我有一张将位置映射到邮政编码的表格。例如,纽约州有大约 2000 个邮政编码。我有另一个表将邮件映射到它发送到的邮政编码,但这个表有大约 500 万行。我想找到所有发往纽约州的邮件,这看起来很简单,但查询速度慢得令人难以置信。我什至无法等待足够长的时间来完成它。有500万行的问题吗?我不禁想,现在 500 万对于一台计算机来说不应该是这么大的数字……哦,一切都被索引了。SQL 只是不是设计用于处理如此大的连接吗?
更新:正如人们所问的,我已经用表定义和我正在使用的查询更新了这个问题。
-- Roughly 70,000 rows
CREATE TABLE `mail_zip` (
`mail_id` int(11) default NULL,
`zip` int(11) default NULL,
KEY `index_mail_zip_on_mail_id` (`mail_id`),
KEY `index_mail_zip_on_zip` (`zip`)
) ENGINE=InnoDB DEFAULT CHARSET=latin1
-- Roughly 5,000,000 rows
CREATE TABLE `geographies` (
`city_id` int(11) default NULL,
`postal_code` int(11) default NULL,
KEY `index_geographies_on_city_id` (`city_id`),
KEY `index_geographies_on_postal_code` (`postal_code`)
) ENGINE=InnoDB DEFAULT CHARSET=latin1
-- Query
select mz.mail_id from mail_zip mz join geographies g on mz.zip = g.postal_code where g.city_id = 36 limit 10;
更新2:好吧,我撒谎了。使用适当的索引,上述查询可以正常工作。问题实际上是 order by 子句。请参阅下面的两个几乎相同的查询:唯一的区别是“order by m.sent_on desc”,它为查询增加了额外的 4 分 30 秒!此外,使用解释,通过使用文件排序添加订单,这一定是减慢速度的原因。但是,sent_on 是有索引的,那为什么不使用索引呢?我一定不能正确地制作索引。
-- Roughly 350,000 rows
CREATE TABLE `mail` (
`id` int(11) NOT NULL auto_increment,
`sent_on` datetime default NULL,
`title` varchar(255) default NULL,
PRIMARY KEY (`id`),
KEY `index_mail_on_sent_on` (`sent_on`),
) ENGINE=InnoDB DEFAULT CHARSET=latin1
-- Runs in 0.19 seconds
-- Query
select distinct(m.id), m.title from mail m join mail_zip mz on mz.mail_id = m.id join geographies g on g.postal_code = mz.zip where g.city_id = 36 limit 10;
+----+-------------+-------+--------+--------------------------------------------------------+---------+---------+----------------------+---------+-----------------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+-------+--------+--------------------------------------------------------+---------+---------+----------------------+---------+-----------------------+
| 1 | SIMPLE | mz | ALL | index_mail_zip_on_com_id,index_mail_zip_on_zip | NULL | NULL | NULL | 5260053 | Using temporary |
| 1 | SIMPLE | m | eq_ref | PRIMARY | PRIMARY | 4 | mz.com_id | 1 | |
| 1 | SIMPLE | g | ref | index_geographies_on_city_id,zip | zip | 5 | mz.zip | 1 | Using where; Distinct |
+----+-------------+-------+--------+--------------------------------------------------------+---------+---------+----------------------+---------+-----------------------+
-- Runs in 4 minutes and 30 seconds
-- Query
select distinct(m.id), m.title from mail m join mail_zip mz on mz.mail_id = m.id join geographies g on g.postal_code = mz.zip where g.city_id = 36 order by m.sent_on desc limit 10;
+----+-------------+-------+--------+--------------------------------------------------------+---------+---------+----------------------+---------+---------------------------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+-------+--------+--------------------------------------------------------+---------+---------+----------------------+---------+---------------------------------+
| 1 | SIMPLE | mz | ALL | index_mail_zip_on_com_id,index_mail_zip_on_zip | NULL | NULL | NULL | 5260053 | Using temporary; Using filesort |
| 1 | SIMPLE | m | eq_ref | PRIMARY | PRIMARY | 4 | mz.com_id | 1 | |
| 1 | SIMPLE | g | ref | index_geographies_on_city_id,zip | zip | 5 | mz.zip | 1 | Using where; Distinct |
+----+-------------+-------+--------+--------------------------------------------------------+---------+---------+----------------------+---------+---------------------------------+