转自：http://www.blogjava.net/dunkbird/archive/2011/01/28/343718.html


表特征：
mysql> select * from test;
+------+------+
| a    | b    |
+------+------+
|    1 |   20 |
|    1 |   21 |
|    1 |   24 |
|    2 |   20 |
|    2 |   32 |
|    2 |   14 |
+------+------+
6 rows in set (0.00 sec)

现在，我们以a分组，查询b列最大的2个值。 这条sql要怎么写了？




1.创建表
Create Table: CREATE TABLE `sam` (
  `a` int(11) DEFAULT NULL,
  `b` int(11) DEFAULT NULL
) ENGINE=MyISAM DEFAULT CHARSET=utf8

2.插入模拟数据
INSERT INTO `sam` VALUES (1,10),(1,15),(1,20),(1,25),(2,20),(2,22),(2,33),(2,45);
+------+------+
| a    | b    |
+------+------+
|    1 |   10 |
|    1 |   15 |
|    1 |   20 |
|    1 |   25 |
|    2 |   20 |
|    2 |   22 |
|    2 |   33 |
|    2 |   45 |
+------+------+

3.SQL实现
select a,b,rownum,rank from
    (select ff.a,ff.b,@rownum:=@rownum+1 rownum,if(@pa=ff.a,@rank:=@rank+1,@rank:=1) as rank,@pa:=ff.a
    FROM
       (select a,b from sam group by a,b order by a asc,b desc) ff,(select @rank:=0,@rownum:=0,@pa=null) tt) result
    having rank <=2;

4.结果：
+------+------+--------+------+
| a    | b    | rownum | rank |
+------+------+--------+------+
|    1 |   25 |      1 |    1 |
|    1 |   20 |      2 |    2 |
|    2 |   45 |      5 |    1 |
|    2 |   33 |      6 |    2 |
+------+------+--------+------+
4 rows in set (0.00 sec)