使用Rank（）更新，Row_Number排除重复值

Question

我有一个类似于下面的数据集。

我需要基于updated_CustomerId列中提供的值来更新基本查找表。基本表与数据集相同，但是没有updated_CustomerId列。

基于以下三列的组合，基本表具有唯一约束的挑战： Current_CustomerID
Order_ID OrderCategory

期望的输出：

更新后，可以将Old_customerId中的任何一个（17360410-Pk 8、21044488-Pk = 9）重新分配给Update_CustomerID。 主键2不会更新，因为这会导致违反唯一性约束，但是它将与上面的8或9中的一个主键一起被删除，具体取决于哪个被更新（重新分配给新的ID）< / strong>

在基表上的所有内容更新之后，我然后从基表中删除未将Current_CustomerID重新分配给updated_CustomerId（如果有）的所有记录

    IF OBJECT_ID('tempdb..#DataSet') IS NOT NULL
    DROP TABLE #DataSet


    IF OBJECT_ID('tempdb..#BaseTable') IS NOT NULL
    DROP TABLE #BaseTable

CREATE TABLE #DataSet
(
    PrimaryKey INT NOT NULL CONSTRAINT [PK_dataset_ID] PRIMARY KEY,
    Current_CustomerID INT NOT NULL,
    Order_ID INT NOT NULL,
    OrderCategory VARCHAR(50) NOT NULL,
    Updated_CustomerId INT NOT NULL
)
INSERT INTO #DataSet (PrimaryKey,   Current_CustomerID, Order_ID,   OrderCategory,  updated_CustomerId)
VALUES
(1, 17395001,   4451784,    'Kitchen',  25693110),
(2, 25693110,   4451784,    'Kitchen',  25693110),

(3, 25693110,   2083059,    'Kitchen',  25693110),
(4, 25693110,   2163679,    'Kitchen',  25693110),
(5, 25693110,   2171466,    'Kitchen',  25693110),

(6, 25693110,   2163679,    'Bathroom', 25693110),
(7, 25693110,   2171466,    'Bathroom', 25693110),

(8, 17360410,   3377931,    'Furniture',    16303984),
(9, 21044488,   3377931,    'Furniture',    16303984),

(10,    1534323,    2641714,    'Furniture',    16303984),
(11,    16303984,   2641726,    'Furniture',    16303984),
(12,    16303984,   2641793,    'Furniture',    16303984),
(13,    16303984,   2641816,    'Furniture',    16303984),

(14,    16303345,   2641816,    'Garden',   16301239),
(15,    12345678,   1239065,    'Medicine', 1075432)

CREATE TABLE #BaseTable
(
    PrimaryKey INT NOT NULL CONSTRAINT [PK_baseTable_ID] PRIMARY KEY,
    CustomerID INT NOT NULL,
    Order_ID INT NOT NULL,
    OrderCategory VARCHAR(50) NOT NULL,
)

CREATE UNIQUE NONCLUSTERED INDEX [IDX_LookUp] ON #BaseTable
(
    CustomerID ASC,
    Order_ID ASC,
    OrderCategory ASC
) ON [PRIMARY]


INSERT INTO #BaseTable (PrimaryKey, CustomerID, Order_ID,   OrderCategory)
VALUES
(1, 17395001,   4451784,    'Kitchen'),
(2, 25693110,   4451784,    'Kitchen'),

(3, 25693110,   2083059,    'Kitchen'),
(4, 25693110,   2163679,    'Kitchen'),
(5, 25693110,   2171466,    'Kitchen'),

(6, 25693110,   2163679,    'Bathroom'),
(7, 25693110,   2171466,    'Bathroom'),

(8, 17360410,   3377931,    'Furniture'),
(9, 21044488,   3377931,    'Furniture'),

(10,    1534323,    2641714,    'Furniture'),
(11,    16303984,   2641726,    'Furniture'),
(12,    16303984,   2641793,    'Furniture'),
(13,    16303984,   2641816,    'Furniture'),

(14,    16303345,   2641816,    'Garden'),
(15,    12345678,   1239065,    'Medicine')

-- select * from #BaseTable
-- select * from #DataSet

    ; with CTE AS (
    select a.*
        ,rank() over (partition by a.updated_CustomerId, a.Order_ID, a.OrderCategory
                                                 order by a.Current_CustomerID) as flag
    from #DataSet a
    )

with CTE AS (
select a.*
    ,rank() over (partition by a.updated_CustomerId, a.Order_ID, a.OrderCategory order by a.Current_CustomerID) as flag
from #DataSet a
)

update b
set CustomerID = a.Updated_CustomerId
from #BaseTable b 
inner join CTE a on b.PrimaryKey = a.PrimaryKey
where flag <> 2

消息2601，第14层，状态1，第82行 无法在具有唯一索引“ IDX_LookUp”的对象“ dbo。＃BaseTable”中插入重复的键行。重复的键值为（25693110，4451784，厨房）。 该声明已终止。

Answer 1

我认为您只想获取#DataTable的row_number，然后根据唯一键删除存在多个的行：

//...

DELETE  bt
FROM    #BaseTable bt
    INNER JOIN (
                SELECT  a.PrimaryKey,
                    a.Updated_CustomerId,
                    a.Order_ID,
                    a.OrderCategory,
                    row = ROW_NUMBER() OVER (PARTITION BY a.Updated_CustomerId, a.Order_ID, a.OrderCategory ORDER BY a.Current_CustomerID)
                FROM    #BaseTable b
                    INNER JOIN #DataSet a
                        ON b.PrimaryKey = a.PrimaryKey
            ) x
        ON bt.PrimaryKey = x.PrimaryKey
        AND x.row > 1