SickGear/lib/enzyme/mp4.py
echel0n 0d9fbc1ad7 Welcome to our SickBeard-TVRage Edition ...
This version of SickBeard uses both TVDB and TVRage to search and gather it's series data from allowing you to now have access to and download shows that you couldn't before because of being locked into only what TheTVDB had to offer.

Also this edition is based off the code we used in our XEM editon so it does come with scene numbering support as well as all the other features our XEM edition has to offer.

Please before using this with your existing database (sickbeard.db) please make a backup copy of it and delete any other database files such as cache.db and failed.db if present, we HIGHLY recommend starting out with no database files at all to make this a fresh start but the choice is at your own risk!

Enjoy!
2014-03-09 22:39:12 -07:00

474 lines
16 KiB
Python

# -*- coding: utf-8 -*-
# enzyme - Video metadata parser
# Copyright 2011-2012 Antoine Bertin <diaoulael@gmail.com>
# Copyright 2003-2007 Thomas Schueppel <stain@acm.org>
# Copyright 2003-2007 Dirk Meyer <dischi@freevo.org>
#
# This file is part of enzyme.
#
# enzyme is free software; you can redistribute it and/or modify it under
# the terms of the GNU General Public License as published by
# the Free Software Foundation; either version 3 of the License, or
# (at your option) any later version.
#
# enzyme is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with enzyme. If not, see <http://www.gnu.org/licenses/>.
__all__ = ['Parser']
import zlib
import logging
import StringIO
import struct
from exceptions import ParseError
import core
# get logging object
log = logging.getLogger(__name__)
# http://developer.apple.com/documentation/QuickTime/QTFF/index.html
# http://developer.apple.com/documentation/QuickTime/QTFF/QTFFChap4/\
# chapter_5_section_2.html#//apple_ref/doc/uid/TP40000939-CH206-BBCBIICE
# Note: May need to define custom log level to work like ATOM_DEBUG did here
QTUDTA = {
'nam': 'title',
'aut': 'artist',
'cpy': 'copyright'
}
QTLANGUAGES = {
0: "en",
1: "fr",
2: "de",
3: "it",
4: "nl",
5: "sv",
6: "es",
7: "da",
8: "pt",
9: "no",
10: "he",
11: "ja",
12: "ar",
13: "fi",
14: "el",
15: "is",
16: "mt",
17: "tr",
18: "hr",
19: "Traditional Chinese",
20: "ur",
21: "hi",
22: "th",
23: "ko",
24: "lt",
25: "pl",
26: "hu",
27: "et",
28: "lv",
29: "Lappish",
30: "fo",
31: "Farsi",
32: "ru",
33: "Simplified Chinese",
34: "Flemish",
35: "ga",
36: "sq",
37: "ro",
38: "cs",
39: "sk",
40: "sl",
41: "yi",
42: "sr",
43: "mk",
44: "bg",
45: "uk",
46: "be",
47: "uz",
48: "kk",
49: "az",
50: "AzerbaijanAr",
51: "hy",
52: "ka",
53: "mo",
54: "ky",
55: "tg",
56: "tk",
57: "mn",
58: "MongolianCyr",
59: "ps",
60: "ku",
61: "ks",
62: "sd",
63: "bo",
64: "ne",
65: "sa",
66: "mr",
67: "bn",
68: "as",
69: "gu",
70: "pa",
71: "or",
72: "ml",
73: "kn",
74: "ta",
75: "te",
76: "si",
77: "my",
78: "Khmer",
79: "lo",
80: "vi",
81: "id",
82: "tl",
83: "MalayRoman",
84: "MalayArabic",
85: "am",
86: "ti",
87: "om",
88: "so",
89: "sw",
90: "Ruanda",
91: "Rundi",
92: "Chewa",
93: "mg",
94: "eo",
128: "cy",
129: "eu",
130: "ca",
131: "la",
132: "qu",
133: "gn",
134: "ay",
135: "tt",
136: "ug",
137: "Dzongkha",
138: "JavaneseRom",
}
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)
elif atomtype == 'mvhd':
# movie header
mvhd = struct.unpack('>6I2h', file.read(28))
self.length = max(self.length, mvhd[4] / mvhd[3])
self.volume = mvhd[6]
file.seek(atomsize - 8 - 28, 1)
elif atomtype == 'cmov':
# compressed movie
datasize, atomtype = struct.unpack('>I4s', file.read(8))
if not atomtype == 'dcom':
return atomsize
method = struct.unpack('>4s', file.read(datasize - 8))[0]
datasize, atomtype = struct.unpack('>I4s', file.read(8))
if not atomtype == 'cmvd':
return atomsize
if method == 'zlib':
data = file.read(datasize - 8)
try:
decompressed = zlib.decompress(data)
except Exception, e:
try:
decompressed = zlib.decompress(data[4:])
except Exception, e:
log.exception(u'There was a proble decompressiong atom')
return atomsize
decompressedIO = StringIO.StringIO(decompressed)
while self._readatom(decompressedIO):
pass
else:
log.info(u'unknown compression %r' % method)
# unknown compression method
file.seek(datasize - 8, 1)
elif atomtype == 'moov':
# decompressed movie info
while self._readatom(file):
pass
elif atomtype == 'mdat':
pos = file.tell() + atomsize - 8
# maybe there is data inside the mdat
log.info(u'parsing mdat')
while self._readatom(file):
pass
log.info(u'end of mdat')
file.seek(pos, 0)
elif atomtype == 'rmra':
# reference list
while self._readatom(file):
pass
elif atomtype == 'rmda':
# reference
atomdata = file.read(atomsize - 8)
pos = 0
url = ''
quality = 0
datarate = 0
while pos < atomsize - 8:
(datasize, datatype) = struct.unpack('>I4s', atomdata[pos:pos + 8])
if datatype == 'rdrf':
rflags, rtype, rlen = struct.unpack('>I4sI', atomdata[pos + 8:pos + 20])
if rtype == 'url ':
url = atomdata[pos + 20:pos + 20 + rlen]
if url.find('\0') > 0:
url = url[:url.find('\0')]
elif datatype == 'rmqu':
quality = struct.unpack('>I', atomdata[pos + 8:pos + 12])[0]
elif datatype == 'rmdr':
datarate = struct.unpack('>I', atomdata[pos + 12:pos + 16])[0]
pos += datasize
if url:
self._references.append((url, quality, datarate))
else:
if not atomtype in ['wide', 'free']:
log.info(u'unhandled base atom %r' % atomtype)
# Skip unknown atoms
try:
file.seek(atomsize - 8, 1)
except IOError:
return 0
return atomsize
Parser = MPEG4