Question

我有一个包含很多重复项目列表的表格。我正在研究一个存储过程，将它们合并为一个记录。每个重复项都有许多子表，应该删除这些子表，或者重新键入以指向结果记录。我的表有一个Id，但ReadableIdentifier是我需要进行重复数据删除的列。

Id | ReadableIdentifier | Name        | UpdatedOn
1  | ABC1234            | Product X   | 2014-04-25 16:00:08.000
2  | ABC1234            | Product X   | 2014-04-28 16:00:08.000
3  | ABC1234            | Product X   | 2014-04-21 16:00:08.000
4  | ABDD9945           | Widget R    | 2014-04-25 16:00:08.000
5  | ABDD9945           | Widget R    | 2014-04-25 18:45:08.000

如您所见，记录1-3是重复的，具有不同的ID和更新的日期。 4-5相同。我需要将这些合并到一个记录中，而不是使用最新的UpdatedOn日期记录。

结束目标（不显示子表）：

Id | ReadableIdentifier | Name        | UpdatedOn
2  | ABC1234            | Product X   | 2014-04-28 16:00:08.000
5  | ABDD9945           | Widget R    | 2014-04-25 18:45:08.000

我正在使用CURSOR来执行此操作，但我想知道是否有更好的解决方案。

DECLARE dupeCursor CURSOR 
    FAST_FORWARD
FOR 
    WITH Counts AS (
        SELECT
            COUNT(1) Count,
            ReadableIdentifier
        FROM dbo.Item WITH (NOLOCK)
        WHERE ReadableIdentifier IS NOT NULL
        GROUP BY ReadableIdentifier)
    SELECT
        Counts.Count, 
        Counts.ReadableIdentifier,
        Counts.CompanyId
    FROM
        Counts
    WHERE Counts.Count > 1;

OPEN dupeCursor;
DECLARE @readableId VARCHAR(50);
DECLARE @itemToPersistId INT, @itemToDeleteId INT;
FETCH NEXT FROM dupeCursor INTO @readableId;

WHILE @@FETCH_STATUS = 0
BEGIN

   WITH V AS (
      SELECT Id, ROW_NUMBER() OVER (PARTITION BY ReadableId ORDER BY UpdatedOn DESC) as Row
        FROM dbo.Item WITH (NOLOCK) WHERE ReadableId = @readableId
   )
   SELECT @itemToPersistId = Id
   FROM V
   WHERE V.Row = 1

   CREATE TABLE #itemsToDelete (Id UNIQUEIDENTIFIER)
   INSERT INTO #itemsToDelete
   SELECT Id
   FROM dbo.Item WITH (NOLOCK)
   WHERE ReadableId = @readableId AND Id != @itemToPersistId;

   --UPDATE CHILDREN TABLES
   DELETE FROM dbo.ItemDetails WHERE ItemId IN (SELECT Id FROM #itemsToDelete);

   UPDATE dbo.ItemPurchases SET ItemId = @itemToPersistId 
   WHERE ItemId IN (SELECT Id FROM #itemsToDelete); 

   UPDATE dbo.PurchaseOrders SET ItemId = @itemToPersistId 
   WHERE ItemId IN (SELECT Id FROM #itemsToDelete); 

   DELETE FROM dbo.ItemMetadata WHERE ItemId IN (SELECT Id FROM #itemsToDelete); 

   --delete Duplicated Items
   DELETE FROM dbo.Item WHERE Id IN (SELECT Id FROM #itemsToDelete);

   DROP TABLE #itemsToDelete

   FETCH NEXT FROM dupeCursor INTO @readableId;
END

CLOSE dupeCursor;
DEALLOCATE dupeCursor;

我意识到光标很可能是问题所在，但我不确定如何在不使用光标图的情况下更新所有子表。

Answer 1

好的我没有数据来测试这个子表，但它应该工作：

WITH V
     AS (SELECT *,
                ROW_NUMBER() OVER(PARTITION BY ReadableId ORDER BY UpdatedOn DESC) AS Row
         FROM dbo.Item WITH (NOLOCK))
     SELECT *
     INTO #itemsToDelete
     FROM V;

--UPDATE CHILDREN TABLES
DELETE FROM dbo.ItemDetails
WHERE ItemId IN
(
    SELECT Id
    FROM #itemsToDelete
    WHERE Row > 1
);
UPDATE IP
  SET
      IP.ItemId = itk.ID
FROM dbo.ItemPurchases AS IP
     INNER JOIN #itemsToDelete AS itd ON IP.ItemId = itd.ID
                                         AND itd.Row > 1
     INNER JOIN #itemsToDelete AS itk ON itk.ReadableIdentifier = itd.ReadableIdentifier
                                         AND itk.Row = 1
                                         AND itd.Row > 1;
UPDATE po
  SET
      po.ItemId = itk.ID
FROM dbo.PurchaseOrders AS po
     INNER JOIN #itemsToDelete AS itd ON po.ItemId = itd.ID
                                         AND itd.Row > 1
     INNER JOIN #itemsToDelete AS itk ON itk.ReadableIdentifier = itd.ReadableIdentifier
                                         AND itk.Row = 1
                                         AND itd.Row > 1;
DELETE FROM dbo.ItemMetadata
WHERE ItemId IN
(
    SELECT Id
    FROM #itemsToDelete
    WHERE Row > 1
); 

--delete Duplicated Items
DELETE FROM dbo.Item
WHERE Id IN
(
    SELECT Id
    FROM #itemsToDelete
    WHERE Row > 1
);

删除没有游标

1 个答案: