关于去除 部分列相同的行

[SKYingX]

去除表中多列相等的行
已知表A 有五个列，a1,a2,a3,a4,a5
当有多行数据的a1,a2,a3,a4,相等时，取a5最大的那一行

[304的的哥]

[Quote=引用 1 楼 zhuomingwang 的回复:]
select a1,a2,a3,a4,a5
from a ta1
where not exists(select 1 from a ta2
where ta1.a1=ta2.a1
and ta1.a2=ta2.a2
and ta1.a3=ta2.a3
and ta1.a4=ta2.a4
and ta1.a5<ta2.a5)
[/Quote]
--正解

[真的不知道取什么名]

select a1,a2,a3,a4,a5
from a ta1
where not exists(select 1 from a ta2 
                 where ta1.a1=ta2.a1 
                   and ta1.a2=ta2.a2
                   and ta1.a3=ta2.a3
                   and ta1.a4=ta2.a4
                   and ta1.a5<ta2.a5)

[dawugui]

[Quote=引用楼主 skyingx 的回复:]
去除表中多列相等的行
已知表A 有五个列，a1,a2,a3,a4,a5
当有多行数据的a1,a2,a3,a4,相等时，取a5最大的那一行
[/Quote]
select a1,a2,a3,a4,max(a5) a5 from tb group by a1,a2,a3,a4

[njlywy]

[Quote=引用 1 楼 zhuomingwang 的回复:]
SQL code

select a1,a2,a3,a4,a5
from a ta1
where not exists(select 1 from a ta2
where ta1.a1=ta2.a1
and ta1.a2=ta2.a2
and ta1.a3=ta2.a3
……
[/Quote]
+1

[304的的哥]

create table T(
  a1 number(5),
  a2 number(5),
  a3 number(5),
  a4 number(5),
  a5 number(5)
  );
insert into T(a1,a2,a3,a4,a5)
values(101,1,2,3,4);
insert into T(a1,a2,a3,a4,a5)
values(102,2,3,4,6);
insert into T(a1,a2,a3,a4,a5)
values(102,2,3,4,51);
insert into T(a1,a2,a3,a4,a5)
values(103,3,4,5,70);
insert into T(a1,a2,a3,a4,a5)
values(103,3,4,5,7);
insert into T(a1,a2,a3,a4,a5)
values(103,3,4,5,12);
insert into T(a1,a2,a3,a4,a5)
values(104,4,5,6,71);
insert into T(a1,a2,a3,a4,a5)
values(104,4,5,6,98);
insert into T(a1,a2,a3,a4,a5)
values(104,4,5,6,101);
insert into T(a1,a2,a3,a4,a5)
values(104,4,5,6,11);
insert into T(a1,a2,a3,a4,a5)
values(105,5,6,7,78);
insert into T(a1,a2,a3,a4,a5)
values(105,5,6,7,74);
insert into T(a1,a2,a3,a4,a5)
values(105,5,6,7,75);
insert into T(a1,a2,a3,a4,a5)
values(105,5,6,7,55);
insert into T(a1,a2,a3,a4,a5)
values(105,5,6,7,4);

SQL> select * from T;

    A1     A2     A3     A4     A5
------ ------ ------ ------ ------
   101      1      2      3      4
   102      2      3      4      6
   102      2      3      4     51
   103      3      4      5     70
   103      3      4      5      7
   103      3      4      5     12
   104      4      5      6     71
   104      4      5      6     98
   104      4      5      6    101
   104      4      5      6     11
   105      5      6      7     78
   105      5      6      7     74
   105      5      6      7     75
   105      5      6      7     55
   105      5      6      7      4

--方法1：Aspen
SQL> select a1,a2,a3,a4,a5
  2  from T ta1
  3  where not exists(select 1 from T ta2
  4                   where ta1.a1=ta2.a1
  5                     and ta1.a2=ta2.a2
  6                     and ta1.a3=ta2.a3
  7                     and ta1.a4=ta2.a4
  8                     and ta1.a5<ta2.a5);
/*
    A1     A2     A3     A4     A5
------ ------ ------ ------ ------
   101      1      2      3      4
   103      3      4      5     70
   102      2      3      4     51
   105      5      6      7     78
   104      4      5      6    101
   
Executed in 0.016 seconds
*/
--方法2:子查询
SQL> select a1,a2,a3,a4,a5
  2  from T
  3  where (a1,a2,a3,a4,a5) in
  4        (select a1,a2,a3,a4,max(a5)
  5        from T
  6        group by a1,a2,a3,a4);
/*
    A1     A2     A3     A4     A5
------ ------ ------ ------ ------
   101      1      2      3      4
   102      2      3      4     51
   103      3      4      5     70
   104      4      5      6    101
   105      5      6      7     78

Executed in 0.016 seconds
*/

--比较：
/*
总觉得子查询的效率要高点，这里或许是数据量小的缘故吧，这里执行时间都差不多
子查询中：先将a1,a2,a3,a4相同的行分为一组，在取出a5最大的那一行
而且始终在同一张表中进行的
*/
/*
方法1：将一张表看做两张，再将两张表连接起来，然后再将两表中a1,a2,a3,a4相同的数据行进行比较
取出a5最大的一行，如果数据量大的话，这种无休止的比较可不是一种好的选择；
方法2也将一张表当做两张，但没将两张表连接起来，子查询select明显将数据量减小了：只取a5最大的行。
*/

[leoliu0822]

SELECT A.*,ROW_NUMBER() OVER(PARTITION BY a1,a2,a3,a4 ORDER BY a5 DESC）RN
FROM  A;

取RN=1的记录