在 SQL 中 DENSE_RANK() 的潜在用途

Question

在 SQL 中，我正在寻找一个潜在的函数，每次满足特定条件时我都可以使用它来应用行号。

每次唯一 ID 相同（即同一个人）且

Dep_Group

为

WES

时，如果它们彼此连续，则为该行提供相同的编号。我想做的就是最终将这些行压缩成一行，希望通过在末尾添加 GROUP BY

New_Col

子句来实现。所以本质上

Line

1 和 2 成为一行，

Line

5 和 6.

|Line| Uniq_ID | Start_DT      | End_DT        | Sequence | Department | Dep_Group | New_Col|
| 1  |   1     | 2023-01-02    | 2023-01-03    |    1     |     WES    |    WES    |     1  |
| 2  |   1     | 2023-01-02    | 2023-01-03    |    2     |     WES    |    WES    |     1  |
| 3  |   1     | 2023-01-02    | 2023-01-03    |    3     |     DEM    |    NULL   |     2  |
| 4  |   2     | 2023-01-02    | 2023-01-03    |    1     |     WES    |    WES    |     3  |
| 5  |   3     | 2023-01-02    | 2023-01-03    |    1     |     WES    |    WES    |     4  |
| 6  |   3     | 2023-01-02    | 2023-01-03    |    2     |     WES    |    WES    |     4  |
| 7  |   4     | 2023-01-02    | 2023-01-03    |    1     |     DEM    |    NULL   |     5  |

到目前为止，我的代码示例在 DENSE_RANK() 中并不能完全按照我想要的方式工作：

SELECT Line
       ,Uniq_ID
       ,Start_DT
       ,End_DT
       ,Sequence
       ,Department
       ,Dep_Group
       ,DENSE_RANK() OVER (PARTITION BY Uniq_ID, Dep_Group ORDER BY Sequence) AS New_Col

FROM TblA

Answer 1

这相当简单：使用运行总和，如果没有则递增（前一组= wes，当前组= wes）：

with cte1 as (
  select *
       , case when lag(dep_group) over (partition by uniq_id order by sequence) = 'WES' and dep_group = 'WES' then 0 else 1 end as incr
  from t
), cte2 as (
  select *
       , sum(incr) over (partition by uniq_id order by sequence) as grp_num
   from cte1
)
select *
from cte2

这将为您提供以下结果：

线	uniq_id	开始_dt	结束_dt	顺序	部门	dep_group	增加	grp_num
1	1	2023-01-02	2023-01-03	1	WES	WES	1	1
2	1	2023-01-02	2023-01-03	2	WES	WES	0	1
3	1	2023-01-02	2023-01-03	3	DEM	空	1	2
4	2	2023-01-02	2023-01-03	1	WES	WES	1	1
5	3	2023-01-02	2023-01-03	1	WES	WES	1	1
6	3	2023-01-02	2023-01-03	2	WES	WES	0	1
7	4	2023-01-02	2023-01-03	1	DEM	空	1	1

grp_num

列可与

uniq_id

一起用于对行进行分组。

Answer 2

SELECT Line,
       Uniq_ID,
       Start_DT,
       End_DT,
       Sequence,
       Department,
       Dep_Group,
       SUM(Flag) OVER (PARTITION BY Uniq_ID ORDER BY Sequence) AS New_Col
FROM (
  SELECT *,
         IIF(LAG(Dep_Group) OVER (PARTITION BY Uniq_ID ORDER BY Sequence) = 'WES' AND Dep_Group = 'WES', 0, 1) AS Flag
  FROM TblA
) AS subquery;