pandora/pandora/item/models.py
2011-09-04 20:36:10 +02:00

1186 lines
43 KiB
Python

# -*- coding: utf-8 -*-
# vi:si:et:sw=4:sts=4:ts=4
from __future__ import division, with_statement
from datetime import datetime
import os.path
import re
import subprocess
from glob import glob
import shutil
import uuid
import unicodedata
from urllib import quote
from django.db import models
from django.db.models import Count, Q, Sum
from django.core.files.base import ContentFile
from django.utils import simplejson as json
from django.conf import settings
from django.contrib.auth.models import User, Group
from django.db.models.signals import pre_delete
from django.contrib.sites.models import Site
import ox
from ox.django import fields
from ox.normalize import canonicalTitle
import ox.web.imdb
import ox.image
import managers
import utils
import tasks
from .timelines import join_timelines
from archive import extract
from annotation.models import Annotation, Layer
import archive.models
from person.models import get_name_sort
from app.models import site_config
def get_item(info, user=None, async=False):
'''
info dict with:
imdbId, title, director, episode_title, season, series
'''
if settings.USE_IMDB:
if 'imdbId' in info and info['imdbId']:
try:
item = Item.objects.get(itemId=info['imdbId'])
except Item.DoesNotExist:
item = Item(itemId=info['imdbId'])
if 'title' in info and 'director' in info:
item.external_data = {
'title': info['title'],
'director': info['director'],
'year': info.get('year', '')
}
item.user = user
item.save()
if async:
tasks.update_external.delay(item.itemId)
else:
item.update_external()
else:
q = Item.objects.all()
for key in ('title', 'director', 'year'):
if key in info and info[key]:
if isinstance(info[key], list):
q = q.filter(find__key=key, find__value='\n'.join(info[key]))
else:
q = q.filter(find__key=key, find__value=info[key])
if q.count() >= 1:
item = q[0]
elif not 'oxdbId' in info:
item = Item()
item.data = {
'title': info['title'],
'director': info['director'],
'year': info.get('year', '')
}
for key in ('episode_title', 'series_title', 'season', 'episode'):
if key in info and info[key]:
item.data[key] = info[key]
item.oxdbId = item.oxdb_id()
item.save()
else:
try:
item = Item.objects.get(itemId=info['oxdbId'])
except Item.DoesNotExist:
item = Item()
item.data = {
'title': info['title'],
'director': info['director'],
'year': info.get('year', '')
}
item.itemId = info['oxdbId']
for key in ('episode_title', 'series_title', 'season', 'episode'):
if key in info and info[key]:
item.data[key] = info[key]
try:
existing_item = Item.objects.get(oxdbId=item.oxdb_id())
item = existing_item
except Item.DoesNotExist:
item.save()
else:
qs = Item.objects.filter(find__key='title', find__value=info['title'])
if qs.count() == 1:
item = qs[0]
else:
item = Item()
item.data = {
'title': info['title']
}
item.user = user
item.save()
return item
class Item(models.Model):
created = models.DateTimeField(auto_now_add=True)
modified = models.DateTimeField(auto_now=True)
published = models.DateTimeField(default=datetime.now, editable=False)
user = models.ForeignKey(User, null=True, related_name='items')
groups = models.ManyToManyField(Group, blank=True, related_name='items')
#while metadata is updated, files are set to rendered=False
rendered = models.BooleanField(default=False, db_index=True)
public = models.BooleanField(default=False, db_index=True)
itemId = models.CharField(max_length=128, unique=True, blank=True)
oxdbId = models.CharField(max_length=42, unique=True, blank=True, null=True)
external_data = fields.DictField(default={}, editable=False)
data = fields.DictField(default={}, editable=False)
json = fields.DictField(default={}, editable=False)
poster = models.ImageField(default=None, blank=True,
upload_to=lambda i, x: i.path("poster.jpg"))
poster_source = models.TextField(blank=True)
poster_height = models.IntegerField(default=0)
poster_width = models.IntegerField(default=0)
poster_frame = models.FloatField(default=-1)
icon = models.ImageField(default=None, blank=True,
upload_to=lambda i, x: i.path("icon.jpg"))
torrent = models.FileField(default=None, blank=True,
upload_to=lambda i, x: i.path('torrent.torrent'))
stream_info = fields.DictField(default={}, editable=False)
#stream related fields
stream_aspect = models.FloatField(default=4/3)
objects = managers.ItemManager()
def get(self, key, default=None):
if self.data and key in self.data:
return self.data[key]
if self.external_data and key in self.external_data:
return self.external_data[key]
return default
def access(self, user):
#check rights level
if self.public:
return True
elif user.is_authenticated() and \
(user.is_staff or self.user == user or \
self.groups.filter(id__in=user.groups.all()).count() > 0):
return True
return False
def editable(self, user):
if user.is_anonymous():
return False
if user.is_staff or \
self.user == user or \
self.groups.filter(id__in=user.groups.all()).count() > 0:
return True
return False
def edit(self, data):
#FIXME: how to map the keys to the right place to write them to?
if 'id' in data:
#FIXME: check if id is valid and exists and move/merge items accordingly
del data['id']
if 'groups' in data:
groups = data.pop('groups')
self.groups.exclude(name__in=groups).delete()
for g in groups:
group, created = Group.objects.get_or_create(name=g)
self.groups.add(group)
for key in data:
if key != 'id':
self.data[key] = data[key]
self.save()
def reviews(self):
reviews = self.get('reviews', [])
_reviews = []
for r in reviews:
for url in settings.REVIEW_WHITELIST:
if url in r[0]:
_reviews.append({
'source': settings.REVIEW_WHITELIST[url],
'url': r[0]
})
return _reviews
def update_external(self):
if len(self.itemId) == 7:
data = ox.web.imdb.Imdb(self.itemId)
#FIXME: all this should be in ox.web.imdb.Imdb
for key in ('directors', 'writers', 'editors', 'producers',
'cinematographers', 'languages', 'genres', 'keywords',
'episode_directors'):
if key in data:
data[key[:-1]] = data.pop(key)
if 'countries' in data:
data['country'] = data.pop('countries')
if 'release date' in data:
data['releasedate'] = data.pop('release date')
if isinstance(data['releasedate'], list):
data['releasedate'] = min(data['releasedate'])
if 'plot' in data:
data['summary'] = data.pop('plot')
if 'cast' in data:
if isinstance(data['cast'][0], basestring):
data['cast'] = [data['cast']]
data['actor'] = [c[0] for c in data['cast']]
data['cast'] = map(lambda x: {'actor': x[0], 'character': x[1]}, data['cast'])
if 'trivia' in data:
def fix_links(t):
def fix_names(m):
return '<a href="/?find=name:%s">%s</a>' % (
quote(m.group(2).encode('utf-8')), m.group(2)
)
t = re.sub('<a href="(/name/.*?/)">(.*?)</a>', fix_names, t)
def fix_titles(m):
return '<a href="/?find=title:%s">%s</a>' % (
quote(m.group(2).encode('utf-8')), m.group(2)
)
t = re.sub('<a href="(/title/.*?/)">(.*?)</a>', fix_titles, t)
return t
data['trivia'] = [fix_links(t) for t in data['trivia']]
if 'aspectratio' in data:
data['aspectRatio'] = data.pop('aspectratio')
#filter reviews
self.external_data = data
self.save()
def __unicode__(self):
year = self.get('year')
if year:
return u'%s (%s)' % (self.get('title', 'Untitled'), self.get('year'))
return self.get('title', u'Untitled')
def get_absolute_url(self):
return '/%s' % self.itemId
def save(self, *args, **kwargs):
if not self.id:
if not self.itemId:
self.itemId = str(uuid.uuid1())
super(Item, self).save(*args, **kwargs)
if not settings.USE_IMDB:
self.itemId = ox.to32(self.id)
self.oxdbId = self.oxdb_id()
#id changed, what about existing item with new id?
if settings.USE_IMDB and len(self.itemId) != 7 and self.oxdbId != self.itemId:
self.itemId = self.oxdbId
#FIXME: move files to new id here
if self.poster:
self.poster_height = self.poster.height
self.poster_width = self.poster.width
else:
self.poster_height = 128
self.poster_width = 80
self.update_find()
self.update_sort()
self.update_facets()
update_poster = False
if not settings.USE_IMDB:
if self.poster_frame == -1 and self.sort.duration:
self.poster_frame = self.sort.duration/2
update_poster = True
if not self.get('runtime') and self.sort.duration:
self.data['runtime'] = self.sort.duration
self.update_sort()
self.json = self.get_json()
super(Item, self).save(*args, **kwargs)
if update_poster:
tasks.update_poster.delay(self.itemId)
def delete_files(self):
path = os.path.join(settings.MEDIA_ROOT, self.path())
if os.path.exists(path):
shutil.rmtree(path)
def delete(self, *args, **kwargs):
self.delete_files()
super(Item, self).delete(*args, **kwargs)
def merge_with(self, other):
'''
move all related tables to other and delete self
'''
for l in self.lists.all():
l.remove(self)
if l.items.filter(id=other.id) == 0:
l.add(other)
#FIXME: should this really happen for annotations?
for a in self.annotations.all():
a.item = other
if hasattr(self, 'files'):
for f in self.files.all():
f.item = other
f.save()
self.delete()
other.save()
#FIXME: update poster, stills and streams after this
def get_posters(self):
url = self.prefered_poster_url()
index = []
services = [p['service']
for p in self.poster_urls.values("service")
.annotate(Count("id")).order_by()]
for service in settings.POSTER_PRECEDENCE:
if service in services:
index.append(service)
for service in services:
if service not in index:
index.append(service)
if settings.URL not in index:
index.append(settings.URL)
posters = []
poster = self.path('siteposter.jpg')
poster = os.path.abspath(os.path.join(settings.MEDIA_ROOT, poster))
if os.path.exists(poster):
posters.append({
'url': '/%s/siteposter.jpg' % self.itemId,
'width': 640,
'height': 1024,
'source': settings.URL,
'selected': url == None,
'index': index.index(settings.URL)
})
got = {}
for p in self.poster_urls.all().order_by('-height'):
if p.service not in got:
got[p.service] = 1
posters.append({
'url': p.url,
'width': p.width,
'height': p.height,
'source': p.service,
'selected': p.url == url,
'index': index.index(p.service)
})
posters.sort(key=lambda a: a['index'])
return posters
def get_frames(self):
frames = []
pframes = self.poster_frames()
if pframes:
pos = self.poster_frame
if pos < 0:
pos = int(len(pframes) / 2)
p = 0
for f in pframes:
frames.append({
'index': p,
'position': f['position'],
'selected': p == pos,
'url': '/%s/frameposter%d.jpg' %(self.itemId, p),
'height': f['height'],
'width': f['width']
})
p += 1
return frames
def get_stream(self):
for s in self.streams():
return s.json()
def get_layers(self, user=None):
layers = {}
for l in Layer.objects.all():
ll = layers.setdefault(l.name, [])
qs = Annotation.objects.filter(layer=l, item=self)
if l.private:
if user and user.is_anonymous():
user = None
qs = qs.filter(user=user)
for a in qs.order_by('start'):
ll.append(a.json())
return layers
def get_json(self, keys=None):
i = {
'id': self.itemId,
'rendered': self.rendered
}
i.update(self.external_data)
i.update(self.data)
for k in site_config()['itemKeys']:
key = k['id']
if not keys or key in keys:
if key not in i:
value = self.get(key)
#also get values from sort table, i.e. numberof values
if not value and self.sort and hasattr(self.sort, key):
value = getattr(self.sort, key)
if value:
i[key] = value
if 'reviews' in i:
i['reviews'] = self.reviews()
if not i['reviews']:
del i['reviews']
if 'cast' in i and isinstance(i['cast'][0], basestring):
i['cast'] = [i['cast']]
if 'cast' in i and isinstance(i['cast'][0], list):
i['cast'] = map(lambda x: {'actor': x[0], 'character': x[1]}, i['cast'])
if not keys or 'posterRatio' in keys:
i['posterRatio'] = self.poster_width / self.poster_height
streams = self.streams()
i['durations'] = [s.duration for s in streams]
i['duration'] = sum(i['durations'])
i['parts'] = len(i['durations'])
if i['parts']:
i['videoRatio'] = streams[0].aspect_ratio
#only needed by admins
if keys and 'posters' in keys:
i['posters'] = self.get_posters()
if keys and 'frames' in keys:
i['frames'] = self.get_frames()
if keys:
info = {}
for key in keys:
if key in i:
info[key] = i[key]
return info
return i
def oxdb_id(self):
if not settings.USE_IMDB:
return self.itemId
if not self.get('title') and not self.get('director'):
return None
return utils.oxdb_id(self.get('title', ''), self.get('director', []), str(self.get('year', '')),
self.get('season', ''), self.get('episode', ''),
self.get('episode_title', ''), self.get('episode_director', []), self.get('episode_year', ''))
'''
Search related functions
'''
def update_find(self):
def save(key, value):
f, created = ItemFind.objects.get_or_create(item=self, key=key)
if value not in ('', ):
if isinstance(value, basestring):
value = value.strip()
f.value = value
f.save()
else:
f.delete()
#FIXME: use site_config
save('title', u'\n'.join([self.get('title', 'Untitled'),
self.get('original_title', '')]))
for key in self.facet_keys:
if key == 'character':
values = self.get('cast', '')
if values:
if isinstance(values[0], basestring):
values = [values]
if isinstance(values[0], list):
values = map(lambda x: {'actor': x[0], 'character': x[1]}, values)
values = [i['character'] for i in values]
elif key == 'name':
values = []
for k in map(lambda x: x['id'],
filter(lambda x: x.get('sort') == 'person',
config['itemKeys'])):
values += self.get(k, [])
else:
values = self.get(key, '')
if isinstance(values, list):
save(key, '\n'.join(values))
else:
save(key, values)
save('summary', self.get('summary', ''))
save('trivia', ' '.join(self.get('trivia', [])))
#FIXME:
qs = Annotation.objects.filter(layer__name='subtitles', item=self).order_by('start')
save('subtitles', '\n'.join([l.value for l in qs]))
save('filename', '\n'.join([u'%s/%s' % (f.folder, f.name) for f in self.files.all()]))
def update_sort(self):
try:
s = self.sort
except ItemSort.DoesNotExist:
s = ItemSort(item=self)
def sortNames(values):
sort_value = u''
if values:
sort_value = u'; '.join([get_name_sort(name) for name in values])
if not sort_value:
sort_value = u''
return sort_value
def set_value(s, name, value):
if not value:
value = None
setattr(s, name, value)
base_keys = (
'id',
'aspectratio',
'duration',
'hue',
'saturation',
'lightness',
'volume',
'clips',
'cuts',
'cutsperminute',
'words',
'wordsperminute',
'resolution',
'pixels',
'size',
'bitrate',
'numberoffiles',
'parts',
'published',
'modified',
'popularity',
)
for key in filter(lambda k: 'columnWidth' in k, config['itemKeys']):
name = key['id']
source = name
sort_type = key.get('sort', key['type'])
if 'value' in key:
if 'layer' in key['value']:
continue
source = key['value']['key']
sort_type = key['value'].get('type', sort_type)
if name not in base_keys:
if sort_type == 'title':
value = utils.sort_title(canonicalTitle(self.get(source, u'Untitled')))
value = utils.sort_string(value)
set_value(s, name, value)
elif sort_type == 'person':
value = sortNames(self.get(source, []))
value = utils.sort_string(value)[:955]
set_value(s, name, value)
elif sort_type == 'string':
value = self.get(source, u'')
if isinstance(value, list):
value = u','.join(value)
value = utils.sort_string(value)[:955]
set_value(s, name, value)
elif sort_type in ('length', 'integer', 'float'):
#can be length of strings or length of arrays, i.e. keywords
value = self.get(source)
if isinstance(value, list):
value = len(value)
set_value(s, name, value)
elif sort_type == 'words':
value = self.get(source)
if isinstance(value, list):
value = '\n'.join(value)
if value:
value = len(value.split(' '))
set_value(s, name, value)
elif sort_type == 'year':
value = self.get(source)
set_value(s, name, value)
elif sort_type == 'date':
value = self.get(source)
if isinstance(value, basestring):
value = datetime.strptime(value, '%Y-%m-%d')
set_value(s, name, value)
#sort keys based on database, these will always be available
s.itemId = self.itemId.replace('0x', 'xx')
s.modified = self.modified
s.published = self.published
# sort values based on data from videos
s.words = 0 #FIXME: get words from all layers or something
s.wordsperminute = 0
s.clips = 0 #FIXME: get clips from all layers or something
s.popularity = 0 #FIXME: get popularity from somewhere
videos = self.files.filter(active=True, is_video=True)
if videos.count() > 0:
s.duration = sum([v.duration for v in videos])
v = videos[0]
s.resolution = v.width * v.height
s.aspectratio = float(utils.parse_decimal(v.display_aspect_ratio))
s.pixels = sum([v.pixels for v in videos])
s.numberoffiles = self.files.all().count()
s.parts = videos.count()
s.size = sum([v.size for v in videos]) #FIXME: only size of movies?
s.bitrate = s.size * 8 / s.duration
s.volume = 0
else:
s.duration = None
s.resolution = None
s.aspectratio = None
s.bitrate = None
s.pixels = None
s.filename = None
s.files = None
s.size = None
s.volume = None
s.parts = 0
if 'color' in self.data:
s.hue, s.saturation, s.lightness = self.data['color']
else:
s.hue = None
s.saturation = None
s.brighness = None
s.cuts = len(self.data.get('cuts', []))
if s.duration:
s.cutsperminute = s.cuts / (s.duration/60)
else:
s.cutsperminute = None
s.popularity = self.accessed.aggregate(Sum('accessed'))['accessed__sum']
s.save()
def update_facets(self):
#FIXME: what to do with Unkown Director, Year, Country etc.
for key in self.facet_keys + ['title']:
current_values = self.get(key, [])
if key == 'title':
if current_values:
current_values = [current_values]
else:
current_values = []
ot = self.get('original_title')
if ot:
current_values.append(ot)
#FIXME: is there a better way to build name collection?
if key == 'name':
current_values = []
for k in map(lambda x: x['id'],
filter(lambda x: x.get('sort') == 'person',
config['itemKeys'])):
current_values += self.get(k, [])
if not isinstance(current_values, list):
current_values = [unicode(current_values)]
current_values = list(set(current_values))
saved_values = [i.value for i in Facet.objects.filter(item=self, key=key)]
removed_values = filter(lambda i: i not in current_values, saved_values)
if removed_values:
Facet.objects.filter(item=self, key=key, value__in=removed_values).delete()
for value in current_values:
if value not in saved_values:
value_sort = value
if key in self.person_keys:
value_sort = get_name_sort(value)
f = Facet(key=key, value=value, value_sort=value_sort)
f.item = self
f.save()
def path(self, name=''):
h = self.itemId
return os.path.join('items', h[:2], h[2:4], h[4:6], h[6:], name)
'''
Video related functions
'''
def frame(self, position, height=128):
offset = 0
streams = self.streams()
for stream in streams:
if stream.duration + offset < position:
offset += stream.duration
else:
position = position - offset
height = min(height, stream.resolution)
path = os.path.join(settings.MEDIA_ROOT, stream.path(),
'frames', "%dp"%height, "%s.jpg"%position)
if not os.path.exists(path) and stream.video:
extract.frame(stream.video.path, path, position, height)
if not os.path.exists(path):
return None
return path
@property
def timeline_prefix(self):
videos = self.streams()
if len(videos) == 1:
return os.path.join(settings.MEDIA_ROOT, videos[0].path('timeline'))
return os.path.join(settings.MEDIA_ROOT, self.path(), 'timeline')
def get_files(self, user):
#FIXME: limit by user
return [f.json() for f in self.files.all()]
def users_with_files(self):
return User.objects.filter(volumes__files__file__item=self).distinct()
def update_wanted(self):
users = self.users_with_files()
if users.filter(is_superuser=True).count()>0:
files = self.files.filter(instances__volume__user__is_superuser=True)
users = User.objects.filter(volumes__files__file__in=files,
is_superuser=True).distinct()
elif users.filter(is_staff=True).count()>0:
files = self.files.filter(instances__volume__user__is_staff=True)
users = User.objects.filter(volumes__files__file__in=files,
is_staff=True).distinct()
else:
files = self.files.all()
files = files.filter(is_video=True, instances__extra=False, instances__gt=0).order_by('part')
folders = list(set([f.folder for f in files]))
if len(folders) > 1:
files = files.filter(folder=folders[0])
files.update(wanted=True)
self.files.exclude(id__in=files).update(wanted=False)
def update_selected(self):
files = archive.models.File.objects.filter(item=self,
streams__available=True,
streams__source=None)
if files.count() == 0:
return
def get_level(users):
if users.filter(is_superuser=True).count() > 0: level = 0
elif users.filter(is_staff=True).count() > 0: level = 1
else: level = 2
return level
current_users = User.objects.filter(volumes__files__file__in=self.files.filter(active=True)).distinct()
current_level = get_level(current_users)
users = User.objects.filter(volumes__files__file__in=files).distinct()
possible_level = get_level(users)
if possible_level < current_level:
files = self.files.filter(instances__volume__user__in=users).order_by('part')
#FIXME: this should be instance folders
folders = list(set([f.folder
for f in files.filter(is_video=True, instances__extra=False)]))
files = files.filter(folder__startswith=folders[0])
files.update(active=True)
self.rendered = False
self.save()
self.update_timeline()
else:
files = self.files.filter(instances__volume__user__in=current_users).order_by('part')
#FIXME: this should be instance folders
folders = list(set([f.folder
for f in files.filter(is_video=True, instances__extra=False)]))
files = files.filter(folder__startswith=folders[0])
if files.filter(active=False).count() > 0:
files.update(active=True)
self.rendered = False
self.save()
self.update_timeline()
def make_torrent(self):
base = self.path('torrent')
base = os.path.abspath(os.path.join(settings.MEDIA_ROOT, base))
if os.path.exists(base):
shutil.rmtree(base)
os.makedirs(base)
base = self.path('torrent/%s' % self.get('title'))
base = os.path.abspath(os.path.join(settings.MEDIA_ROOT, base))
size = 0
duration = 0.0
streams = self.streams()
if streams.count() == 1:
url = "%s/torrent/%s.webm" % (self.get_absolute_url(),
quote(self.get('title').encode('utf-8')))
video = "%s.webm" % base
v = streams[0]
os.symlink(v.video.path, video)
size = v.video.size
duration = v.duration
else:
url = "%s/torrent/" % self.get_absolute_url()
part = 1
os.makedirs(base)
for v in streams:
video = "%s/%s.Part %d.webm" % (base, self.get('title'), part)
part += 1
os.symlink(v.video.path, video)
size += v.video.size
duration += v.duration
video = base
torrent = '%s.torrent' % base
url = "http://%s%s" % (Site.objects.get_current().domain, url)
meta = {
'target': torrent,
'url-list': url,
}
if duration:
meta['playtime'] = ox.formatDuration(duration*1000)[:-4]
#slightly bigger torrent file but better for streaming
piece_size_pow2 = 15 #1 mbps -> 32KB pieces
if size / duration >= 1000000:
piece_size_pow2 = 16 #2 mbps -> 64KB pieces
meta['piece_size_pow2'] = piece_size_pow2
ox.torrent.createTorrent(video, settings.TRACKER_URL, meta)
self.torrent.name = self.path('torrent/%s.torrent' % self.get('title'))
self.save()
def streams(self):
return archive.models.Stream.objects.filter(source=None, available=True,
file__item=self, file__is_video=True, file__active=True).order_by('file__part')
def update_timeline(self, force=False):
config = site_config()
streams = self.streams()
self.make_timeline()
self.data['cuts'] = extract.cuts(self.timeline_prefix)
self.data['color'] = extract.average_color(self.timeline_prefix)
#extract.timeline_strip(self, self.data['cuts'], stream.info, self.timeline_prefix[:-8])
self.select_frame()
self.make_local_poster()
self.make_poster()
self.make_icon()
if config['video']['download']:
self.make_torrent()
self.load_subtitles()
self.rendered = streams != []
self.save()
'''
Poster related functions
'''
def update_poster_urls(self):
_current = {}
for s in settings.POSTER_SERVICES:
url = '%s?id=%s'%(s, self.itemId)
try:
data = json.loads(ox.net.readUrlUnicode(url))
except:
continue
for service in data:
if service not in _current:
_current[service] = []
for poster in data[service]:
_current[service].append(poster)
#FIXME: remove urls that are no longer listed
for service in _current:
for poster in _current[service]:
p, created = PosterUrl.objects.get_or_create(item=self, url=poster['url'], service=service)
if created:
p.width = poster['width']
p.height = poster['height']
p.save()
def delete_poster(self):
if self.poster:
path = self.poster.path
self.poster.delete()
else:
poster= self.path('poster.jpg')
path = os.path.abspath(os.path.join(settings.MEDIA_ROOT, poster))
for f in glob(path.replace('.jpg', '*.jpg')):
os.unlink(f)
def prefered_poster_url(self):
self.update_poster_urls()
service = self.poster_source
if service and service != settings.URL:
for u in self.poster_urls.filter(service=service).order_by('-height'):
return u.url
if not service:
for service in settings.POSTER_PRECEDENCE:
for u in self.poster_urls.filter(service=service).order_by('-height'):
return u.url
return None
def make_timeline(self):
streams = self.streams()
if len(streams) > 1:
timelines = [s.timeline_prefix for s in self.streams()]
join_timelines(timelines, self.timeline_prefix)
def make_poster(self, force=False):
if not self.poster or force:
url = self.prefered_poster_url()
if url:
data = ox.net.readUrl(url)
self.delete_poster()
self.poster.save('poster.jpg', ContentFile(data))
self.save()
else:
self.delete_poster()
poster = self.make_local_poster()
with open(poster) as f:
self.poster.save('poster.jpg', ContentFile(f.read()))
def make_local_poster(self):
poster = self.path('siteposter.jpg')
poster = os.path.abspath(os.path.join(settings.MEDIA_ROOT, poster))
frame = self.get_poster_frame_path()
timeline = '%s.64.png' % self.timeline_prefix
director = u', '.join(self.get('director', ['Unknown Director']))
cmd = [settings.ITEM_POSTER,
'-t', self.get('title').encode('utf-8'),
'-d', director.encode('utf-8'),
'-y', str(self.get('year', '')),
'-p', poster
]
if frame:
cmd += [
'-f', frame,
]
if os.path.exists(timeline):
cmd += [
'-l', timeline,
]
if settings.USE_IMDB:
if len(self.itemId) == 7:
cmd += ['-i', self.itemId]
cmd += ['-o', self.oxdbId]
else:
cmd += ['-i', self.itemId]
ox.makedirs(os.path.join(settings.MEDIA_ROOT,self.path()))
p = subprocess.Popen(cmd)
p.wait()
for f in glob(poster.replace('.jpg', '*.jpg')):
if f != poster:
os.unlink(f)
return poster
def poster_frames(self):
frames = []
offset = 0
for f in self.files.filter(active=True, is_video=True):
for ff in f.frames.all():
frames.append({
'position': offset + ff.position,
'path': ff.frame.path,
'width': ff.frame.width,
'height': ff.frame.height
})
offset += f.duration
return frames
def select_frame(self):
frames = self.poster_frames()
if frames:
heat = [ox.image.getImageHeat(f['path']) for f in frames]
self.poster_frame = heat.index(max(heat))
self.save()
def get_poster_frame_path(self):
frames = self.poster_frames()
if self.poster_frame >= 0:
if frames and len(frames) > int(self.poster_frame):
return frames[int(self.poster_frame)]['path']
else:
size = site_config()['video']['resolutions'][0]
return self.frame(self.poster_frame, size)
if frames:
return frames[int(len(frames)/2)]['path']
def make_icon(self):
frame = self.get_poster_frame_path()
icon = self.path('icon.jpg')
self.icon.name = icon
timeline = '%s.64.png' % self.timeline_prefix
cmd = [settings.ITEM_ICON,
'-i', self.icon.path
]
if os.path.exists(timeline):
cmd += ['-l', timeline]
if frame:
cmd += ['-f', frame]
p = subprocess.Popen(cmd)
p.wait()
self.save()
icons = os.path.abspath(os.path.join(settings.MEDIA_ROOT, icon))
icons = glob(icons.replace('.jpg', '*.jpg'))
for f in filter(lambda p: not p.endswith('/icon.jpg'), icons):
os.unlink(f)
return icon
def load_subtitles(self):
layer = Layer.objects.get(name='subtitles')
Annotation.objects.filter(layer=layer,item=self).delete()
offset = 0
language = ''
languages = [f.language for f in self.files.filter(active=True, is_subtitle=True,
available=True)]
if languages:
if 'en' in languages:
language = 'en'
elif '' in languages:
language = ''
else:
language = languages[0]
for f in self.files.filter(active=True, is_subtitle=True,
available=True, language=language).order_by('part'):
user = f.instances.all()[0].volume.user
for data in f.srt(offset):
annotation = Annotation(
item=f.item,
layer=layer,
start=data['in'],
end=data['out'],
value=data['value'],
user=user
)
annotation.save()
duration = self.files.filter(Q(is_audio=True)|Q(is_video=True)) \
.filter(active=True, available=True, part=f.part)
if duration:
duration = duration[0].duration
else:
Annotation.objects.filter(layer=layer,item=self).delete()
break
offset += duration
self.update_find()
def delete_item(sender, **kwargs):
i = kwargs['instance']
i.delete_files()
pre_delete.connect(delete_item, sender=Item)
config = site_config()
Item.facet_keys = []
for key in config['itemKeys']:
if 'autocomplete' in key and not 'autocompleteSortKey' in key:
Item.facet_keys.append(key['id'])
Item.person_keys = []
for key in config['itemKeys']:
if 'sort' in key and key['sort'] == 'person':
Item.person_keys.append(key['id'])
class ItemFind(models.Model):
"""
used to find items,
item.update_find populates this table
its used in manager.ItemManager
"""
class Meta:
unique_together = ("item", "key")
item = models.ForeignKey('Item', related_name='find', db_index=True)
key = models.CharField(max_length=200, db_index=True)
value = models.TextField(blank=True)
def __unicode__(self):
return u"%s=%s" % (self.key, self.value)
'''
ItemSort
table constructed based on info in site_config['itemKeys']
'''
attrs = {
'__module__': 'item.models',
'item': models.OneToOneField('Item', related_name='sort', primary_key=True),
}
for key in filter(lambda k: 'columnWidth' in k, config['itemKeys']):
name = key['id']
name = {'id': 'itemId'}.get(name, name)
sort_type = key.get('sort', key['type'])
if isinstance(sort_type, list):
sort_type = sort_type[0]
model = {
'char': (models.CharField, dict(null=True, max_length=1000, db_index=True)),
'year': (models.CharField, dict(null=True, max_length=4, db_index=True)),
'integer': (models.BigIntegerField, dict(null=True, blank=True, db_index=True)),
'float': (models.FloatField, dict(null=True, blank=True, db_index=True)),
'date': (models.DateTimeField, dict(null=True, blank=True, db_index=True))
}[{
'string': 'char',
'title': 'char',
'person': 'char',
'year': 'year',
'words': 'integer',
'length': 'integer',
'date': 'date',
'hue': 'float',
}.get(sort_type, sort_type)]
attrs[name] = model[0](**model[1])
ItemSort = type('ItemSort', (models.Model,), attrs)
ItemSort.fields = [f.name for f in ItemSort._meta.fields]
class Access(models.Model):
class Meta:
unique_together = ("item", "user")
access = models.DateTimeField(auto_now=True)
item = models.ForeignKey(Item, related_name='accessed')
user = models.ForeignKey(User, null=True, related_name='accessed_items')
accessed = models.IntegerField(default=0)
def save(self, *args, **kwargs):
if not self.accessed:
self.accessed = 0
self.accessed += 1
super(Access, self).save(*args, **kwargs)
def __unicode__(self):
if self.user:
return u"%s/%s/%s" % (self.user, self.item, self.access)
return u"%s/%s" % (self.item, self.access)
class Facet(models.Model):
'''
used for keys that can have multiple values like people, languages etc.
does not perform to well if total number of items goes above 10k
this happens for keywords in 0xdb right now
'''
class Meta:
unique_together = ("item", "key", "value")
item = models.ForeignKey('Item', related_name='facets')
key = models.CharField(max_length=200, db_index=True)
value = models.CharField(max_length=1000, db_index=True)
value_sort = models.CharField(max_length=1000, db_index=True)
def __unicode__(self):
return u"%s=%s" % (self.key, self.value)
def save(self, *args, **kwargs):
if not self.value_sort:
self.value_sort = utils.sort_string(self.value)
super(Facet, self).save(*args, **kwargs)
class PosterUrl(models.Model):
class Meta:
unique_together = ("item", "service", "url")
ordering = ('-height', )
item = models.ForeignKey(Item, related_name='poster_urls')
url = models.CharField(max_length=1024)
service = models.CharField(max_length=1024)
width = models.IntegerField(default=80)
height = models.IntegerField(default=128)
def __unicode__(self):
return u'%s %s %dx%d' % (unicode(self.item), self.service, self.width, self.height)