def process_video_track(self, elements):
track = core.VideoStream()
# Defaults
track.codec = u'Unknown'
track.fps = 0
self.set_track_defaults(track)
for elem in elements:
elem_id = elem.get_id()
if elem_id == MATROSKA_CODEC_ID:
track.codec = elem.get_str()
elif elem_id == MATROSKA_FRAME_DURATION_ID:
try:
track.fps = 1 / (pow(10, -9) * (elem.get_value()))
except ZeroDivisionError:
pass
elif elem_id == MATROSKA_VIDEO_SETTINGS_ID:
d_width = d_height = None
for settings_elem in self.process_one_level(elem):
settings_elem_id = settings_elem.get_id()
if settings_elem_id == MATROSKA_VIDEO_WIDTH_ID:
track.width = settings_elem.get_value()
elif settings_elem_id == MATROSKA_VIDEO_HEIGHT_ID:
track.height = settings_elem.get_value()
elif settings_elem_id == MATROSKA_VIDEO_DISPLAY_WIDTH_ID:
d_width = settings_elem.get_value()
elif settings_elem_id == MATROSKA_VIDEO_DISPLAY_HEIGHT_ID:
d_height = settings_elem.get_value()
elif settings_elem_id == MATROSKA_VIDEO_INTERLACED_ID:
value = int(settings_elem.get_value())
self._set('interlaced', value)
if None not in [d_width, d_height]:
track.aspect = float(d_width) / d_height
else:
self.process_track_common(elem, track)
# convert codec information
# http://haali.cs.msu.ru/mkv/codecs.pdf
if track.codec in FOURCCMap:
track.codec = FOURCCMap[track.codec]
elif '/' in track.codec and track.codec.split(
'/')[0] + '/' in FOURCCMap:
track.codec = FOURCCMap[track.codec.split('/')[0] + '/']
elif track.codec.endswith('FOURCC') and len(track.codec_private
or '') == 40:
track.codec = track.codec_private[16:20]
elif track.codec.startswith('V_REAL/'):
track.codec = track.codec[7:]
elif track.codec.startswith('V_'):
# FIXME: add more video codecs here
track.codec = track.codec[2:]
track.id = len(self.video)
self.video.append(track)
return track
def parseDisc(self, device):
type = None
if self.is_disc(device) != 2:
raise core.ParseError()
f = open(device)
try:
# read CD-XA001 at byte 1024 in sector 16
f.seek(2048 * 16 + 1024, 0)
if f.read(8) != 'CD-XA001':
raise core.ParseError()
# read VIDEO_CD at sector 150
f.seek(2048 * 150, 0)
typebuffer = f.read(8)
if typebuffer != 'VIDEO_CD' and typebuffer != 'SUPERVCD':
raise core.ParseError()
# Read some bytes of the ISO9660 part to better guess VCD or SVCD.
# Maybe this code is not needed and VIDEO_CD and SUPERVCD are enough.
f.seek(2048 * 16, 0)
iso9660 = f.read(2048 * 16)
if typebuffer == 'VIDEO_CD' and iso9660.find('MPEGAV') > 0:
type = 'VCD'
elif typebuffer == 'SUPERVCD' and \
(iso9660.find('SVCD') > 0 or iso9660.find('MPEG2') > 0):
type = 'SVCD'
else:
raise core.ParseError()
finally:
f.close()
# read the tracks to generate the title list
device = open(device)
(first, last) = cdrom.audiocd_toc_header(device)
lmin = 0
lsec = 0
num = 0
for i in range(first, last + 2):
if i == last + 1:
min, sec, frames = cdrom.audiocd_leadout(device)
else:
min, sec, frames = cdrom.audiocd_toc_entry(device, i)
if num:
vi = core.VideoStream()
# XXX add more static information here, it's also possible
# XXX to scan for more informations like fps
# XXX Settings to MPEG1/2 is a wild guess, maybe the track
# XXX isn't playable at all (e.g. the menu)
if type == 'VCD':
vi.codec = 'MPEG1'
else:
vi.codec = 'MPEG2'
vi.length = (min - lmin) * 60 + (sec - lsec)
self.tracks.append(vi)
num += 1
lmin, lsec = min, sec
device.close()
def __init__(self, file):
core.AVContainer.__init__(self)
self.sequence_header_offset = 0
self.mpeg_version = 2
# detect TS (fast scan)
if not self.isTS(file):
# detect system mpeg (many infos)
if not self.isMPEG(file):
# detect PES
if not self.isPES(file):
# Maybe it's MPEG-ES
if self.isES(file):
# If isES() succeeds, we needn't do anything further.
return
if file.name.lower().endswith('mpeg') or \
file.name.lower().endswith('mpg'):
# This has to be an mpeg file. It could be a bad
# recording from an ivtv based hardware encoder with
# same bytes missing at the beginning.
# Do some more digging...
if not self.isMPEG(file, force=True) or \
not self.video or not self.audio:
# does not look like an mpeg at all
raise ParseError()
else:
# no mpeg at all
raise ParseError()
self.mime = 'video/mpeg'
if not self.video:
self.video.append(core.VideoStream())
if self.sequence_header_offset <= 0:
return
self.progressive(file)
for vi in self.video:
vi.width, vi.height = self.dxy(file)
vi.fps, vi.aspect = self.framerate_aspect(file)
vi.bitrate = self.bitrate(file)
if self.length:
vi.length = self.length
if not self.type:
self.type = 'MPEG Video'
# set fourcc codec for video and audio
vc, ac = 'MP2V', 'MP2A'
if self.mpeg_version == 1:
vc, ac = 'MPEG', 0x0050
for v in self.video:
v.codec = vc
for a in self.audio:
if not a.codec:
a.codec = ac
def _parseSTRF(self, t, strh):
fccType = strh['fccType']
retval = {}
if fccType == 'auds':
(
retval['wFormatTag'],
retval['nChannels'],
retval['nSamplesPerSec'],
retval['nAvgBytesPerSec'],
retval['nBlockAlign'],
retval['nBitsPerSample'],
) = struct.unpack('<HHHHHH', t[0:12])
ai = core.AudioStream()
ai.samplerate = retval['nSamplesPerSec']
ai.channels = retval['nChannels']
# FIXME: Bitrate calculation is completely wrong.
#ai.samplebits = retval['nBitsPerSample']
#ai.bitrate = retval['nAvgBytesPerSec'] * 8
# TODO: set code if possible
# http://www.stats.uwa.edu.au/Internal/Specs/DXALL/FileSpec/\
# Languages
# ai.language = strh['wLanguage']
ai.codec = retval['wFormatTag']
self.audio.append(ai)
elif fccType == 'vids':
v = struct.unpack('<IIIHH', t[0:16])
(
retval['biSize'],
retval['biWidth'],
retval['biHeight'],
retval['biPlanes'],
retval['biBitCount'],
) = v
v = struct.unpack('IIIII', t[20:40])
(
retval['biSizeImage'],
retval['biXPelsPerMeter'],
retval['biYPelsPerMeter'],
retval['biClrUsed'],
retval['biClrImportant'],
) = v
vi = core.VideoStream()
vi.codec = t[16:20]
vi.width = retval['biWidth']
vi.height = retval['biHeight']
# FIXME: Bitrate calculation is completely wrong.
#vi.bitrate = strh['dwRate']
vi.fps = float(strh['dwRate']) / strh['dwScale']
vi.length = strh['dwLength'] / vi.fps
self.video.append(vi)
return retval
def _read_header(self, object_id, s):
if object_id == 'PROP':
prop = struct.unpack('>9IHH', s)
log.debug('PROP: %r' % prop)
if object_id == 'MDPR':
mdpr = struct.unpack('>H7I', s[:30])
log.debug('MDPR: %r' % mdpr)
self.length = mdpr[7] / 1000.0
(stream_name_size, ) = struct.unpack('>B', s[30:31])
stream_name = s[31:31 + stream_name_size]
pos = 31 + stream_name_size
(mime_type_size, ) = struct.unpack('>B', s[pos:pos + 1])
mime = s[pos + 1:pos + 1 + mime_type_size]
pos += mime_type_size + 1
(type_specific_len, ) = struct.unpack('>I', s[pos:pos + 4])
type_specific = s[pos + 4:pos + 4 + type_specific_len]
pos += 4 + type_specific_len
if mime[:5] == 'audio':
ai = core.AudioStream()
ai.id = mdpr[0]
ai.bitrate = mdpr[2]
self.audio.append(ai)
elif mime[:5] == 'video':
vi = core.VideoStream()
vi.id = mdpr[0]
vi.bitrate = mdpr[2]
self.video.append(vi)
else:
log.debug('Unknown: %r' % mime)
if object_id == 'CONT':
pos = 0
(title_len, ) = struct.unpack('>H', s[pos:pos + 2])
self.title = s[2:title_len + 2]
pos += title_len + 2
(author_len, ) = struct.unpack('>H', s[pos:pos + 2])
self.artist = s[pos + 2:pos + author_len + 2]
pos += author_len + 2
(copyright_len, ) = struct.unpack('>H', s[pos:pos + 2])
self.copyright = s[pos + 2:pos + copyright_len + 2]
pos += copyright_len + 2
(comment_len, ) = struct.unpack('>H', s[pos:pos + 2])
self.comment = s[pos + 2:pos + comment_len + 2]
def isES(self, file):
file.seek(0, 0)
try:
header = struct.unpack('>LL', file.read(8))
except (struct.error, IOError):
return False
if header[0] != 0x1B3:
return False
# Is an mpeg video elementary stream
self.mime = 'video/mpeg'
video = core.VideoStream()
video.width = header[1] >> 20
video.height = (header[1] >> 8) & 0xfff
if header[1] & 0xf < len(FRAME_RATE):
video.fps = FRAME_RATE[header[1] & 0xf]
if (header[1] >> 4) & 0xf < len(ASPECT_RATIO):
# FIXME: Empirically the aspect looks like PAR rather than DAR
video.aspect = ASPECT_RATIO[(header[1] >> 4) & 0xf]
self.video.append(video)
return True
class MPEG4(core.AVContainer):
"""
Parser for the MP4 container format. This format is mostly
identical to Apple Quicktime and 3GP files. It maps to mp4, mov,
qt and some other extensions.
"""
table_mapping = {'QTUDTA': QTUDTA}
def __init__(self, file):
core.AVContainer.__init__(self)
self._references = []
self.mime = 'video/quicktime'
self.type = 'Quicktime Video'
h = file.read(8)
try:
(size, type) = struct.unpack('>I4s', h)
except struct.error:
# EOF.
raise ParseError()
if type == 'ftyp':
# file type information
if size >= 12:
# this should always happen
if file.read(4) != 'qt ':
# not a quicktime movie, it is a mpeg4 container
self.mime = 'video/mp4'
self.type = 'MPEG-4 Video'
size -= 4
file.seek(size - 8, 1)
h = file.read(8)
(size, type) = struct.unpack('>I4s', h)
while type in ['mdat', 'skip']:
# movie data at the beginning, skip
file.seek(size - 8, 1)
h = file.read(8)
(size, type) = struct.unpack('>I4s', h)
if not type in ['moov', 'wide', 'free']:
log.debug(u'invalid header: %r' % type)
raise ParseError()
# Extended size
if size == 1:
size = struct.unpack('>Q', file.read(8))
# Back over the atom header we just read, since _readatom expects the
# file position to be at the start of an atom.
file.seek(-8, 1)
while self._readatom(file):
pass
if self._references:
self._set('references', self._references)
def _readatom(self, file):
s = file.read(8)
if len(s) < 8:
return 0
atomsize, atomtype = struct.unpack('>I4s', s)
if not str(atomtype).decode('latin1').isalnum():
# stop at nonsense data
return 0
log.debug(u'%r [%X]' % (atomtype, atomsize))
if atomtype == 'udta':
# Userdata (Metadata)
pos = 0
tabl = {}
i18ntabl = {}
atomdata = file.read(atomsize - 8)
while pos < atomsize - 12:
(datasize, datatype) = struct.unpack('>I4s', atomdata[pos:pos + 8])
if ord(datatype[0]) == 169:
# i18n Metadata...
mypos = 8 + pos
while mypos + 4 < datasize + pos:
# first 4 Bytes are i18n header
(tlen, lang) = struct.unpack('>HH', atomdata[mypos:mypos + 4])
i18ntabl[lang] = i18ntabl.get(lang, {})
l = atomdata[mypos + 4:mypos + tlen + 4]
i18ntabl[lang][datatype[1:]] = l
mypos += tlen + 4
elif datatype == 'WLOC':
# Drop Window Location
pass
else:
if ord(atomdata[pos + 8:pos + datasize][0]) > 1:
tabl[datatype] = atomdata[pos + 8:pos + datasize]
pos += datasize
if len(i18ntabl.keys()) > 0:
for k in i18ntabl.keys():
if QTLANGUAGES.has_key(k) and QTLANGUAGES[k] == 'en':
self._appendtable('QTUDTA', i18ntabl[k])
self._appendtable('QTUDTA', tabl)
else:
log.debug(u'NO i18')
self._appendtable('QTUDTA', tabl)
elif atomtype == 'trak':
atomdata = file.read(atomsize - 8)
pos = 0
trackinfo = {}
tracktype = None
while pos < atomsize - 8:
(datasize, datatype) = struct.unpack('>I4s', atomdata[pos:pos + 8])
if datatype == 'tkhd':
tkhd = struct.unpack('>6I8x4H36xII', atomdata[pos + 8:pos + datasize])
trackinfo['width'] = tkhd[10] >> 16
trackinfo['height'] = tkhd[11] >> 16
trackinfo['id'] = tkhd[3]
try:
# XXX Timestamp of Seconds is since January 1st 1904!
# XXX 2082844800 is the difference between Unix and
# XXX Apple time. FIXME to work on Apple, too
self.timestamp = int(tkhd[1]) - 2082844800
except Exception, e:
log.exception(u'There was trouble extracting timestamp')
elif datatype == 'mdia':
pos += 8
datasize -= 8
log.debug(u'--> mdia information')
while datasize:
mdia = struct.unpack('>I4s', atomdata[pos:pos + 8])
if mdia[1] == 'mdhd':
# Parse based on version of mdhd header. See
# http://wiki.multimedia.cx/index.php?title=QuickTime_container#mdhd
ver = ord(atomdata[pos + 8])
if ver == 0:
mdhd = struct.unpack('>IIIIIhh', atomdata[pos + 8:pos + 8 + 24])
elif ver == 1:
mdhd = struct.unpack('>IQQIQhh', atomdata[pos + 8:pos + 8 + 36])
else:
mdhd = None
if mdhd:
# duration / time scale
trackinfo['length'] = mdhd[4] / mdhd[3]
if mdhd[5] in QTLANGUAGES:
trackinfo['language'] = QTLANGUAGES[mdhd[5]]
# mdhd[6] == quality
self.length = max(self.length, mdhd[4] / mdhd[3])
elif mdia[1] == 'minf':
# minf has only atoms inside
pos -= (mdia[0] - 8)
datasize += (mdia[0] - 8)
elif mdia[1] == 'stbl':
# stbl has only atoms inside
pos -= (mdia[0] - 8)
datasize += (mdia[0] - 8)
elif mdia[1] == 'hdlr':
hdlr = struct.unpack('>I4s4s', atomdata[pos + 8:pos + 8 + 12])
if hdlr[1] == 'mhlr':
if hdlr[2] == 'vide':
tracktype = 'video'
if hdlr[2] == 'soun':
tracktype = 'audio'
elif mdia[1] == 'stsd':
stsd = struct.unpack('>2I', atomdata[pos + 8:pos + 8 + 8])
if stsd[1] > 0:
codec = atomdata[pos + 16:pos + 16 + 8]
codec = struct.unpack('>I4s', codec)
trackinfo['codec'] = codec[1]
if codec[1] == 'jpeg':
tracktype = 'image'
elif mdia[1] == 'dinf':
dref = struct.unpack('>I4s', atomdata[pos + 8:pos + 8 + 8])
log.debug(u' --> %r, %r (useless)' % mdia)
if dref[1] == 'dref':
num = struct.unpack('>I', atomdata[pos + 20:pos + 20 + 4])[0]
rpos = pos + 20 + 4
for ref in range(num):
# FIXME: do somthing if this references
ref = struct.unpack('>I3s', atomdata[rpos:rpos + 7])
data = atomdata[rpos + 7:rpos + ref[0]]
rpos += ref[0]
else:
if mdia[1].startswith('st'):
log.debug(u' --> %r, %r (sample)' % mdia)
elif mdia[1] == 'vmhd' and not tracktype:
# indicates that this track is video
tracktype = 'video'
elif mdia[1] in ['vmhd', 'smhd'] and not tracktype:
# indicates that this track is audio
tracktype = 'audio'
else:
log.debug(u' --> %r, %r (unknown)' % mdia)
pos += mdia[0]
datasize -= mdia[0]
elif datatype == 'udta':
log.debug(u'udta: %r' % struct.unpack('>I4s', atomdata[:8]))
else:
if datatype == 'edts':
log.debug(u'--> %r [%d] (edit list)' % \
(datatype, datasize))
else:
log.debug(u'--> %r [%d] (unknown)' % \
(datatype, datasize))
pos += datasize
info = None
if tracktype == 'video':
info = core.VideoStream()
self.video.append(info)
if tracktype == 'audio':
info = core.AudioStream()
self.audio.append(info)
if info:
for key, value in trackinfo.items():
setattr(info, key, value)
def _parseHeader(self, header, granule):
headerlen = len(header)
flags = ord(header[0])
if headerlen >= 30 and header[1:7] == 'vorbis':
ai = core.AudioStream()
ai.version, ai.channels, ai.samplerate, bitrate_max, ai.bitrate, \
bitrate_min, blocksize, framing = \
struct.unpack('<IBIiiiBB', header[7:7 + 23])
ai.codec = 'Vorbis'
#ai.granule = granule
#ai.length = granule / ai.samplerate
self.audio.append(ai)
self.all_streams.append(ai)
elif headerlen >= 7 and header[1:7] == 'theora':
# Theora Header
# XXX Finish Me
vi = core.VideoStream()
vi.codec = 'theora'
self.video.append(vi)
self.all_streams.append(vi)
elif headerlen >= 142 and \
header[1:36] == 'Direct Show Samples embedded in Ogg':
# Old Directshow format
# XXX Finish Me
vi = core.VideoStream()
vi.codec = 'dshow'
self.video.append(vi)
self.all_streams.append(vi)
elif flags & PACKET_TYPE_BITS == PACKET_TYPE_HEADER and \
headerlen >= struct.calcsize(STREAM_HEADER_VIDEO) + 1:
# New Directshow Format
htype = header[1:9]
if htype[:5] == 'video':
sh = header[9:struct.calcsize(STREAM_HEADER_VIDEO) + 9]
streamheader = struct.unpack(STREAM_HEADER_VIDEO, sh)
vi = core.VideoStream()
(type, ssize, timeunit, samplerate, vi.length, buffersize, \
vi.bitrate, vi.width, vi.height) = streamheader
vi.width /= 65536
vi.height /= 65536
# XXX length, bitrate are very wrong
vi.codec = type
vi.fps = 10000000 / timeunit
self.video.append(vi)
self.all_streams.append(vi)
elif htype[:5] == 'audio':
sha = header[9:struct.calcsize(STREAM_HEADER_AUDIO) + 9]
streamheader = struct.unpack(STREAM_HEADER_AUDIO, sha)
ai = core.AudioStream()
(type, ssize, timeunit, ai.samplerate, ai.length, buffersize, \
ai.bitrate, ai.channels, bloc, ai.bitrate) = streamheader
self.samplerate = ai.samplerate
log.debug(u'Samplerate %d' % self.samplerate)
self.audio.append(ai)
self.all_streams.append(ai)
elif htype[:4] == 'text':
subtitle = core.Subtitle()
# FIXME: add more info
self.subtitles.append(subtitle)
self.all_streams.append(subtitle)
else:
log.debug(u'Unknown Header')
def __init__(self, file):
core.AVContainer.__init__(self)
self.mime = 'video/flv'
self.type = 'Flash Video'
data = file.read(13)
if len(data) < 13 or struct.unpack('>3sBBII', data)[0] != 'FLV':
raise ParseError()
for _ in range(10):
if self.audio and self.video:
break
data = file.read(11)
if len(data) < 11:
break
chunk = struct.unpack('>BH4BI', data)
size = (chunk[1] << 8) + chunk[2]
if chunk[0] == FLV_TAG_TYPE_AUDIO:
flags = ord(file.read(1))
if not self.audio:
a = core.AudioStream()
a.channels = (flags & FLV_AUDIO_CHANNEL_MASK) + 1
srate = (flags & FLV_AUDIO_SAMPLERATE_MASK)
a.samplerate = (
44100 << (srate >> FLV_AUDIO_SAMPLERATE_OFFSET) >> 3)
codec = (flags & FLV_AUDIO_CODECID_MASK
) >> FLV_AUDIO_CODECID_OFFSET
if codec < len(FLV_AUDIO_CODECID):
a.codec = FLV_AUDIO_CODECID[codec]
self.audio.append(a)
file.seek(size - 1, 1)
elif chunk[0] == FLV_TAG_TYPE_VIDEO:
flags = ord(file.read(1))
if not self.video:
v = core.VideoStream()
codec = (flags & FLV_VIDEO_CODECID_MASK) - 2
if codec < len(FLV_VIDEO_CODECID):
v.codec = FLV_VIDEO_CODECID[codec]
# width and height are in the meta packet, but I have
# no file with such a packet inside. So maybe we have
# to decode some parts of the video.
self.video.append(v)
file.seek(size - 1, 1)
elif chunk[0] == FLV_TAG_TYPE_META:
log.info('metadata %r', str(chunk))
metadata = file.read(size)
try:
while metadata:
length, value = self._parse_value(metadata)
if isinstance(value, dict):
log.info('metadata: %r', value)
if value.get('creator'):
self.copyright = value.get('creator')
if value.get('width'):
self.width = value.get('width')
if value.get('height'):
self.height = value.get('height')
if value.get('duration'):
self.length = value.get('duration')
self._appendtable('FLVINFO', value)
if not length:
# parse error
break
metadata = metadata[length:]
except (IndexError, struct.error, TypeError):
pass
else:
log.info('unkown %r', str(chunk))
file.seek(size, 1)
file.seek(4, 1)
def ReadPESHeader(self, offset, buffer, id=0):
"""
Parse a PES header.
Since it starts with 0x00 0x00 0x01 like 'normal' mpegs, this
function will return (0, None) when it is no PES header or
(packet length, timestamp position (maybe None))
http://dvd.sourceforge.net/dvdinfo/pes-hdr.html
"""
if not buffer[0:3] == '\x00\x00\x01':
return 0, None
packet_length = (ord(buffer[4]) << 8) + ord(buffer[5]) + 6
align = ord(buffer[6]) & 4
header_length = ord(buffer[8])
# PES ID (starting with 001)
if ord(buffer[3]) & 0xE0 == 0xC0:
id = id or ord(buffer[3]) & 0x1F
for a in self.audio:
if a.id == id:
break
else:
self.audio.append(core.AudioStream())
self.audio[-1]._set('id', id)
elif ord(buffer[3]) & 0xF0 == 0xE0:
id = id or ord(buffer[3]) & 0xF
for v in self.video:
if v.id == id:
break
else:
self.video.append(core.VideoStream())
self.video[-1]._set('id', id)
# new mpeg starting
if buffer[header_length + 9:header_length + 13] == \
'\x00\x00\x01\xB3' and not self.sequence_header_offset:
# yes, remember offset for later use
self.sequence_header_offset = offset + header_length + 9
elif ord(buffer[3]) == 189 or ord(buffer[3]) == 191:
# private stream. we don't know, but maybe we can guess later
id = id or ord(buffer[3]) & 0xF
if align and \
buffer[header_length + 9:header_length + 11] == '\x0b\x77':
# AC3 stream
for a in self.audio:
if a.id == id:
break
else:
self.audio.append(core.AudioStream())
self.audio[-1]._set('id', id)
self.audio[-1].codec = 0x2000 # AC3
else:
# unknown content
pass
ptsdts = ord(buffer[7]) >> 6
if ptsdts and ptsdts == ord(buffer[9]) >> 4:
if ord(buffer[9]) >> 4 != ptsdts:
log.warning(u'WARNING: bad PTS/DTS, please contact us')
return packet_length, None
# timestamp = self.ReadPTS(buffer[9:14])
high = ((ord(buffer[9]) & 0xF) >> 1)
med = (ord(buffer[10]) << 7) + (ord(buffer[11]) >> 1)
low = (ord(buffer[12]) << 7) + (ord(buffer[13]) >> 1)
return packet_length, 9
return packet_length, None
def ReadHeader(self, buffer, offset):
"""
Handle MPEG header in buffer on position offset
Return None on error, new offset or 0 if the new offset can't be scanned
"""
if buffer[offset:offset + 3] != '\x00\x00\x01':
return None
id = ord(buffer[offset + 3])
if id == PADDING_PKT:
return offset + (ord(buffer[offset + 4]) << 8) + \
ord(buffer[offset + 5]) + 6
if id == PACK_PKT:
if ord(buffer[offset + 4]) & 0xF0 == 0x20:
self.type = 'MPEG-1 Video'
self.get_time = self.ReadSCRMpeg1
self.mpeg_version = 1
return offset + 12
elif (ord(buffer[offset + 4]) & 0xC0) == 0x40:
self.type = 'MPEG-2 Video'
self.get_time = self.ReadSCRMpeg2
return offset + (ord(buffer[offset + 13]) & 0x07) + 14
else:
# I have no idea what just happened, but for some DVB
# recordings done with mencoder this points to a
# PACK_PKT describing something odd. Returning 0 here
# (let's hope there are no extensions in the header)
# fixes it.
return 0
if 0xC0 <= id <= 0xDF:
# code for audio stream
for a in self.audio:
if a.id == id:
break
else:
self.audio.append(core.AudioStream())
self.audio[-1]._set('id', id)
return 0
if 0xE0 <= id <= 0xEF:
# code for video stream
for v in self.video:
if v.id == id:
break
else:
self.video.append(core.VideoStream())
self.video[-1]._set('id', id)
return 0
if id == SEQ_HEAD:
# sequence header, remember that position for later use
self.sequence_header_offset = offset
return 0
if id in [PRIVATE_STREAM1, PRIVATE_STREAM2]:
# private stream. we don't know, but maybe we can guess later
add = ord(buffer[offset + 8])
# if (ord(buffer[offset+6]) & 4) or 1:
# id = ord(buffer[offset+10+add])
if buffer[offset + 11 + add:offset + 15 +
add].find('\x0b\x77') != -1:
# AC3 stream
for a in self.audio:
if a.id == id:
break
else:
self.audio.append(core.AudioStream())
self.audio[-1]._set('id', id)
self.audio[-1].codec = 0x2000 # AC3
return 0
if id == SYS_PKT:
return 0
if id == EXT_START:
return 0
return 0
def _getnextheader(self, s):
r = struct.unpack('<16sQ', s[:24])
(guidstr, objsize) = r
guid = self._parseguid(guidstr)
if guid == GUIDS['ASF_File_Properties_Object']:
log.debug('File Properties Object')
val = struct.unpack('<16s6Q4I', s[24:24 + 80])
(fileid, size, date, packetcount, duration, senddur, preroll, flags, minpack, maxpack, maxbr) = val
# FIXME: parse date to timestamp
self.length = duration / 10000000.0
elif guid == GUIDS['ASF_Stream_Properties_Object']:
log.debug('Stream Properties Object [%d]' % objsize)
streamtype = self._parseguid(s[24:40])
errortype = self._parseguid(s[40:56])
offset, typelen, errorlen, flags = struct.unpack('<QIIH', s[56:74])
strno = flags & 0x7f
encrypted = flags >> 15
if encrypted:
self._set('encrypted', True)
if streamtype == GUIDS['ASF_Video_Media']:
vi = core.VideoStream()
vi.width, vi.height, depth, codec, = struct.unpack('<4xII2xH4s', s[89:89 + 20])
vi.codec = codec
vi.id = strno
self.video.append(vi)
elif streamtype == GUIDS['ASF_Audio_Media']:
ai = core.AudioStream()
twocc, ai.channels, ai.samplerate, bitrate, block, \
ai.samplebits, = struct.unpack('<HHIIHH', s[78:78 + 16])
ai.bitrate = 8 * bitrate
ai.codec = twocc
ai.id = strno
self.audio.append(ai)
self._apply_extinfo(strno)
elif guid == GUIDS['ASF_Extended_Stream_Properties_Object']:
streamid, langid, frametime = struct.unpack('<HHQ', s[72:84])
(bitrate,) = struct.unpack('<I', s[40:40 + 4])
if streamid not in self._extinfo:
self._extinfo[streamid] = [None, None, None, {}]
if frametime == 0:
# Problaby VFR, report as 1000fps (which is what MPlayer does)
frametime = 10000.0
self._extinfo[streamid][:3] = [bitrate, 10000000.0 / frametime, langid]
self._apply_extinfo(streamid)
elif guid == GUIDS['ASF_Header_Extension_Object']:
log.debug('ASF_Header_Extension_Object %d' % objsize)
size = struct.unpack('<I', s[42:46])[0]
data = s[46:46 + size]
while len(data):
log.debug('Sub:')
h = self._getnextheader(data)
data = data[h[1]:]
elif guid == GUIDS['ASF_Codec_List_Object']:
log.debug('List Object')
pass
elif guid == GUIDS['ASF_Error_Correction_Object']:
log.debug('Error Correction')
pass
elif guid == GUIDS['ASF_Content_Description_Object']:
log.debug('Content Description Object')
val = struct.unpack('<5H', s[24:24 + 10])
pos = 34
strings = []
for i in val:
ss = s[pos:pos + i].replace('\0', '').lstrip().rstrip()
strings.append(ss)
pos += i
# Set empty strings to None
strings = [x or None for x in strings]
self.title, self.artist, self.copyright, self.caption, rating = strings
elif guid == GUIDS['ASF_Extended_Content_Description_Object']:
(count,) = struct.unpack('<H', s[24:26])
pos = 26
descriptor = {}
for i in range(0, count):
# Read additional content descriptors
d = self._parsekv(s[pos:])
pos += d[0]
descriptor[d[1]] = d[2]
self._appendtable('ASFDESCRIPTOR', descriptor)
elif guid == GUIDS['ASF_Metadata_Object']:
(count,) = struct.unpack('<H', s[24:26])
pos = 26
streams = {}
for i in range(0, count):
# Read additional content descriptors
size, key, value, strno = self._parsekv2(s[pos:])
if strno not in streams:
streams[strno] = {}
streams[strno][key] = value
pos += size
for strno, metadata in streams.items():
if strno not in self._extinfo:
self._extinfo[strno] = [None, None, None, {}]
self._extinfo[strno][3].update(metadata)
self._apply_extinfo(strno)
elif guid == GUIDS['ASF_Language_List_Object']:
count = struct.unpack('<H', s[24:26])[0]
pos = 26
for i in range(0, count):
idlen = struct.unpack('<B', s[pos:pos + 1])[0]
idstring = s[pos + 1:pos + 1 + idlen]
idstring = unicode(idstring, 'utf-16').replace('\0', '')
log.debug('Language: %d/%d: %r' % (i + 1, count, idstring))
self._languages.append(idstring)
pos += 1 + idlen
elif guid == GUIDS['ASF_Stream_Bitrate_Properties_Object']:
# This record contains stream bitrate with payload overhead. For
# audio streams, we should have the average bitrate from
# ASF_Stream_Properties_Object. For video streams, we get it from
# ASF_Extended_Stream_Properties_Object. So this record is not
# used.
pass
elif guid == GUIDS['ASF_Content_Encryption_Object'] or \
guid == GUIDS['ASF_Extended_Content_Encryption_Object']:
self._set('encrypted', True)
else:
# Just print the type:
for h in GUIDS.keys():
if GUIDS[h] == guid:
log.debug('Unparsed %r [%d]' % (h, objsize))
break
else:
u = "%.8X-%.4X-%.4X-%.2X%.2X-%s" % guid
log.debug('unknown: len=%d [%d]' % (len(u), objsize))
return r
