optimize the sql statement in get_click_times_to_count_uid_df()

eda/recommended_indexs_v2/utils.py

 # -*- coding: UTF-8 -*-
 import pymysql
 import datetime
-import pandas as pd
 
 def con_sql(sql):
 	#从数据库的表里获取数据
@@ -72,15 +71,16 @@ def get_activate_uid_ctr(platform, ndays=1):
 	return [platform, clk_count, imp_count, clk_rate]
 
 
-#获取 {点击次数 : 独立用户数}
+#获取 (点击次数 : 独立用户数)
 def get_click_times_to_count_uid_df():
 	"""
-	rtype : pandas.DataFrame
+	rtype : tuple
 	"""
-	sql = "select device_id,count(cid_type) click_times from data_feed_click where from_unixtime(time,'%Y-%m-%d')=date_add(curdate(), interval -1 day) group by device_id order by click_times desc"
-	uid_click_times = con_sql(sql)
-	uid_lst = [i[0] for i in uid_click_times]
-	click_times_lst = [i[1] for i in uid_click_times]
-	uid_click_times_df = pd.DataFrame({"uid":uid_lst,"click_times":click_times_lst})
-	df = uid_click_times_df.groupby(by="click_times",as_index=False).count()
-	return df
+	sql = "select times,count(device_id) \
+	from (select device_id,count(cid_type) as times \
+		from data_feed_click \
+		where from_unixtime(time,'%Y-%m-%d')=date_add(curdate(), interval -1 day) \
+		group by device_id) as t \
+	group by times order by times"
+	result = con_sql(sql)
+	return result