mirror of
https://github.com/moparisthebest/SickRage
synced 2024-12-15 04:22:17 -05:00
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!
191 lines
7.5 KiB
Python
191 lines
7.5 KiB
Python
"""
|
|
Extract metadata from RIFF file format: AVI video and WAV sound.
|
|
"""
|
|
|
|
from lib.hachoir_metadata.metadata import Metadata, MultipleMetadata, registerExtractor
|
|
from lib.hachoir_metadata.safe import fault_tolerant, getValue
|
|
from lib.hachoir_parser.container.riff import RiffFile
|
|
from lib.hachoir_parser.video.fourcc import UNCOMPRESSED_AUDIO
|
|
from lib.hachoir_core.tools import humanFilesize, makeUnicode, timedelta2seconds
|
|
from lib.hachoir_core.i18n import _
|
|
from lib.hachoir_metadata.audio import computeComprRate as computeAudioComprRate
|
|
from datetime import timedelta
|
|
|
|
class RiffMetadata(MultipleMetadata):
|
|
TAG_TO_KEY = {
|
|
"INAM": "title",
|
|
"IART": "artist",
|
|
"ICMT": "comment",
|
|
"ICOP": "copyright",
|
|
"IENG": "author", # (engineer)
|
|
"ISFT": "producer",
|
|
"ICRD": "creation_date",
|
|
"IDIT": "creation_date",
|
|
}
|
|
|
|
def extract(self, riff):
|
|
type = riff["type"].value
|
|
if type == "WAVE":
|
|
self.extractWAVE(riff)
|
|
size = getValue(riff, "audio_data/size")
|
|
if size:
|
|
computeAudioComprRate(self, size*8)
|
|
elif type == "AVI ":
|
|
if "headers" in riff:
|
|
self.extractAVI(riff["headers"])
|
|
self.extractInfo(riff["headers"])
|
|
elif type == "ACON":
|
|
self.extractAnim(riff)
|
|
if "info" in riff:
|
|
self.extractInfo(riff["info"])
|
|
|
|
def processChunk(self, chunk):
|
|
if "text" not in chunk:
|
|
return
|
|
value = chunk["text"].value
|
|
tag = chunk["tag"].value
|
|
if tag not in self.TAG_TO_KEY:
|
|
self.warning("Skip RIFF metadata %s: %s" % (tag, value))
|
|
return
|
|
key = self.TAG_TO_KEY[tag]
|
|
setattr(self, key, value)
|
|
|
|
@fault_tolerant
|
|
def extractWAVE(self, wav):
|
|
format = wav["format"]
|
|
|
|
# Number of channel, bits/sample, sample rate
|
|
self.nb_channel = format["nb_channel"].value
|
|
self.bits_per_sample = format["bit_per_sample"].value
|
|
self.sample_rate = format["sample_per_sec"].value
|
|
|
|
self.compression = format["codec"].display
|
|
if "nb_sample/nb_sample" in wav \
|
|
and 0 < format["sample_per_sec"].value:
|
|
self.duration = timedelta(seconds=float(wav["nb_sample/nb_sample"].value) / format["sample_per_sec"].value)
|
|
if format["codec"].value in UNCOMPRESSED_AUDIO:
|
|
# Codec with fixed bit rate
|
|
self.bit_rate = format["nb_channel"].value * format["bit_per_sample"].value * format["sample_per_sec"].value
|
|
if not self.has("duration") \
|
|
and "audio_data/size" in wav \
|
|
and self.has("bit_rate"):
|
|
duration = float(wav["audio_data/size"].value)*8 / self.get('bit_rate')
|
|
self.duration = timedelta(seconds=duration)
|
|
|
|
def extractInfo(self, fieldset):
|
|
for field in fieldset:
|
|
if not field.is_field_set:
|
|
continue
|
|
if "tag" in field:
|
|
if field["tag"].value == "LIST":
|
|
self.extractInfo(field)
|
|
else:
|
|
self.processChunk(field)
|
|
|
|
@fault_tolerant
|
|
def extractAVIVideo(self, header, meta):
|
|
meta.compression = "%s (fourcc:\"%s\")" \
|
|
% (header["fourcc"].display, makeUnicode(header["fourcc"].value))
|
|
if header["rate"].value and header["scale"].value:
|
|
fps = float(header["rate"].value) / header["scale"].value
|
|
meta.frame_rate = fps
|
|
if 0 < fps:
|
|
self.duration = meta.duration = timedelta(seconds=float(header["length"].value) / fps)
|
|
|
|
if "../stream_fmt/width" in header:
|
|
format = header["../stream_fmt"]
|
|
meta.width = format["width"].value
|
|
meta.height = format["height"].value
|
|
meta.bits_per_pixel = format["depth"].value
|
|
else:
|
|
meta.width = header["right"].value - header["left"].value
|
|
meta.height = header["bottom"].value - header["top"].value
|
|
|
|
@fault_tolerant
|
|
def extractAVIAudio(self, format, meta):
|
|
meta.nb_channel = format["channel"].value
|
|
meta.sample_rate = format["sample_rate"].value
|
|
meta.bit_rate = format["bit_rate"].value * 8
|
|
if format["bits_per_sample"].value:
|
|
meta.bits_per_sample = format["bits_per_sample"].value
|
|
if "../stream_hdr" in format:
|
|
header = format["../stream_hdr"]
|
|
if header["rate"].value and header["scale"].value:
|
|
frame_rate = float(header["rate"].value) / header["scale"].value
|
|
meta.duration = timedelta(seconds=float(header["length"].value) / frame_rate)
|
|
if header["fourcc"].value != "":
|
|
meta.compression = "%s (fourcc:\"%s\")" \
|
|
% (format["codec"].display, header["fourcc"].value)
|
|
if not meta.has("compression"):
|
|
meta.compression = format["codec"].display
|
|
|
|
self.computeAudioComprRate(meta)
|
|
|
|
@fault_tolerant
|
|
def computeAudioComprRate(self, meta):
|
|
uncompr = meta.get('bit_rate', 0)
|
|
if not uncompr:
|
|
return
|
|
compr = meta.get('nb_channel') * meta.get('sample_rate') * meta.get('bits_per_sample', default=16)
|
|
if not compr:
|
|
return
|
|
meta.compr_rate = float(compr) / uncompr
|
|
|
|
@fault_tolerant
|
|
def useAviHeader(self, header):
|
|
microsec = header["microsec_per_frame"].value
|
|
if microsec:
|
|
self.frame_rate = 1000000.0 / microsec
|
|
total_frame = getValue(header, "total_frame")
|
|
if total_frame and not self.has("duration"):
|
|
self.duration = timedelta(microseconds=total_frame * microsec)
|
|
self.width = header["width"].value
|
|
self.height = header["height"].value
|
|
|
|
def extractAVI(self, headers):
|
|
audio_index = 1
|
|
for stream in headers.array("stream"):
|
|
if "stream_hdr/stream_type" not in stream:
|
|
continue
|
|
stream_type = stream["stream_hdr/stream_type"].value
|
|
if stream_type == "vids":
|
|
if "stream_hdr" in stream:
|
|
meta = Metadata(self)
|
|
self.extractAVIVideo(stream["stream_hdr"], meta)
|
|
self.addGroup("video", meta, "Video stream")
|
|
elif stream_type == "auds":
|
|
if "stream_fmt" in stream:
|
|
meta = Metadata(self)
|
|
self.extractAVIAudio(stream["stream_fmt"], meta)
|
|
self.addGroup("audio[%u]" % audio_index, meta, "Audio stream")
|
|
audio_index += 1
|
|
if "avi_hdr" in headers:
|
|
self.useAviHeader(headers["avi_hdr"])
|
|
|
|
# Compute global bit rate
|
|
if self.has("duration") and "/movie/size" in headers:
|
|
self.bit_rate = float(headers["/movie/size"].value) * 8 / timedelta2seconds(self.get('duration'))
|
|
|
|
# Video has index?
|
|
if "/index" in headers:
|
|
self.comment = _("Has audio/video index (%s)") \
|
|
% humanFilesize(headers["/index"].size/8)
|
|
|
|
@fault_tolerant
|
|
def extractAnim(self, riff):
|
|
if "anim_rate/rate[0]" in riff:
|
|
count = 0
|
|
total = 0
|
|
for rate in riff.array("anim_rate/rate"):
|
|
count += 1
|
|
if 100 < count:
|
|
break
|
|
total += rate.value / 60.0
|
|
if count and total:
|
|
self.frame_rate = count / total
|
|
if not self.has("frame_rate") and "anim_hdr/jiffie_rate" in riff:
|
|
self.frame_rate = 60.0 / riff["anim_hdr/jiffie_rate"].value
|
|
|
|
registerExtractor(RiffFile, RiffMetadata)
|
|
|