我正在编写一个脚本来对用户的消息进行部分单词搜索。每个对话都有一个 mail_id,每个单独的消息都有一个 msg_id。
我有一个表 mail_word_index,其中包含消息中每个单词的一行。
CREATE TABLE IF NOT EXISTS `mail_word_index` (
`id` int(10) unsigned NOT NULL AUTO_INCREMENT,
`sender_id` int(10) unsigned NOT NULL DEFAULT '0',
`dest_id` int(10) unsigned NOT NULL DEFAULT '0',
`mail_id` int(10) unsigned NOT NULL DEFAULT '0',
`msg_id` int(10) unsigned NOT NULL DEFAULT '0',
`word` varchar(15) NOT NULL DEFAULT '',
PRIMARY KEY (`id`),
KEY `dest_id` (`dest_id`,`word`),
KEY `sender_id` (`sender_id`,`word`),
KEY `multiple_words` (`mail_id`,`msg_id`,`word`)
) ENGINE=MyISAM ;
我有一个查询需要 0.01 秒才能完成
SELECT DISTINCT w1.mail_id FROM mail_word_index AS w1,
mail_word_index AS w2
WHERE w1.sender_id=1
AND w1.word LIKE 'str%'
AND w1.mail_id=w2.mail_id
AND w1.msg_id=w2.msg_id
AND w2.word LIKE 'con%' LIMIT 20
但是,一次搜索一个单词只需要 0.002 秒即可完成每个单词,总共需要 0.004 秒:
SELECT DISTINCT w1.mail_id FROM mail_word_index AS w1
WHERE w1.sender_id=1 AND w1.word LIKE 'str%' LIMIT 20
SELECT DISTINCT w1.mail_id FROM mail_word_index AS w1
WHERE w1.sender_id=1 AND w1.word LIKE 'con%' LIMIT 20
内连接似乎减慢了第一个查询。我将如何更改第一个查询以使其更快?
解释告诉我这个:
id select_type table type possible_keys key key_len ref rows Extra
1 SIMPLE w1 range sender_id,multiple_words sender_id 21 NULL 1 Using where; Using temporary
1 SIMPLE w2 ref multiple_words multiple_words 8 game-node4.w1.mail_id,game-node4.w1.msg_id 8 Using where; Using index; Distinct