基于累积价值的排名

问题描述

我想根据UID的升序对ID和值列进行排名。一旦值列的值与先前值不同，则预期输出必须更改。排名必须在每个新ID上重新启动

UID ID  Value Expected Output
1   1   0         1
2   1   0         1
3   1   1         2
4   1   1         2
5   1   1         2
6   1   0         3
7   1   1         4
8   1   0         5
9   1   0         5
10  1   0         5
11  2   1         1
12  2   1         1
13  2   0         2
14  2   0         2
15  2   1         3

这是我创建的样本数据集：

CREATE TABLE [dbo].[Data] (
    [UID] [int] NOT NULL,[ID] [int] NULL,[Value] [int] NULL
);

INSERT [dbo].[Data] ([UID],[ID],[Value]) VALUES (1,1,0);
INSERT [dbo].[Data] ([UID],[Value]) VALUES (2,[Value]) VALUES (3,1);
INSERT [dbo].[Data] ([UID],[Value]) VALUES (4,[Value]) VALUES (5,[Value]) VALUES (6,[Value]) VALUES (7,[Value]) VALUES (8,[Value]) VALUES (9,[Value]) VALUES (10,[Value]) VALUES (11,2,[Value]) VALUES (12,[Value]) VALUES (13,[Value]) VALUES (14,[Value]) VALUES (15,1);

解决方法

我认为，解决这个“缺岛”问题的最简单方法是使用select uid,id,value,1 + sum(case when value <> lag_value then 1 else 0 end) over(partition by id order by uid) grp from ( select d.*,lag(value,1,value) over(partition by id order by uid) lag_value from data d ) d order by uid 来检索“先前”值，然后使用窗口总和在每次值更改时增加。

id date variable value
 1 2019    x      100
 1 2019    y      50.5
 1 2020    x      10.0
 1 2020    y      NA

Demo on DB Fiddle ：

uid | id | value | grp
--: | -: | ----: | --:
  1 |  1 |     0 |   1
  2 |  1 |     0 |   1
  3 |  1 |     1 |   2
  4 |  1 |     1 |   2
  5 |  1 |     1 |   2
  6 |  1 |     0 |   3
  7 |  1 |     1 |   4
  8 |  1 |     0 |   5
  9 |  1 |     0 |   5
 10 |  1 |     0 |   5
 11 |  2 |     1 |   1
 12 |  2 |     1 |   1
 13 |  2 |     0 |   2
 14 |  2 |     0 |   2
 15 |  2 |     1 |   3

,

这是一个空白和孤岛的问题。我认为最简单的方法是使用行数差异方法：

WITH cte AS (
    SELECT *,ROW_NUMBER() OVER (PARTITION BY ID ORDER BY UID) rn1,ROW_NUMBER() OVER (PARTITION BY ID,[Value] ORDER BY UID) rn2
    FROM Data
)

SELECT *,DENSE_RANK() OVER (PARTITION BY ID ORDER BY rn1 - rn2,[Value]) AS output
FROM cte
ORDER BY UID;

Demo