如何改进 SQL 中不同时间范围的聚合查询

Posted 2023-03-29

【中文标题】如何改进 SQL 中不同时间范围的聚合查询

【英文标题】：How to improve this aggregation queries for different timeframe in SQL

【发布时间】：2021-08-20 22:07:39

【问题描述】：

我使用的是 SQL Server 2017

创建表的脚本：

Create table tb(
  ID INT Identity(1,1),
  MUID INT ,
  VisitDTM DateTime not null,
  EventAction Varchar(20)
  );
  
  Insert into tb(MUID,VisitDTM,EventAction)
  values
  (5001,'2021-03-08 13:18:06.000','Open'),
  (5001,'2021-03-12 13:05:13.000','Ad'),
  (5001,'2021-03-18 13:18:06.000','Ad'),
  (5001,'2021-05-29 13:18:06.000','Open'),
  (5001,'2021-05-31 13:18:06.000','Open'),
  (5001,'2021-06-02 00:48:49.000','Open'),
  
  (5005,'2021-01-01 13:18:06.000','Open'),
  (5005,'2021-02-11 13:05:13.000','Ad'),
  (5005,'2021-03-18 13:18:06.000','Ad'),
  (5005,'2021-05-02 13:18:06.000','Open'),
  (5005,'2021-05-31 13:18:06.000','Open'),
  (5005,'2021-05-31 23:58:47.000','Open'),
  (5005,'2021-05-31 23:58:47.000','Ad'),
  (5005,'2021-05-31 23:58:47.000','Ad'),
  (5005,'2021-05-31 23:59:43.000','Open')

我正在尝试为不同的时间范围（last7days、last30days、last60days、last90days）进行聚合

我试过在这里加入：

select *
from
(select MUID, count(VisitDTM) as VisitsinLast7days
from tb
where VisitDTM > DATEADD(day, -7, GETDATE())
group by MUID
)a
Join
(
select MUID, count(VisitDTM) as VisitsinLast30days
from tb
where VisitDTM > DATEADD(day, -30, GETDATE())
group by MUID
  ) b on a.MUID = b.MUID

有没有更好的方法？

Updated Fiddle link here

我有 1250 万条记录。

【参考方案1】：

我猜，你可以使用“条件聚合”

SELECT X.[MUID],
SUM(
     CASE
        WHEN VisitDTM > DATEADD(day, -7, GETDATE()) THEN 1
        ELSE 0
     END
   )AS Last7Days,
SUM(
   CASE
      WHEN VisitDTM > DATEADD(day, -30, GETDATE()) THEN 1
      ELSE 0
   END
  )AS Last30Days
 FROM [dbo].[tb] AS X
 GROUP BY X.MUID