class Spider(object):
def __init__(self,seed,depth,pool_size=10):
self.seed = seed
self.depth = depth
self.all_url_list = [seed]
self.finished_url_list = []
self.failure_url_list = []
self.pool = ThreadPool(pool_size)
def crawl(self):
base_deep_size = 0
while base_deep_size <= self.depth:
for url in self.all_url_list:
if url not in self.finished_url_list:
self.pool.add_task(self.download,url)
self.pool.close()
self.depth-=1
def download(self,url):
try:
data = urllib2.urlopen(url)
page = data.read()
self.finished_url_list.append(url)
links = self.get_urls(page)
return page,links
except Exception as e:
print 'open url:%s raise exception(%s)'%(url,e)
return None
def get_urls(self,page):
soup = BeautifulSoup(page,fromEncoding="gb18030")
if soup.title:
print soup.title.string
links = []
for item in soup.findAll('a'):
link=item.get('href')
if link and link.startswith('http://') and link not in self.finished_url_list:
links.append(link)
print links
return links
def get_next_url(self):
pass
def worker_routine(context, idx):
""" Worker routine """
global worker_exe_path
# Socket to talk to dispatcher
socket0 = context.socket(zmq.REP)
socket0.connect(url_worker)
worker_id = 'Worker-%s' % idx
socket0.setsockopt(zmq.IDENTITY, worker_id)
socket_exit = context.socket(zmq.SUB)
socket_exit.connect(url_worker_cmd)
socket_exit.setsockopt(zmq.SUBSCRIBE, '')
poller = zmq.Poller()
poller.register(socket0, zmq.POLLIN)
poller.register(socket_exit, zmq.POLLIN)
from threadpool import ThreadPool, WorkRequest
thread_pool = ThreadPool(thread_num_of_worker)
def thread_work(client_exe_path, msg):
# create a process to execute the client_exe
# TODO: add execute timeout
_cmd = '%s %s' % (client_exe_path, msg)
cmd = WrapCommand(_cmd)
cmd.start()
logger.debug('Start command start %s' % _cmd)
cmd.join()
logger.debug('Start command end %s' % _cmd)
return cmd
def thread_work_cb(request, ret):
if ret.returncode == 0:
# run cmd process in thread success
logger.debug("execute %s success %s" %
(request.args[0], ret.returncode))
else:
logger.error(
"execute %s failed %s \n%s\n" %
(request.args[0], ret.returncode, ' '.join(ret.results)))
def handle_exception(request, exc_info):
if not isinstance(exc_info, tuple):
# Something is seriously wrong...
logger.debug(request)
logger.debug(exc_info)
raise SystemExit
logger.debug("**** Exception occured in request #%s: %s" % \
(request.requestID, exc_info))
def register_task(client_exe_path, msg):
request = WorkRequest(thread_work, (client_exe_path, msg), {},
callback=thread_work_cb,
exc_callback=handle_exception)
thread_pool.putRequest(request)
def get_client_exe_paths(client_name):
client_exe_paths = []
for client_id in CONF.clients.clients_list:
client_conf = getattr(CONF, client_id)
client_n = client_conf.name
if client_n == client_name:
client_exe_paths.append(client_conf.exe_path)
return client_exe_paths
while True:
events = dict(poller.poll(2000))
if events.get(socket0) == zmq.POLLIN:
# Deal with message
_msg = socket0.recv()
print("Received request: [%s]\n" % (_msg))
client_name = _msg.split()[0]
msg = _msg[len(client_name):].strip()
exe_paths = get_client_exe_paths(client_name)
logger.debug("Client name \"%s\" message: %s" % (client_name, msg))
logger.debug("Client exe path %s" % exe_paths)
if exe_paths is not None:
for exe_path in exe_paths:
register_task(exe_path, msg)
#send reply back to client
socket0.send("OK")
else:
logger.debug(
"Can not execute client, because there have none clients in config file(%s), please check"
% str(CONF.clients.clients_list))
socket0.send('FAILED')
if events.get(socket_exit) == zmq.POLLIN:
cmd = socket_exit.recv()
logger.debug('%s CMD %s' % (worker_id, cmd))
sys.stdout.flush()
if cmd == 'EXIT':
break
elif cmd == 'CONFIG_CHANGED':
reload_config()
else:
pass
thread_pool.close()
thread_pool.join()
socket0.close()
socket_exit.close()
def worker_routine(context, idx):
""" Worker routine """
global worker_exe_path
# Socket to talk to dispatcher
socket0 = context.socket(zmq.REP)
socket0.connect(url_worker)
worker_id = 'Worker-%s' % idx
socket0.setsockopt(zmq.IDENTITY, worker_id)
socket_exit = context.socket(zmq.SUB)
socket_exit.connect(url_worker_cmd)
socket_exit.setsockopt(zmq.SUBSCRIBE, '')
poller = zmq.Poller()
poller.register(socket0, zmq.POLLIN)
poller.register(socket_exit, zmq.POLLIN)
from threadpool import ThreadPool, WorkRequest
thread_pool = ThreadPool(thread_num_of_worker)
def thread_work(client_exe_path, msg):
# create a process to execute the client_exe
# TODO: add execute timeout
_cmd = '%s %s' % (client_exe_path, msg)
cmd = WrapCommand(_cmd)
cmd.start()
logger.debug('Start command start %s' % _cmd)
cmd.join()
logger.debug('Start command end %s' % _cmd)
return cmd
def thread_work_cb(request, ret):
if ret.returncode == 0:
# run cmd process in thread success
logger.debug("execute %s success %s" % (request.args[0], ret.returncode))
else:
logger.error("execute %s failed %s \n%s\n" % (request.args[0], ret.returncode, ' '.join(ret.results)))
def handle_exception(request, exc_info):
if not isinstance(exc_info, tuple):
# Something is seriously wrong...
logger.debug(request)
logger.debug(exc_info)
raise SystemExit
logger.debug("**** Exception occured in request #%s: %s" % \
(request.requestID, exc_info))
def register_task(client_exe_path, msg):
request = WorkRequest(thread_work, (client_exe_path, msg), {}, callback=thread_work_cb,
exc_callback=handle_exception)
thread_pool.putRequest(request)
def get_client_exe_paths(client_name):
client_exe_paths = []
for client_id in CONF.clients.clients_list:
client_conf = getattr(CONF, client_id)
client_n = client_conf.name
if client_n == client_name:
client_exe_paths.append(client_conf.exe_path)
return client_exe_paths
while True:
events = dict(poller.poll(2000))
if events.get(socket0) == zmq.POLLIN:
# Deal with message
_msg = socket0.recv()
print("Received request: [%s]\n" % (_msg))
client_name = _msg.split()[0]
msg = _msg[len(client_name):].strip()
exe_paths = get_client_exe_paths(client_name)
logger.debug("Client name \"%s\" message: %s" %(client_name, msg))
logger.debug("Client exe path %s" % exe_paths)
if exe_paths is not None:
for exe_path in exe_paths:
register_task(exe_path, msg)
#send reply back to client
socket0.send("OK")
else:
logger.debug("Can not execute client, because there have none clients in config file(%s), please check" % str(CONF.clients.clients_list))
socket0.send('FAILED')
if events.get(socket_exit) == zmq.POLLIN:
cmd = socket_exit.recv()
logger.debug('%s CMD %s' % (worker_id, cmd))
sys.stdout.flush()
if cmd == 'EXIT':
break
elif cmd == 'CONFIG_CHANGED':
reload_config()
else:
pass
thread_pool.close()
thread_pool.join()
socket0.close()
socket_exit.close()
