SQL Server CTE查找从一个ID到另一个ID的路径

Question

我在SQL Server 2008 R2数据库中有一个表，该表定义了不同状态之间的转换。

相关表格列示例：

TransitionID | SourceStateID | DestinationStateID | WorkflowID
--------------------------------------------------------------
1            | 17            | 18                 | 3
2            | 17            | 21                 | 3
3            | 17            | 24                 | 3
4            | 17            | 172                | 3
5            | 18            | 17                 | 3
6            | 18            | 24                 | 3
7            | 18            | 31                 | 3
8            | 21            | 17                 | 3
9            | 21            | 20                 | 3
10           | 21            | 141                | 3
11           | 21            | 184                | 3

...等

我的目标是能够定义起始 StateID， 终止 StateID和WorkflowID，并希望在该工作流中找到从起始StateID到终止StateID的逻辑路径。

例如，对于上表，如果我提供的起始StateID为17，结束的StateID为184，而WorkflowID为3，则可能以“路径”为17> 21> 184（或更理想的是，TransitionID 2> TransitionID 11）

好：

• 定义的起始和结束StateID在定义的WorkflowID中始终具有可能的路径

坏处：

• 实际上，几乎每个源/目标StateID都有循环引用（即，可能从SourceStateID 1到DestinationStateID 2，也可能是从SourceStateID 2到DestinationStateID 1的过渡）

• 从几乎任何已定义的开始StateID和结束StateID肯定都有多种可能的路径

我意识到这是我所追求的CTE方式，但是诚然，我发现CTE总体上很难掌握，而且在循环引用是一个有保证的问题时更是如此。

理想的解决方案是选择从起始StateID到结束StateID的最短路径，但是老实说，如果我能得到可以可靠地给我两个状态之间任何有效路径的东西，那么我将是这样。快乐。

那里的任何SQL专家都有您可以指出的方向吗？ 老实说，我什至不知道从哪里开始预防循环问题，例如沿着17> 18> 17> 18> 17> 18的路线前进……

令人不快的更新我提出了这个查询，它在情感上给我带来了伤害（在这篇文章中提供了一些帮助： https : //stackoverflow.com/a/11042012/3253311 ），但似乎可以正常工作。

DECLARE @sourceStatusId INT = 17
DECLARE @destinationStatusId INT = 24
DECLARE @workflowId INT = 3

DECLARE @sourceStatusIdString NVARCHAR(MAX) = CAST(@sourceStatusId AS             NVARCHAR(MAX))
DECLARE @destinationStatusIdString NVARCHAR(MAX) = CAST(@destinationStatusId     AS NVARCHAR(MAX))
DECLARE @workflowIdString NVARCHAR(MAX) = CAST(@workflowId AS NVARCHAR(MAX))

;WITH CTE ([Source], [Destination], [Sentinel]) AS 
(
SELECT
    CAST(t.[Source] AS NVARCHAR(MAX)) AS [Source],
    CAST(t.[Destination] AS NVARCHAR(MAX)) AS [Destination],
    [Sentinel] = CAST([Source] AS NVARCHAR(MAX))
FROM
    Transitions t
WHERE
    CAST([Source] AS NVARCHAR(MAX)) = @sourceStatusIdString AND 
    CAST([WorkflowID] AS NVARCHAR(MAX)) = @workflowIdString

UNION ALL

SELECT
    CAST(CTE.[Destination] AS NVARCHAR(MAX)),
    CAST(t.[Destination] AS NVARCHAR(MAX)),
    CAST([Sentinel] AS NVARCHAR(MAX)) + CAST('|' AS NVARCHAR(MAX)) + CAST(CTE.[Destination] AS NVARCHAR(MAX))
FROM
    CTE
    JOIN Transitions t
        ON CAST(t.[Source] AS NVARCHAR(MAX)) = CAST(CTE.[Destination] AS     NVARCHAR(MAX))
WHERE
    CHARINDEX(CTE.[Destination], Sentinel)=0
)

SELECT TOP 1
[Sentinel]
FROM
CTE
WHERE
LEFT([Sentinel], LEN(@sourceStatusIdString)) = @sourceStatusIdString AND 
RIGHT([Sentinel], LEN(@destinationStatusIdString)) =     @destinationStatusIdString 
ORDER BY
LEN([Sentinel])

Answer 1

与Quassnoi的答案类似，但可防止在第一个元素之后开始的循环引用：

DECLARE @src int = 18, @dst int = 184;
WITH cte  AS 
(Select TransitionId, SourceStateId, DestinationStateID, SourceStateID as StartingState, 0 as Depth
  , cast(TransitionId as varchar(max)) + ','  as IDPath  
  , cast(SourceStateID as varchar(max)) + ',' as States
 FROM Transitions
 WHERE SourceStateId = @src
 UNION ALL
 Select t.TransitionId, t.SourceStateId, t.DestinationStateID, StartingState, cte.Depth + 1
  , cte.IDPath + cast(t.TransitionId as varchar(max)) + ',' 
  , cte.States + cast(t.SourceStateID as varchar(max)) + ','
 FROM cte JOIN Transitions t on cte.DestinationStateID = t.SourceStateId 
 and CHARINDEX(',' + cast(t.SourceStateID as varchar(max)) + ',', States) = 0 --prevent loop starting after first element
 and t.DestinationStateID <> StartingState --prevent loop starting with first element
 where cte.Depth < 10 -- prevent going too deep
 )
 select TransitionId, SourceStateId, DestinationStateID, Depth, left(IDPath, len(IDPath) - 1) IDPath, left(States, len(States) - 1) States
 from cte
 where DestinationStateID = @dst
 order by Depth

Answer 2

WITH    q (id, src, dst, sid, cnt, path) AS
        (
        SELECT  transitionId, sourceStateId, destinationStateId, sourceStateId, 1,
                CAST(transitionId AS VARCHAR(MAX)) path
        FROM    mytable
        WHERE   sourceStateId = 18
        UNION ALL
        SELECT  m.transitionId, m.sourceStateId, m.destinationStateId,
                CASE WHEN sid < sourceStateId THEN sid ELSE sourceStateId END, cnt + 1,
                path + ', ' + CAST(transitionId AS VARCHAR(MAX))
        FROM    q
        JOIN    mytable m
        ON      m.sourceStateId = q.dst
                AND m.sourceStateId <> q.sid
        )
SELECT  TOP 1
        *
FROM    q
WHERE   dst = 184
ORDER BY
        cnt DESC

看到小提琴： http ://www.sqlfiddle.com/#!6 / 9342e /17

Answer 3

这是我的看法...在Quassnoi的基础上，但比Kateract短;-)

http://www.sqlfiddle.com/#!6/322d3/4

WITH data AS (
    SELECT  TransitionId, SourceStateId, DestinationStateId,
        '|' + CAST(transitionId AS VARCHAR(MAX)) as Path
    FROM    States
    WHERE   sourceStateId = 17
    AND WorkflowID = 3
    UNION ALL
    SELECT  m.TransitionId, m.sourceStateId, m.DestinationStateId,
        Path + '|' + CAST(m.TransitionId AS VARCHAR(MAX))
    FROM data d
        JOIN States m ON
            m.sourceStateId = d.DestinationStateId
        AND CHARINDEX( '|' + convert(varchar(10),m.TransitionID) + '|', path) = 0
        AND WorkflowID = 3
)
SELECT TOP 1 *
FROM data
WHERE DestinationStateId = 184
ORDER BY LEN(Path)
-- If you want the longest path... ORDER BY LEN(Path) DESC