def process_group(hg, crawl_id, requesters, processed_groups, dbpool):
"""Gevent worker that should process single hitgroup.
This should write some data into database and do not return any important
data.
"""
hg['keywords'] = ', '.join(hg['keywords'])
# for those hit goups that does not contain hash group, create one and
# setup apropiate flag
hg['qualifications'] = ', '.join(hg['qualifications'])
conn = dbpool.getconn(thread.get_ident())
db = DB(conn)
try:
hit_group_content_id = db.hit_group_content_id(hg['group_id'])
if hit_group_content_id is None:
# check if there's profile for current requester and if does
# exists with non-public status, then setup non public status for
# current hitsgroup content
profile = requesters.get(hg['requester_id'], None)
if profile and profile.is_public is False:
hg['is_public'] = False
else:
hg['is_public'] = True
# fresh hitgroup - create group content entry, but first add some data
# required by hitgroup content table
hg['occurrence_date'] = datetime.datetime.now()
hg['first_crawl_id'] = crawl_id
if not hg['group_id_hashed']:
# if group_id is hashed, we cannot fetch details because we
# don't know what the real hash is
hg.update(hits_group_info(hg['group_id']))
else:
hg['html'] = ''
hit_group_content_id = db.insert_hit_group_content(hg)
log.debug('new hit group content: %s;;%s',
hit_group_content_id, hg['group_id'])
hg['hit_group_content_id'] = hit_group_content_id
hg['crawl_id'] = crawl_id
hg['now'] = datetime.datetime.now()
db.insert_hit_group_status(hg)
conn.commit()
except Exception:
processed_groups.remove(hg['group_id'])
log.exception('process_group fail - rollback')
conn.rollback()
finally:
db.curr.close()
dbpool.putconn(conn, thread.get_ident())
msg = ('This really should not happen, Hitgroupstatus was processed but'
' is not on the list, race condition?')
assert hg['group_id'] in processed_groups, msg
return True
def close(self):
tid = get_ident()
sock = self.in_use.pop(tid, None)
if sock:
sock.close()
return True
return False
def checkin(self):
tid = get_ident()
if tid in self.in_use:
sock = self.in_use.pop(tid)
if not sock.is_closed:
heapq.heappush(self.free, (time.time(), sock))
return True
return False
def checkout(self):
now = time.time()
tid = get_ident()
if tid in self.in_use:
sock = self.in_use[tid]
if sock.is_closed:
del self.in_use[tid]
else:
return self.in_use[tid]
while self.free:
ts, sock = heapq.heappop(self.free)
if ts < now - self.max_age:
sock.close()
else:
self.in_use[tid] = sock
return sock
sock = self.create_socket()
self.in_use[tid] = sock
return sock
def process_group(hg, crawl_id, requesters, processed_groups):
"""Gevent worker that should process single hitgroup.
This should write some data into database and do not return any important
data.
"""
hg['keywords'] = ', '.join(hg['keywords'])
# for those hit goups that does not contain hash group, create one and
# setup apropiate flag
hg['group_id_hashed'] = not bool(hg.get('group_id', None))
hg['qualifications'] = ', '.join(hg['qualifications'])
if hg['group_id_hashed']:
composition = ';'.join(
map(str, (hg['title'], hg['requester_id'], hg['time_alloted'],
hg['reward'], hg['description'], hg['keywords'],
hg['qualifications']))) + ';'
hg['group_id'] = hashlib.md5(composition).hexdigest()
log.debug('group_id not found, creating hash: %s %s', hg['group_id'],
composition)
if hg['group_id'] in processed_groups:
# this higroup was already processed
log.info('duplicated group: %s;;%s', crawl_id, hg['group_id'])
return False
conn = dbpool.getconn(thread.get_ident())
db = DB(conn)
try:
hit_group_content_id = db.hit_group_content_id(hg['group_id'])
if hit_group_content_id is None:
# check if there's profile for current requester and if does
# exists with non-public status, then setup non public status for
# current hitsgroup content
profile = requesters.get(hg['requester_id'], None)
if profile and profile.is_public is False:
hg['is_public'] = False
else:
hg['is_public'] = True
# fresh hitgroup - create group content entry, but first add some data
# required by hitgroup content table
hg['occurrence_date'] = datetime.datetime.now()
hg['first_crawl_id'] = crawl_id
if not hg['group_id_hashed']:
# if group_id is hashed, we cannot fetch details because we
# don't know what the real hash is
hg.update(hits_group_info(hg['group_id']))
else:
hg['html'] = ''
hit_group_content_id = db.insert_hit_group_content(hg)
log.debug('new hit group content: %s;;%s', hit_group_content_id,
hg['group_id'])
hg['hit_group_content_id'] = hit_group_content_id
hg['crawl_id'] = crawl_id
db.insert_hit_group_status(hg)
conn.commit()
except Exception:
log.exception('process_group fail - rollback')
conn.rollback()
finally:
db.curr.close()
dbpool.putconn(conn, thread.get_ident())
processed_groups.add(hg['group_id'])
return True
def inject_greenlet_id(record):
"""
Sets the greenlet ID on `record.extra`.
:param record: logbook record
"""
record.extra[GREENLET_ID_KEY] = get_ident()
def thread(self):
"""The ident of the thread. This is evaluated late and means that
if the log record is passed to another thread, :meth:`pull_information`
was called in the old thread.
"""
return thread.get_ident()
def get(self,
strategy,
persist=None,
wait=True,
request_ident=None,
**proxy_params):
if not callable(strategy):
if isinstance(strategy, str):
strategy, *strategy_params = strategy.split(':')
strategy = getattr(self, GET_STRATEGY[strategy].value)
if strategy_params:
strategy_ = strategy
def strategy(proxies):
return strategy_(proxies, *strategy_params)
elif isinstance(strategy, enum.Enum):
strategy = getattr(self, strategy.value)
if not len(self.active_proxies) and not self.fetcher:
raise InsufficientProxies('No proxies and no fetcher {}'.format(
self._stats_str))
self.maybe_update()
ident = get_ident() # unique integer id for greenlet
while True:
ready_proxies = self.get_ready_proxies(**proxy_params)
if ready_proxies:
break
elif not wait or ((not self.fetcher or self.fetcher.ready)
and not self.in_use):
# fetcher.ready also returns false on checker processing
raise InsufficientProxies('No ready proxies {} {}{}'.format(
proxy_params, request_ident and request_ident + ' ' or '',
self._stats_str))
else:
# logger.info('Wait proxy (thread %s) %s', ident, self._stats_str)
self.proxy_ready.clear()
if ident not in self.waiting:
# Storing extra data for superproxy monitoring
self.waiting[ident] = dict(since=datetime.utcnow(),
request_ident=request_ident,
params=proxy_params)
delta = 0
elif wait is not True:
delta = (datetime.utcnow() -
self.waiting[ident]['since']).total_seconds()
if delta >= wait:
del self.waiting[ident]
raise InsufficientProxies(
'Ready proxies wait timeout({}) {} {}{}'.format(
wait, proxy_params,
request_ident and request_ident + ' ' or '',
self._stats_str))
try:
self.proxy_ready.wait(None if wait is True else wait -
delta)
except Timeout:
continue
except BaseException:
del self.waiting[ident]
raise
if ident in self.waiting:
del self.waiting[ident]
if persist:
proxy = ready_proxies.get(persist, None)
if proxy:
proxy.in_use += 1
return proxy
proxy = strategy(ready_proxies)
if proxy:
proxy.in_use += 1
return proxy
raise InsufficientProxies(
'No proxies from {} ready with {} strategy {}{}'.format(
len(ready_proxies), strategy,
request_ident and request_ident + ' ' or '', self._stats_str))
def process_group(hg, crawl_id, requesters, processed_groups):
"""Gevent worker that should process single hitgroup.
This should write some data into database and do not return any important
data.
"""
hg['keywords'] = ', '.join(hg['keywords'])
# for those hit goups that does not contain hash group, create one and
# setup apropiate flag
hg['group_id_hashed'] = not bool(hg.get('group_id', None))
hg['qualifications'] = ', '.join(hg['qualifications'])
if hg['group_id_hashed']:
composition = ';'.join(map(str, (
hg['title'], hg['requester_id'], hg['time_alloted'],
hg['reward'], hg['description'], hg['keywords'],
hg['qualifications']))) + ';'
hg['group_id'] = hashlib.md5(composition).hexdigest()
log.debug('group_id not found, creating hash: %s %s',
hg['group_id'], composition)
if hg['group_id'] in processed_groups:
# this higroup was already processed
log.info('duplicated group: %s;;%s', crawl_id, hg['group_id'])
return False
conn = dbpool.getconn(thread.get_ident())
db = DB(conn)
try:
hit_group_content_id = db.hit_group_content_id(hg['group_id'])
if hit_group_content_id is None:
# check if there's profile for current requester and if does
# exists with non-public status, then setup non public status for
# current hitsgroup content
profile = requesters.get(hg['requester_id'], None)
if profile and profile.is_public is False:
hg['is_public'] = False
else:
hg['is_public'] = True
# fresh hitgroup - create group content entry, but first add some data
# required by hitgroup content table
hg['occurrence_date'] = datetime.datetime.now()
hg['first_crawl_id'] = crawl_id
if not hg['group_id_hashed']:
# if group_id is hashed, we cannot fetch details because we
# don't know what the real hash is
hg.update(hits_group_info(hg['group_id']))
else:
hg['html'] = ''
hit_group_content_id = db.insert_hit_group_content(hg)
log.debug('new hit group content: %s;;%s',
hit_group_content_id, hg['group_id'])
hg['hit_group_content_id'] = hit_group_content_id
hg['crawl_id'] = crawl_id
db.insert_hit_group_status(hg)
conn.commit()
except Exception:
log.exception('process_group fail - rollback')
conn.rollback()
finally:
db.curr.close()
dbpool.putconn(conn, thread.get_ident())
processed_groups.add(hg['group_id'])
return True
