def __init__(self):
monkey.patch_all()
self.queue = queue.Queue()
self.pool = pool.Pool(
int(CoreConfigure().get_config_section_map("spider")
['concurrency']))
self.url_table = UrlTable()
self.timer = Timer(
int(CoreConfigure().get_config_section_map("spider")['timeout']),
self.stop)
self._stop = event.Event()
self.greenlet_finished = event.Event()
self.root = None # url_object
self.initialize_db()
def __init__(self):
monkey.patch_all()
self.queue = queue.Queue()
self.pool = pool.Pool(int(CoreConfigure().get_config_section_map("spider")['concurrency']))
self.url_table = UrlTable()
self.timer = Timer(int(CoreConfigure().get_config_section_map("spider")['timeout']), self.stop)
self._stop = event.Event()
self.greenlet_finished = event.Event()
self.root = None # url_object
self.initialize_db()
class Spider(object):
def __init__(self):
monkey.patch_all()
self.queue = queue.Queue()
self.pool = pool.Pool(int(CoreConfigure().get_config_section_map("spider")['concurrency']))
self.url_table = UrlTable()
self.timer = Timer(int(CoreConfigure().get_config_section_map("spider")['timeout']), self.stop)
self._stop = event.Event()
self.greenlet_finished = event.Event()
self.root = None # url_object
self.initialize_db()
def initialize_db(self):
host = CoreConfigure().get_config_section_map('db')['host']
port = CoreConfigure().get_config_section_map('db')['port']
db_name = CoreConfigure().get_config_section_map('db')['database']
collect_name = CoreConfigure().get_config_section_map('db')['collection']
self.db_cli = MongoClient(host, int(port))
# mongodb collection
self.collection = self.db_cli[db_name][collect_name]
def set_root(self, url):
if isinstance(url, basestring):
url = UrlObj(url, type = 0)
self.root = url
self.push_task(self.root)
self.url_table.insert(self.root)
def push_task(self, url):
# type(url) is UrlObj
if url not in self.url_table:
self.queue.put(url)
def run(self, url = None):
begin = time.time()
if url is None:
# read from configure file for default value
url = CoreConfigure().get_config_section_map('content')['root_url']
self.set_root(url)
self.timer.start()
logger.info("spider begin crawl")
while not self.stopped() and self.timer.isAlive():
for greenlet in list(self.pool):
if greenlet.dead:
self.pool.discard(greenlet)
try:
url = self.queue.get_nowait()
except queue.Empty:
if self.pool.free_count() != self.pool.size:
# wait until one greenlet finish to flash queue
self.greenlet_finished.wait()
self.greenlet_finished.clear()
continue
else:
self.stop()
greenlet = Handler(url, self)
self.pool.start(greenlet)
logger.info("total time elapsed %0.2f" % (time.time() - begin))
def stopped(self):
return self._stop.is_set()
def stop(self):
logger.info("spider finish, totally catched (%d) urls" % len(self.url_table))
self.timer.cancel()
self._stop.set()
self.pool.join()
self.queue.put(StopIteration)
class Spider(object):
def __init__(self):
monkey.patch_all()
self.queue = queue.Queue()
self.pool = pool.Pool(
int(CoreConfigure().get_config_section_map("spider")
['concurrency']))
self.url_table = UrlTable()
self.timer = Timer(
int(CoreConfigure().get_config_section_map("spider")['timeout']),
self.stop)
self._stop = event.Event()
self.greenlet_finished = event.Event()
self.root = None # url_object
self.initialize_db()
def initialize_db(self):
host = CoreConfigure().get_config_section_map('db')['host']
port = CoreConfigure().get_config_section_map('db')['port']
db_name = CoreConfigure().get_config_section_map('db')['database']
collect_name = CoreConfigure().get_config_section_map(
'db')['collection']
self.db_cli = MongoClient(host, int(port))
# mongodb collection
self.collection = self.db_cli[db_name][collect_name]
def set_root(self, url):
if isinstance(url, basestring):
url = UrlObj(url, type=0)
self.root = url
self.push_task(self.root)
self.url_table.insert(self.root)
def push_task(self, url):
# type(url) is UrlObj
if url not in self.url_table:
self.queue.put(url)
def run(self, url=None):
begin = time.time()
if url is None:
# read from configure file for default value
url = CoreConfigure().get_config_section_map('content')['root_url']
self.set_root(url)
self.timer.start()
logger.info("spider begin crawl")
while not self.stopped() and self.timer.isAlive():
for greenlet in list(self.pool):
if greenlet.dead:
self.pool.discard(greenlet)
try:
url = self.queue.get_nowait()
except queue.Empty:
if self.pool.free_count() != self.pool.size:
# wait until one greenlet finish to flash queue
self.greenlet_finished.wait()
self.greenlet_finished.clear()
continue
else:
self.stop()
greenlet = Handler(url, self)
self.pool.start(greenlet)
logger.info("total time elapsed %0.2f" % (time.time() - begin))
def stopped(self):
return self._stop.is_set()
def stop(self):
logger.info("spider finish, totally catched (%d) urls" %
len(self.url_table))
self.timer.cancel()
self._stop.set()
self.pool.join()
self.queue.put(StopIteration)
