def first():
session = Session()
if session.query(Url).filter_by(type=URL_TYPE.BEST_SELL_CATEGORY,
has_crawled=False).count() < 1:
b = BestSell() # string_proxy='socks4://192.168.1.188:1080')
# b.product('B01FQN3W5U')
b.categories_first(categories=config.spider_categories)
# b.categories(url_obj=url_obj)
session.close()
def urlopt(queue):
session = Session(autocommit=True)
while 1:
try:
opt, urls = queue.get(timeout=30)
if opt == 'add':
url_objs = []
for url in urls:
if session.query(Url).filter_by(key=url.key).count() < 1:
logger.info('Opt: Add %s' % url.url)
url_objs.append(url)
session.bulk_save_objects(url_objs)
elif opt == 'error':
for url in urls:
logger.info('Opt: Error %s' % url.url)
session.query(Url).filter_by(id=url.id).update({Url.tries: url.tries+1})
elif opt == 'over':
for url in urls:
logger.info('Opt: Over %s' % url.url)
session.query(Url).filter_by(id=url.id).update({Url.has_crawled: True})
# session.commit()
except Empty:
logger.info('Empty')
except:
logger.error(traceback.format_exc())
def category():
session = Session()
while session.query(Url).filter_by(type=URL_TYPE.BEST_SELL_CATEGORY,
has_crawled=False).count() > 0:
for i, url_obj in enumerate(
session.query(Url).filter_by(
type=URL_TYPE.BEST_SELL_CATEGORY,
has_crawled=False)): # .limit(20*len(string_proxies))):
executor.submit(bs[i % len(bs)].categories, url_obj=url_obj)
time.sleep(0.1)
time.sleep(30)
def product():
try:
session = Session()
while session.query(Url).filter_by(type=URL_TYPE.PRODUCT_URL,
has_crawled=False).count() > 0:
for i, url_obj in enumerate(
session.query(Url).filter_by(
type=URL_TYPE.PRODUCT_URL,
has_crawled=False)): #.limit(20*len(string_proxies))):
executor.submit(bs[i % len(bs)].product, url_obj=url_obj)
time.sleep(0.1)
time.sleep(30)
except:
logger.error(traceback.format_exc())
def category_next():
try:
session = Session()
while session.query(Url).filter_by(
type=URL_TYPE.BEST_SELL_CATEGORY_NEXT,
has_crawled=False).count() > 0:
for i, url_obj in enumerate(
session.query(Url).filter_by(
type=URL_TYPE.BEST_SELL_CATEGORY_NEXT,
has_crawled=False)): #.limit(20*len(string_proxies))):
executor.submit(bs[i % len(bs)].category_next, url_obj=url_obj)
time.sleep(0.1)
time.sleep(30)
except:
logger.error(traceback.format_exc())
def do():
input = Queue(maxsize=2000)
output = Queue(maxsize=2000)
bs = []
for i in range(num):
string_proxy = string_proxies[i%len(string_proxies)]
b = BestSell(string_proxy=string_proxy, input=input, output=output, domain='amazon.es', cache_root='I:\\cache_amazon_es')
bs.append(b)
b.start()
t = Thread(target=urlopt, args=[output, ])
t.start()
session = Session()
if session.query(Url).count() < 1:
bs[0].categories_first(categories=spider_categories)
session.commit()
while session.query(Url).filter(and_(Url.has_crawled==False, Url.tries<3)).count() > 0:
for url_obj in session.query(Url).filter(and_(Url.has_crawled==False, Url.tries<3)).limit(1000*len(string_proxies)):
try:
# m = Map(url_obj.as_dict())
logger.info('put: %s' % url_obj.url)
input.put(url_obj.as_dict())
except:
logger.error(traceback.format_exc())
logger.info('wait 5 seconds...')
time.sleep(5)
logger.info('oooooooooooooooooooooooooooooooover')