我分区表被描述为这样的划分:Mysql的,用聚合函数
CREATE TABLE `event_data` (id int(11) NOT NULL AUTO_INCREMENT,
...
Timestamp int(10) unsigned NOT NULL,
...
...
CurrentDate date NOT NULL,
KEY `id_index` (`id`),
KEY `ix_filter` (`Action`,`Location`),
KEY `ix_time` (`Timestamp`)
)ENGINE=MyISAM AUTO_INCREMENT=1176568 DEFAULT CHARSET=latin1
/*!50500 PARTITION BY RANGE COLUMNS(CurrentDate)
(PARTITION p20130106 VALUES LESS THAN ('2013-01-06') ENGINE = MyISAM,
PARTITION p20130113 VALUES LESS THAN ('2013-01-13') ENGINE = MyISAM,
PARTITION p20130120 VALUES LESS THAN ('2013-01-20') ENGINE = MyISAM) */
我试图执行以下查询:
explain partitions select min(Timestamp) from event_data where CurrentDate < "2013-01-06";
+----+-------------+------------+------------+------+---------------+------+---------+------+--------+-------------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+------------+------------+------+---------------+------+---------+------+--------+-------------+
| 1 | SIMPLE | event_data | p20130106 | ALL | NULL | NULL | NULL | NULL | 512983 | Using where |
+----+-------------+------------+------------+------+---------------+------+---------+------+--------+-------------+
而且
explain partitions select min(Timestamp) from event_data;
+----+-------------+-------+------------+------+---------------+------+---------+------+------+------------------------------+
| id | select_type | table | partitions | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+-------+------------+------+---------------+------+---------+------+------+------------------------------+
| 1 | SIMPLE | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | Select tables optimized away |
+----+-------------+-------+------------+------+---------------+------+---------+------+------+------------------------------+
看来,如果没有指定分区查询更快(我知道的最小总是在第一个分区)
select min(Timestamp) from event_data where CurrentDate < "2013-01-06";
+----------------+
| min(Timestamp) |
+----------------+
| 1321747200 |
+----------------+
1 row in set (0.16 sec)
而且
select min(Timestamp) from event_data ;
+----------------+
| min(Timestamp) |
+----------------+
| 1321747200 |
+----------------+
1 row in set (0.00 sec)
指定分区的查询不应该更快,因为它只需要在单个分区中查看最小值,而不是在所有分区上查找最小值?
看起来,当指示分区时没有使用时间戳索引,但为什么?我有每个分区文件的MYI文件,我确信索引是为每个这样的文件构建的...
我也知道,索引用于不具有聚合函数(基准测试)的不同选择查询。
UPDATE
我发现这个错误报告http://bugs.mysql.com/bug.php?id=66187,这是有关我的问题。
@ypercube,是它出现在我的表格定义中(附在问题中) – Michael
我没有玩过分区,但答案很可能是这样。在第二个查询中,索引被使用,最小值可以被发现非常快。在第一个查询中,它不使用索引,而是扫描整个(第一个)分区。 –
@ypercube,我也这么认为,但为什么我们有索引每个分区?我以为我们有索引树每个分区文件或至少为什么我有每个分区的MYI文件 – Michael