1
0
mirror of https://github.com/moparisthebest/SickRage synced 2024-11-18 15:25:07 -05:00
SickRage/sickbeard/tvcache.py
2014-12-24 09:19:25 -08:00

375 lines
14 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/>.
from __future__ import with_statement
import time
import datetime
import itertools
import traceback
import sickbeard
from sickbeard import db
from sickbeard import logger
from sickbeard.common import Quality
from sickbeard import helpers
from sickbeard.exceptions import ex
from sickbeard.exceptions import AuthException
from sickbeard.rssfeeds import RSSFeeds
from name_parser.parser import NameParser, InvalidNameException, InvalidShowException
from sickbeard import encodingKludge as ek
class CacheDBConnection(db.DBConnection):
def __init__(self, providerName):
db.DBConnection.__init__(self, "cache.db")
# Create the table if it's not already there
try:
if not self.hasTable(providerName):
self.action(
"CREATE TABLE [" + providerName + "] (name TEXT, season NUMERIC, episodes TEXT, indexerid NUMERIC, url TEXT, time NUMERIC, quality TEXT, release_group TEXT)")
else:
sqlResults = self.select("SELECT url, COUNT(url) AS count FROM [" + providerName + "] GROUP BY url HAVING count > 1")
for cur_dupe in sqlResults:
self.action("DELETE FROM [" + providerName + "] WHERE url = ?", [cur_dupe["url"]])
# add unique index to prevent further dupes from happening if one does not exist
self.action("CREATE UNIQUE INDEX IF NOT EXISTS idx_url ON [" + providerName + "] (url)")
# add release_group column to table if missing
if not self.hasColumn(providerName, 'release_group'):
self.addColumn(providerName, 'release_group', "TEXT", "")
# add version column to table if missing
if not self.hasColumn(providerName, 'version'):
self.addColumn(providerName, 'version', "NUMERIC", "-1")
except Exception, e:
if str(e) != "table [" + providerName + "] already exists":
raise
# Create the table if it's not already there
try:
if not self.hasTable('lastUpdate'):
self.action("CREATE TABLE lastUpdate (provider TEXT, time NUMERIC)")
except Exception, e:
if str(e) != "table lastUpdate already exists":
raise
class TVCache():
def __init__(self, provider):
self.provider = provider
self.providerID = self.provider.getID()
self.providerDB = None
self.minTime = 10
def _getDB(self):
# init provider database if not done already
if not self.providerDB:
self.providerDB = CacheDBConnection(self.providerID)
return self.providerDB
def _clearCache(self):
if self.shouldClearCache():
myDB = self._getDB()
myDB.action("DELETE FROM [" + self.providerID + "] WHERE 1")
def _get_title_and_url(self, item):
return self.provider._get_title_and_url(item)
def _getRSSData(self):
return None
def _checkAuth(self, data):
return True
def _checkItemAuth(self, title, url):
return True
def updateCache(self):
# check if we should update
if not self.shouldUpdate():
return
try:
data = self._getRSSData()
if self._checkAuth(data):
# clear cache
self._clearCache()
# set updated
self.setLastUpdate()
cl = []
for item in data['entries'] or []:
ci = self._parseItem(item)
if ci is not None:
cl.append(ci)
if len(cl) > 0:
myDB = self._getDB()
myDB.mass_action(cl)
except AuthException, e:
logger.log(u"Authentication error: " + ex(e), logger.ERROR)
except Exception, e:
logger.log(u"Error while searching " + self.provider.name + ", skipping: " + ex(e), logger.ERROR)
logger.log(traceback.format_exc(), logger.DEBUG)
def getRSSFeed(self, url, post_data=None, items=[]):
if self.provider.proxy.isEnabled():
self.provider.headers.update({'Referer': self.provider.proxy.getProxyURL()})
return RSSFeeds(self.providerID).getFeed(self.provider.proxy._buildURL(url), post_data, self.provider.headers, items)
def _translateTitle(self, title):
return u'' + title.replace(' ', '.')
def _translateLinkURL(self, url):
return url.replace('&amp;', '&')
def _parseItem(self, item):
title, url = self._get_title_and_url(item)
self._checkItemAuth(title, url)
if title and url:
title = self._translateTitle(title)
url = self._translateLinkURL(url)
logger.log(u"Attempting to add item to cache: " + title, logger.DEBUG)
return self._addCacheEntry(title, url)
else:
logger.log(
u"The data returned from the " + self.provider.name + " feed is incomplete, this result is unusable",
logger.DEBUG)
def _getLastUpdate(self):
myDB = self._getDB()
sqlResults = myDB.select("SELECT time FROM lastUpdate WHERE provider = ?", [self.providerID])
if sqlResults:
lastTime = int(sqlResults[0]["time"])
if lastTime > int(time.mktime(datetime.datetime.today().timetuple())):
lastTime = 0
else:
lastTime = 0
return datetime.datetime.fromtimestamp(lastTime)
def _getLastSearch(self):
myDB = self._getDB()
sqlResults = myDB.select("SELECT time FROM lastSearch WHERE provider = ?", [self.providerID])
if sqlResults:
lastTime = int(sqlResults[0]["time"])
if lastTime > int(time.mktime(datetime.datetime.today().timetuple())):
lastTime = 0
else:
lastTime = 0
return datetime.datetime.fromtimestamp(lastTime)
def setLastUpdate(self, toDate=None):
if not toDate:
toDate = datetime.datetime.today()
myDB = self._getDB()
myDB.upsert("lastUpdate",
{'time': int(time.mktime(toDate.timetuple()))},
{'provider': self.providerID})
def setLastSearch(self, toDate=None):
if not toDate:
toDate = datetime.datetime.today()
myDB = self._getDB()
myDB.upsert("lastSearch",
{'time': int(time.mktime(toDate.timetuple()))},
{'provider': self.providerID})
lastUpdate = property(_getLastUpdate)
lastSearch = property(_getLastSearch)
def shouldUpdate(self):
# if we've updated recently then skip the update
if datetime.datetime.today() - self.lastUpdate < datetime.timedelta(minutes=self.minTime):
logger.log(u"Last update was too soon, using old cache: today()-" + str(self.lastUpdate) + "<" + str(
datetime.timedelta(minutes=self.minTime)), logger.DEBUG)
return False
return True
def shouldClearCache(self):
# if daily search hasn't used our previous results yet then don't clear the cache
if self.lastUpdate > self.lastSearch:
return False
return True
def _addCacheEntry(self, name, url, parse_result=None, indexer_id=0):
# check if we passed in a parsed result or should we try and create one
if not parse_result:
# create showObj from indexer_id if available
showObj = None
if indexer_id:
showObj = helpers.findCertainShow(sickbeard.showList, indexer_id)
try:
myParser = NameParser(showObj=showObj, convert=True)
parse_result = myParser.parse(name)
except InvalidNameException:
logger.log(u"Unable to parse the filename " + name + " into a valid episode", logger.DEBUG)
return None
except InvalidShowException:
logger.log(u"Unable to parse the filename " + name + " into a valid show", logger.DEBUG)
return None
if not parse_result or not parse_result.series_name:
return None
# if we made it this far then lets add the parsed result to cache for usager later on
season = parse_result.season_number if parse_result.season_number else 1
episodes = parse_result.episode_numbers
if season and episodes:
# store episodes as a seperated string
episodeText = "|" + "|".join(map(str, episodes)) + "|"
# get the current timestamp
curTimestamp = int(time.mktime(datetime.datetime.today().timetuple()))
# get quality of release
quality = parse_result.quality
name = ek.ss(name)
# get release group
release_group = parse_result.release_group
# get version
version = parse_result.version
logger.log(u"Added RSS item: [" + name + "] to cache: [" + self.providerID + "]", logger.DEBUG)
return [
"INSERT OR IGNORE INTO [" + self.providerID + "] (name, season, episodes, indexerid, url, time, quality, release_group, version) VALUES (?,?,?,?,?,?,?,?,?)",
[name, season, episodeText, parse_result.show.indexerid, url, curTimestamp, quality, release_group, version]]
def searchCache(self, episode, manualSearch=False):
neededEps = self.findNeededEpisodes(episode, manualSearch)
return neededEps[episode] if len(neededEps) > 0 else []
def listPropers(self, date=None, delimiter="."):
myDB = self._getDB()
sql = "SELECT * FROM [" + self.providerID + "] WHERE name LIKE '%.PROPER.%' OR name LIKE '%.REPACK.%'"
if date != None:
sql += " AND time >= " + str(int(time.mktime(date.timetuple())))
return filter(lambda x: x['indexerid'] != 0, myDB.select(sql))
def findNeededEpisodes(self, episode, manualSearch=False):
neededEps = {}
cl = []
myDB = self._getDB()
if type(episode) != list:
sqlResults = myDB.select(
"SELECT * FROM [" + self.providerID + "] WHERE indexerid = ? AND season = ? AND episodes LIKE ?",
[episode.show.indexerid, episode.season, "%|" + str(episode.episode) + "|%"])
else:
for epObj in episode:
cl.append([
"SELECT * FROM [" + self.providerID + "] WHERE indexerid = ? AND season = ? AND episodes LIKE ? AND quality IN (" + ",".join(
[str(x) for x in epObj.wantedQuality]) + ")",
[epObj.show.indexerid, epObj.season, "%|" + str(epObj.episode) + "|%"]])
sqlResults = myDB.mass_action(cl, fetchall=True)
sqlResults = list(itertools.chain(*sqlResults))
# for each cache entry
for curResult in sqlResults:
# get the show object, or if it's not one of our shows then ignore it
showObj = helpers.findCertainShow(sickbeard.showList, int(curResult["indexerid"]))
if not showObj:
continue
# skip if provider is anime only and show is not anime
if self.provider.anime_only and not showObj.is_anime:
logger.log(u"" + str(showObj.name) + " is not an anime, skiping", logger.DEBUG)
continue
# get season and ep data (ignoring multi-eps for now)
curSeason = int(curResult["season"])
if curSeason == -1:
continue
curEp = curResult["episodes"].split("|")[1]
if not curEp:
continue
curEp = int(curEp)
curQuality = int(curResult["quality"])
curReleaseGroup = curResult["release_group"]
curVersion = curResult["version"]
# if the show says we want that episode then add it to the list
if not showObj.wantEpisode(curSeason, curEp, curQuality, manualSearch):
logger.log(u"Skipping " + curResult["name"] + " because we don't want an episode that's " +
Quality.qualityStrings[curQuality], logger.DEBUG)
continue
epObj = showObj.getEpisode(curSeason, curEp)
# build a result object
title = curResult["name"]
url = curResult["url"]
logger.log(u"Found result " + title + " at " + url)
result = self.provider.getResult([epObj])
result.show = showObj
result.url = url
result.name = title
result.quality = curQuality
result.release_group = curReleaseGroup
result.version = curVersion
result.content = None
# add it to the list
if epObj not in neededEps:
neededEps[epObj] = [result]
else:
neededEps[epObj].append(result)
# datetime stamp this search so cache gets cleared
self.setLastSearch()
return neededEps