update

task/spark_test.py

@@ -91,8 +91,11 @@ for t in range(2, task_days):
     yesterday_str_format = (now + datetime.timedelta(days=-1)).strftime("%Y-%m-%d")
     one_week_age_str = (now + datetime.timedelta(days=-7)).strftime("%Y%m%d")
     new_urser_device_id_sql = r"""
-         select cl_type,app_version,count(app_version) as count_num from (select cl_type,app_version,cl_id from online.bl_hdfs_maidian_updates where partition_date >= "20200201" group by cl_type,app_version,cl_id) group by cl_type,app_version
-    """.format(start_date="20210201", end_date="20200222", today_str_format=today_str_format,tomorrow_str=tomorrow_str)
+         select cl_type,app_version,count(app_version) as count_num from 
+         ( select *,rank() over(partition by cl_id order by partition_date desc) as ranking from
+         (select cl_type,app_version,cl_id,partition_date from online.bl_hdfs_maidian_updates where partition_date >= "20200201" group by cl_type,app_version,cl_id,partition_date) 
+         ) where ranking =1 group by cl_type,app_version
+    """.format(start_date="20210101", end_date="20200222", today_str_format=today_str_format,tomorrow_str=tomorrow_str)
 
   #   print(new_urser_device_id_sql)
     new_urser_device_id_df = spark.sql(new_urser_device_id_sql)