首页指标

task/home_feed_recommend_strategy_d.py

@@ -62,6 +62,7 @@ HIVE_SQL = """
 SELECT partition_date as `日期`
 	  ,a.device_os_type as `设备类型`
 	  ,b.active_type as `活跃类型`
+	  ,c.rank_type as `排序类型`
 	  ,d.card_content_type as `卡片类型`
 	  ,e.is_cpc as `是否cpc`
 	  ,sum(exp_pv) as `卡片曝光pv`
@@ -72,6 +73,7 @@ FROM
 		  ,array(active_type,'合计') as active_type
 		  ,array(t2.card_content_type,'合计') as card_content_type
 		  ,array(t2.is_cpc,'合计') as is_cpc 
+		  ,array(t1.rank_type,'合计') as rank_type 
 		  ,t1.device_id
 		  ,sum(exp_pv) as exp_pv
 		  ,sum(click_pv) as click_pv
@@ -82,10 +84,12 @@ FROM
 			  ,device_id
 			  ,device_os_type
 			  ,active_type
+			  ,rank_type
 		FROM
 		(	--渠道，新老
 			SELECT partition_date
 				  ,device_id
+				  ,case when substr(md5('androidid_1aefc7584e57983e'),-1) in ('0','1', '2', '3', '4', '5', '6', '7') then 'lr' else 'smr' end as rank_type
 		          ,device_os_type
 		          ,CASE WHEN active_type = '4' THEN '老活'  when active_type in ('1','2') then '新增' END as active_type
 			FROM online.ml_device_day_active_status
@@ -184,11 +188,13 @@ FROM
 )t 
 lateral view explode(t.device_os_type ) a as device_os_type
 lateral view explode(t.active_type ) b as active_type
+lateral view explode(t.rank_type ) c as rank_type
 lateral view explode(t.card_content_type ) d as card_content_type 
 lateral view explode(t.is_cpc ) e as is_cpc
 group by partition_date 
 	  ,a.device_os_type 
 	  ,b.active_type 
+	  ,c.rank_type
 	  ,d.card_content_type 
 	  ,e.is_cpc
 """.format(START_PARTITION_DATE,START_PARTITION_DATE,START_PARTITION_DATE)
@@ -204,18 +210,19 @@ print(pandas_df)
 for row in pandas_df.iterrows():
     partition_date = row[1]['日期']
     device_os_type = row[1]['设备类型']
+    rank_type = row[1]['排序类型']
     active_type = row[1]['活跃类型']
     card_content_type = row[1]['卡片类型']
     is_cpc = row[1]['是否cpc']
     exp_pv = row[1]['卡片曝光pv']
     click_pv = row[1]['卡片点击pv']
     ctr = row[1]['卡片ctr']
-    pid = hashlib.md5((partition_date + device_os_type + active_type + card_content_type + is_cpc).encode("utf8")).hexdigest()
+    pid = hashlib.md5((partition_date + device_os_type + rank_type + active_type + card_content_type + is_cpc).encode("utf8")).hexdigest()
     instert_sql = """replace into home_feed_recommend_strategy_d(
-    partition_date,device_os_type,active_type,card_content_type,is_cpc,exp_pv,click_pv,ctr,pid
-    ) VALUES('{partition_date}','{device_os_type}','{active_type}','{card_content_type}','{is_cpc}',{exp_pv},{click_pv},
+    partition_date,device_os_type,rank_type,active_type,card_content_type,is_cpc,exp_pv,click_pv,ctr,pid
+    ) VALUES('{partition_date}','{device_os_type}','{rank_type}','{active_type}','{card_content_type}','{is_cpc}',{exp_pv},{click_pv},
     {ctr},'{pid}');""".format(
-        partition_date=partition_date, device_os_type=device_os_type, active_type=active_type, card_content_type=card_content_type,
+        partition_date=partition_date, device_os_type=device_os_type,rank_type=rank_type, active_type=active_type, card_content_type=card_content_type,
         is_cpc=is_cpc, exp_pv=exp_pv, click_pv=click_pv, ctr=ctr,pid=pid
     )
     print(instert_sql)