删除SQL中一对元素的顶行

我在数据库中有一个表，它将两个对象

id（a，B）

配对，每对对象都有一个标志

（标志1）

ObjectA和ObjectB对是主键

表示例：

Object_A | Object_B | Flag_1 809790 809144 1 814231 809144 1 809790 510127 1 809022 809144 1 809790 803745 1 801250 782873 1 809022 623154 1 809022 911310 1 809022 573158 1 909022 573158 1 示例表中的结果：

Object_A 809790 809022 Object_A | Object_B | Flag_1 809790 809144 1 809790 803745 1 809790 510127 1 809022 911310 1 809022 809144 1 809022 623154 1 809022 573158 1 Object_A | Object_B | Flag_1 809790 809144 1 809790 803745 1 809790 510127 0 809022 911310 1 809022 809144 1 809022 623154 0 809022 573158 0 示例表中的结果：

Object_A 809790 809022 Object_A | Object_B | Flag_1 809790 809144 1 809790 803745 1 809790 510127 1 809022 911310 1 809022 809144 1 809022 623154 1 809022 573158 1 Object_A | Object_B | Flag_1 809790 809144 1 809790 803745 1 809790 510127 0 809022 911310 1 809022 809144 1 809022 623154 0 809022 573158 0 这有点复杂，所以我希望我对这个问题解释得足够好。我可能会放弃，在

C#

中以编程方式完成它，但我仍然想知道是否可以直接在SQL server中完成。也许创建临时表？不知道。

你可以用这个：

update t1
set t1.Flag_1 = 0
from table1 t1
inner join 
      (select t2.Object_A, t2.Object_B, 
             ROW_NUMBER() OVER(PARTITION BY Object_A ORDER BY Object_A, Object_B DESC) AS RowNo
       from table1 t2             
                ) t2 on t1.Object_B = t2.Object_B and t1.Object_A = t2.Object_A
where t2.RowNo > 2

MS SQL Server 2012架构设置：

CREATE TABLE Table1
    ([Object_A] int, [Object_B] int, [Flag_1] int)
;

INSERT INTO Table1
    ([Object_A], [Object_B], [Flag_1])
VALUES
    (809790, 809144, 1),
    (814231, 809144, 1),
    (809790, 510127, 1),
    (809022, 809144, 1),
    (809790, 803745, 1),
    (801250, 782873, 1),
    (809022, 623154, 1),
    (809022, 911310, 1),
    (809022, 573158, 1)
;

update t1
set t1.Flag_1 = 0
from table1 t1
inner join (SELECT t2.Object_A, t2.Object_B, ROW_NUMBER() OVER(PARTITION BY Object_A ORDER BY Object_A, Object_B DESC) AS RowNo
              FROM table1 t2              
            ) t2 on t1.Object_B = t2.Object_B and t1.Object_A = t2.Object_A
where t2.RowNo > 2

select * 
from table1
order by 1,2

| OBJECT_A | OBJECT_B | FLAG_1 |
|----------|----------|--------|
|   801250 |   782873 |      1 |
|   809022 |   573158 |      0 |
|   809022 |   623154 |      0 |
|   809022 |   809144 |      1 |
|   809022 |   911310 |      1 |
|   809790 |   510127 |      0 |
|   809790 |   803745 |      1 |
|   809790 |   809144 |      1 |
|   814231 |   809144 |      1 |

查询1：

CREATE TABLE Table1
    ([Object_A] int, [Object_B] int, [Flag_1] int)
;

INSERT INTO Table1
    ([Object_A], [Object_B], [Flag_1])
VALUES
    (809790, 809144, 1),
    (814231, 809144, 1),
    (809790, 510127, 1),
    (809022, 809144, 1),
    (809790, 803745, 1),
    (801250, 782873, 1),
    (809022, 623154, 1),
    (809022, 911310, 1),
    (809022, 573158, 1)
;

update t1
set t1.Flag_1 = 0
from table1 t1
inner join (SELECT t2.Object_A, t2.Object_B, ROW_NUMBER() OVER(PARTITION BY Object_A ORDER BY Object_A, Object_B DESC) AS RowNo
              FROM table1 t2              
            ) t2 on t1.Object_B = t2.Object_B and t1.Object_A = t2.Object_A
where t2.RowNo > 2

select * 
from table1
order by 1,2

| OBJECT_A | OBJECT_B | FLAG_1 |
|----------|----------|--------|
|   801250 |   782873 |      1 |
|   809022 |   573158 |      0 |
|   809022 |   623154 |      0 |
|   809022 |   809144 |      1 |
|   809022 |   911310 |      1 |
|   809790 |   510127 |      0 |
|   809790 |   803745 |      1 |
|   809790 |   809144 |      1 |
|   814231 |   809144 |      1 |

：

CREATE TABLE Table1
    ([Object_A] int, [Object_B] int, [Flag_1] int)
;

INSERT INTO Table1
    ([Object_A], [Object_B], [Flag_1])
VALUES
    (809790, 809144, 1),
    (814231, 809144, 1),
    (809790, 510127, 1),
    (809022, 809144, 1),
    (809790, 803745, 1),
    (801250, 782873, 1),
    (809022, 623154, 1),
    (809022, 911310, 1),
    (809022, 573158, 1)
;

update t1
set t1.Flag_1 = 0
from table1 t1
inner join (SELECT t2.Object_A, t2.Object_B, ROW_NUMBER() OVER(PARTITION BY Object_A ORDER BY Object_A, Object_B DESC) AS RowNo
              FROM table1 t2              
            ) t2 on t1.Object_B = t2.Object_B and t1.Object_A = t2.Object_A
where t2.RowNo > 2

select * 
from table1
order by 1,2

| OBJECT_A | OBJECT_B | FLAG_1 |
|----------|----------|--------|
|   801250 |   782873 |      1 |
|   809022 |   573158 |      0 |
|   809022 |   623154 |      0 |
|   809022 |   809144 |      1 |
|   809022 |   911310 |      1 |
|   809790 |   510127 |      0 |
|   809790 |   803745 |      1 |
|   809790 |   809144 |      1 |
|   814231 |   809144 |      1 |

---

示例数据中是否缺少此值：

809022 573158

，因为它在结果中？是的，抱歉。修正了。后来我添加了这一点，以显示每对上可能有3行以上的行，我想更改前两行上每对上的标志。尝试在结果集中创建一个

row\u number（）

，并更新

flag\u 1

的值，其中row\u number>2？@NickyvV没有想到这一点。我可以使用“row_number（）OVER（按对象划分，按对象排序）作为行”，然后更新行数大于2的每一对。@Kesty已经更新了我的答案。我认为这是正确的方法，但可能仍然存在一些问题。我的数据库中有4万行。如果我对计数（*）大于2的对象进行分组，我会得到4k的结果。我得到的排号>2的版本20k@Kesty但是GROUPBY将合并这些行，因此如果结果有4行，它们将被滚动为1行。其中，行号将是重复行的列表，因为它们具有不同的行号。是的，您是100%正确的。我盯着这个问题看得太久了，需要休息一下：）谢谢你的帮助。我可以做更新，但是更新中你的答案有问题。对象b不是唯一的，因为您可以看到id 809144与多个对象A成对出现。因此，您的更新将使用select中的objectB更新每一对。您可以在新字段中看到错误。我所做的是在表上添加一个自动递增索引，并对索引而不是ObjectB进行更新。也许有更好的选择。我们能解决这个问题吗，这样我就不必拒绝了？如果你加入

t1.Object\u B=t2.Object\u B和t1.Object\u A=t2.Object\u A

像这样：