update

task/spark_test.py

@@ -196,8 +196,74 @@ FROM (
                     AND action = 'on_click_card'
                     AND params['card_type'] = 'highlight_word'
               ) click
+                  LEFT JOIN
+    (
+        select distinct device_id
+        from ml.ml_d_ct_dv_devicespam_d  --去除机构刷单设备，即作弊设备（浏览和曝光事件去除）
+        WHERE partition_day=regexp_replace(DATE_SUB(current_date,1) ,'-','')
+
+        union all
+
+        select distinct device_id
+        from dim.dim_device_user_staff   --去除内网用户
+    )spam_pv
+    on spam_pv.device_id=t2.cl_id
+        LEFT JOIN
+    (
+        SELECT partition_date,device_id
+        FROM
+            (--找出user_id当天活跃的第一个设备id
+          SELECT user_id,partition_date,
+                  if(size(device_list) > 0, device_list [ 0 ], '') AS device_id
+            FROM online.ml_user_updates
+            WHERE partition_date=regexp_replace(DATE_SUB(current_date,1) ,'-','')
+        )t1
+                JOIN
+            (  --医生账号
+          SELECT distinct user_id
+          FROM online.tl_hdfs_doctor_view
+          WHERE partition_date = regexp_replace(DATE_SUB(current_date,1) ,'-','')
+
+          --马甲账号/模特用户
+          UNION ALL
+          SELECT user_id
+          FROM ml.ml_c_ct_ui_user_dimen_d
+          WHERE partition_day = regexp_replace(DATE_SUB(current_date,1) ,'-','')
+          AND (is_puppet = 'true' or is_classifyuser = 'true')
+
+          UNION ALL
+          --公司内网覆盖用户
+          select distinct user_id
+          from dim.dim_device_user_staff
+
+          UNION ALL
+          --登陆过医生设备
+          SELECT distinct t1.user_id
+            FROM
+            (
+                SELECT user_id, v.device_id as device_id
+                FROM online.ml_user_history_detail
+                    LATERAL VIEW EXPLODE(device_history_list) v AS device_id
+                WHERE partition_date = regexp_replace(DATE_SUB(current_date,1) ,'-','')
+            ) t1
+            JOIN
+            (
+                SELECT device_id
+                FROM online.ml_device_history_detail
+                WHERE partition_date = regexp_replace(DATE_SUB(current_date,1) ,'-','')
+                AND is_login_doctor = '1'
+            ) t2
+                ON t1.device_id = t2.device_id
+        )t2
+            on t1.user_id=t2.user_id
+        group by partition_date,device_id
+    )dev
+    on click.partition_date=dev.partition_date and click.cl_id=dev.device_id
+        WHERE spam_pv.device_id IS NULL
+          and dev.device_id is null
+              
          GROUP BY click.partition_date, query
-     ) t3
+     ) t3 order by all_search_uv asc
 
 """.format(start_date='20201116',end_date='20201117')