SQL Server 2008的更新表和開關值查詢優化

Question

我有值

ID   Son  Father 
----------- ---------- ---------- 
1   Mark  Gerard 
2   Gerard  Ivan 
3   Leo  Samuel 
4   Samuel  Johan 
5   Ivan  Carles 

我需要改變的表像這樣的表：

ID   Son  Father 
----------- ---------- ---------- 
1   Mark  Carles 
2   Gerard  Carles 
3   Leo  Johan 
4   Samuel  Johan 
5   Ivan  Carles 

的目標是找到一個主要'Father'和用此值更新所有'Son'記錄。主要'Father'可以不同。

我的代碼是未來：

DECLARE @CNT INT 
DECLARE @CH_1 NVARCHAR(10) 
DECLARE @CH_2 NVARCHAR(10) 

CREATE TABLE #PPL (ID INT, Son NVARCHAR(10), Father NVARCHAR(10)) 

INSERT INTO #PPL VALUES (1, 'Mark', 'Gerard') 
INSERT INTO #PPL VALUES (2, 'Gerard', 'Ivan') 
INSERT INTO #PPL VALUES (3, 'Leo', 'Samuel') 
INSERT INTO #PPL VALUES (4, 'Samuel', 'Johan') 
INSERT INTO #PPL VALUES (5, 'Ivan', 'Carles') 

SET @I = 1 
SET @CNT = (SELECT COUNT(ID) FROM #PPL) 

WHILE @I <= @CNT 
BEGIN 
    SET @J = 1 

     WHILE @J <= @CNT 
     BEGIN 
      SET @CH_1 = (SELECT Son FROM #PPL WHERE ID = @J) 
      SET @CH_2 = (SELECT Father FROM #PPL WHERE ID = @J) 
      UPDATE #PPL SET Father = @CH_2 WHERE Father = @CH_1 
      SET @J = @J + 1 
     END; 

    SET @I = @I + 1 
END; 

SELECT * FROM #PPL 

DROP TABLE #PPL 

此代碼工作正確的，但對於低數量的記錄。這個代碼如何優化？

謝謝！

Answer 1

下面是你如何使用遞歸CTE來做到這一點。

CREATE TABLE #PPL (ID INT, Son NVARCHAR(10), Father NVARCHAR(10)) 

INSERT INTO #PPL VALUES (1, 'Mark', 'Gerard') 
INSERT INTO #PPL VALUES (2, 'Gerard', 'Ivan') 
INSERT INTO #PPL VALUES (3, 'Leo', 'Samuel') 
INSERT INTO #PPL VALUES (4, 'Samuel', 'Johan') 
INSERT INTO #PPL VALUES (5, 'Ivan', 'Carles') 

;WITH CTE_FamilyGenealogy 
AS 
(
    SELECT ID 
      ,Son 
      ,Father 
      ,1 AS [Level] 
    FROM #PPL Ancor 
    UNION ALL 
    SELECT CTE_FamilyGenealogy.ID 
      ,CTE_FamilyGenealogy.Son 
      ,Fathers.Father AS Father 
      ,CTE_FamilyGenealogy.[Level] + 1 AS [Level] 
    FROM #PPL Fathers 
    INNER JOIN CTE_FamilyGenealogy ON CTE_FamilyGenealogy.Father = Fathers.Son 
), 
CTE_MajorFathers 
AS 
(
    SELECT ID 
      ,Son 
      ,Father 
      ,ROW_NUMBER() OVER (PARTITION BY Son ORDER BY [Level] DESC) AS RowRank 
    FROM CTE_FamilyGenealogy 
) 
SELECT ID 
     ,Son 
     ,Father 
FROM CTE_MajorFathers 
WHERE RowRank = 1 
ORDER BY ID 

的遞歸CTE CTE_FamilyGenealogy找到所有的父子組合，並確定家譜中的水平。 CTE使用ROW_NUMBER根據FamilyGenealogy中的等級對可能的組合進行排序以確定主要父親。

Answer 2

請嘗試使用基於遞歸的方法（請參閱遞歸公用表表達式）和HIERARCHYID（SQL2008 +）數據類型。基本思想是爲每一行構建一個從「第一個」父親開始的層次結構值:-)並以「last」子結束:-)。例如：對於第一行(1, 'Mark', 'Gerard')該節點/家族樹是/ 5/2/1 /其中/ 5 /是「第一」父親;-)和/ 1 /是「最後」的兒子。接下來，將這些值轉換爲hiearchyid值，並使用GetLevel和GetAncestor方法計算「第一個」父親：Father1ID：Johan或Carles。

IF OBJECT_ID('tempdb.dbo.#Results') IS NOT NULL 
BEGIN 
    DROP TABLE #Results; 
END 
CREATE TABLE #Results (ID INT NOT NULL PRIMARY KEY, Father1ID INT); 

WITH CteRec 
AS (
    -- It returns Father only rows 
    SELECT l1.ID, l1.Son, l1.Father, CONVERT(VARCHAR(900), '/'+LTRIM(l1.ID)+'/') AS Node -- FamilyTree 
    FROM #PPL AS l1 -- First level 
    WHERE NOT EXISTS(SELECT * FROM #PPL p WHERE p.Son = l1.Father) 
    UNION ALL 
    -- It returns Son only and Son-Father rows 
    SELECT ln.ID, ln.Son, ln.Father, CONVERT(VARCHAR(900), prt.Node+LTRIM(ln.ID)+'/') AS Node -- FamilyTree 
    FROM #PPL AS ln -- Next level 
    JOIN CteRec AS prt ON prt.Son = ln.Father 
) 
INSERT #Results (ID, Father1ID) 
SELECT ID, 
     Father1ID = CONVERT(INT,REPLACE(CONVERT(HIERARCHYID, Node).GetAncestor(CONVERT(HIERARCHYID, Node).GetLevel()-1).ToString(),'/','')) 
FROM CteRec; 

SELECT p.*, r.Father1ID, rp.Father AS Father1Name 
FROM #PPL p 
INNER JOIN #Results r ON p.ID = r.ID 
INNER JOIN #PPL rp ON r.Father1ID = rp.ID 
-- Also you ca use #Result with UPDATE statement but I would store this values within new column Father1 

Answer 3

遞歸CTE的是高估=）

這種簡單的方法將通過一樣快（在正常數據運行），永遠不會抱怨最大遞歸和易於閱讀。我能看到的唯一缺點是，當數據被破壞時，它可能會進入一個永恆的循環。

CREATE TABLE #PPL (ID INT, Son NVARCHAR(10), Father NVARCHAR(10)) 

INSERT INTO #PPL VALUES (1, 'Mark', 'Gerard') 
INSERT INTO #PPL VALUES (2, 'Gerard', 'Ivan') 
INSERT INTO #PPL VALUES (3, 'Leo', 'Samuel') 
INSERT INTO #PPL VALUES (4, 'Samuel', 'Johan') 
INSERT INTO #PPL VALUES (5, 'Ivan', 'Carles') 

DECLARE @rowcount int = -1 
WHILE @rowcount <> 0 
    BEGIN 
     UPDATE upd 
      SET Father = new.Father 
     FROM #PPL upd 
     JOIN #PPL new 
      ON new.Son = upd.Father 
     WHERE upd.Father <> new.Father 

     SELECT @rowcount = @@ROWCOUNT 
    END 

SELECT * FROM #PPL 

PS：在大型數據集上運行時，可能有助於在子列上創建索引。