mirror of
https://github.com/moparisthebest/SickRage
synced 2024-10-31 23:45:02 -04:00
0d9fbc1ad7
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!
413 lines
15 KiB
Python
413 lines
15 KiB
Python
from lib.hachoir_core.field import MissingField
|
|
from lib.hachoir_metadata.metadata import (registerExtractor,
|
|
Metadata, RootMetadata, MultipleMetadata)
|
|
from lib.hachoir_metadata.metadata_item import QUALITY_GOOD
|
|
from lib.hachoir_metadata.safe import fault_tolerant
|
|
from lib.hachoir_parser.video import MovFile, AsfFile, FlvFile
|
|
from lib.hachoir_parser.video.asf import Descriptor as ASF_Descriptor
|
|
from lib.hachoir_parser.container import MkvFile
|
|
from lib.hachoir_parser.container.mkv import dateToDatetime
|
|
from lib.hachoir_core.i18n import _
|
|
from lib.hachoir_core.tools import makeUnicode, makePrintable, timedelta2seconds
|
|
from datetime import timedelta
|
|
|
|
class MkvMetadata(MultipleMetadata):
|
|
tag_key = {
|
|
"TITLE": "title",
|
|
"URL": "url",
|
|
"COPYRIGHT": "copyright",
|
|
|
|
# TODO: use maybe another name?
|
|
# Its value may be different than (...)/Info/DateUTC/date
|
|
"DATE_RECORDED": "creation_date",
|
|
|
|
# TODO: Extract subtitle metadata
|
|
"SUBTITLE": "subtitle_author",
|
|
}
|
|
|
|
def extract(self, mkv):
|
|
for segment in mkv.array("Segment"):
|
|
self.processSegment(segment)
|
|
|
|
def processSegment(self, segment):
|
|
for field in segment:
|
|
if field.name.startswith("Info["):
|
|
self.processInfo(field)
|
|
elif field.name.startswith("Tags["):
|
|
for tag in field.array("Tag"):
|
|
self.processTag(tag)
|
|
elif field.name.startswith("Tracks["):
|
|
self.processTracks(field)
|
|
elif field.name.startswith("Cluster["):
|
|
if self.quality < QUALITY_GOOD:
|
|
return
|
|
|
|
def processTracks(self, tracks):
|
|
for entry in tracks.array("TrackEntry"):
|
|
self.processTrack(entry)
|
|
|
|
def processTrack(self, track):
|
|
if "TrackType/enum" not in track:
|
|
return
|
|
if track["TrackType/enum"].display == "video":
|
|
self.processVideo(track)
|
|
elif track["TrackType/enum"].display == "audio":
|
|
self.processAudio(track)
|
|
elif track["TrackType/enum"].display == "subtitle":
|
|
self.processSubtitle(track)
|
|
|
|
def trackCommon(self, track, meta):
|
|
if "Name/unicode" in track:
|
|
meta.title = track["Name/unicode"].value
|
|
if "Language/string" in track \
|
|
and track["Language/string"].value not in ("mis", "und"):
|
|
meta.language = track["Language/string"].value
|
|
|
|
def processVideo(self, track):
|
|
video = Metadata(self)
|
|
self.trackCommon(track, video)
|
|
try:
|
|
video.compression = track["CodecID/string"].value
|
|
if "Video" in track:
|
|
video.width = track["Video/PixelWidth/unsigned"].value
|
|
video.height = track["Video/PixelHeight/unsigned"].value
|
|
except MissingField:
|
|
pass
|
|
self.addGroup("video[]", video, "Video stream")
|
|
|
|
def getDouble(self, field, parent):
|
|
float_key = '%s/float' % parent
|
|
if float_key in field:
|
|
return field[float_key].value
|
|
double_key = '%s/double' % parent
|
|
if double_key in field:
|
|
return field[double_key].value
|
|
return None
|
|
|
|
def processAudio(self, track):
|
|
audio = Metadata(self)
|
|
self.trackCommon(track, audio)
|
|
if "Audio" in track:
|
|
frequency = self.getDouble(track, "Audio/SamplingFrequency")
|
|
if frequency is not None:
|
|
audio.sample_rate = frequency
|
|
if "Audio/Channels/unsigned" in track:
|
|
audio.nb_channel = track["Audio/Channels/unsigned"].value
|
|
if "Audio/BitDepth/unsigned" in track:
|
|
audio.bits_per_sample = track["Audio/BitDepth/unsigned"].value
|
|
if "CodecID/string" in track:
|
|
audio.compression = track["CodecID/string"].value
|
|
self.addGroup("audio[]", audio, "Audio stream")
|
|
|
|
def processSubtitle(self, track):
|
|
sub = Metadata(self)
|
|
self.trackCommon(track, sub)
|
|
try:
|
|
sub.compression = track["CodecID/string"].value
|
|
except MissingField:
|
|
pass
|
|
self.addGroup("subtitle[]", sub, "Subtitle")
|
|
|
|
def processTag(self, tag):
|
|
for field in tag.array("SimpleTag"):
|
|
self.processSimpleTag(field)
|
|
|
|
def processSimpleTag(self, tag):
|
|
if "TagName/unicode" not in tag \
|
|
or "TagString/unicode" not in tag:
|
|
return
|
|
name = tag["TagName/unicode"].value
|
|
if name not in self.tag_key:
|
|
return
|
|
key = self.tag_key[name]
|
|
value = tag["TagString/unicode"].value
|
|
setattr(self, key, value)
|
|
|
|
def processInfo(self, info):
|
|
if "TimecodeScale/unsigned" in info:
|
|
duration = self.getDouble(info, "Duration")
|
|
if duration is not None:
|
|
try:
|
|
seconds = duration * info["TimecodeScale/unsigned"].value * 1e-9
|
|
self.duration = timedelta(seconds=seconds)
|
|
except OverflowError:
|
|
# Catch OverflowError for timedelta (long int too large
|
|
# to be converted to an int)
|
|
pass
|
|
if "DateUTC/date" in info:
|
|
try:
|
|
self.creation_date = dateToDatetime(info["DateUTC/date"].value)
|
|
except OverflowError:
|
|
pass
|
|
if "WritingApp/unicode" in info:
|
|
self.producer = info["WritingApp/unicode"].value
|
|
if "MuxingApp/unicode" in info:
|
|
self.producer = info["MuxingApp/unicode"].value
|
|
if "Title/unicode" in info:
|
|
self.title = info["Title/unicode"].value
|
|
|
|
class FlvMetadata(MultipleMetadata):
|
|
def extract(self, flv):
|
|
if "video[0]" in flv:
|
|
meta = Metadata(self)
|
|
self.extractVideo(flv["video[0]"], meta)
|
|
self.addGroup("video", meta, "Video stream")
|
|
if "audio[0]" in flv:
|
|
meta = Metadata(self)
|
|
self.extractAudio(flv["audio[0]"], meta)
|
|
self.addGroup("audio", meta, "Audio stream")
|
|
# TODO: Computer duration
|
|
# One technic: use last video/audio chunk and use timestamp
|
|
# But this is very slow
|
|
self.format_version = flv.description
|
|
|
|
if "metadata/entry[1]" in flv:
|
|
self.extractAMF(flv["metadata/entry[1]"])
|
|
if self.has('duration'):
|
|
self.bit_rate = flv.size / timedelta2seconds(self.get('duration'))
|
|
|
|
@fault_tolerant
|
|
def extractAudio(self, audio, meta):
|
|
if audio["codec"].display == "MP3" and "music_data" in audio:
|
|
meta.compression = audio["music_data"].description
|
|
else:
|
|
meta.compression = audio["codec"].display
|
|
meta.sample_rate = audio.getSampleRate()
|
|
if audio["is_16bit"].value:
|
|
meta.bits_per_sample = 16
|
|
else:
|
|
meta.bits_per_sample = 8
|
|
if audio["is_stereo"].value:
|
|
meta.nb_channel = 2
|
|
else:
|
|
meta.nb_channel = 1
|
|
|
|
@fault_tolerant
|
|
def extractVideo(self, video, meta):
|
|
meta.compression = video["codec"].display
|
|
|
|
def extractAMF(self, amf):
|
|
for entry in amf.array("item"):
|
|
self.useAmfEntry(entry)
|
|
|
|
@fault_tolerant
|
|
def useAmfEntry(self, entry):
|
|
key = entry["key"].value
|
|
if key == "duration":
|
|
self.duration = timedelta(seconds=entry["value"].value)
|
|
elif key == "creator":
|
|
self.producer = entry["value"].value
|
|
elif key == "audiosamplerate":
|
|
self.sample_rate = entry["value"].value
|
|
elif key == "framerate":
|
|
self.frame_rate = entry["value"].value
|
|
elif key == "metadatacreator":
|
|
self.producer = entry["value"].value
|
|
elif key == "metadatadate":
|
|
self.creation_date = entry.value
|
|
elif key == "width":
|
|
self.width = int(entry["value"].value)
|
|
elif key == "height":
|
|
self.height = int(entry["value"].value)
|
|
|
|
class MovMetadata(RootMetadata):
|
|
def extract(self, mov):
|
|
for atom in mov:
|
|
if "movie" in atom:
|
|
self.processMovie(atom["movie"])
|
|
|
|
@fault_tolerant
|
|
def processMovieHeader(self, hdr):
|
|
self.creation_date = hdr["creation_date"].value
|
|
self.last_modification = hdr["lastmod_date"].value
|
|
self.duration = timedelta(seconds=float(hdr["duration"].value) / hdr["time_scale"].value)
|
|
self.comment = _("Play speed: %.1f%%") % (hdr["play_speed"].value*100)
|
|
self.comment = _("User volume: %.1f%%") % (float(hdr["volume"].value)*100//255)
|
|
|
|
@fault_tolerant
|
|
def processTrackHeader(self, hdr):
|
|
width = int(hdr["frame_size_width"].value)
|
|
height = int(hdr["frame_size_height"].value)
|
|
if width and height:
|
|
self.width = width
|
|
self.height = height
|
|
|
|
def processTrack(self, atom):
|
|
for field in atom:
|
|
if "track_hdr" in field:
|
|
self.processTrackHeader(field["track_hdr"])
|
|
|
|
def processMovie(self, atom):
|
|
for field in atom:
|
|
if "track" in field:
|
|
self.processTrack(field["track"])
|
|
if "movie_hdr" in field:
|
|
self.processMovieHeader(field["movie_hdr"])
|
|
|
|
|
|
class AsfMetadata(MultipleMetadata):
|
|
EXT_DESC_TO_ATTR = {
|
|
"Encoder": "producer",
|
|
"ToolName": "producer",
|
|
"AlbumTitle": "album",
|
|
"Track": "track_number",
|
|
"TrackNumber": "track_total",
|
|
"Year": "creation_date",
|
|
"AlbumArtist": "author",
|
|
}
|
|
SKIP_EXT_DESC = set((
|
|
# Useless informations
|
|
"WMFSDKNeeded", "WMFSDKVersion",
|
|
"Buffer Average", "VBR Peak", "EncodingTime",
|
|
"MediaPrimaryClassID", "UniqueFileIdentifier",
|
|
))
|
|
|
|
def extract(self, asf):
|
|
if "header/content" in asf:
|
|
self.processHeader(asf["header/content"])
|
|
|
|
def processHeader(self, header):
|
|
compression = []
|
|
is_vbr = None
|
|
|
|
if "ext_desc/content" in header:
|
|
# Extract all data from ext_desc
|
|
data = {}
|
|
for desc in header.array("ext_desc/content/descriptor"):
|
|
self.useExtDescItem(desc, data)
|
|
|
|
# Have ToolName and ToolVersion? If yes, group them to producer key
|
|
if "ToolName" in data and "ToolVersion" in data:
|
|
self.producer = "%s (version %s)" % (data["ToolName"], data["ToolVersion"])
|
|
del data["ToolName"]
|
|
del data["ToolVersion"]
|
|
|
|
# "IsVBR" key
|
|
if "IsVBR" in data:
|
|
is_vbr = (data["IsVBR"] == 1)
|
|
del data["IsVBR"]
|
|
|
|
# Store data
|
|
for key, value in data.iteritems():
|
|
if key in self.EXT_DESC_TO_ATTR:
|
|
key = self.EXT_DESC_TO_ATTR[key]
|
|
else:
|
|
if isinstance(key, str):
|
|
key = makePrintable(key, "ISO-8859-1", to_unicode=True)
|
|
value = "%s=%s" % (key, value)
|
|
key = "comment"
|
|
setattr(self, key, value)
|
|
|
|
if "file_prop/content" in header:
|
|
self.useFileProp(header["file_prop/content"], is_vbr)
|
|
|
|
if "codec_list/content" in header:
|
|
for codec in header.array("codec_list/content/codec"):
|
|
if "name" in codec:
|
|
text = codec["name"].value
|
|
if "desc" in codec and codec["desc"].value:
|
|
text = "%s (%s)" % (text, codec["desc"].value)
|
|
compression.append(text)
|
|
|
|
audio_index = 1
|
|
video_index = 1
|
|
for index, stream_prop in enumerate(header.array("stream_prop")):
|
|
if "content/audio_header" in stream_prop:
|
|
meta = Metadata(self)
|
|
self.streamProperty(header, index, meta)
|
|
self.streamAudioHeader(stream_prop["content/audio_header"], meta)
|
|
if self.addGroup("audio[%u]" % audio_index, meta, "Audio stream #%u" % audio_index):
|
|
audio_index += 1
|
|
elif "content/video_header" in stream_prop:
|
|
meta = Metadata(self)
|
|
self.streamProperty(header, index, meta)
|
|
self.streamVideoHeader(stream_prop["content/video_header"], meta)
|
|
if self.addGroup("video[%u]" % video_index, meta, "Video stream #%u" % video_index):
|
|
video_index += 1
|
|
|
|
if "metadata/content" in header:
|
|
info = header["metadata/content"]
|
|
try:
|
|
self.title = info["title"].value
|
|
self.author = info["author"].value
|
|
self.copyright = info["copyright"].value
|
|
except MissingField:
|
|
pass
|
|
|
|
@fault_tolerant
|
|
def streamAudioHeader(self, audio, meta):
|
|
if not meta.has("compression"):
|
|
meta.compression = audio["twocc"].display
|
|
meta.nb_channel = audio["channels"].value
|
|
meta.sample_rate = audio["sample_rate"].value
|
|
meta.bits_per_sample = audio["bits_per_sample"].value
|
|
|
|
@fault_tolerant
|
|
def streamVideoHeader(self, video, meta):
|
|
meta.width = video["width"].value
|
|
meta.height = video["height"].value
|
|
if "bmp_info" in video:
|
|
bmp_info = video["bmp_info"]
|
|
if not meta.has("compression"):
|
|
meta.compression = bmp_info["codec"].display
|
|
meta.bits_per_pixel = bmp_info["bpp"].value
|
|
|
|
@fault_tolerant
|
|
def useExtDescItem(self, desc, data):
|
|
if desc["type"].value == ASF_Descriptor.TYPE_BYTE_ARRAY:
|
|
# Skip binary data
|
|
return
|
|
key = desc["name"].value
|
|
if "/" in key:
|
|
# Replace "WM/ToolName" with "ToolName"
|
|
key = key.split("/", 1)[1]
|
|
if key in self.SKIP_EXT_DESC:
|
|
# Skip some keys
|
|
return
|
|
value = desc["value"].value
|
|
if not value:
|
|
return
|
|
value = makeUnicode(value)
|
|
data[key] = value
|
|
|
|
@fault_tolerant
|
|
def useFileProp(self, prop, is_vbr):
|
|
self.creation_date = prop["creation_date"].value
|
|
self.duration = prop["play_duration"].value
|
|
if prop["seekable"].value:
|
|
self.comment = u"Is seekable"
|
|
value = prop["max_bitrate"].value
|
|
text = prop["max_bitrate"].display
|
|
if is_vbr is True:
|
|
text = "VBR (%s max)" % text
|
|
elif is_vbr is False:
|
|
text = "%s (CBR)" % text
|
|
else:
|
|
text = "%s (max)" % text
|
|
self.bit_rate = (value, text)
|
|
|
|
def streamProperty(self, header, index, meta):
|
|
key = "bit_rates/content/bit_rate[%u]/avg_bitrate" % index
|
|
if key in header:
|
|
meta.bit_rate = header[key].value
|
|
|
|
# TODO: Use codec list
|
|
# It doesn't work when the video uses /header/content/bitrate_mutex
|
|
# since the codec list are shared between streams but... how is it
|
|
# shared?
|
|
# key = "codec_list/content/codec[%u]" % index
|
|
# if key in header:
|
|
# codec = header[key]
|
|
# if "name" in codec:
|
|
# text = codec["name"].value
|
|
# if "desc" in codec and codec["desc"].value:
|
|
# meta.compression = "%s (%s)" % (text, codec["desc"].value)
|
|
# else:
|
|
# meta.compression = text
|
|
|
|
registerExtractor(MovFile, MovMetadata)
|
|
registerExtractor(AsfFile, AsfMetadata)
|
|
registerExtractor(FlvFile, FlvMetadata)
|
|
registerExtractor(MkvFile, MkvMetadata)
|
|
|