类别中的领导和滞后

Question

我正在编写一个查询，以便从数据转储中获得SCD 2类型的数据。我的数据和代码如下：

create table promotions 
(
start_date date,
end_date date,
promotion_name varchar(50));

填充表的插入语句：

insert into promotions values ('9/1/2017','9/2/2017','P1');
insert into promotions values ('9/2/2017','9/3/2017','P1');
insert into promotions values ('9/3/2017','9/4/2017','P1');
insert into promotions values ('9/4/2017','9/5/2017','P1');
insert into promotions values ('9/5/2017','9/6/2017','P2');
insert into promotions values ('9/6/2017','9/7/2017','P2');
insert into promotions values ('9/7/2017','9/8/2017','P2');
insert into promotions values ('9/8/2017','9/9/2017','P2');
insert into promotions values ('9/9/2017','9/10/2017','P2');
insert into promotions values ('9/10/2017','9/11/2017','P2');
insert into promotions values ('9/11/2017','9/12/2017','P3');
insert into promotions values ('9/12/2017','9/13/2017','P3');
insert into promotions values ('9/13/2017','9/14/2017','P3');
insert into promotions values ('9/14/2017','9/15/2017','P3');

预期结果：

Date_Start      Date_End        Promotion Name
9/1/2017        9/4/2017        P1
9/5/2017        9/10/2017       P2
9/11/2017       9/13/2017       P3

我编写的查询：

with cte as (select rank() over (partition by promotion_name order by start_date asc) as "Rank"
,start_date
,dateadd(day,-1,start_date) as EndDate
,promotion_name
--first_name, last_name
from dbo.promotions)
select * from cte where rank=1;

查询输出

start_date  EndDate promotion_name
2017-09-01  2017-08-31  P1
2017-09-05  2017-09-04  P2
2017-09-11  2017-09-10  P3

上面的查询的问题是，与上面的输出表相比，EndDate的显示方式是错误的。

在server上，领导函数和滞后函数解决了这个问题，但是在TERADATA上无法得到等价的领导/lag函数。

我该怎么做。我不想创建任何易失性/临时表，它只是对ETL的简单查询。

Answer 1

LAG和LEAD只是较短的语法，您可以这样重写它：

LAG(col1, n) OVER (PARTITION BY ... ORDER BY col2)
=
MIN(col1) OVER (PARTITION BY ... ORDER BY col2
                ROWS BETWEEN n PRECEDING AND n PRECEDING), 0)

LEAD(col1, n) OVER (PARTITION BY ... ORDER BY col2)
=
MIN(col1) OVER (PARTITION BY ... ORDER BY col2
                ROWS BETWEEN n FOLLOWING AND n FOLLOWING), 0)

要获得默认值，只需使用COALESCE

LAG(col1, n, default) OVER (PARTITION BY ... ORDER BY col2)
=
COALESCE(MIN(col1) OVER (PARTITION BY ... ORDER BY col2
                         ROWS BETWEEN n PRECEDING AND n PRECEDING), 0)
        ,default)