我正在使用唯一列对连接进行索引扫描;它声称正在检查大量行,即使它只查找一行。
这是查询:
select t.id,
t.twitter_id,
t.screen_name,
t.text
from tweets t
inner join twitter_handle th on th.handle = t.screen_name
order by t.created_at desc
limit 1;
添加/删除限制子句不会更改查询计划。我希望它会在tweet的created_at索引上扫描多个等于limit子句中的行的行,然后对twitter_handle进行eq_ref查找。
然而,根据解释的查询计划是:
+----+-------------+-------+-------+---------------+-------------+---------+------+--------+----------------------------------------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+-------+-------+---------------+-------------+---------+------+--------+----------------------------------------------+
| 1 | SIMPLE | th | index | NULL | handle | 32 | NULL | 100126 | Using index; Using temporary; Using filesort |
| 1 | SIMPLE | t | ref | screen_name | screen_name | 17 | func | 2 | Using where |
+----+-------------+-------+-------+---------------+-------------+---------+------+--------+----------------------------------------------+
请注意索引扫描的 100126 行数和连接顺序中第二个表的ref = func 。
这个查询出现在我的慢查询日志中,我对于为什么mysql选择以这种方式执行查询感到非常困惑。
这两个表的架构:
CREATE TABLE `twitter_handle` (
`handle_id` int(11) NOT NULL AUTO_INCREMENT,
`handle` varchar(30) CHARACTER SET ascii NOT NULL,
`twitter_token_id` int(11) DEFAULT NULL,
`name` varchar(255) CHARACTER SET utf8 DEFAULT NULL,
`twitter_user_id` int(11) unsigned DEFAULT NULL,
`location` varchar(100) CHARACTER SET utf8 DEFAULT NULL,
`profile_image_url` varchar(255) CHARACTER SET utf8 DEFAULT NULL,
`followers_count` int(11) DEFAULT NULL,
`twitter_list_id` int(4) DEFAULT NULL,
`last_update` timestamp NULL DEFAULT CURRENT_TIMESTAMP,
`bio` varchar(160) CHARACTER SET utf8 DEFAULT NULL,
PRIMARY KEY (`handle_id`),
UNIQUE KEY `handle` (`handle`),
KEY `twitter_token_id` (`twitter_token_id`),
KEY `twitter_user_id` (`twitter_user_id`)
) ENGINE=InnoDB;
CREATE TABLE `tweets` (
`id` int(11) NOT NULL AUTO_INCREMENT,
`twitter_id` char(15) DEFAULT NULL,
`screen_name` varchar(15) NOT NULL,
`logged_at` timestamp NOT NULL DEFAULT CURRENT_TIMESTAMP,
`text` char(200) NOT NULL,
`created_at` timestamp NOT NULL DEFAULT '0000-00-00 00:00:00',
`status` enum('pending','processed','ignored','pending_delete','deleted','pending_tweet','preview') NOT NULL DEFAULT 'pending',
`interaction_id` int(11) DEFAULT NULL,
PRIMARY KEY (`id`),
UNIQUE KEY `twitter_id_UNIQUE` (`twitter_id`),
UNIQUE KEY `interaction_id_idx` (`interaction_id`),
UNIQUE KEY `interaction_id` (`interaction_id`,`status`),
KEY `screen_name` (`screen_name`,`created_at`),
KEY `status_2` (`status`,`created_at`),
KEY `created_at_2` (`created_at`)
) ENGINE=InnoDB;
答案 0 :(得分:0)
原因是twitter_handle中的句柄是charset ascii,但推文中的screen_name是latin1!