update

task/search_strategy_d.py

@@ -78,7 +78,24 @@ for t in range(0, task_days):
               FROM online.ml_user_updates
               WHERE partition_date>='{yesterday_str}' AND partition_date<'{today_str}'
         )t1
-        group by partition_date,device_id
+        JOIN
+        (  
+            SELECT distinct user_id
+            FROM online.tl_hdfs_doctor_view
+            WHERE partition_date = '{yesterday_str}'
+    
+            UNION ALL
+            SELECT user_id
+            FROM ml.ml_c_ct_ui_user_dimen_d
+            WHERE partition_day = '{yesterday_str}'
+            AND (is_puppet = 'true' or is_classifyuser = 'true')
+    
+            UNION ALL
+            select distinct user_id
+            from dim.dim_device_user_staff
+           
+        )t3
+        on t1.user_id=t3.user_id group by partition_date,device_id
 """.format(today_str=today_str, yesterday_str=yesterday_str)
     print(sql_distinct_device_id)
     distinct_device_id_df = spark.sql(sql_distinct_device_id)