGroup::GetKind() const
{
return kBlockGroup;
}
const Block* BlockGroup::GetBlock() const
{
return &m_block;
}
long long BlockGroup::GetPrevTimeCode() const
{
return m_prev;
}
long long BlockGroup::GetNextTimeCode() const
{
return m_next;
}
Block::Block(long long start, long long size_) :
m_start(start),
m_size(size_),
m_track(0),
m_timecode(-1),
m_flags(0),
m_frames(NULL),
m_frame_count(-1)
{
}
Block::~Block()
{
delete[] m_frames;
}
long Block::Parse(IMkvReader* pReader)
{
assert(pReader);
assert(m_start >= 0);
assert(m_size >= 0);
assert(m_track <= 0);
assert(m_frames == NULL);
assert(m_frame_count <= 0);
long long pos = m_start;
const long long stop = m_start + m_size;
long len;
m_track = ReadUInt(pReader, pos, len);
if (m_track <= 0)
return E_FILE_FORMAT_INVALID;
if ((pos + len) > stop)
return E_FILE_FORMAT_INVALID;
pos += len; //consume track number
if ((stop - pos) < 2)
return E_FILE_FORMAT_INVALID;
long status;
long long value;
status = UnserializeInt(pReader, pos, 2, value);
if (status)
return E_FILE_FORMAT_INVALID;
if (value < SHRT_MIN)
return E_FILE_FORMAT_INVALID;
if (value > SHRT_MAX)
return E_FILE_FORMAT_INVALID;
m_timecode = static_cast<short>(value);
pos += 2;
if ((stop - pos) <= 0)
return E_FILE_FORMAT_INVALID;
status = pReader->Read(pos, 1, &m_flags);
if (status)
return E_FILE_FORMAT_INVALID;
const int lacing = int(m_flags & 0x06) >> 1;
++pos; //consume flags byte
if (lacing == 0) //no lacing
{
if (pos > stop)
return E_FILE_FORMAT_INVALID;
m_frame_count = 1;
m_frames = new Frame[m_frame_count];
Frame& f = m_frames[0];
f.pos = pos;
const long long frame_size = stop - pos;
if (frame_size > LONG_MAX)
return E_FILE_FORMAT_INVALID;
f.len = static_cast<long>(frame_size);
return 0; //success
}
if (pos >= stop)
return E_FILE_FORMAT_INVALID;
unsigned char biased_count;
status = pReader->Read(pos, 1, &biased_count);
if (status)
return E_FILE_FORMAT_INVALID;
++pos; //consume frame count
assert(pos <= stop);
m_frame_count = int(biased_count) + 1;
m_frames = new Frame[m_frame_count];
assert(m_frames);
if (lacing == 1) //Xiph
{
Frame* pf = m_frames;
Frame* const pf_end = pf + m_frame_count;
long size = 0;
int frame_count = m_frame_count;
while (frame_count > 1)
{
long frame_size = 0;
for (;;)
{
unsigned char val;
if (pos >= stop)
return E_FILE_FORMAT_INVALID;
status = pReader->Read(pos, 1, &val);
if (status)
return E_FILE_FORMAT_INVALID;
++pos; //consume xiph size byte
frame_size += val;
if (val < 255)
break;
}
Frame& f = *pf++;
assert(pf < pf_end);
f.pos = 0; //patch later
f.len = frame_size;
size += frame_size; //contribution of this frame
--frame_count;
}
assert(pf < pf_end);
assert(pos <= stop);
{
Frame& f = *pf++;
if (pf != pf_end)
return E_FILE_FORMAT_INVALID;
f.pos = 0; //patch later
const long long total_size = stop - pos;
if (total_size < size)
return E_FILE_FORMAT_INVALID;
const long long frame_size = total_size - size;
if (frame_size > LONG_MAX)
return E_FILE_FORMAT_INVALID;
f.len = static_cast<long>(frame_size);
}
pf = m_frames;
while (pf != pf_end)
{
Frame& f = *pf++;
assert((pos + f.len) <= stop);
f.pos = pos;
pos += f.len;
}
assert(pos == stop);
}
else if (lacing == 2) //fixed-size lacing
{
const long long total_size = stop - pos;
if ((total_size % m_frame_count) != 0)
return E_FILE_FORMAT_INVALID;
const long long frame_size = total_size / m_frame_count;
if (frame_size > LONG_MAX)
return E_FILE_FORMAT_INVALID;
Frame* pf = m_frames;
Frame* const pf_end = pf + m_frame_count;
while (pf != pf_end)
{
assert((pos + frame_size) <= stop);
Frame& f = *pf++;
f.pos = pos;
f.len = static_cast<long>(frame_size);
pos += frame_size;
}
assert(pos == stop);
}
else
{
assert(lacing == 3); //EBML lacing
if (pos >= stop)
return E_FILE_FORMAT_INVALID;
long size = 0;
int frame_count = m_frame_count;
long long frame_size = ReadUInt(pReader, pos, len);
if (frame_size < 0)
return E_FILE_FORMAT_INVALID;
if (frame_size > LONG_MAX)
return E_FILE_FORMAT_INVALID;
if ((pos + len) > stop)
return E_FILE_FORMAT_INVALID;
pos += len; //consume length of size of first frame
if ((pos + frame_size) > stop)
return E_FILE_FORMAT_INVALID;
Frame* pf = m_frames;
Frame* const pf_end = pf + m_frame_count;
{
Frame& curr = *pf;
curr.pos = 0; //patch later
curr.len = static_cast<long>(frame_size);
size += curr.len; //contribution of this frame
}
--frame_count;
while (frame_count > 1)
{
if (pos >= stop)
return E_FILE_FORMAT_INVALID;
assert(pf < pf_end);
const Frame& prev = *pf++;
assert(prev.len == frame_size);
assert(pf < pf_end);
Frame& curr = *pf;
curr.pos = 0; //patch later
const long long delta_size_ = ReadUInt(pReader, pos, len);
if (delta_size_ < 0)
return E_FILE_FORMAT_INVALID;
if ((pos + len) > stop)
return E_FILE_FORMAT_INVALID;
pos += len; //consume length of (delta) size
assert(pos <= stop);
const int exp = 7*len - 1;
const long long bias = (1LL << exp) - 1LL;
const long long delta_size = delta_size_ - bias;
frame_size += delta_size;
if (frame_size < 0)
return E_FILE_FORMAT_INVALID;
if (frame_size > LONG_MAX)
return E_FILE_FORMAT_INVALID;
curr.len = static_cast<long>(frame_size);
size += curr.len; //contribution of this frame
--frame_count;
}
{
assert(pos <= stop);
assert(pf < pf_end);
const Frame& prev = *pf++;
assert(prev.len == frame_size);
assert(pf < pf_end);
Frame& curr = *pf++;
assert(pf == pf_end);
curr.pos = 0; //patch later
const long long total_size = stop - pos;
if (total_size < size)
return E_FILE_FORMAT_INVALID;
frame_size = total_size - size;
if (frame_size > LONG_MAX)
return E_FILE_FORMAT_INVALID;
curr.len = static_cast<long>(frame_size);
}
pf = m_frames;
while (pf != pf_end)
{
Frame& f = *pf++;
assert((pos + f.len) <= stop);
f.pos = pos;
pos += f.len;
}
assert(pos == stop);
}
return 0; //success
}
long long Block::GetTimeCode(const Cluster* pCluster) const
{
if (pCluster == 0)
return m_timecode;
const long long tc0 = pCluster->GetTimeCode();
assert(tc0 >= 0);
const long long tc = tc0 + m_timecode;
assert(tc >= 0);
return tc; //unscaled timecode units
}
long long Block::GetTime(const Cluster* pCluster) const
{
assert(pCluster);
const long long tc = GetTimeCode(pCluster);
const Segment* const pSegment = pCluster->m_pSegment;
const SegmentInfo* const pInfo = pSegment->GetInfo();
assert(pInfo);
const long long scale = pInfo->GetTimeCodeScale();
assert(scale >= 1);
const long long ns = tc * scale;
return ns;
}
long long Block::GetTrackNumber() const
{
return m_track;
}
bool Block::IsKey() const
{
return ((m_flags & static_cast<unsigned char>(1 << 7)) != 0);
}
void Block::SetKey(bool bKey)
{
if (bKey)
m_flags |= static_cast<unsigned char>(1 << 7);
else
m_flags &= 0x7F;
}
bool Block::IsInvisible() const
{
return bool(int(m_flags & 0x08) != 0);
}
Block::Lacing Block::GetLacing() const
{
const int value = int(m_flags & 0x06) >> 1;
return static_cast<Lacing>(value);
}
int Block::GetFrameCount() const
{
return m_frame_count;
}
const Block::Frame& Block::GetFrame(int idx) const
{
assert(idx >= 0);
assert(idx < m_frame_count);
const Frame& f = m_frames[idx];
assert(f.pos > 0);
assert(f.len > 0);
return f;
}
long Block::Frame::Read(IMkvReader* pReader, unsigned char* buf) const
{
assert(pReader);
assert(buf);
const long status = pReader->Read(pos, len, buf);
return status;
}
} //end namespace mkvparser