def process_item(self, item, spider):
"""保存数据"""
demo = item.get('demo')
if not demo:
raise DropItem("item data type error")
# self.put_queue(item)
data = copy.deepcopy(dict(item))
if not data:
raise DropItem("item data is empty")
# info = self.mongo.find_one({'demo': data['demo']})
demo_test = item.get('demo_test', '')
if not demo_test:
raise DropItem("demo_test is empty")
# return
condition = {'demo': demo}
try:
info = self.mysql.select(demo_test, condition=condition, limit=1)
if not info:
# self.mongo.insert(data)
item['create_time'] = util.date()
item['update_time'] = util.date()
self.mysql.insert(demo_test, data=item)
# _logger.info('success insert mysql : %s' % data['demo'])
else:
item['create_time'] = info['create_time']
item['update_time'] = util.date()
# self.mongo.update({'_id': info['_id']}, {"$set": data})
self.mysql.update(demo_test, condition=condition, data=item)
# _logger.info('success update mysql : %s' % data['demo'])
except Exception as e:
_logger.info('error op mysql : {0} : e {1}'.format(
data['demo'], e))
raise DropItem('success process')
def save_data(url, db_name, item):
'''
数据保存
'''
info = None
if not info:
item['create_time'] = util.date()
mysql.insert(db_name, data=item)
_logger.info('INFO: DB:%s 数据保存成功, 期号%s ; URL:%s' % (db_name, item['demo'], url))
else:
item['update_time'] = util.date()
del item['open_time']
del item['create_time']
mysql.update(db_name, condition=[('demo', '=', item['demo'])], data=item)
_logger.info('INFO: DB:%s 数据已存在 更新成功, 期号: %s ; URL:%s' % (db_name, item['demo'], url))
def write_update_info(self, num_list):
'''记录更新信息
@param num_list 记录每次更新数目信息
@param name 记录类型值，默认count为成功值
'''
if not num_list:
return None
mq.put('crawler_update_stats', {'data': num_list, 'time': util.date()})
def _init_args(self, **kwargs):
start_url = kwargs.get('START_URL', '')
self.abbreviation = kwargs.get('ABBREVIATION', '')
self.start_date = kwargs.get('START_DATE', '')
self.end_date = kwargs.get('END_DATE', '')
self.end_date = self.end_date if self.end_date else util.date()
if start_url:
self.start_urls = [start_url]
self.rules = (Rule(LinkExtractor(allow=filter_rules),
callback='parse_resp',
follow=True), )
def exception_notice(etype=''):
"""异常通知"""
now_minuter = util.date(format='%Y-%m-%d %H:%M')
subject = '【HQChip】合作库存 %s 数据更新异常通知 %s' % (PN2, now_minuter)
if etype == 'mysql':
except_msg = 'mysql数据库连接异常'
elif etype == 'mongo':
except_msg = 'mongodb 数据库连接异常'
else:
except_msg = '数据获取异常'
body = "合作库存 %s 数据更新数据获取异常, 异常原因：%s,请注意检查！" % (PN2, except_msg)
util.sendmail(config.EMAIL_NOTICE.get(
'accept_list'), subject=subject, body=body)
def update_data(self, queue_name=None):
"""更新指定队列数据"""
if not queue_name:
return 0
qsize = mq.qsize(queue_name)
self.limit = self.limit if qsize > self.limit else qsize # 每次更新的数量
queue_list = []
for i in range(self.limit):
queue_data = mq.get(queue_name)
if queue_data and queue_data not in queue_list:
queue_list.append(queue_data)
if not queue_list:
print('等待中，队列 %s 为空' % queue_name)
return 0
proxy = None
if not self.no_proxy:
proxy = self.get_prolist()
tlist = []
data_list = []
total_num = 0
for data in queue_list:
# 无效队列数据
if 'id' not in data:
continue
if 'proxy' in data:
del data['proxy']
try:
if len(tlist) > 30:
for t in tlist:
t.join(45)
except (KeyboardInterrupt, SystemExit):
mq.put(queue_name, queue_data)
return 0
# 有效队列的总数（非型号总数）
total_num += 1
t = threading.Thread(target=self.fetch_update_data,
args=(data_list, proxy), kwargs=data)
tlist.append(t)
t.start()
time.sleep(1)
del data, queue_list
valid_num = 0
delete_list = []
# 所有线程执行完毕后 再进行数据处理
for data in data_list:
if not data:
continue
if data['status'] == 200:
mq.put(config.WAIT_UPDATE_QUEUE, data['dlist']) # 等待提交数据
valid_num += 1
id = data.get('dlist').get('id', )
lottery_name = data.get('dlist').get('lottery_name', )
status = data.get('status')
config.LOG.info('ID：{0} ;产品: {1} ;数据获取成功：{2} ;提交到入库队列: {3} !'.format(id, lottery_name, status,
config.WAIT_UPDATE_QUEUE))
continue
else:
delete_list.append(data)
count = data.get('count', '')
if count and count < self.exception_threshold: # 重复更新的次数
config.LOG.info('ID：%s，更新状态：%s, 重新入队中!' % (data.get('id', ), data['status']))
# update_list.append(data)
mq.put(queue_name, data)
else:
config.LOG.error('ID：%s，更新状态：%s, 重试次数超过阀值,保存日志中!' % (data.get('id', ), data['status']))
if 'count' in data:
del data['count']
if 'time' not in data:
data['time'] = util.date()
# db.mongo['update_exception_logs'].insert(data)
mq.put('update_exception_logs', data)
self.write_update_info(valid_num)
print('队列 %s 本次共有 %s 条数据更新成功，成功率：%s %%' %
(queue_name, valid_num, valid_num * 1.0 / total_num * 100 if total_num > 0 else 0))
print('完成 , 等待下一个队列!')