Question

我正在尝试编写一个查询来删除下表（valid_columns）中的重复记录，并仅保留具有最低[订单]编号的记录。

例如，在下表中，我想删除重复的行，区域2,3和作业3，并使记录尽可能低[订单]。

E.g。输入表，valid_columns如下所示：

name    col_order
-------------
job     1   
job     3   
status  2   
cust    2   
county  1   
state   1   
region  1
region  2
region  3
so      4

期望的输出：

name    col_order
-------------
job     1       
status  2   
cust    2   
county  1   
state   1   
region  1
so      4

我正在尝试修复错误，但我无法弄清楚SQL。目前它使用删除语句和子查询。目前使用的查询如下所示：

- 3）删除重复的列

DELETE 
FROM valid_columns
WHERE   NOT ( col_order = ( SELECT  TOP 1 col_order 
            FROM    valid_columns   firstValid
            WHERE   name = firstValid.name
            AND col_order = firstValid.col_order
            ORDER BY col_order ASC ))

但是，这只会返回以下内容，这是不正确的：

name    col_order
-------------
job     1
county  1
state   1
region  1

非常感谢

Answer 1

-- Test table
declare @T table(Name varchar(10), col_order int)

-- Sample data
insert into @T
select 'job',     1 union all
select 'job',     3 union all
select 'status',  2 union all
select 'cust',    2 union all
select 'county',  1 union all
select 'state',   1 union all
select 'region',  1 union all
select 'region',  2 union all
select 'region',  3 union all
select 'so',      4

-- Delete using CTE and row_number()
;with cte as
(
  select row_number() over(partition by Name order by col_order) as rn
  from @T
)
delete from cte
where rn > 1

-- Result
select *
from @T

或使用子查询而不是CTE

delete vc
from (select row_number() over(partition by Name order by col_order) as rn
      from valid_columns) as vc
where vc.rn > 1

Answer 2

DELETE FROM t1
FROM valid_columns t1
WHERE col_order > 
    (SELECT MIN(col_order) from valid_columns t2 WHERE t1.name = t2.name)

修改可简化为：

DELETE FROM valid_columns WHERE col_order > (SELECT MIN(col_order) from valid_columns t2 WHERE valid_columns.name = t2.name)

DELETE语句可以使用FROM子句根据第二个表中相关记录的值删除记录。在这种情况下，实际上并不需要FROM（我有时使用FROM来为表名设置别名，因为我不喜欢额外的输入。）

DELETE FROM TableA FROM TableA JOIN TableB On TableA.CriteriaA = TableB.CriteriaA

你也可以尝试这个例子（如果你必须这么做的话可能会更快）：

DELETE FROM valid_columns WHERE EXISTS (SELECT * FROM valid_columns t1 WHERE t1.name = valid_columns.name AND valid_columns.col_order > t1.col_order);

Answer 3

试试这个（你可以用select取代删除，以确保在删除之前得到正确的结果）。

DELETE FROM [valid_columns] t1
WHERE col_order > (SELECT MIN(col_order) from [valid_columns] t2 
        WHERE t1.name = t2.name)

Answer 4

这应该做你需要的：

DELETE FROM valid_columns a
WHERE (SELECT MAX(col_order)
    FROM valid_columns b
    WHERE a.name = b.name) > a.col_order;

我建议您在测试之前备份数据。

Answer 5

或者您可以使用游标遍历表并插入临时表中项目遇到的第一个值（确保临时表具有为名称列指定的唯一约束）。

编辑：为方便起见，我添加了一段代码段

declare @Ti table(name varchar(10), col_order int);
declare @Tf table(name varchar(10) unique not null, col_order int not null);

declare @name varchar(10);
declare @col_order int;

-- Sample data
insert into @Ti
select 'job',     1 union all
select 'job',     3 union all
select 'status',  2 union all
select 'cust',    2 union all
select 'county',  1 union all
select 'state',   1 union all
select 'region',  1 union all
select 'region',  2 union all
select 'region',  3 union all
select 'so',      4

select * from @Ti

declare i cursor for
    select * from @Ti;

open i;
fetch next from i into @name, @col_order;

while @@FETCH_STATUS = 0
begin
    if not exists( select * from @Tf where name = @name )
    begin
        insert into @Tf(name, col_order)
            select @name, @col_order;
    end

    fetch next from i into @name, @col_order;
end

close i;
deallocate i;

select * from @Tf;

Answer 6

使用二进制校验和删除记录（这适用于任何sql server版本）


CREATE TABLE #t1(ID INT NULL, VALUE VARCHAR(2))
INSERT INTO #t1(ID, VALUE) VALUES (1,'aa')
INSERT INTO #t1(ID, VALUE) VALUES (2,'bb')
INSERT INTO #t1(ID, VALUE) VALUES (1,'aa')
INSERT INTO #t1(ID, VALUE) VALUES (1,'aa')
INSERT INTO #t1(ID, VALUE) VALUES (3,'cc')
INSERT INTO #t1(ID, VALUE) VALUES (3,'cc')
GO

-- BINARY_CHECKSUM():  are columns that we want to compare duplicates for
-- if you want to compare the full row then change BINARY_CHECKSUM() -> BINARY_CHECKSUM(*)

-- for SQL Server 2000+ a loop
-- save checksums and rowcounts for duplicates

SELECT BINARY_CHECKSUM(ID, VALUE) AS ChkSum, COUNT(*) AS Cnt 
INTO #t2 
FROM #t1 
GROUP BY BINARY_CHECKSUM(ID, VALUE) HAVING COUNT(*)>1

DECLARE @ChkSum BIGINT, @rc INT

-- get the first checksum and set the rowcount to the count - 1 
-- because we want to leave one duplicate

SELECT TOP 1 @ChkSum = ChkSum, @rc = Cnt-1 FROM #t2

WHILE EXISTS (SELECT * FROM #t2)
BEGIN    
    -- rowcount is one less than the duplicate rows count
    SET ROWCOUNT @rc
    DELETE FROM #t1 WHERE BINARY_CHECKSUM(ID, VALUE) = @ChkSum 
    -- remove the processed duplicate from the checksum table
    DELETE #t2 WHERE ChkSum = @ChkSum 
    -- select the next duplicate rows to delete
    SELECT TOP 1 @ChkSum = ChkSum, @rc = Cnt-1 FROM #t2    
END 
SET ROWCOUNT 0
GO

SELECT * FROM #t1 

-- for SQL Server 2005+ a cool CTE
;WITH Numbered 
AS 
(
    SELECT ROW_NUMBER() OVER (PARTITION BY ChkSum ORDER BY ChkSum) AS RN, *
    FROM (
             SELECT BINARY_CHECKSUM(ID, VALUE) AS ChkSum
             FROM #t1
         ) t
) 
DELETE FROM Numbered WHERE RN > 1;
GO

SELECT * FROM #t1 

DROP TABLE #t1;
DROP TABLE #t2;

SQL删除子查询，删除重复记录

6 个答案: