def deal_with_data(self, *params):
'''
处理并存储相关拼团商品的数据
:param params: 待传参数
:return:
'''
goods_list = params[0]
mogujie = MoGuJieParse()
my_pipeline = SqlServerMyPageInfoSaveItemPipeline()
if my_pipeline.is_connect_success:
db_goods_id_list = [item[0] for item in list(my_pipeline._select_table(sql_str=mg_select_str_1))]
print(db_goods_id_list)
for item in goods_list:
if item.get('goods_id', '') in db_goods_id_list:
print('该goods_id已经存在于数据库中, 此处跳过')
pass
else:
goods_id = str(item.get('goods_id', ''))
tmp_url = 'https://shop.mogujie.com/detail/' + str(goods_id)
mogujie.get_goods_data(goods_id=str(goods_id))
goods_data = mogujie.deal_with_data()
if goods_data == {}: # 返回的data为空则跳过
pass
else: # 否则就解析并且插入
# 规范化
goods_data['price_info_list'] = _get_mogujie_pintuan_price_info_list(goods_data['price_info_list'])
goods_data['goods_url'] = tmp_url
goods_data['goods_id'] = str(goods_id)
goods_data['pintuan_time'] = item.get('pintuan_time', {})
goods_data['pintuan_begin_time'], goods_data['pintuan_end_time'] = get_miaosha_begin_time_and_miaosha_end_time(miaosha_time=item.get('pintuan_time', {}))
goods_data['all_sell_count'] = item.get('all_sell_count', '')
goods_data['fcid'] = str(item.get('fcid'))
goods_data['page'] = str(item.get('page'))
goods_data['sort'] = str(item.get('sort', ''))
# pprint(goods_data)
# print(goods_data)
_r = mogujie.insert_into_mogujie_pintuan_table(data=goods_data, pipeline=my_pipeline)
if _r: # 更新
db_goods_id_list.append(goods_id)
db_goods_id_list = list(set(db_goods_id_list))
sleep(MOGUJIE_SLEEP_TIME) # 放慢速度
else:
print('数据库连接失败，此处跳过!')
pass
try:
del mogujie
except:
pass
gc.collect()
def run_forever(self):
'''
实时更新数据
:return:
'''
tmp_sql_server = SqlServerMyPageInfoSaveItemPipeline()
try:
result = list(tmp_sql_server.select_mogujie_pintuan_all_goods_id())
except TypeError:
print('TypeError错误, 原因数据库连接失败...(可能维护中)')
result = None
if result is None:
pass
else:
print('------>>> 下面是数据库返回的所有符合条件的goods_id <<<------')
print(result)
print('--------------------------------------------------------')
print('即将开始实时更新数据, 请耐心等待...'.center(100, '#'))
index = 1
self.my_phantomjs = MyPhantomjs()
for item in result: # 实时更新数据
pintuan_end_time = json.loads(item[1]).get('end_time')
pintuan_end_time = int(
str(
time.mktime(
time.strptime(pintuan_end_time,
'%Y-%m-%d %H:%M:%S')))[0:10])
# print(miaosha_end_time)
data = {}
mogujie_pintuan = MoGuJieParse()
if index % 8 == 0:
try:
del self.my_phantomjs
except:
pass
gc.collect()
self.my_phantomjs = MyPhantomjs()
if index % 50 == 0: # 每50次重连一次，避免单次长连无响应报错
print('正在重置，并与数据库建立新连接中...')
tmp_sql_server = SqlServerMyPageInfoSaveItemPipeline()
print('与数据库的新连接成功建立...')
if tmp_sql_server.is_connect_success:
if self.is_recent_time(pintuan_end_time) == 0:
tmp_sql_server.delete_mogujie_pintuan_expired_goods_id(
goods_id=item[0])
print(
'过期的goods_id为(%s)' % item[0],
', 拼团开始时间为(%s), 删除成功!' %
json.loads(item[1]).get('begin_time'))
elif self.is_recent_time(pintuan_end_time) == 2:
# break # 跳出循环
pass # 此处应该是pass,而不是break，因为数据库传回的goods_id不都是按照顺序的
else: # 返回1，表示在待更新区间内
print(
'------>>>| 正在更新的goods_id为(%s) | --------->>>@ 索引值为(%d)'
% (item[0], index))
data['goods_id'] = item[0]
tmp_url = 'http://list.mogujie.com/search?page={0}&fcid={1}&algoKey=pc_tuan_book_pop&cKey=pc-tuan'.format(
item[3], item[2])
# print(tmp_url)
# requests请求不到数据，涉及证书认证，直接用phantomjs
# body = MyRequests.get_url_body(url=tmp_url, headers=self.headers, had_referer=True)
body = self.my_phantomjs.use_phantomjs_to_get_url_body(
url=tmp_url)
# print(body)
if body == '':
print('获取到的body为空值! 此处跳过')
else:
try:
body = re.compile(
r'<pre.*?>(.*?)</pre>').findall(body)[0]
tmp_data = json.loads(body)
# pprint(tmp_data)
except:
print('json.loads转换body时出错, 请检查')
tmp_data = {}
if tmp_data.get('result',
{}).get('wall',
{}).get('docs', []) == []:
print('得到的docs为[]!')
print('该商品已被下架限时秒杀活动，此处将其删除')
tmp_sql_server.delete_mogujie_pintuan_expired_goods_id(
goods_id=item[0])
print('下架的goods_id为(%s)' % item[0], ', 删除成功!')
pass
else:
tmp_item_list = tmp_data.get('result', {}).get(
'wall', {}).get('docs', [])
# print(tmp_item_list)
# pprint(tmp_item_list)
begin_time_timestamp = int(
time.time()) # 开始拼团的时间戳
item_list = [{
'goods_id':
item.get('tradeItemId', ''),
'pintuan_time': {
'begin_time':
self.timestamp_to_regulartime(
timestamp=begin_time_timestamp),
'end_time':
self.timestamp_to_regulartime(
self.get_pintuan_end_time(
begin_time_timestamp,
item.get('leftTimeOrg', ''))),
},
'all_sell_count':
str(item.get('salesVolume', 0)),
} for item in tmp_item_list]
# print(item_list)
pintuan_goods_all_goods_id = [
item_1.get('goods_id', '')
for item_1 in item_list
]
# print(pintuan_goods_all_goods_id)
'''
内部已经下架的(内部下架的其实并未真实下架，还在卖的，所以我就更新其商品信息数据，不更新上下架时间)
'''
if item[0] not in pintuan_goods_all_goods_id:
# print('该商品已被下架限时秒杀活动，此处将其删除')
# tmp_sql_server.delete_mogujie_pintuan_expired_goods_id(goods_id=item[0])
# print('下架的goods_id为(%s)' % item[0], ', 删除成功!')
# pass
mogujie_pintuan.get_goods_data(
goods_id=item[0])
goods_data = mogujie_pintuan.deal_with_data(
)
if goods_data == {}:
pass
else:
# 规范化
print('+++ 内部下架，其实还在售卖的商品更新')
tmp_price_info_list = goods_data[
'price_info_list']
price_info_list = [{
'spec_value':
item_4.get('spec_value'),
'pintuan_price':
item_4.get('detail_price'),
'normal_price':
item_4.get('normal_price'),
'img_url':
item_4.get('img_url'),
'rest_number':
item_4.get('rest_number'),
} for item_4 in tmp_price_info_list]
goods_data['goods_id'] = item[0]
goods_data[
'price_info_list'] = price_info_list
# pprint(goods_data)
# print(goods_data)
mogujie_pintuan.update_mogujie_pintuan_table_2(
data=goods_data,
pipeline=tmp_sql_server)
sleep(MOGUJIE_SLEEP_TIME) # 放慢速度
else: # 未下架的
for item_2 in item_list:
if item_2.get('goods_id',
'') == item[0]:
mogujie_pintuan.get_goods_data(
goods_id=item[0])
goods_data = mogujie_pintuan.deal_with_data(
)
if goods_data == {}: pass
else:
# 规范化
tmp_price_info_list = goods_data[
'price_info_list']
price_info_list = [
{
'spec_value':
item_4.get(
'spec_value'),
'pintuan_price':
item_4.get(
'detail_price'),
'normal_price':
item_4.get(
'normal_price'),
'img_url':
item_4.get('img_url'),
'rest_number':
item_4.get(
'rest_number'),
} for item_4 in
tmp_price_info_list
]
goods_data['goods_id'] = item[
0]
goods_data[
'price_info_list'] = price_info_list
goods_data[
'pintuan_time'] = item_2.get(
'pintuan_time', {})
goods_data['pintuan_begin_time'], goods_data[
'pintuan_end_time'] = self.get_pintuan_begin_time_and_pintuan_end_time(
pintuan_time=goods_data[
'pintuan_time'])
goods_data[
'all_sell_count'] = item_2.get(
'all_sell_count', '')
# pprint(goods_data)
# print(goods_data)
mogujie_pintuan.update_mogujie_pintuan_table(
data=goods_data,
pipeline=tmp_sql_server)
sleep(
MOGUJIE_SLEEP_TIME) # 放慢速度
else:
pass
else:
print('数据库连接失败，此处跳过!')
pass
index += 1
gc.collect()
print('全部数据更新完毕'.center(100, '#')) # sleep(60*60)
if get_shanghai_time().hour == 0: # 0点以后不更新
sleep(60 * 60 * 5.5)
else:
sleep(5)
gc.collect()
def __init__(self):
MoGuJieParse.__init__(self)
def run_forever(self):
'''
实时更新数据
:return:
'''
sql_cli = SqlServerMyPageInfoSaveItemPipeline()
try:
sql_cli._delete_table(sql_str=mg_delete_str_2)
result = list(sql_cli._select_table(sql_str=mg_select_str_2))
except TypeError:
print('TypeError错误, 原因数据库连接失败...(可能维护中)')
result = None
if result is None:
pass
else:
_block_print_db_old_data(result=result)
index = 1
self.my_phantomjs = BaseDriver(
executable_path=PHANTOMJS_DRIVER_PATH,
ip_pool_type=self.ip_pool_type)
for item in result: # 实时更新数据
goods_id = item[0]
pintuan_end_time = json.loads(item[1]).get('end_time')
pintuan_end_time = int(
str(
time.mktime(
time.strptime(pintuan_end_time,
'%Y-%m-%d %H:%M:%S')))[0:10])
# print(miaosha_end_time)
data = {}
mogujie_pintuan = MoGuJieParse()
if index % 8 == 0:
try:
del self.my_phantomjs
except:
pass
gc.collect()
self.my_phantomjs = BaseDriver(
executable_path=PHANTOMJS_DRIVER_PATH,
ip_pool_type=self.ip_pool_type)
sql_cli = _block_get_new_db_conn(db_obj=sql_cli,
index=index,
remainder=50)
if sql_cli.is_connect_success:
if self.is_recent_time(pintuan_end_time) == 0:
_handle_goods_shelves_in_auto_goods_table(
goods_id=goods_id,
update_sql_str=mg_update_str_5,
sql_cli=sql_cli,
)
print(
'过期的goods_id为(%s)' % goods_id,
', 拼团开始时间为(%s), 逻辑删除成功!' %
json.loads(item[1]).get('begin_time'))
sleep(.3)
elif self.is_recent_time(pintuan_end_time) == 2:
# break # 跳出循环
pass # 此处应该是pass,而不是break，因为数据库传回的goods_id不都是按照顺序的
else: # 返回1，表示在待更新区间内
print(
'------>>>| 正在更新的goods_id为(%s) | --------->>>@ 索引值为(%d)'
% (goods_id, index))
data['goods_id'] = goods_id
tmp_url = 'http://list.mogujie.com/search?page={0}&fcid={1}&algoKey=pc_tuan_book_pop&cKey=pc-tuan'.format(
item[3], item[2])
# print(tmp_url)
# requests请求不到数据，涉及证书认证，直接用phantomjs
# body = MyRequests.get_url_body(url=tmp_url, headers=self.headers, had_referer=True)
body = self.my_phantomjs.use_phantomjs_to_get_url_body(
url=tmp_url)
# print(body)
if body == '':
print('获取到的body为空值! 此处跳过')
else:
try:
body = re.compile(
r'<pre.*?>(.*?)</pre>').findall(body)[0]
tmp_data = json.loads(body)
# pprint(tmp_data)
except:
print('json.loads转换body时出错, 请检查')
tmp_data = {}
if tmp_data.get('result',
{}).get('wall',
{}).get('docs', []) == []:
print('得到的docs为[]!')
_handle_goods_shelves_in_auto_goods_table(
goods_id=goods_id,
update_sql_str=mg_update_str_5,
sql_cli=sql_cli,
)
sleep(.3)
else:
tmp_item_list = tmp_data.get('result', {}).get(
'wall', {}).get('docs', [])
# pprint(tmp_item_list)
begin_time_timestamp = int(
time.time()) # 开始拼团的时间戳
item_list = [{
'goods_id':
item.get('tradeItemId', ''),
'pintuan_time': {
'begin_time':
timestamp_to_regulartime(
timestamp=begin_time_timestamp),
'end_time':
timestamp_to_regulartime(
self.get_pintuan_end_time(
begin_time_timestamp,
item.get('leftTimeOrg', ''))),
},
'all_sell_count':
str(item.get('salesVolume', 0)),
} for item in tmp_item_list]
# pprint(item_list)
pintuan_goods_all_goods_id = [
item_1.get('goods_id', '')
for item_1 in item_list
]
# print(pintuan_goods_all_goods_id)
'''
内部已经下架的(内部下架的其实并未真实下架，还在卖的，所以我就更新其商品信息数据，不更新上下架时间)
'''
if goods_id not in pintuan_goods_all_goods_id:
mogujie_pintuan.get_goods_data(
goods_id=goods_id)
goods_data = mogujie_pintuan.deal_with_data(
)
if goods_data == {}:
pass
else:
# 规范化
print('+++ 内部下架，其实还在售卖的商品更新')
goods_data['goods_id'] = goods_id
goods_data[
'price_info_list'] = _get_mogujie_pintuan_price_info_list(
goods_data['price_info_list'])
# pprint(goods_data)
mogujie_pintuan.update_mogujie_pintuan_table_2(
data=goods_data, pipeline=sql_cli)
sleep(MOGUJIE_SLEEP_TIME) # 放慢速度
else: # 未下架的
for item_2 in item_list:
if item_2.get('goods_id',
'') == goods_id:
mogujie_pintuan.get_goods_data(
goods_id=goods_id)
goods_data = mogujie_pintuan.deal_with_data(
)
if goods_data == {}: pass
else:
# 规范化
goods_data[
'goods_id'] = goods_id
goods_data[
'price_info_list'] = _get_mogujie_pintuan_price_info_list(
goods_data[
'price_info_list'])
goods_data[
'pintuan_time'] = item_2.get(
'pintuan_time', {})
goods_data[
'pintuan_begin_time'], goods_data[
'pintuan_end_time'] = get_miaosha_begin_time_and_miaosha_end_time(
miaosha_time=
goods_data[
'pintuan_time']
)
goods_data[
'all_sell_count'] = item_2.get(
'all_sell_count', '')
# pprint(goods_data)
mogujie_pintuan.update_mogujie_pintuan_table(
data=goods_data,
pipeline=sql_cli)
sleep(
MOGUJIE_SLEEP_TIME) # 放慢速度
else:
pass
else:
print('数据库连接失败，此处跳过!')
pass
index += 1
gc.collect()
print('全部数据更新完毕'.center(100, '#')) # sleep(60*60)
if get_shanghai_time().hour == 0: # 0点以后不更新
sleep(60 * 60 * 5.5)
else:
sleep(10 * 60)
gc.collect()
def deal_with_data(self, *params):
'''
处理并存储相关拼团商品的数据
:param params: 待传参数
:return:
'''
goods_list = params[0]
mogujie = MoGuJieParse()
my_pipeline = SqlServerMyPageInfoSaveItemPipeline()
if my_pipeline.is_connect_success:
db_goods_id_list = [
item[0] for item in list(
my_pipeline.select_mogujie_pintuan_all_goods_id())
]
print(db_goods_id_list)
for item in goods_list:
if item.get('goods_id', '') in db_goods_id_list:
print('该goods_id已经存在于数据库中, 此处跳过')
pass
else:
goods_id = str(item.get('goods_id', ''))
tmp_url = 'https://shop.mogujie.com/detail/' + str(
goods_id)
mogujie.get_goods_data(goods_id=str(goods_id))
goods_data = mogujie.deal_with_data()
if goods_data == {}: # 返回的data为空则跳过
pass
else: # 否则就解析并且插入
# 规范化
tmp_price_info_list = goods_data['price_info_list']
price_info_list = [{
'spec_value':
item_4.get('spec_value'),
'pintuan_price':
item_4.get('detail_price'),
'normal_price':
item_4.get('normal_price'),
'img_url':
item_4.get('img_url'),
'rest_number':
item_4.get('rest_number'),
} for item_4 in tmp_price_info_list]
goods_data['price_info_list'] = price_info_list
goods_data['goods_url'] = tmp_url
goods_data['goods_id'] = str(goods_id)
goods_data['pintuan_time'] = item.get(
'pintuan_time', {})
goods_data['pintuan_begin_time'], goods_data[
'pintuan_end_time'] = self.get_pintuan_begin_time_and_pintuan_end_time(
pintuan_time=item.get('pintuan_time', {}))
goods_data['all_sell_count'] = item.get(
'all_sell_count', '')
goods_data['fcid'] = str(item.get('fcid'))
goods_data['page'] = str(item.get('page'))
goods_data['sort'] = str(item.get('sort', ''))
# pprint(goods_data)
# print(goods_data)
mogujie.insert_into_mogujie_pintuan_table(
data=goods_data, pipeline=my_pipeline)
sleep(MOGUJIE_SLEEP_TIME) # 放慢速度
else:
print('数据库连接失败，此处跳过!')
pass
try:
del mogujie
except:
pass
gc.collect()