mirror of
https://github.com/moparisthebest/SickRage
synced 2024-11-14 21:35:03 -05:00
831248b412
Fixed issues with Backup/Restore. Fixed issues with banners, posters, images via showPoster method. Fixed issues with WebAPI apikeys. Fixed issues with WebUI and url pattern matching.
560 lines
22 KiB
Python
560 lines
22 KiB
Python
# Author: Nic Wolfe <nic@wolfeden.ca>
|
|
# URL: http://code.google.com/p/sickbeard/
|
|
#
|
|
# This file is part of SickRage.
|
|
#
|
|
# SickRage 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.
|
|
#
|
|
# SickRage 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 SickRage. If not, see <http://www.gnu.org/licenses/>.
|
|
|
|
import datetime
|
|
import os
|
|
import re
|
|
|
|
import sickbeard
|
|
|
|
import generic
|
|
|
|
from sickbeard import logger, exceptions, helpers
|
|
from sickbeard import encodingKludge as ek
|
|
|
|
from sickbeard.exceptions import ex
|
|
|
|
import xml.etree.cElementTree as etree
|
|
|
|
|
|
class MediaBrowserMetadata(generic.GenericMetadata):
|
|
"""
|
|
Metadata generation class for Media Browser 2.x/3.x - Standard Mode.
|
|
|
|
The following file structure is used:
|
|
|
|
show_root/series.xml (show metadata)
|
|
show_root/folder.jpg (poster)
|
|
show_root/backdrop.jpg (fanart)
|
|
show_root/Season ##/folder.jpg (season thumb)
|
|
show_root/Season ##/filename.ext (*)
|
|
show_root/Season ##/metadata/filename.xml (episode metadata)
|
|
show_root/Season ##/metadata/filename.jpg (episode thumb)
|
|
"""
|
|
|
|
def __init__(self,
|
|
show_metadata=False,
|
|
episode_metadata=False,
|
|
fanart=False,
|
|
poster=False,
|
|
banner=False,
|
|
episode_thumbnails=False,
|
|
season_posters=False,
|
|
season_banners=False,
|
|
season_all_poster=False,
|
|
season_all_banner=False):
|
|
|
|
generic.GenericMetadata.__init__(self,
|
|
show_metadata,
|
|
episode_metadata,
|
|
fanart,
|
|
poster,
|
|
banner,
|
|
episode_thumbnails,
|
|
season_posters,
|
|
season_banners,
|
|
season_all_poster,
|
|
season_all_banner)
|
|
|
|
self.name = 'MediaBrowser'
|
|
|
|
self._ep_nfo_extension = 'xml'
|
|
self._show_metadata_filename = 'series.xml'
|
|
|
|
self.fanart_name = "backdrop.jpg"
|
|
self.poster_name = "folder.jpg"
|
|
|
|
# web-ui metadata template
|
|
self.eg_show_metadata = "series.xml"
|
|
self.eg_episode_metadata = "Season##\\metadata\\<i>filename</i>.xml"
|
|
self.eg_fanart = "backdrop.jpg"
|
|
self.eg_poster = "folder.jpg"
|
|
self.eg_banner = "banner.jpg"
|
|
self.eg_episode_thumbnails = "Season##\\metadata\\<i>filename</i>.jpg"
|
|
self.eg_season_posters = "Season##\\folder.jpg"
|
|
self.eg_season_banners = "Season##\\banner.jpg"
|
|
self.eg_season_all_poster = "<i>not supported</i>"
|
|
self.eg_season_all_banner = "<i>not supported</i>"
|
|
|
|
# Override with empty methods for unsupported features
|
|
def retrieveShowMetadata(self, folder):
|
|
# while show metadata is generated, it is not supported for our lookup
|
|
return (None, None, None)
|
|
|
|
def create_season_all_poster(self, show_obj):
|
|
pass
|
|
|
|
def create_season_all_banner(self, show_obj):
|
|
pass
|
|
|
|
def get_episode_file_path(self, ep_obj):
|
|
"""
|
|
Returns a full show dir/metadata/episode.xml path for MediaBrowser
|
|
episode metadata files
|
|
|
|
ep_obj: a TVEpisode object to get the path for
|
|
"""
|
|
|
|
if ek.ek(os.path.isfile, ep_obj.location):
|
|
xml_file_name = helpers.replaceExtension(ek.ek(os.path.basename, ep_obj.location), self._ep_nfo_extension)
|
|
metadata_dir_name = ek.ek(os.path.join, ek.ek(os.path.dirname, ep_obj.location), 'metadata')
|
|
xml_file_path = ek.ek(os.path.join, metadata_dir_name, xml_file_name)
|
|
else:
|
|
logger.log(u"Episode location doesn't exist: " + str(ep_obj.location), logger.DEBUG)
|
|
return ''
|
|
|
|
return xml_file_path
|
|
|
|
def get_episode_thumb_path(self, ep_obj):
|
|
"""
|
|
Returns a full show dir/metadata/episode.jpg path for MediaBrowser
|
|
episode thumbs.
|
|
|
|
ep_obj: a TVEpisode object to get the path from
|
|
"""
|
|
|
|
if ek.ek(os.path.isfile, ep_obj.location):
|
|
tbn_file_name = helpers.replaceExtension(ek.ek(os.path.basename, ep_obj.location), 'jpg')
|
|
metadata_dir_name = ek.ek(os.path.join, ek.ek(os.path.dirname, ep_obj.location), 'metadata')
|
|
tbn_file_path = ek.ek(os.path.join, metadata_dir_name, tbn_file_name)
|
|
else:
|
|
return None
|
|
|
|
return tbn_file_path
|
|
|
|
def get_season_poster_path(self, show_obj, season):
|
|
"""
|
|
Season thumbs for MediaBrowser go in Show Dir/Season X/folder.jpg
|
|
|
|
If no season folder exists, None is returned
|
|
"""
|
|
|
|
dir_list = [x for x in ek.ek(os.listdir, show_obj.location) if
|
|
ek.ek(os.path.isdir, ek.ek(os.path.join, show_obj.location, x))]
|
|
|
|
season_dir_regex = '^Season\s+(\d+)$'
|
|
|
|
season_dir = None
|
|
|
|
for cur_dir in dir_list:
|
|
# MediaBrowser 1.x only supports 'Specials'
|
|
# MediaBrowser 2.x looks to only support 'Season 0'
|
|
# MediaBrowser 3.x looks to mimic KODI/Plex support
|
|
if season == 0 and cur_dir == "Specials":
|
|
season_dir = cur_dir
|
|
break
|
|
|
|
match = re.match(season_dir_regex, cur_dir, re.I)
|
|
if not match:
|
|
continue
|
|
|
|
cur_season = int(match.group(1))
|
|
|
|
if cur_season == season:
|
|
season_dir = cur_dir
|
|
break
|
|
|
|
if not season_dir:
|
|
logger.log(u"Unable to find a season dir for season " + str(season), logger.DEBUG)
|
|
return None
|
|
|
|
logger.log(u"Using " + str(season_dir) + "/folder.jpg as season dir for season " + str(season), logger.DEBUG)
|
|
|
|
return ek.ek(os.path.join, show_obj.location, season_dir, 'folder.jpg')
|
|
|
|
def get_season_banner_path(self, show_obj, season):
|
|
"""
|
|
Season thumbs for MediaBrowser go in Show Dir/Season X/banner.jpg
|
|
|
|
If no season folder exists, None is returned
|
|
"""
|
|
|
|
dir_list = [x for x in ek.ek(os.listdir, show_obj.location) if
|
|
ek.ek(os.path.isdir, ek.ek(os.path.join, show_obj.location, x))]
|
|
|
|
season_dir_regex = '^Season\s+(\d+)$'
|
|
|
|
season_dir = None
|
|
|
|
for cur_dir in dir_list:
|
|
# MediaBrowser 1.x only supports 'Specials'
|
|
# MediaBrowser 2.x looks to only support 'Season 0'
|
|
# MediaBrowser 3.x looks to mimic KODI/Plex support
|
|
if season == 0 and cur_dir == "Specials":
|
|
season_dir = cur_dir
|
|
break
|
|
|
|
match = re.match(season_dir_regex, cur_dir, re.I)
|
|
if not match:
|
|
continue
|
|
|
|
cur_season = int(match.group(1))
|
|
|
|
if cur_season == season:
|
|
season_dir = cur_dir
|
|
break
|
|
|
|
if not season_dir:
|
|
logger.log(u"Unable to find a season dir for season " + str(season), logger.DEBUG)
|
|
return None
|
|
|
|
logger.log(u"Using " + str(season_dir) + "/banner.jpg as season dir for season " + str(season), logger.DEBUG)
|
|
|
|
return ek.ek(os.path.join, show_obj.location, season_dir, 'banner.jpg')
|
|
|
|
def _show_data(self, show_obj):
|
|
"""
|
|
Creates an elementTree XML structure for a MediaBrowser-style series.xml
|
|
returns the resulting data object.
|
|
|
|
show_obj: a TVShow instance to create the NFO for
|
|
"""
|
|
|
|
indexer_lang = show_obj.lang
|
|
# There's gotta be a better way of doing this but we don't wanna
|
|
# change the language value elsewhere
|
|
lINDEXER_API_PARMS = sickbeard.indexerApi(show_obj.indexer).api_params.copy()
|
|
|
|
lINDEXER_API_PARMS['actors'] = True
|
|
|
|
if indexer_lang and not indexer_lang == 'en':
|
|
lINDEXER_API_PARMS['language'] = indexer_lang
|
|
|
|
if show_obj.dvdorder != 0:
|
|
lINDEXER_API_PARMS['dvdorder'] = True
|
|
|
|
t = sickbeard.indexerApi(show_obj.indexer).indexer(**lINDEXER_API_PARMS)
|
|
|
|
tv_node = etree.Element("Series")
|
|
|
|
try:
|
|
myShow = t[int(show_obj.indexerid)]
|
|
except sickbeard.indexer_shownotfound:
|
|
logger.log(u"Unable to find show with id " + str(show_obj.indexerid) + " on " + sickbeard.indexerApi(
|
|
show_obj.indexer).name + ", skipping it", logger.ERROR)
|
|
raise
|
|
|
|
except sickbeard.indexer_error:
|
|
logger.log(
|
|
u"" + sickbeard.indexerApi(show_obj.indexer).name + " is down, can't use its data to make the NFO",
|
|
logger.ERROR)
|
|
raise
|
|
|
|
# check for title and id
|
|
if getattr(myShow, 'seriesname', None) is None or getattr(myShow, 'id', None) is None:
|
|
logger.log(u"Incomplete info for show with id " + str(show_obj.indexerid) + " on " + sickbeard.indexerApi(
|
|
show_obj.indexer).name + ", skipping it", logger.ERROR)
|
|
return False
|
|
|
|
indexerid = etree.SubElement(tv_node, "id")
|
|
if getattr(myShow, 'id', None) is not None:
|
|
indexerid.text = str(myShow['id'])
|
|
|
|
indexer = etree.SubElement(tv_node, "indexer")
|
|
if show_obj.indexer != None:
|
|
indexer.text = str(show_obj.indexer)
|
|
|
|
SeriesName = etree.SubElement(tv_node, "SeriesName")
|
|
if getattr(myShow, 'seriesname', None) is not None:
|
|
SeriesName.text = myShow['seriesname']
|
|
|
|
Status = etree.SubElement(tv_node, "Status")
|
|
if getattr(myShow, 'status', None) is not None:
|
|
Status.text = myShow['status']
|
|
|
|
Network = etree.SubElement(tv_node, "Network")
|
|
if getattr(myShow, 'network', None) is not None:
|
|
Network.text = myShow['network']
|
|
|
|
Airs_Time = etree.SubElement(tv_node, "Airs_Time")
|
|
if getattr(myShow, 'airs_time', None) is not None:
|
|
Airs_Time.text = myShow['airs_time']
|
|
|
|
Airs_DayOfWeek = etree.SubElement(tv_node, "Airs_DayOfWeek")
|
|
if getattr(myShow, 'airs_dayofweek', None) is not None:
|
|
Airs_DayOfWeek.text = myShow['airs_dayofweek']
|
|
|
|
FirstAired = etree.SubElement(tv_node, "FirstAired")
|
|
if getattr(myShow, 'firstaired', None) is not None:
|
|
FirstAired.text = myShow['firstaired']
|
|
|
|
ContentRating = etree.SubElement(tv_node, "ContentRating")
|
|
MPAARating = etree.SubElement(tv_node, "MPAARating")
|
|
certification = etree.SubElement(tv_node, "certification")
|
|
if getattr(myShow, 'contentrating', None) is not None:
|
|
ContentRating.text = myShow['contentrating']
|
|
MPAARating.text = myShow['contentrating']
|
|
certification.text = myShow['contentrating']
|
|
|
|
MetadataType = etree.SubElement(tv_node, "Type")
|
|
MetadataType.text = "Series"
|
|
|
|
Overview = etree.SubElement(tv_node, "Overview")
|
|
if getattr(myShow, 'overview', None) is not None:
|
|
Overview.text = myShow['overview']
|
|
|
|
PremiereDate = etree.SubElement(tv_node, "PremiereDate")
|
|
if getattr(myShow, 'firstaired', None) is not None:
|
|
PremiereDate.text = myShow['firstaired']
|
|
|
|
Rating = etree.SubElement(tv_node, "Rating")
|
|
if getattr(myShow, 'rating', None) is not None:
|
|
Rating.text = myShow['rating']
|
|
|
|
ProductionYear = etree.SubElement(tv_node, "ProductionYear")
|
|
if getattr(myShow, 'firstaired', None) is not None:
|
|
try:
|
|
year_text = str(datetime.datetime.strptime(myShow['firstaired'], '%Y-%m-%d').year)
|
|
if year_text:
|
|
ProductionYear.text = year_text
|
|
except:
|
|
pass
|
|
|
|
RunningTime = etree.SubElement(tv_node, "RunningTime")
|
|
Runtime = etree.SubElement(tv_node, "Runtime")
|
|
if getattr(myShow, 'runtime', None) is not None:
|
|
RunningTime.text = myShow['runtime']
|
|
Runtime.text = myShow['runtime']
|
|
|
|
IMDB_ID = etree.SubElement(tv_node, "IMDB_ID")
|
|
IMDB = etree.SubElement(tv_node, "IMDB")
|
|
IMDbId = etree.SubElement(tv_node, "IMDbId")
|
|
if getattr(myShow, 'imdb_id', None) is not None:
|
|
IMDB_ID.text = myShow['imdb_id']
|
|
IMDB.text = myShow['imdb_id']
|
|
IMDbId.text = myShow['imdb_id']
|
|
|
|
Zap2ItId = etree.SubElement(tv_node, "Zap2ItId")
|
|
if getattr(myShow, 'zap2it_id', None) is not None:
|
|
Zap2ItId.text = myShow['zap2it_id']
|
|
|
|
Genres = etree.SubElement(tv_node, "Genres")
|
|
for genre in myShow['genre'].split('|'):
|
|
if genre:
|
|
cur_genre = etree.SubElement(Genres, "Genre")
|
|
cur_genre.text = genre
|
|
|
|
Genre = etree.SubElement(tv_node, "Genre")
|
|
if getattr(myShow, 'genre', None) is not None:
|
|
Genre.text = "|".join([x for x in myShow["genre"].split('|') if x])
|
|
|
|
Studios = etree.SubElement(tv_node, "Studios")
|
|
Studio = etree.SubElement(Studios, "Studio")
|
|
if getattr(myShow, 'network', None) is not None:
|
|
Studio.text = myShow['network']
|
|
|
|
Persons = etree.SubElement(tv_node, "Persons")
|
|
if getattr(myShow, 'actors', None) is not None:
|
|
for actor in myShow['_actors']:
|
|
cur_actor = etree.SubElement(Persons, "Person")
|
|
|
|
cur_actor_name = etree.SubElement(cur_actor, "Name")
|
|
if getattr(actor, 'name', None):
|
|
cur_actor_name.text = actor['name'].strip()
|
|
|
|
cur_actor_type = etree.SubElement(cur_actor, "Type")
|
|
cur_actor_type.text = "Actor"
|
|
|
|
cur_actor_role = etree.SubElement(cur_actor, "Role")
|
|
if getattr(actor, 'role', None):
|
|
cur_actor_role.text = actor['role']
|
|
|
|
helpers.indentXML(tv_node)
|
|
|
|
data = etree.ElementTree(tv_node)
|
|
|
|
return data
|
|
|
|
def _ep_data(self, ep_obj):
|
|
"""
|
|
Creates an elementTree XML structure for a MediaBrowser style episode.xml
|
|
and returns the resulting data object.
|
|
|
|
show_obj: a TVShow instance to create the NFO for
|
|
"""
|
|
|
|
eps_to_write = [ep_obj] + ep_obj.relatedEps
|
|
|
|
persons_dict = {}
|
|
persons_dict['Director'] = []
|
|
persons_dict['GuestStar'] = []
|
|
persons_dict['Writer'] = []
|
|
|
|
indexer_lang = ep_obj.show.lang
|
|
|
|
try:
|
|
lINDEXER_API_PARMS = sickbeard.indexerApi(ep_obj.show.indexer).api_params.copy()
|
|
|
|
lINDEXER_API_PARMS['actors'] = True
|
|
|
|
if indexer_lang and not indexer_lang == 'en':
|
|
lINDEXER_API_PARMS['language'] = indexer_lang
|
|
|
|
if ep_obj.show.dvdorder != 0:
|
|
lINDEXER_API_PARMS['dvdorder'] = True
|
|
|
|
t = sickbeard.indexerApi(ep_obj.show.indexer).indexer(**lINDEXER_API_PARMS)
|
|
|
|
myShow = t[ep_obj.show.indexerid]
|
|
except sickbeard.indexer_shownotfound, e:
|
|
raise exceptions.ShowNotFoundException(e.message)
|
|
except sickbeard.indexer_error, e:
|
|
logger.log(u"Unable to connect to " + sickbeard.indexerApi(
|
|
ep_obj.show.indexer).name + " while creating meta files - skipping - " + ex(e), logger.ERROR)
|
|
return False
|
|
|
|
rootNode = etree.Element("Item")
|
|
|
|
# write an MediaBrowser XML containing info for all matching episodes
|
|
for curEpToWrite in eps_to_write:
|
|
|
|
try:
|
|
myEp = myShow[curEpToWrite.season][curEpToWrite.episode]
|
|
except (sickbeard.indexer_episodenotfound, sickbeard.indexer_seasonnotfound):
|
|
logger.log(u"Unable to find episode " + str(curEpToWrite.season) + "x" + str(
|
|
curEpToWrite.episode) + " on " + sickbeard.indexerApi(
|
|
ep_obj.show.indexer).name + ".. has it been removed? Should I delete from db?")
|
|
return None
|
|
|
|
if curEpToWrite == ep_obj:
|
|
# root (or single) episode
|
|
|
|
# default to today's date for specials if firstaired is not set
|
|
if getattr(myEp, 'firstaired', None) is None and ep_obj.season == 0:
|
|
myEp['firstaired'] = str(datetime.date.fromordinal(1))
|
|
|
|
if getattr(myEp, 'episodename', None) is None or getattr(myEp, 'firstaired', None) is None:
|
|
return None
|
|
|
|
episode = rootNode
|
|
|
|
EpisodeName = etree.SubElement(episode, "EpisodeName")
|
|
if curEpToWrite.name != None:
|
|
EpisodeName.text = curEpToWrite.name
|
|
else:
|
|
EpisodeName.text = ""
|
|
|
|
EpisodeNumber = etree.SubElement(episode, "EpisodeNumber")
|
|
EpisodeNumber.text = str(ep_obj.episode)
|
|
|
|
if ep_obj.relatedEps:
|
|
EpisodeNumberEnd = etree.SubElement(episode, "EpisodeNumberEnd")
|
|
EpisodeNumberEnd.text = str(curEpToWrite.episode)
|
|
|
|
SeasonNumber = etree.SubElement(episode, "SeasonNumber")
|
|
SeasonNumber.text = str(curEpToWrite.season)
|
|
|
|
if not ep_obj.relatedEps:
|
|
absolute_number = etree.SubElement(episode, "absolute_number")
|
|
if getattr(myEp, 'absolute_number', None) is not None:
|
|
absolute_number.text = myEp['absolute_number']
|
|
|
|
FirstAired = etree.SubElement(episode, "FirstAired")
|
|
if curEpToWrite.airdate != datetime.date.fromordinal(1):
|
|
FirstAired.text = str(curEpToWrite.airdate)
|
|
else:
|
|
FirstAired.text = ""
|
|
|
|
MetadataType = etree.SubElement(episode, "Type")
|
|
MetadataType.text = "Episode"
|
|
|
|
Overview = etree.SubElement(episode, "Overview")
|
|
if curEpToWrite.description != None:
|
|
Overview.text = curEpToWrite.description
|
|
else:
|
|
Overview.text = ""
|
|
|
|
if not ep_obj.relatedEps:
|
|
Rating = etree.SubElement(episode, "Rating")
|
|
if getattr(myEp, 'rating', None) is not None:
|
|
Rating.text = myEp['rating']
|
|
|
|
IMDB_ID = etree.SubElement(episode, "IMDB_ID")
|
|
IMDB = etree.SubElement(episode, "IMDB")
|
|
IMDbId = etree.SubElement(episode, "IMDbId")
|
|
if getattr(myShow, 'imdb_id', None) is not None:
|
|
IMDB_ID.text = myShow['imdb_id']
|
|
IMDB.text = myShow['imdb_id']
|
|
IMDbId.text = myShow['imdb_id']
|
|
|
|
indexerid = etree.SubElement(episode, "id")
|
|
indexerid.text = str(curEpToWrite.indexerid)
|
|
|
|
indexer = etree.SubElement(episode, "indexer")
|
|
indexer.text = str(curEpToWrite.show.indexer)
|
|
|
|
Persons = etree.SubElement(episode, "Persons")
|
|
|
|
Language = etree.SubElement(episode, "Language")
|
|
try:
|
|
Language.text = myEp['language']
|
|
except:
|
|
Language.text = 'en' # tvrage api doesn't provide language so we must assume a value here
|
|
|
|
thumb = etree.SubElement(episode, "filename")
|
|
# TODO: See what this is needed for.. if its still needed
|
|
# just write this to the NFO regardless of whether it actually exists or not
|
|
# note: renaming files after nfo generation will break this, tough luck
|
|
thumb_text = self.get_episode_thumb_path(ep_obj)
|
|
if thumb_text:
|
|
thumb.text = thumb_text
|
|
|
|
else:
|
|
# append data from (if any) related episodes
|
|
EpisodeNumberEnd.text = str(curEpToWrite.episode)
|
|
|
|
if curEpToWrite.name:
|
|
if not EpisodeName.text:
|
|
EpisodeName.text = curEpToWrite.name
|
|
else:
|
|
EpisodeName.text = EpisodeName.text + ", " + curEpToWrite.name
|
|
|
|
if curEpToWrite.description:
|
|
if not Overview.text:
|
|
Overview.text = curEpToWrite.description
|
|
else:
|
|
Overview.text = Overview.text + "\r" + curEpToWrite.description
|
|
|
|
# collect all directors, guest stars and writers
|
|
if getattr(myEp, 'director', None) is not None:
|
|
persons_dict['Director'] += [x.strip() for x in myEp['director'].split('|') if x]
|
|
if getattr(myEp, 'gueststars', None) is not None:
|
|
persons_dict['GuestStar'] += [x.strip() for x in myEp['gueststars'].split('|') if x]
|
|
if getattr(myEp, 'writer', None) is not None:
|
|
persons_dict['Writer'] += [x.strip() for x in myEp['writer'].split('|') if x]
|
|
|
|
# fill in Persons section with collected directors, guest starts and writers
|
|
for person_type, names in persons_dict.iteritems():
|
|
# remove doubles
|
|
names = list(set(names))
|
|
for cur_name in names:
|
|
Person = etree.SubElement(Persons, "Person")
|
|
cur_person_name = etree.SubElement(Person, "Name")
|
|
cur_person_name.text = cur_name
|
|
cur_person_type = etree.SubElement(Person, "Type")
|
|
cur_person_type.text = person_type
|
|
|
|
helpers.indentXML(rootNode)
|
|
data = etree.ElementTree(rootNode)
|
|
|
|
return data
|
|
|
|
|
|
# present a standard "interface" from the module
|
|
metadata_class = MediaBrowserMetadata
|