def hilton_to_database(tid, used_times, source, keyword, extra, spider_tag, need_cache=True):
task = Task()
task.content = keyword
task.extra = extra
spider = factory.get_spider_by_old_source(spider_tag)
spider.task = task
if need_cache:
error_code = spider.crawl(required=['suggest'], cache_config=cache_config)
else:
error_code = spider.crawl(required=['suggest'], cache_config=none_cache_config)
logger.info(
str(len(spider.result['suggest'])) + ' -- ' + keyword)
return error_code, spider.result['suggest']
def poidetail_to_database(tid, used_times, source, url, need_cache=True):
task = Task()
task.content = url
task.ticket_info = {
'tid': tid,
'used_times': used_times
}
print (source + '_detail')
spider = factory.get_spider_by_old_source(source+'_detail')
spider.task = task
if need_cache:
error_code = spider.crawl(required=['POIdetail'], cache_config=cache_config)
else:
error_code = spider.crawl(required=['POIdetail'], cache_config=none_cache_config)
print(error_code)
logger.info(str(spider.result['POIdetail']) + ' -- ' + task.content)
return error_code, spider.result['POIdetail'], spider.page_store_key_list
def hilton_to_database(tid, used_times, source, source_id, city_id, check_in, need_cache=True):
task = Task()
task.content = 'NULL&' + str(city_id) + '&' + str(source_id) + '&' + '2&{0}'.format(check_in)
task.ticket_info = {
'tid': tid,
'used_times': used_times,
'room_info': [{"occ": 2, "num": 1}]
}
spider = factory.get_spider_by_old_source('hiltonHotel2')
spider.task = task
if need_cache:
error_code = spider.crawl(required=['list', 'room'], cache_config=cache_config)
else:
error_code = spider.crawl(required=['list', 'room'], cache_config=none_cache_config)
print(error_code)
logger.info(str(spider.result['room']) + ' -- ' + task.content)
return error_code, spider.result['room'], spider.page_store_key_list
def qyer_list_to_database(tid,
used_times,
source,
city_id,
check_in,
city_url,
need_cache=True):
task = Task()
task.content = city_url
task.ticket_info = {'tid': tid, 'used_times': used_times}
spider = factory.get_spider_by_old_source('qyerList')
spider.task = task
if need_cache:
error_code = spider.crawl(required=['list'], cache_config=cache_config)
else:
error_code = spider.crawl(required=['list'],
cache_config=none_cache_config)
print(error_code)
logger.info(str(spider.result['list']) + ' -- ' + task.content)
return error_code, spider.result[
'list'], spider.page_store_key_list, spider.types_result_num
def hotel_rest_list_task(self, source, url, city_id, **kwargs):
try:
self.task_source = source.title()
self.task_type = 'DaodaoListInfo'
logger.info("任务进行中。。。")
code, result = hotel_list_database(source, url)
logger.info("code : %s" % str(code))
if int(code) != 0:
logger.info("=======================0=========================\n")
logger.info(str(code) + ' | ' + str(result))
logger.info("\n=======================1=========================")
raise Exception
self.error_code = str(code)
for one in result:
for key, view in one.items():
rest = HotelRestList()
rest.source = source
rest.source_id = int(view['source_id'])
rest.city_id = int(city_id)
rest.url = view['view_url']
rest.name = view['view_name'].strip('\n').strip()
try:
ss = DBSession_mb4()
ss.merge(rest)
ss.commit()
except Exception as e:
logger.info(
"======================= sql 异常========================="
)
logger.exception(traceback.format_exc(e))
return True
except Exception as e:
logger.exception('================== 异常 0==================')
logger.exception(source + ' | ' + str(city_id) + ' | ' + url)
logger.exception(traceback.format_exc(e))
logger.exception('================== 异常 1==================')
raise Exception(e)
def _execute(self, **kwargs):
source = self.task.kwargs['source']
city_id = self.task.kwargs['city_id']
country_id = self.task.kwargs['country_id']
fla = self.task.kwargs.get('list_more', False)
@func_time_logger
def hotel_list_crawl():
error_code, result, page_store_key = hotel_list_database(
tid=self.task.task_id,
used_times=self.task.used_times,
source=source,
city_id=city_id,
check_in=self.task.kwargs['check_in'],
is_new_type=self.task.kwargs.get('is_new_type', False),
suggest_type=self.task.kwargs.get('suggest_type', '1'),
suggest=self.task.kwargs.get('suggest', ''),
need_cache=self.task.used_times == 0,
flag=fla)
return error_code, result, page_store_key
error_code, result, page_store_key = hotel_list_crawl()
print(result)
# more_list
if fla:
for line in result['filter']:
line['country_id'] = country_id
line['source'] = source
filter_collections.insert_many(result['filter'])
if len(result['filter']) > 0:
self.task.error_code = 0
elif int(error_code) == 0:
raise ServiceStandardError(ServiceStandardError.EMPTY_TICKET)
else:
raise ServiceStandardError(error_code=error_code)
return result, error_code, self.task.error_code, self.task.task_name, self.task.kwargs[
'suggest']
if source == 'starwood' and error_code == 29:
self.task.error_code = 109
error_code = 109
else:
self.task.error_code = error_code
res_data = []
if source in ('ctrip', 'ctripcn', 'starwood', 'gha'):
for line in result['hotel']:
sid = line[3]
hotel_url = line[-1]
res_data.append((source, sid, city_id, country_id, hotel_url))
elif source in ('bestwest'):
for sr, sid, city_id, hotel_url in result['hotel']:
res_data.append((source, sid, city_id, country_id, hotel_url))
elif source in ('fourseasons'):
for line in result['hotel']:
sid = line[-1]
hotel_url = line[0]
res_data.append((source, sid, city_id, country_id, hotel_url))
elif source in ('hyatt'):
for line in result['hotel']:
sid = line[-1]
hotel_url = line[1]
res_data.append((source, sid, city_id, country_id, hotel_url))
elif source == 'hilton':
for dict_obj in result['hotel']:
line = dict_obj.values()
res_data.append(
(source, line[2], city_id, country_id, line[0]))
else:
for sid, hotel_url in result['hotel']:
res_data.append((source, sid, city_id, country_id, hotel_url))
@func_time_logger
def hotel_list_insert_db():
try:
service_platform_conn = service_platform_pool.connection()
cursor = service_platform_conn.cursor()
sql = "INSERT IGNORE INTO {} (source, source_id, city_id, country_id, hotel_url) VALUES (%s,%s,%s,%s,%s)".format(
self.task.task_name)
_res = cursor.executemany(sql, res_data)
service_platform_conn.commit()
cursor.close()
service_platform_conn.close()
self.task.list_task_insert_db_count = _res
self.task.get_data_per_times = len(res_data)
except Exception as e:
self.logger.exception(msg="[mysql error]", exc_info=e)
raise ServiceStandardError(
error_code=ServiceStandardError.MYSQL_ERROR,
wrapped_exception=e)
hotel_list_insert_db()
try:
data_collections = mongo_data_client['ServicePlatform'][
self.task.task_name]
data_collections.create_index([('source', 1), ('source_id', 1),
('city_id', 1)],
unique=True,
background=True)
data = []
if data:
for line in res_data:
data.append({
'list_task_token': self.task.list_task_token,
'task_id': self.task.task_id,
'source': line[0],
'source_id': line[1],
'city_id': line[2],
'country_id': line[3],
'hotel_url': line[4]
})
data_collections.insert(data, continue_on_error=True)
except pymongo.errors.DuplicateKeyError:
logger.info("[Duplicate Key]")
except Exception as exc:
raise ServiceStandardError(
error_code=ServiceStandardError.MONGO_ERROR,
wrapped_exception=exc)
# 由于错误都是 raise 的，
# 所以当出现此种情况是，return 的内容均为正确内容
# 对于抓取平台来讲，当出现此中情况时，数据均应该入库
# 用 res_data 判断，修改 self.error_code 的值
if len(res_data) > 0:
self.task.error_code = 0
elif int(error_code) == 0:
raise ServiceStandardError(ServiceStandardError.EMPTY_TICKET)
else:
raise ServiceStandardError(error_code=error_code)
return len(
res_data), error_code, self.task.error_code, self.task.task_name
