SQL按字段分组取最大(小)值记录（重复记录）--资料收集未整理

http://raozhiyong11.iteye.com/blog/1559226

SQL Server 按某一字段分组取最大 (小)值所在行的数据
数据如下：
name val memo
a    2   a2(a的第二个值)
a    1   a1--a的第一个值
a    3   a3:a的第三个值
b    1   b1--b的第一个值
b    3   b3:b的第三个值
b    2   b2b2b2b2
b    4   b4b4
b    5   b5b5b5b5b5
*/
-- 创建表并插入数据：
create table tb(name varchar ( 10 ),val int ,memo varchar ( 20 ))
insert intotb values ( ' a ' ,    2,   'a2(a的第二个值) ' )
insert intotb values ( ' a ' ,    1,   'a1--a的第一个值 ' )
insert intotb values ( ' a ' ,    3,   'a3:a的第三个值 ' )
insert intotb values ( ' b ' ,    1,   'b1--b的第一个值 ' )
insert intotb values ( ' b ' ,    3,   'b3:b的第三个值 ' )
insert intotb values ( ' b ' ,    2,   'b2b2b2b2 ' )
insert intotb values ( ' b ' ,    4,   'b4b4 ' )
insert intotb values ( ' b ' ,    5,   'b5b5b5b5b5 ' )
go

-- 一、按name分组取 val最大的值所在行的数据。
-- 方法1：
select a. * from tba where val = ( select max (val) from tb where name = a.name) order by a.name
-- 方法2：
select a. * from tba where notexists ( select 1 from tb where name = a.name and val > a.val)
-- 方法3：
select a. * from tba,( select name,max (val) val from tb group by name) b where a.name = b.name and a.val = b.val order by a.name
-- 方法4：
select a. * from tba inner join( select name , max (val) val from tb group by name) b on a.name = b.name and a.val = b.val order by a.name
-- 方法5
select a.* from tb a where 1 > ( select count ( * ) from tb where name = a.name and val > a.val ) order by a.name
/*
name      val        memo
---------- ----------- --------------------
a         3           a3:a的第三个值
b         5           b5b5b5b5b5
*/

-- 二、按name分组取 val最小的值所在行的数据。
-- 方法1：
select a. * from tba where val = ( select min (val) from tb where name = a.name) order by a.name
-- 方法2：
select a. * from tba where notexists ( select 1 from tb where name = a.name and val < a.val)
-- 方法3：
select a. * from tba,( select name,min (val) val from tb group by name) b where a.name = b.name and a.val = b.val order by a.name
-- 方法4：
select a. * from tba inner join( select name , min (val) val from tb group by name) b on a.name = b.name and a.val = b.val order by a.name
-- 方法5
select a.* from tb a where 1 > ( select count ( * ) from tb where name = a.name and val < a.val) order by a.name
/*
name      val        memo
---------- ----------- --------------------
a         1           a1--a的第一个值
b         1           b1--b的第一个值
*/

-- 三、按name分组取第一次出现的行所在的数据。
select a. * from tba where val = ( select top 1 val from tb where name = a.name) order by a.name
/*
name      val        memo
---------- ----------- --------------------
a         2           a2(a的第二个值)
b         1           b1--b的第一个值
*/

-- 四、按name分组随机取一条数据。
select a. * from tba where val = ( select top 1 val from tb where name = a.name order by newid ()) order by a.name
/*
name      val        memo
---------- ----------- --------------------
a         1           a1--a的第一个值
b         5           b5b5b5b5b5
*/

-- 五、按name分组取最小的两个(N个)val
select a.* from tb a where 2 > ( select count ( * ) from tb where name = a.name and val < a.val ) order by a.name,a.val
select a.* from tb a where val in ( select top 2 val from tb where name = a.name order by val) order by a.name,a.val
select a.* from tb a where exists ( select count ( * ) from tb where name = a.name and val < a.val having Count ( * ) < 2 ) order by a.name
/*
name      val        memo
---------- ----------- --------------------
a         1           a1--a的第一个值
a         2           a2(a的第二个值)
b         1           b1--b的第一个值
b         2           b2b2b2b2
*/

--六、按name分组取最大的两个(N个)val
select a.* from tb a where 2 > (select count(*) from tb where name = a.name and val > a.val ) order bya.name,a.val

--提取出每行的 name 和val 去数据库中找比自己小的个数。如果小于2个，代表自己就是满足条件的

select a.* from tb a where val in (select top 2 val from tb where name=a.name order by val desc) order bya.name,a.val

详细说下这个方法

select a.* fromtb a where val in --1段
(select top 2 val from tb where name=a.name order by val) --2段
order by a.name,a.val --3段

思路先看2段。里面是按每个名称排序，然后选重最小的val的值
然后1段是从tb表里面选出符合val值的记录
3段是把提出出来的数据排下序

开始担心之限制val会会出现其他name的值也有相同的情况，然后提起的不是很准确
类似于
select a.* from tb a where val in (1,2,4,3)
开始认为是和这个类似，如果要是和这个类似的话，结果确实不对。
后来仔细思考后得到结果，当2段里面提取val的时候已经限制了名字，也就是说
当name为一个值的时候，只会出项和他对应的2个最小val

select a.* from tb a where exists (select count(*) from tb where name = a.name and val > a.val having Count(*)< 2) order by a.name
/*
name      val        memo
---------- ----------- --------------------
a         2           a2(a的第二个值)
a         3           a3:a的第三个值
b         4           b4b4
b         5           b5b5b5b5b5
*/
--七，如果整行数据有重复，所有的列都相同。
/*
数据如下：
name val memo
a    2   a2(a的第二个值)
a    1   a1--a的第一个值
a    1   a1--a的第一个值
a    3   a3:a的第三个值
a    3   a3:a的第三个值
b    1   b1--b的第一个值
b    3   b3:b的第三个值
b    2   b2b2b2b2
b    4   b4b4
b    5   b5b5b5b5b5
*/
--在sql server 2000中只能用一个临时表来解决，生成一个自增列，先对val取最大或最小，然后再通过自增列来取数据。
--创建表并插入数据：
create table tb(name varchar(10),val int,memo varchar(20))
insert into tb values('a',    2,   'a2(a的第二个值)')
insert into tb values('a',    1,   'a1--a的第一个值')
insert into tb values('a',    1,   'a1--a的第一个值')
insert into tb values('a',    3,   'a3:a的第三个值')
insert into tb values('a',    3,   'a3:a的第三个值')
insert into tb values('b',    1,   'b1--b的第一个值')
insert into tb values('b',    3,   'b3:b的第三个值')
insert into tb values('b',    2,   'b2b2b2b2')
insert into tb values('b',    4,   'b4b4')
insert into tb values('b',    5,   'b5b5b5b5b5')
go

select * , px = identity(int,1,1) into tmp from tb

select m.name,m.val,m.memo from
(
  select t.* from tmp t where val = (select min(val) from tmp where name = t.name)
) m where px = (select min(px) from
(
  select t.* from tmp t where val = (select min(val) from tmp where name = t.name)
) n where n.name = m.name)

drop table tb,tmp

/*
name      val         memo
---------- ----------- --------------------
a         1           a1--a的第一个值
b         1           b1--b的第一个值

(2 行受影响)
*/
--在sql server 2005中可以使用row_number函数，不需要使用临时表。
--创建表并插入数据：
create table tb(name varchar(10),val int,memo varchar(20))
insert into tb values('a',    2,   'a2(a的第二个值)')
insert into tb values('a',    1,   'a1--a的第一个值')
insert into tb values('a',    1,   'a1--a的第一个值')
insert into tb values('a',    3,   'a3:a的第三个值')
insert into tb values('a',    3,   'a3:a的第三个值')
insert into tb values('b',    1,   'b1--b的第一个值')
insert into tb values('b',    3,   'b3:b的第三个值')
insert into tb values('b',    2,   'b2b2b2b2')
insert into tb values('b',    4,   'b4b4')
insert into tb values('b',    5,   'b5b5b5b5b5')
go

select m.name,m.val,m.memo from
(
  select * , px = row_number() over(order by name ,val) from tb
) m where px = (select min(px) from
(
  select * , px = row_number() over(order by name ,val) from tb
) n where n.name = m.name)

drop table tb

/*
name      val         memo
---------- ----------- --------------------
a         1           a1--a的第一个值
b         1           b1--b的第一个值

(2 行受影响)
*/

我们首先再来确认一下我们要做什么事情. 只有在清楚我们需要做什么事情的情况下, 我们才能得到正确的结论. 我们要根据某一个字段(例如: name), 进行分组, 并在分组的基础上, 选出某列(例如: val)上的最值. 并最终返回name, val值在表tb中所有字段信息.

由于表tb上没有主键, 因此name, val, memo这个三元组, 不一定能从表tb中返回唯一的元组. 由于我们要根据name进行分组, 并选出val上的最值, 我们可以发现name, val的二元组, 也不一定从表tb中返回唯一的元组.

根据"爱新觉罗.毓华"设计的用例(testcase)来看, 并没有考虑当name, val相同时, memo不同时应该返回何种结果. 通过"爱新觉罗.毓华"给出的求解方案, 我们知道无论memo为何值, 利用临时表或row_number都只返回"第一个"元组.

我们已经确认了"爱新觉罗.毓华"这位朋友的文章中第七种问题的情况, 我们除了要完成"爱新觉罗.毓华"已经完成的操作(问题1). 还要实现一个能够对不同memo返回多条结果的SQL语句(问题2). 实际上问题2比问题1要容易得多.

针对问题1来说, "爱新觉罗.毓华"给出2种方法分别利用:

1. 临时表+自增列;

2. row_number函数;

我们是否还能给出第三种方法呢? 在继续讨论问题1之前, 我们先来解决问题2, 问题2更加简单. 我们首先追加一些数据:

insert into tb values('a', 1, '问题2 ADD BY Edengundam')

insert into tb values('b', 1, '问题2 ADD BY Edengundam')

我们只对a, b加上两条数据, 来补充一下测试用例(我们不再讨论NULL值问题).

我们首先根据name字段进行分组, 并取出val中最小值:

SELECT name, MIN(val) AS minval FROM tb GROUP BY name;

输出结果:

name      minval
---------- -----------
a          1
b          1

(2 row(s) affected)

接下来, 我们只需要利用 "相关子查询" 或者 "JOIN" 来选出最小值对应的元组:

SELECT DISTINCT tb.name, tb.val, tb.memo

FROM

tb,

(SELECT name, MIN(val) AS minval FROM tb GROUP BY name) AS tmp

WHERE tb.name = tmp.name AND tb.val = tmp.minval;

输出结果:

name      val         memo
---------- ----------- ----------------------------------------
a         1           a1--a的第一个值
a         1           问题2 ADD BY Edengundam
b         1           b1--b的第一个值
b         1           问题2 ADD BY Edengundam

(4 row(s) affected)

注意, 这里的DISTINCT用来将重复的a, 1, 'a1--a的第一个值'列消去. 现在我们已经完成了问题2, 下面我们需要思考如何能够不依赖临时表和row_number这类函数来针对不同的name, MIN(val)返回一条元组.

我们应该从tb表中选择出name, MIN(val)的列, 并且这一列的memo是通过TOP语句唯一选择的. 这就是我们的思路, 利用相关子查询, TOP语句:

SELECT DISTINCT tb.name, tb.val, tb.memo

FROM

tb,

(SELECT name, MIN(val) AS minval FROM tb GROUP BY name) AS tmp

WHERE tb.name = tmp.name

AND tb.val = tmp.minval

AND tb.memo = (SELECT TOP 1 memo FROM tb AS d WHERE tb.name = d.name AND tb.val = d.val);

我们看看输出结果:

name      val         memo
---------- ----------- ----------------------------------------
a         1           a1--a的第一个值
b         1           b1--b的第一个值

(2 row(s) affected)

我们已经完成任务了, 但是我必须提醒大家效率上来说, 最好的是"爱新觉罗.毓华"利用row_number函数实现的办法. 我的讨论只是帮助大家开头思路, 并且将原有讨论情况进行扩充.

本文转载于：http://hi.baidu.com/lr02/blog/item/d9262e300683ac95a9018eb7.html

问题2：

常用的文章系统，如何用一条sql语句提取出每个分类的第一条并按OrderCol字段排序。分类大概几十个，一共只要取top 10 这样

http://topic.csdn.net/u/20080122/17/8820a852-b106-4096-be23-3878e1bf825d.html?1753015752

方法1：
--取小

select top 10 t.* from tb t where OrderCol = (select min(OrderCol) from tb where 分类 = t.分类) order by OrderCol

--取大

select top 10 t.* from tb t where OrderCol = (select max(OrderCol) from tb where 分类 = t.分类) order by OrderCol desc

方法2：

select * from 文章表 a

where not exists(select 1 from 文章表 where 类别=a.类别 and OrderCol>a.OrderCol)

order byOrderCol desc

含有重复值：

--取小

select top 10 m.* from

(

select t.* from tb t whereOrderCol = (select min(OrderCol) from tb where 分类 = t.分类)

) m where id = (select min(id) from

(

select t.* from tb t whereOrderCol = (select min(OrderCol) from tb where 分类 = t.分类)

) n where m.分类 = n.分类

)

--取大

select top 10 m.* from

(

select t.* from tb t whereOrderCol = (select max(OrderCol) from tb where 分类 = t.分类)

) m where id = (select max(id) from

(

select t.* from tb t where OrderCol= (select max(OrderCol) from tb where 分类 = t.分类)

) n where m.分类 = n.分类

)

方法3：

[code=SQL]declare @t table(type varchar(4),detail varchar(10))

insert @t select

'001','a' union select

'001','sss' union select

'001','gssfa' union select

'002','fdsf' union select

'002','fs' union select

'002','gahas' union select

'003','sdga' union select

'003','hahasg' union select

'003','gdsag'

select a.type,a.detail

from @t a

join @t b

on a.type= b.type

group by a.type,a.detail

having count(case when a.detail <= b.detail then 1 else null end) < = 2 --可动态修改

order by a.type asc

/*

type detail

---- ----------

001  gssfa

001  sss

002  fs

002  gahas

003  hahasg

003  sdga

（所影响的行数为 6 行）

*/

select a.* from @t a

where exists (select count(*)

              from @t

              where type = a.type and detail > a.detail having Count(*) < 2)

order by a.type

/*

type detail

---- ----------

001  gssfa

001  sss

002  fs

002  gahas

003  hahasg

003  sdga

（所影响的行数为 6 行）

*/

select a.* from @t a

where exists (select count(*)

              from @t

              where type = a.type and detail < a.detail having Count(*) < 2)

order by a.type

/*

type detail

---- ----------

001  a

001  gssfa

002  fdsf

002  fs

003  gdsag

003  hahasg

（所影响的行数为 6 行）

*/

select a.*

from @t a

where detail in (select top 2 detail

              from @t where type=a.type

              order by detail desc)

order by a.type,a.detail

/*

type detail

---- ----------

001  gssfa

001  sss

002  fs

002  gahas

003  hahasg

003  sdga

（所影响的行数为 6 行）

*/

select a.*

from @t a

where detail in (select top 2 detail

              from @t where type=a.type

              order by detail asc)

order by a.type,a.detail

/*

type detail

---- ----------

001  a

001  gssfa

002  fdsf

002  fs

003  gdsag

003  hahasg

（所影响的行数为 6 行）

*/

方法4：

select top 10 m.id as id, rtrim(left(title,11)) as titles,Clicks,AddTime,Auther,meno,Classes,s.phone as phone,s.name as name from

(select t.* from tb t where refreshtime = (select max(OrderCol) from tb where userid = t.userid) ) m

inner joinUserTable on s.userid=m.useridwhere m.id = (select max(id) from

(select t.* from tb t where OrderCol = (select max(OrderCol) from tb where isshow=0 and userid =t.userid)) n

where m.userid =n.userid)

order by OrderColdesc

--处理表重复记录(查询和删除)

/******************************************************************************************************************************************************

1、Num、Name相同的重复值记录,没有大小关系只保留一条

2、Name相同,ID有大小关系时,保留大或小其中一个记录

整理人：中国风(Roy)

日期:2008.06.06

******************************************************************************************************************************************************/

--1、用于查询重复处理记录(如果列没有大小关系时2000用生成自增列和临时表处理,SQL2005用row_number函数处理)

--> --> (Roy)生成測試數據

if not object_id('Tempdb..#T') is null

drop table #T

Create table #T([ID] int,[Name] nvarchar(1),[Memo] nvarchar(2))

Insert #T

select 1,N'A',N'A1' union all

select 2,N'A',N'A2' union all

select 3,N'A',N'A3' union all

select 4,N'B',N'B1' union all

select 5,N'B',N'B2'

--I、Name相同ID最小的记录(推荐用1,2,3),方法3在SQl05时，效率高于1、2

方法1:

Select * from #T a where not exists(select 1 from #T where Name=a.Name and ID<a.ID)

方法2:

select a.* from #T a join (select min(ID)ID,Name from #T group by Name) b on a.Name=b.Name and a.ID=b.ID

方法3:

select * from #T a where ID=(select min(ID) from #T where Name=a.Name)

方法4:

select a.* from #T a join #T b on a.Name=b.Name and a.ID>=b.ID group by a.ID,a.Name,a.Memo having count(1)=1

方法5:

select * from #T a group by ID,Name,Memo having ID=(select min(ID)from #T where Name=a.Name)

方法6:

select * from #T a where (select count(1) from #T where Name=a.Name and ID<a.ID)=0

方法7:

select * from #T a where ID=(select top 1 ID from #T where Name=a.name order by ID)

方法8:

select * from #T a where ID!>all(select ID from #T where Name=a.Name)

方法9(注:ID为唯一时可用):

select * from #T a where ID in(select min(ID) from #T group by Name)

--SQL2005:

方法10:

select ID,Name,Memo from (select *,min(ID)over(partition by Name) as MinID from #T a)T where ID=MinID

方法11:

select ID,Name,Memo from (select *,row_number()over(partition by Name order by ID) as MinID from #T a)T where MinID=1

生成结果:

ID Name Memo

----------- ---- ----

1 A A1

4 B B1

(2 行受影响)

--II、Name相同ID最大的记录,与min相反:

方法1:

Select * from #T a where not exists(select 1 from #T where Name=a.Name and ID>a.ID)

方法2:

select a.* from #T a join (select max(ID)ID,Name from #T group by Name) b on a.Name=b.Name and a.ID=b.ID order by ID

方法3:

select * from #T a where ID=(select max(ID) from #T where Name=a.Name) order by ID

方法4:

select a.* from #T a join #T b on a.Name=b.Name and a.ID<=b.ID group by a.ID,a.Name,a.Memo having count(1)=1

方法5:

select * from #T a group by ID,Name,Memo having ID=(select max(ID)from #T where Name=a.Name)

方法6:

select * from #T a where (select count(1) from #T where Name=a.Name and ID>a.ID)=0

方法7:

select * from #T a where ID=(select top 1 ID from #T where Name=a.name order by ID desc)

方法8:

select * from #T a where ID!<all(select ID from #T where Name=a.Name)

方法9(注:ID为唯一时可用):

select * from #T a where ID in(select max(ID) from #T group by Name)

--SQL2005:

方法10:

select ID,Name,Memo from (select *,max(ID)over(partition by Name) as MinID from #T a)T where ID=MinID

方法11:

select ID,Name,Memo from (select *,row_number()over(partition by Name order by ID desc) as MinID from #T a)T where MinID=1

生成结果2:

ID Name Memo

----------- ---- ----

3 A A3

5 B B2

(2 行受影响)

--2、删除重复记录有大小关系时,保留大或小其中一个记录

--> --> (Roy)生成測試數據

if not object_id('Tempdb..#T') is null

drop table #T

Create table #T([ID] int,[Name] nvarchar(1),[Memo] nvarchar(2))

Insert #T

select 1,N'A',N'A1' union all

select 2,N'A',N'A2' union all

select 3,N'A',N'A3' union all

select 4,N'B',N'B1' union all

select 5,N'B',N'B2'

--I、Name相同ID最小的记录(推荐用1,2,3),保留最小一条

方法1:

delete a from #T a where exists(select 1 from #T where Name=a.Name and ID<a.ID)

方法2:

delete a from #T a left join (select min(ID)ID,Name from #T group by Name) b on a.Name=b.Name and a.ID=b.ID where b.Id is null

方法3:

delete a from #T a where ID not in (select min(ID) from #T where Name=a.Name)

方法4(注:ID为唯一时可用):

delete a from #T a where ID not in(select min(ID)from #T group by Name)

方法5:

delete a from #T a where (select count(1) from #T where Name=a.Name and ID<a.ID)>0

方法6:

delete a from #T a where ID<>(select top 1 ID from #T where Name=a.name order by ID)

方法7:

delete a from #T a where ID>any(select ID from #T where Name=a.Name)

select * from #T

生成结果:

ID Name Memo

----------- ---- ----

1 A A1

4 B B1

(2 行受影响)

--II、Name相同ID保留最大的一条记录:

方法1:

delete a from #T a where exists(select 1 from #T where Name=a.Name and ID>a.ID)

方法2:

delete a from #T a left join (select max(ID)ID,Name from #T group by Name) b on a.Name=b.Name and a.ID=b.ID where b.Id is null

方法3:

delete a from #T a where ID not in (select max(ID) from #T where Name=a.Name)

方法4(注:ID为唯一时可用):

delete a from #T a where ID not in(select max(ID)from #T group by Name)

方法5:

delete a from #T a where (select count(1) from #T where Name=a.Name and ID>a.ID)>0

方法6:

delete a from #T a where ID<>(select top 1 ID from #T where Name=a.name order by ID desc)

方法7:

delete a from #T a where ID<any(select ID from #T where Name=a.Name)

select * from #T

ID Name Memo

----------- ---- ----

3 A A3

5 B B2

(2 行受影响)

--3、删除重复记录没有大小关系时，处理重复值

--> --> (Roy)生成測試數據

if not object_id('Tempdb..#T') is null

drop table #T

Create table #T([Num] int,[Name] nvarchar(1))

Insert #T

select 1,N'A' union all

select 2,N'B' union all

select 2,N'B'

方法1:

if object_id('Tempdb..#') is not null

drop table #

Select distinct * into # from #T--排除重复记录结果集生成临时表#

truncate table #T--清空表

insert #T select * from # --把临时表#插入到表#T中

--查看结果

select * from #T

Num Name

----------- ----

1 A

2 B

(2 行受影响)

--重新执行测试数据后用方法2

方法2:

alter table #T add ID int identity--新增标识列

delete a from #T a where exists(select 1 from #T where Num=a.Num and Name=a.Name and ID>a.ID)--只保留一条记录

alter table #T drop column ID--删除标识列

--查看结果

select * from #T

Num Name

----------- ----

1 A

2 B

(2 行受影响)

--重新执行测试数据后用方法3

方法3:

declare Roy_Cursor cursor local for

select count(1)-1,Num,Name from #T group by Num,Name having count(1)>1

declare @con int,@Num int,@Name nvarchar(1)

open Roy_Cursor

fetch next from Roy_Cursor into @con,@Num,@Name

while @@Fetch_status=0

begin

set rowcount @con;

delete #T where Num=@Num and Name=@Name

set rowcount 0;

fetch next from Roy_Cursor into @con,@Num,@Name

end

close Roy_Cursor

deallocate Roy_Cursor

--查看结果

select * from #T

Num Name

----------- ----

1 A

2 B

(2 行受影响)

本站仅提供存储服务，所有内容均由用户发布，如发现有害或侵权内容，请点击举报。