我有一个50行的Innodb表main。
CREATE table main(
id INT AUTO_INCREMENT NOT NULL,
main_id INT NOT NULL,
feature1 INT NOT NULL,
feature2 INT NOT NULL,
feature3 INT NOT NULL,
time_repetition TIMESTAMP NOT NULL,
rating SMALLINT NOT NULL,
PRIMARY KEY (id),
INDEX search(range_col1,range_col2,range_col3)
) engine=innodb;
要点1：
feature*字段是一些特征。因此，在搜索过程中，用户可以检查他是否需要表中的行，这些行满足多个值，因此我需要在IN中使用WHERE：feature1 IN (val1,val2,..) and feature2 IN (val1,val2,..) and feature3 IN (val1,val2,...)。
POINT2：根据时间的重复对表进行非规范化。因此，我需要在GROUP_CONCAT(time_repetition)中选择GROUP BY main_id。
要点3：首先要获得最佳价值评级。所以，我需要ORDER BY rating DESC。我还考虑每天重新计算一次评级，并根据顺序递减评级将数据插入表中。在这种情况下，我需要ORDER BY id。不幸的是，它没有解决问题。
所以，整个查询是：
SELECT
main_id, feature1, feature2, feature3, GROUP_CONCAT(time_repetition), rating
FROM main WHERE feature1 IN (val1,val2,..) and feature2 IN (val1,val2,..) and feature3 IN (val1,val2,...) GROUP BY main_id ORDER BY rating LIMIT 50。
问题是查询使用临时和文件排序平均为5K行。例如，下面是解释的结果：
explain(select SQL_NO_CACHE * from main where feature1 in (4,5,6,7)
and feature2 IN (2,3,4,5) and feature3 IN (1,2,3,4,5) group by main_id
order by id limit 10);
+----+-------------+--------+-------+---------------+----------+---------+------+------+----------------------------------------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+--------+-------+---------------+----------+---------+------+------+----------------------------------------------+
| 1 | SIMPLE | main | range | search | search | 12 | NULL | 7333 | Using where; Using temporary; Using filesort |
+----+-------------+--------+-------+---------------+----------+---------+------+------+----------------------------------------------+
如果我在索引中添加main_id则无效。你能推荐什么？