WHERE Column NOT LIKE 不能正确使用递归 cte

Question

这是SQL服务器代码。

假设您有一个包含三列的 table。第 1 列名为 Monster，第 2 列名为 Level，第 3 列名为 BodyType。 Level表示怪物的强大程度，BodyType表示它拥有的body类型。

我的架构：

CREATE TABLE YourTable
    ([Monster] nvarchar(max), [Level] int, [BodyType] nvarchar(max))
;

INSERT INTO YourTable
    ([Monster], [Level], [BodyType])
VALUES
    ('Small Beast', 300, 'Scaly'),
    ('Large Beast', 700, 'Slimy'),
    ('Small Dragon', 350, 'Fiery'),
    ('Large Dragon', 800, 'Slimy')
;

我有一个 sql 命令来查找所有可能的怪物组合。它使用递归 cte，因为 table 中的怪物数量可能会波动（因此我可以稍后添加更多怪物）。该命令还获取正在组合的怪物等级的总和值。该命令也只输出低于某个总和值的怪物组合。在此示例中，总和值为 1500。到目前为止，一切正常。

我的sql命令：

;WITH cte AS (
SELECT  Monster, 
        [Level],
        BodyType,
        1 as l
FROM YourTable
UNION ALL


SELECT  c1.Monster+','+c2.Monster,
        c1.[Level]+c2.[Level],
        c1.BodyType+','+c2.BodyType,
        c1.l+1
FROM cte c1
CROSS JOIN YourTable c2
WHERE c1.Monster NOT LIKE '%'+c2.Monster+'%'
)


SELECT *
FROM cte
WHERE cte.Level < 1500
ORDER BY l
OPTION (MAXRECURSION 0)

以及正确的输出：

1   Small Beast 300 Scaly   1
2   Large Beast 700 Slimy   1
3   Small Dragon    350 Fiery   1
4   Large Dragon    800 Slimy   1
5   Large Dragon,Small Beast    1100    Slimy,Scaly 2
6   Large Dragon,Small Dragon   1150    Slimy,Fiery 2
7   Small Dragon,Small Beast    650 Fiery,Scaly 2
8   Small Dragon,Large Beast    1050    Fiery,Slimy 2
9   Small Dragon,Large Dragon   1150    Fiery,Slimy 2
10  Large Beast,Small Beast 1000    Slimy,Scaly 2
11  Large Beast,Small Dragon    1050    Slimy,Fiery 2
12  Small Beast,Large Beast 1000    Scaly,Slimy 2
13  Small Beast,Small Dragon    650 Scaly,Fiery 2
14  Small Beast,Large Dragon    1100    Scaly,Slimy 2
15  Small Beast,Large Dragon,Small Dragon   1450    Scaly,Slimy,Fiery   3
16  Small Beast,Small Dragon,Large Beast    1350    Scaly,Fiery,Slimy   3
17  Small Beast,Small Dragon,Large Dragon   1450    Scaly,Fiery,Slimy   3
18  Small Beast,Large Beast,Small Dragon    1350    Scaly,Slimy,Fiery   3
19  Large Beast,Small Dragon,Small Beast    1350    Slimy,Fiery,Scaly   3
20  Large Beast,Small Beast,Small Dragon    1350    Slimy,Scaly,Fiery   3
21  Small Dragon,Large Dragon,Small Beast   1450    Fiery,Slimy,Scaly   3
22  Small Dragon,Large Beast,Small Beast    1350    Fiery,Slimy,Scaly   3
23  Small Dragon,Small Beast,Large Beast    1350    Fiery,Scaly,Slimy   3
24  Small Dragon,Small Beast,Large Dragon   1450    Fiery,Scaly,Slimy   3
25  Large Dragon,Small Dragon,Small Beast   1450    Slimy,Fiery,Scaly   3
26  Large Dragon,Small Beast,Small Dragon   1450    Slimy,Scaly,Fiery   3

我遇到的问题是当我添加一个 Where 子句来只带回不属于某种 body 类型（BodyType 列）的怪物时。上面的那部分代码在修改后是：

;WITH cte AS (
SELECT  Monster, 
        [Level],
        BodyType,
        1 as l
FROM YourTable
WHERE BodyType NOT LIKE 'Fiery' AND BodyType NOT LIKE 'Slimy'
UNION ALL

输出如下，这是不正确的，因为它仍然包含 body 类型的 Slimy 和 Fiery：

    Monster Level   BodyType    l
1   Small Beast 300 Scaly   1
2   Small Beast,Large Beast 1000    Scaly,Slimy 2
3   Small Beast,Small Dragon    650 Scaly,Fiery 2
4   Small Beast,Large Dragon    1100    Scaly,Slimy 2
5   Small Beast,Large Dragon,Small Dragon   1450    Scaly,Slimy,Fiery   3
6   Small Beast,Small Dragon,Large Beast    1350    Scaly,Fiery,Slimy   3
7   Small Beast,Small Dragon,Large Dragon   1450    Scaly,Fiery,Slimy   3
8   Small Beast,Large Beast,Small Dragon    1350    Scaly,Slimy,Fiery   3

输出似乎部分起作用，因为 Large Beast 是 Slimy，它第一次忽略了它，但我怀疑它在 BodyType 级别移动时忽略了 NOT LIKE 子句，这就是它不忽略 Large 的原因野兽在随后的发现。

Answer 1

你的逻辑需要AND，而不是OR:

WHERE BodyType NOT LIKE 'Fiery' AND BodyType NOT LIKE 'Slimy'

"Fiery" 的 BodyType 不像 "Slimy"。所以，它符合第二个条件。请注意，如果您使用 LIKE，那么您需要 AND.

Answer 2

我想我找到了解决办法。输出数据似乎是正确的：

SELECT  c1.Monster+','+c2.Monster,
        c1.[Level]+c2.[Level],
        c1.BodyType+','+c2.BodyType,
        c1.l+1
FROM cte c1
CROSS JOIN YourTable c2
WHERE (c1.Monster NOT LIKE '%'+c2.Monster+'%') AND (c1.BodyType NOT LIKE 'Scaly' AND c2.BodyType NOT LIKE 'Scaly') AND (c1.BodyType NOT LIKE 'Fiery' AND c2.BodyType NOT LIKE 'Fiery') 
)

我去掉了损坏的 Where 子句，只是在原始代码的 CROSS JOIN 之后的 Where 子句中添加了 NOT LIKE。

只是不确定这是否不是最佳做法或者是否会破坏某些东西，有人想插话吗？谢谢。

编辑：我的解决方案有一个问题，即第一次迭代将包括被忽略的 'BodyTypes'

Answer 3

如果我没理解错的话，您可以尝试以下两种方法：

1.Filter血型在recursive cte

之后

;WITH cte AS (
SELECT  Monster, 
        [Level],
        BodyType,
        1 as l
FROM YourTable
UNION ALL

SELECT  c1.Monster+','+c2.Monster,
        c1.[Level]+c2.[Level],
        c1.BodyType+','+c2.BodyType,
        c1.l+1
FROM cte c1
CROSS JOIN YourTable c2
WHERE c1.Monster NOT LIKE '%'+c2.Monster+'%'
)
SELECT *
FROM cte
WHERE cte.Level < 1500
AND ',' + cte.BodyType ',' + NOT LIKE '%,Fiery,%' 
AND ',' + cte.BodyType ','  + NOT LIKE '%,Slimy,%'
ORDER BY l
OPTION (MAXRECURSION 0)

2.Use 在递归 cte 之前要过滤的第二个 cte

;WITH temp AS
(
    SELECT * 
    FROM YourTable
    WHERE   BodyType != 'Fiery' 
            AND BodyType != 'Slimy'
)
,cte AS (
SELECT  Monster, 
        [Level],
        BodyType,
        1 as l
FROM temp
UNION ALL

SELECT  c1.Monster+','+c2.Monster,
        c1.[Level]+c2.[Level],
        c1.BodyType+','+c2.BodyType,
        c1.l+1
FROM cte c1
CROSS JOIN temp c2
WHERE c1.Monster NOT LIKE '%'+c2.Monster+'%'
)
SELECT *
FROM cte
WHERE cte.Level < 1500   
ORDER BY l
OPTION (MAXRECURSION 0)

WHERE Column NOT LIKE 不能正确使用递归 cte

WHERE Column NOT LIKE isn't working correctly with recursive cte

sql

sql-server

recursive-cte