Question

我有这样的数据：

+--------------+------------+----------+
| name  |  id  |    data    |   date   |
+-------+------+-----------------------+
| host1 |   1  |   data1    |07/20/2017|
| host1 |   1  |   data2    |07/20/2017|
| host2 |   2  |   data1    |07/20/2017|
| host2 |   3  |   data1    |07/19/2017|
| host3 |   4  |   data1    |07/20/2017|
| host3 |   4  |   data2    |07/20/2017|
| host3 |   4  |   data3    |07/20/2017|
| host3 |   4  |   data4    |07/20/2017|
| host4 |   5  |   data3    |07/20/2017|
| host4 |   6  |   data4    |07/17/2017|
+-------+------+------------+----------+

我需要能够从表中删除一些陈旧的数据。每个对象的数据都有多行。但是我需要能够搜索表并查找“名称”匹配但“ID”不匹配的实例。然后删除旧记录。在上面的数据示例中，我想delete from tableName where ID = '3' or '6'

我已经能够使用以下查询来隔离哪些不匹配，但我无法弄清楚如何逐步执行（特别是如果有多个'名称'需要删除）

SELECT * 
FROM   tableName
WHERE  NAME IN (SELECT T1.NAME 
                     FROM   aid2245.tableName T1 
                            INNER JOIN aid2245.tableName T2 
                                    ON T1.NAME = T2.NAME 
                     WHERE  T1.ID <> T2.ID)

任何帮助将不胜感激。

谢谢！

Answer 1

删除3和6。

    create table tableName
    (id int,
    name varchar(50),
    [date] date)

    GO

-- Sample data

    insert into tableName
    (name, id, [date])
    values('host1' ,   1   , '07/20/2017'),
    ( 'host1' ,   1   , '07/20/2017'),
    (  'host2' ,   2  ,   '07/20/2017'),
    ( 'host2' ,   3  ,    '07/19/2017'),
    ( 'host3' ,   4  ,    '07/20/2017'),
    (  'host3' ,   4  ,   '07/20/2017'),
    (  'host3' ,   4  ,   '07/20/2017'),
    (  'host3' ,   4  ,   '07/20/2017'),
    (  'host4' ,   5  ,   '07/20/2017'),
    (  'host4' ,   6  ,   '07/17/2017')

    GO

-- Look at what we are deleting first.

    SELECT b.* 
    FROM   (SELECT name, max(date) maxDate
            FROM tableName
            GROUP BY name) AS a
    JOIN tableName AS b
    ON a.name = b.name
    AND b.date < a.maxDate
    AND b.Id not in
    (SELECT max(id)
        FROM tableName
        GROUP BY [date]
        HAVING date = maxDate)

    GO

-- delete the older rows

    DELETE b
    FROM   (SELECT name, max(date) maxDate
            FROM tableName
            GROUP BY name) AS a
    JOIN tableName AS b
    ON a.name = b.name
    AND b.date < a.maxDate
    AND b.Id not in
        (SELECT max(id)
        FROM tableName
        GROUP BY [date]
        HAVING date = maxDate)

    GO

Answer 2

以下应该可以解决问题。

IF OBJECT_ID('tempdb..#TestData', 'U') IS NULL
BEGIN   -- DROP TABLE #TestData
    CREATE TABLE #TestData (
        [name] CHAR(5) NOT NULL,
        id INT NOT NULL,
        [data] CHAR(5) NOT NULL,
        [date] DATE NOT NULL 
        );

    INSERT #TestData (name, id, data, date) VALUES
        ('host1', 1, 'data1', '07/20/2017'),
        ('host1', 1, 'data2', '07/20/2017'),
        ('host2', 2, 'data1', '07/20/2017'),
        ('host2', 3, 'data1', '07/19/2017'),
        ('host3', 4, 'data1', '07/20/2017'),
        ('host3', 4, 'data2', '07/20/2017'),
        ('host3', 4, 'data3', '07/20/2017'),
        ('host3', 4, 'data4', '07/20/2017'),
        ('host4', 5, 'data3', '07/20/2017'),
        ('host4', 6, 'data4', '07/17/2017');
END;

--==========================================================

-- 1) View before DELETE...
SELECT
    *,
    NeedsDelete = CASE WHEN td.id = LAG(td.id, 1, td.id) OVER (PARTITION BY td.name ORDER BY td.date) THEN '' ELSE 'Delete' END 
FROM
    #TestData td
ORDER BY 
    td.name,
    td.date DESC;

-- 2) Do the actual DELETE...
WITH 
    cte_FindDelete AS (
        SELECT
            *,
            NeedsDelete = CASE WHEN td.id = LAG(td.id, 1, td.id) OVER (PARTITION BY td.name ORDER BY td.date) THEN '' ELSE 'Delete' END 
        FROM
            #TestData td
        )
DELETE fd
FROM 
    cte_FindDelete fd
WHERE 
    fd.NeedsDelete = 'Delete';

HTH，杰森

Answer 3

我会使用窗口函数执行此操作：

with todelete as (
      select t.*,
             min(id) over (partition by name) as minid,
             max(id) over (partition by name) as maxid,
             row_number() over (partition by name order by date desc) as seqnum
      from tablename t
     )
delete from todelete
    where minid <> maxid and seqnum > 1;

查找COL1匹配但COL2不匹配的行并删除最旧的条目

3 个答案: