Return 特定行,列中的级联结果 - 如果没有 1 则多个 2,否则如果没有 2 则 3,否则如果没有 3 则多个 4

Return specific rows, cascading results in column - If no 1 then multiple 2s, else if no 2s then 3, else if no 3, then multiple 4s

我在 MS SQL 中处理学生数据,需要遵循一些非常具体的规则。

样本Table

CREATE TABLE students (
    encounterId INT,
    studentId INT,
    positionId INT
);

INSERT INTO students
VALUES
(100,20,1),
(100,32,2),
(100,14,2),
(101,18,1),
(101,87,2),
(101,78,3),
(102,67,2),
(102,20,2),
(103,33,3),
(103,78,4),
(104,16,1),
(104,18,4),
(105,67,4),
(105,18,4),
(105,20,4);

Table 规则

table 显示学生遇到的情况,其中学生被置于 1 和 4 之间的位置。
一次遭遇可以有多个学生。
一场遭遇战只能有1名学生。
一场遭遇战中3号位只能有一个学生
但是一次遭遇战中2、4号位可以有多个同学。

商业规则

每次相遇业务规则如下:

不太好用

studentId 值的串联是acceptable,但并不理想。我有这个半工作与一系列不稳定的工会和 string_aggs。 positionId=3 的行有问题,正如我在代码中所说的那样。

此外,这种类似于 union/not 的架构适用于我的小型开发数据库,​​但在生产数据库中会出现严重的性能问题:

WITH tAll
AS (   SELECT
           encounterId,
           studentId,
           positionId
       FROM
           students)

SELECT
    encounterId,
    CAST(studentId AS VARCHAR) AS [studentId],
    1                          AS [ord]
FROM
    tAll
WHERE
    positionId = 1
UNION
SELECT
    encounterId,
    CAST(studentId AS VARCHAR),
    2 AS [ord]
FROM
    (
        SELECT
            encounterId,
            STRING_AGG(studentId, ',')  AS [studentId],
            STRING_AGG(positionId, ',') AS [positionId]
        FROM
            tAll
        GROUP BY
            encounterId
    ) t2
WHERE
    positionId NOT LIKE '%1%'
    AND positionId NOT LIKE '%3%'
    AND positionId NOT LIKE '%4%'
UNION
SELECT
    encounterId,
    CAST(studentId AS VARCHAR),
    3 AS [ord]
FROM
    --tAll WHERE positionId=3 
    --Limiting to positionId=3 includes results (101,18,1) AND (101,78,3).. I just want (101,18,1)
    --Using the below code instead, but this creates other problems
    (
        SELECT
            encounterId,
            STRING_AGG(studentId, ',')  AS [studentId],
            STRING_AGG(positionId, ',') AS [positionId]
        FROM
            tAll
        GROUP BY
            encounterId
    ) t3
WHERE
    positionId NOT LIKE '%1%'
    AND positionId NOT LIKE '%2%'
    AND positionId NOT LIKE '%4%'
--This excludes 103 entirely since it has both positionId values of 3 AND 4... I just want (103,33,3)
UNION
SELECT
    encounterId,
    CAST(studentId AS VARCHAR),
    4 AS [ord]
FROM
    (
        SELECT
            encounterId,
            STRING_AGG(studentId, ',')  AS [studentId],
            STRING_AGG(positionId, ',') AS [positionId]
        FROM
            tAll
        GROUP BY
            encounterId
    ) t4
WHERE
    positionId NOT LIKE '%1%'
    AND positionId NOT LIKE '%2%'
    AND positionId NOT LIKE '%3%';

我想要什么returned

encounterId studentId ord
100 20 1
101 18 1
102 67 2
102 20 2
103 33 3
104 16 1
105 67 4
105 18 4
105 20 4

这是每组前 1 名的问题...有平局。

您可以在子查询中使用 window 函数 rank() 对每次遭遇中的学生进行排名,然后在外部查询中过滤每组的顶部记录:

select *
from (
    select s.*,
        rank() over(partition by encounterid order by positionid) rn
    from students s
) s
where rn = 1
order by encounterid 

另一个选项使用 with ties - 但您无法控制结果集中行的顺序:

select top (1) with ties *
from students s
order by rank() over(partition by encounterid order by positionid)

另一种典型的解决方案是使用相关子查询进行过滤:

select *
from students s
where positionid = (select min(s1.positionid) from students s1 where s1.encounterid  = s.encounterid)

感谢您获取测试数据。以下查询工作正常。

;with cte_minposition as
(
SELECT encounterId, min(positionid) as min_position FROM students
group by encounterId
)
SELECT * FROM students as s
inner join cte_minposition as m
on s.positionId <= m.min_position and s.encounterId = m.encounterId
encounterId studentId positionId encounterId min_position
100 20 1 100 1
101 18 1 101 1
102 67 2 102 2
102 20 2 102 2
103 33 3 103 3
104 16 1 104 1
105 67 4 105 4
105 18 4 105 4
105 20 4 105 4