Question

CTE会在下一次递归中使用作为CTE一部分更新的数据吗？我试图尝试这个CTE，因为WHILE循环内部类似UPDATE逻辑的性能表现不佳，我希望使用CTE可以更好地设置并且性能更好。

我在使用递归CTE更新表时遇到问题，直到父行都被正确标记为止。

SQL小提琴显示表格和基本CTE。在添加任何AND / OR或级别逻辑之前，我似乎无法让CTE爬上层次结构并将父母标记为＆＃34;遇到＆＃34;。

以下是示例表：

| LogicID   | ParentLogic   | Depth     | Type  | Description                           | Met   |
|---------  |-------------  |-------    |------ |-------------------------------------  |-----  |
| 1         | NULL          | NULL      | NULL  | Conditions All Met                    | 0     |
| 2         | 1             | 1         | AND   | The sky or ocean is blue              | 0     |
| 3         | 2             | 2         | OR    | The sky is blue                       | 0     |
| 4         | 2             | 2         | OR    | The ocean is blue                     | 1     |
| 5         | 1             | 1         | AND   | The grass is green or road is black   | 0     |
| 6         | 5             | 2         | OR    | The grass is green                    | 1     |
| 7         | 5             | 2         | OR    | The road is black                     | 0     |
| 8         | 1             | 1         | AND   | Birds, bugs or the 4 below            | 0     |
| 9         | 8             | 2         | OR    | There are birds                       | 0     |
| 10        | 8             | 2         | OR    | There are bugs                        | 0     |
| 11        | 8             | 2         | OR    | All 4 below                           | 0     |
| 12        | 11            | 3         | AND   | There are dogs                        | 1     |
| 13        | 11            | 3         | AND   | There are cats                        | 1     |
| 14        | 11            | 3         | AND   | There are people                      | 1     |
| 15        | 11            | 3         | AND   | There are chairs                      | 1     |

DROP TABLE MyLogic
CREATE TABLE MyLogic
    (
        LogicID int
        ,ParentLogic int
        ,Depth int
        ,Type varchar(4)
        ,Description varchar(35)
        ,Met int
    );

INSERT  INTO MyLogic
        ( LogicID, ParentLogic, Depth, Type, Description, Met )
VALUES
        ( 1, NULL, NULL, NULL, 'Conditions All Met', 0 ),
        ( 2, 1, 1, 'AND', 'The sky or ocean is blue', 0 ),
        ( 3, 2, 2, 'OR', 'The sky is blue', 0 ),
        ( 4, 2, 2, 'OR', 'The ocean is blue', 1 ),
        ( 5, 1, 1, 'AND', 'The grass is green or road is black', 0 ),
        ( 6, 5, 2, 'OR', 'The grass is green', 1 ),
        ( 7, 5, 2, 'OR', 'The road is black', 0 ),
        ( 8, 1, 1, 'AND', 'Birds, bugs or the 4 below', 0 ),
        ( 9, 8, 2, 'OR', 'There are birds', 0 ),
        ( 10, 8, 2, 'OR', 'There are bugs', 0 ),
        ( 11, 8, 2, 'OR', 'All 4 below', 0 ),
        ( 12, 11, 3, 'AND', 'There are dogs', 1 ),
        ( 13, 11, 3, 'AND', 'There are cats', 1 ),
        ( 14, 11, 3, 'AND', 'There are people', 1 ),
        ( 15, 11, 3, 'AND', 'There are chairs', 1 )

这只是一组更复杂的逻辑示例。基本上我的想法是我需要每一组孩子来“累积”＃34;使用表中的逻辑到父级。深度可变，但很可能是7深。

因此，逻辑ID 12,13,14,15被AND在一起，然后将11标记为Met。然后评估9,10,11，如果有的话（OR）则将8标记为Met。依此类推，直到最高级别的父级逻辑ID 1要么遇到要么不符合。

这可以通过CTE来完成，如果是这样，有人可以帮助我实现吗？

EDIT :: 感谢您的帮助 - 这里要求的是更新声明。



DECLARE @maxdepth AS int = ( SELECT MAX (Depth) FROM MyLogic)   
DECLARE @counter AS int = 0

WHILE ( @counter < @maxdepth )
    BEGIN

        UPDATE
            UP
        SET 
            UP.Met = 
            --SELECT *,
            CASE 
            WHEN ORIG.Type = 'AND' AND ORIG.Met = 0 AND COUNTS.CountMet = 2 THEN 0
            WHEN ORIG.Type = 'AND' AND ORIG.Met = 0 AND COUNTS.CountMet = 1 THEN 0
            WHEN ORIG.Type = 'AND' AND ORIG.Met = 1 AND COUNTS.CountMet = 2 THEN 0
            WHEN ORIG.Type = 'AND' AND ORIG.Met = 1 AND COUNTS.CountMet = 1 THEN 1

            WHEN ORIG.Type = 'OR'  AND ORIG.Met = 1 AND COUNTS.CountMet = 1 THEN 1
            WHEN ORIG.Type = 'OR'  AND ORIG.Met = 0 AND COUNTS.CountMet = 2 THEN 1
            WHEN ORIG.Type = 'OR'  AND ORIG.Met = 1 AND COUNTS.CountMet = 2 THEN 1
            WHEN ORIG.Type = 'OR'  AND ORIG.Met = 0 AND COUNTS.CountMet = 1 THEN 0
            END
        FROM
            MyLogic UP
        INNER JOIN dbo.MyLogic ORIG
            ON UP.LogicID = ORIG.ParentLogic
        INNER JOIN ( SELECT
                        DIST.ParentLogic
                        ,COUNT(DISTINCT DIST.Met) AS CountMet
                    FROM
                        MyLogic DIST
                    GROUP BY
                        DIST.ParentLogic
                    ) COUNTS
            ON ORIG.ParentLogic = COUNTS.ParentLogic

        SET @counter = @counter + 1

    END

Answer 1

我不相信你能用CTE在这个场景中完成你的目标。我尝试使用CTE的每种方法都取决于能否在CTE的递归部分使用GROUP BY，这是不允许的。

这是一个可以加快更新速度的替代方案。它仍然有点迭代，但在我的快速测试中看起来好一点，主要是因为更多基于集合的方法和更少的连接需要。它应该比当前流程更有效地扩展。

DECLARE @D AS INT = ( SELECT MAX (Depth) FROM MyLogic)
WHILE @D > 0 BEGIN
    UPDATE P SET
        Met = C.Met
    FROM MyLogic P
        INNER JOIN (
            SELECT

                ParentLogic,
                CASE
                    WHEN SUM(CASE WHEN Type = 'OR' THEN Met ELSE 0 END) > 0
                        OR SUM(CASE WHEN Type = 'AND' THEN Met ELSE 0 END) = COUNT(*)
                    THEN 1 ELSE 0 END AS Met
            FROM MyLogic
            WHERE Depth = @D
            GROUP BY
                ParentLogic
        ) C -- Only update parents with children
            ON P.LogicID = C.ParentLogic
    WHERE COALESCE(P.Depth, 0) = @D - 1 -- Get next level up
    SET @D = @D - 1
END
SELECT * FROM MyLogic

递归CTE通过多个级别更新父记录

1 个答案: