根据主键去重（针对修改数据，数据重复的场景）

SELECT id,follow_record_id,customer_id,create_time,modify_time,__is_delete__
FROM (
   SELECT id,follow_record_id,customer_id,create_time,modify_time,__is_delete__,
     ROW_NUMBER() OVER (PARTITION BY id
       ORDER BY proctime desc ) AS rownum 
   FROM itcrm_follow_record_customer_relation)
WHERE rownum = 1

数据过滤

select * from itcrm_follow_record_customer_relation_distinct where __is_delete__ = 0

做计算

上面的计算流程在数据被删除时会发出retract的消息，回收下游算子的一些状态，两个SQL是可以合并在一起的，效果等同分开写