根据记录（即行）之间经过的时间长度重新计算开始和结束日期

Question

我被难住了。我在 SQL 服务器数据库中有时间序列数据。每行代表客户入住酒店的时间。但是，有时客户会延长住宿时间、更换房间等，这些都会作为单独的预订记录在系统中。出于我们的目的，我需要将那些连续的 'bookings' 视为持续停留。考虑数据：

| CUSTOMERID | STARTDATE  | ENDDATE  |
| --------   | -----------| -------- |
| 1          |2021-07-02  |2021-07-14|
| 1          |2021-07-19  |2021-07-27|
| 2          |2018-11-12  |2018-11-16|
| 2          |2018-11-17  |2018-11-19|
| 2          |2018-11-19  |2018-11-25|
| 2          |2019-01-10  |2019-01-15|

我需要做的是汇总任何两条记录，其中客户下次访问的 STARTDATE 与当前记录的 ENDDATE 相差 <=1 天。换句话说，两次访问之间必须有一个完整的日历日才能将它们视为不同的。我需要最后的 table 看起来像这样：

| CUSTOMERID | STARTDATE  | ENDDATE  |NEWSTARTDATE|NEWENDDATE|
| --------   | -----------| -------- |------------|----------|
| 1          |2021-07-02  |2021-07-14|2021-07-02  |2021-07-14|
| 1          |2021-07-19  |2021-07-27|2021-07-19  |2021-07-27|
| 2          |2018-11-12  |2018-11-16|2018-11-12  |2018-11-25|
| 2          |2018-11-17  |2018-11-19|2018-11-12  |2018-11-25|
| 2          |2018-11-19  |2018-11-25|2018-11-12  |2018-11-25|
| 2          |2019-01-10  |2019-01-15|2019-01-10  |2019-01-15|

Answer 1

感谢@lptr 的回答，在子查询的 case 语句中包含一个 window 函数。请参阅下面的 link：

https://dbfiddle.uk/?rdbms=sqlserver_2019&fiddle=f2c905f2c31675c179d9f2100d74f44f

create table t(CUSTOMERID int, STARTDATE date, ENDDATE date);
insert into t(CUSTOMERID , STARTDATE , ENDDATE )
values
(1, '2021-07-02', '2021-07-14'),
(1, '2021-07-19', '2021-07-27'),
(2, '2018-11-12', '2018-11-16'),
(2, '2018-11-17', '2018-11-19'),
(2, '2018-11-19', '2018-11-25'),
(2, '2019-01-10', '2019-01-15');


select CUSTOMERID , STARTDATE , ENDDATE,
 max(sdate) over(partition by CUSTOMERID order by STARTDATE ) as NewStartDate,
 min(edate) over(partition by CUSTOMERID order by STARTDATE rows between current row and unbounded following) as NewEndDate
from
(
select *, 
 case when STARTDATE <= lag(dateadd(day, 1, ENDDATE)) over(partition by CUSTOMERID order by STARTDATE ) then null 
      else STARTDATE 
 end as sdate,
 case when ENDDATE>= lead(dateadd(day, -1, STARTDATE )) over(partition by CUSTOMERID order by STARTDATE ) then null 
      else ENDDATE
 end as edate
from t
) as t
order by STARTDATE ;

根据记录（即行）之间经过的时间长度重新计算开始和结束日期

Re-calculate start and end dates based on length of time elapsed between records (i.e., rows)

sql

sql-server

time-series

window-functions

aggregation