def map(self, items):
self.twitter = TwitterResource()
Model.database = MongoDB(name=self.db_name, host=settings.db_host)
for user in items:
try:
self.crawl(user)
self.twitter.sleep_if_needed()
except Exception:
logging.exception("exception for user %s" % user.to_d())
yield None
class CrawlProcess(SplitProcess):
def __init__(self, db_name, **kwargs):
SplitProcess.__init__(self, **kwargs)
self.db_name = db_name
self.waiting = set()
def produce(self):
Model.database = MongoDB(name=self.db_name, host=settings.db_host)
endtime = datetime.utcnow()
return User.find(User.next_crawl_date < endtime, sort=User.next_crawl_date, timeout=False)
def map(self, items):
self.twitter = TwitterResource()
Model.database = MongoDB(name=self.db_name, host=settings.db_host)
for user in items:
try:
self.crawl(user)
self.twitter.sleep_if_needed()
except Exception:
logging.exception("exception for user %s" % user.to_d())
yield None
def crawl(self, user):
logging.info("visiting %s - %s", user._id, user.screen_name)
tweets = self.twitter.save_timeline(user._id, user.last_tid)
if tweets:
user.last_tid = tweets[0]._id
logging.info("saved %d for %s", len(tweets), user.screen_name)
now = datetime.utcnow()
last = user.last_crawl_date if user.last_crawl_date is not None else datetime(2010, 11, 12)
delta = now - last
seconds = delta.seconds + delta.days * 24 * 3600
tph = (3600.0 * len(tweets) / seconds + user.tweets_per_hour) / 2
user.tweets_per_hour = tph
hours = min(settings.tweets_per_crawl / tph, settings.max_hours)
user.next_crawl_date = now + timedelta(hours=hours)
user.last_crawl_date = now
user.save()
def __init__(self,slave_id):
LocalProc.__init__(self,'lookup',slave_id)
self.twitter = TwitterResource()
self.gisgraphy = GisgraphyResource()
class LookupSlave(LocalProc):
def __init__(self,slave_id):
LocalProc.__init__(self,'lookup',slave_id)
self.twitter = TwitterResource()
self.gisgraphy = GisgraphyResource()
def run(self):
while True:
jobs = []
for x in xrange(100):
try:
# reserve blocks to wait when x is 0, but returns None for 1-99
j = self.stalk.reserve(0 if x else None)
except beanstalkc.DeadlineSoon:
break
if j is None:
break
jobs.append(j)
bodies = [LookupJobBody.from_job(j) for j in jobs]
try:
users =self.twitter.user_lookup([b._id for b in bodies])
except ResourceNotFound:
logging.info("no profile for %r",[b._id for b in bodies])
continue
logging.info("looking at %r"%[getattr(u,'screen_name','') for u in users])
for job,body,user in zip(jobs,bodies,users):
if user is None:
logging.info("no profile for %d",body._id)
job.delete()
continue
try:
self.twitter.sleep_if_needed()
logging.info("look at %s",user.screen_name)
if (not body.force) and User.in_db(user._id):
job.delete()
continue
self.crawl_user(user,body.force)
user.save()
job.delete()
except:
logging.exception("exception for job %s"%job.body)
job.bury()
logging.info("api calls remaining: %d",self.twitter.remaining)
def crawl_user(self,user,force):
user.local_prob = guess_location(user,self.gisgraphy)
if (user.local_prob != 1.0 and not force) or user.protected:
return
rels=None
tweets=None
if user.followers_count>0 and user.friends_count>0:
rels = self.twitter.get_edges(user._id)
rels.attempt_save()
if user.statuses_count>0:
tweets = self.twitter.save_timeline(user._id,last_tid=settings.min_tweet_id)
if tweets:
user.next_crawl_date = datetime.utcnow()
user.last_crawl_date = datetime.utcnow()
user.tweets_per_hour = settings.tweets_per_hour
user.last_tid = tweets[0]._id
user.lookup_done = True
if user.local_prob == 1.0 and not force:
self.score_new_users(user, rels, tweets)
def score_new_users(self, user, rels, tweets):
jobs = defaultdict(LookupJobBody)
jobs[user._id].done = True
if rels:
rfriends = rels.rfriends()
if len(rfriends) < RFRIEND_POINTS:
for u in rfriends:
jobs[u].rfriends_score = RFRIEND_POINTS/len(rfriends)
if tweets:
ats = defaultdict(int)
for tweet in tweets:
for uid in tweet.mentions:
ats[uid]+=1
for u,c in ats.iteritems():
points = c*MENTION_POINTS
if points >0:
jobs[u].mention_score = points
for k,j in jobs.iteritems():
j._id = k
j.put(self.stalk)
