1 DELIMITER $$
 2
 3 DROP PROCEDURE IF EXISTS `testMERGE`.`unionSelect`$$
 4 CREATE PROCEDURE `testMERGE`.`unionSelect` (IN startTime CHAR(16), IN endTime CHAR(16))
 5 BEGIN
 6         /* 表test1结果集的总数，结果集开始ID， 结束ID */
 7         DECLARE t1_count BIGINT DEFAULT 0;    
 8         DECLARE t1_startId BIGINT DEFAULT 0;  
 9         DECLARE t1_endId BIGINT DEFAULT 0;            
10         DECLARE t2_count BIGINT DEFAULT 0;
11         DECLARE t2_startId BIGINT DEFAULT 0;  
12         DECLARE t2_endId BIGINT DEFAULT 0;
13
14         /* 先确定两个结果集的开始、结束ID，判断小集合; */
15         SELECT object_id INTO t1_startId FROM t1 WHERE date_format(updated, ‘%Y-%m-%d’) >= startTime ORDER BY object_id LIMIT 1;
16         SELECT object_id INTO t1_endId FROM t1 WHERE date_format(updated, ‘%Y-%m-%d’) <= endTime ORDER BY object_id DESC LIMIT 1;
17         SET t1_count = t1_endId – t1_startId;
18
19         SELECT object_id INTO t2_startId FROM t2 WHERE date_format(updated, ‘%Y-%m-%d’) >= startTime ORDER BY object_id LIMIT 1;
20         SELECT object_id INTO t2_endId FROM t2 WHERE date_format(updated, ‘%Y-%m-%d’) <= endTime ORDER BY object_id DESC LIMIT 1;
21         SET t2_count = t2_endId – t2_startId;
22
23         /* 小集合驱动大集合;  利用主键object_id作为搜索条件！ */
24         IF t1_count < t2_count THEN
25                 SELECT * FROM t1 WHERE object_id BETWEEN t1_startId AND (t1_endId+1)
26                         UNION ALL SELECT * FROM t2 WHERE object_id BETWEEN t2_startId AND (t2_endId+1)
27                         ORDER BY updated DESC LIMIT 0,50;
28         ELSE
29                 SELECT * FROM t2 WHERE object_id BETWEEN t2_startId AND (t2_endId+1)
30                         UNION ALL SELECT * FROM t1 WHERE object_id BETWEEN t1_startId AND (t1_endId+1)
31                         ORDER BY updated DESC LIMIT 0,50;
32         END IF;
33
34 END$$
35
36 DELIMITER ;

mysql> call unionSelect(’2009-10-01′, ’2009-10-31′);
 ——->
Query OK, 0 rows affected (0.97 sec)
多次执行，执行时间为维持在1s左右; 执行时间是原来的一半不到;

LD不是DBA，不过开发中消耗资源比较大的SELECT语句，尽量使用索引，尤其是唯一索引。