前言：
同事的业务场景是，按照cid、author分组，再按照id倒叙，取出前2条记录出来。
oracle里面可以通过row_number() OVER (PARTITION BY cid,author ORDER BY id DESC) 表示根据cid,author分组，在分组内部根据id排序，而此函数计算的值就表示每组内部排序后的顺序编号（组内连续的唯一的)，而mysql数据库就没有这样的统计函数，需要自己写复杂的sql来实现。
1，录入测试数据
USE csdn;
DROP TABLE IF EXISTS test;
CREATE TABLE test (
id INT PRIMARY KEY,
cid INT,
author VARCHAR(30)
) ENGINE=INNODB;
INSERT INTO test VALUES
(1,1,\'test1\'),
(2,1,\'test1\'),
(3,1,\'test2\'),
(4,1,\'test2\'),
(5,1,\'test2\'),
(6,1,\'test3\'),
(7,1,\'test3\'),
(8,1,\'test3\'),
(9,1,\'test3\'),
(10,2,\'test11\'),
(11,2,\'test11\'),
(12,2,\'test22\'),
(13,2,\'test22\'),
(14,2,\'test22\'),
(15,2,\'test33\'),
(16,2,\'test33\'),
(17,2,\'test33\'),
(18,2,\'test33\');
INSERT INTO test VALUES (200,200,\'200test_nagios\');
2，原始的效率比较低下的子查询实现方式
SELECT * FROM test a
WHERE
N>(
SELECT COUNT(*)
FROM test b
WHERE a.cid=b.cid AND a.`author`=b.`author` AND a.id
)ORDER BY cid,author,id DESC;
只要将N换成你要的数字比如2，就表示查询出每个分组的前2条记录，如下所示：
mysql> SELECT * FROM test a
-> WHERE
-> 2>(
-> SELECT COUNT(*)
-> FROM test b
-> WHERE a.cid=b.cid AND a.`author`=b.`author` AND a.id
-> )ORDER BY cid,author,id DESC;
+-----+------+----------------+
| id | cid | author |
+-----+------+----------------+
| 2 | 1 | test1 |
| 1 | 1 | test1 |
| 5 | 1 | test2 |
| 4 | 1 | test2 |
| 9 | 1 | test3 |
| 8 | 1 | test3 |
| 11 | 2 | test11 |
| 10 | 2 | test11 |
| 14 | 2 | test22 |
| 13 | 2 | test22 |
| 18 | 2 | test33 |
| 17 | 2 | test33 |
| 200 | 200 | 200test_nagios |
+-----+------+----------------+
13 ROWS IN SET (0.00 sec)
mysql>
3，使用动态sql来实现
先构造序列号码，引入一个@row来做rownumber
SET @row=0;SET @mid='';SELECT cid, author, @row:=@row+1 rownum FROM test ORDER BY cid, author LIMIT 10;
序列号码已经出来了，再加一个@mid来进行分组，重点在于CASE WHEN @mid = author THEN @row:=@row+1 ELSE @row:=1 END rownum，表示分组的时候会自动从1计数指导这个分组数据遍历结束。
好了，再外面加一层inner JOIN 再对 rownumber 做限制 就可以拿到目标数据了。
SET @row=0;
执行结果如下所示：
mysql> SET @row=0;
QUERY OK, 0 ROWS affected (0.00 sec)
mysql> SET @mid=\'\';
QUERY OK, 0 ROWS affected (0.00 sec)
mysql> SELECT a.*,b.rownum FROM test a
-> INNER JOIN (
-> SELECT cid, author, id, CASE WHEN @mid = author THEN @row:=@row+1 ELSE @row:=1 END rownum, @mid:=author MID
-> FROM test
-> ORDER BY cid,author,id DESC
-> ) b ON b.author=a.author AND b.cid=a.cid AND b.id=a.id WHERE b.rownum<3;
+-----+------+----------------+--------+
| id | cid | author | rownum |
+-----+------+----------------+--------+
| 2 | 1 | test1 | 1 |
| 1 | 1 | test1 | 2 |
| 5 | 1 | test2 | 1 |
| 4 | 1 | test2 | 2 |
| 9 | 1 | test3 | 1 |
| 8 | 1 | test3 | 2 |
| 11 | 2 | test11 | 1 |
| 10 | 2 | test11 | 2 |
| 14 | 2 | test22 | 1 |
| 13 | 2 | test22 | 2 |
| 18 | 2 | test33 | 1 |
| 17 | 2 | test33 | 2 |
| 200 | 200 | 200test_nagios | 1 |
+-----+------+----------------+--------+
13 ROWS IN SET (0.01 sec)
mysql>
参考文章地址：
http://blog.csdn.net/mchdba/article/details/22163223
