SQL Server查询以选择除在特定时间间隔内重复的记录之外的所有记录

Question

我想选择表中的所有记录，除了一些在给定时间间隔内为 CustId 重复的记录。

以下是理解需求的简单示例，假设我想跳过相同 CustId 在 60 分钟内重复的记录（60 分钟是时间间隔）。 正如您在下面的数据中看到的，“Cust1”重复第 2 行和第 3 行，“Cust2”重复第 6 行。

+------------------------------------------------+ |身份证 |客户 ID |捕获时间 | +------------------------------------------------+ | 1 | '客户1' | '2017-10-01 10:30' | | 2 | '客户1' | '2017-10-01 10:45' | -- 重复行 | 3 | '客户1' | '2017-10-01 11:35' | -- 重复行 | 4 | '客户1' | '2017-10-01 12:40' | | 5 | 'Cust2' | '2017-10-01 16:15' | | 6 | 'Cust2' | '2017-10-01 17:00' | -- 重复行 | 7 | 'Cust4' | '2017-10-02 08:00' | | 8 | '客户1' | '2017-10-02 10:45' | | 9 | 'Cust3' | '2017-10-02 13:00' | | 10 | 'Cust3' | '2017-10-02 14:05' | +----+----------+---------+

所以结果应该是

+------------------------------------------------+ |身份证 |客户 ID |捕获时间 | +------------------------------------------------+ | 1 | '客户1' | '2017-10-01 10:30' | | 4 | '客户1' | '2017-10-01 12:40' | | 5 | 'Cust2' | '2017-10-01 16:15' | | 7 | 'Cust4' | '2017-10-02 08:00' | | 8 | '客户1' | '2017-10-02 10:45' | | 9 | 'Cust3' | '2017-10-02 13:00' | | 10 | 'Cust3' | '2017-10-02 14:05' | +----+----------+---------+

感谢任何帮助。

Answer 1

将表加入自身并使用内置函数 Lag

SELECT t1.Id, t1.CastId, t1.capturedtime from Customer t1 INNER JOIN 
(SELECT *,LAG(capturedtime,1,0) OVER(PARTITION BY CastId,CAST(capturedtime as date) ORDER BY CastId,capturedtime) as tempTime FROM Customer) t2
ON t1.Id = t2.Id AND DATEDIFF(MINUTE,tempTime, t1.capturedtime) >= 60

Answer 2

您可以这样做，使用LAG 比较时间

declare @myt table (id int,custid nvarchar(50),capturedtime datetime
)

insert into @myt
values
( 1 , 'Cust1', '2017-10-01 10:30'),
( 2 , 'Cust1', '2017-10-01 10:45'),
( 3 , 'Cust1', '2017-10-01 11:35'),
( 4 , 'Cust1', '2017-10-01 12:40'),
( 5 , 'Cust2', '2017-10-01 16:15'),
( 6 , 'Cust2', '2017-10-01 17:00'),
( 7 , 'Cust4', '2017-10-02 08:00'),
( 8 , 'Cust1', '2017-10-02 10:45'),
( 9 , 'Cust3', '2017-10-02 13:00'),
( 10, 'Cust3', '2017-10-02 14:05')

select ID,Custid,CapturedTime from (
select * ,datediff(MINUTE,odtime,capturedtime) mm from (
select *,LAG(capturedtime,1,0) over(partition by custid,cast(capturedtime as date )order by custid,capturedtime) as odtime from @myt
)x
)z where mm >= 60
order by id

结果