我有一个使用group by和order by的查询，但它很慢！我需要一些帮助才能使索引正确。这是我正在运行的查询：
select *
from puresen_mv_shop.cache_deals
where feature_ids REGEXP 'i,t,d'
and phone_cost > 100.00
group by handset_numeric_id
order by popularity
LIMIT 0,10;
解释
1 SIMPLE cache_deals index popularity 5 10 6635320.00 Using where; Using temporary
这在大约3秒内运行，但我需要一秒钟。表中有超过600k行。这是表结构：
mysql> describe cache_deals
-> ;
+-------------------------+---------------+------+-----+---------+-------+
| Field | Type | Null | Key | Default | Extra |
+-------------------------+---------------+------+-----+---------+-------+
| deal_id | varchar(100) | NO | | NULL | |
| deal_id_replication | varchar(150) | NO | PRI | | |
| handset_numeric_id | int(10) | YES | MUL | NULL | |
| handset_url | varchar(100) | YES | | NULL | |
| image_url | varchar(50) | YES | MUL | NULL | |
| handset_id | varchar(50) | YES | | NULL | |
| phone_cost | decimal(10,2) | YES | | NULL | |
| tariff_numeric_id | int(10) | YES | MUL | NULL | |
| tariff_id | varchar(100) | YES | MUL | NULL | |
| tariff_name | varchar(100) | YES | | NULL | |
| network_name | varchar(20) | YES | | NULL | |
| network_numeric_id | int(5) | YES | MUL | NULL | |
| term | int(5) | YES | MUL | NULL | |
| minutes | int(5) | YES | MUL | NULL | |
| texts | int(5) | YES | MUL | NULL | |
| data | int(5) | YES | MUL | NULL | |
| org_line_rental | decimal(10,2) | YES | MUL | NULL | |
| effective_monthly_cost | decimal(10,2) | YES | MUL | NULL | |
| free_gift_id | int(10) | YES | MUL | NULL | |
| free_gift_name | varchar(50) | YES | | NULL | |
| cashback | int(5) | YES | MUL | NULL | |
| free_lr | int(5) | YES | MUL | NULL | |
| half_lr | int(5) | YES | MUL | NULL | |
| clearance_flag | int(5) | YES | MUL | NULL | |
| manufacturer_numeric_id | int(5) | YES | MUL | NULL | |
| manufacturer_name | varchar(50) | YES | | NULL | |
| full_handset_name | varchar(100) | YES | | NULL | |
| popularity | int(20) | YES | MUL | NULL | |
| handset_colour | varchar(50) | YES | MUL | NULL | |
| feature_ids | varchar(100) | YES | | NULL | |
| operating_system | varchar(30) | YES | MUL | NULL | |
+-------------------------+---------------+------+-----+---------+-------+
以下是我目前在此表中获得的索引：
mysql> show index from cache_deals;
+-------------+------------+-------------------------+--------------+-------------------------+-----------+-------------+----------+--------+------+------------
| Table | Non_unique | Key_name | Seq_in_index | Column_name | Collation | Cardinality | Sub_part | Packed | Null | Index_type
+-------------+------------+-------------------------+--------------+-------------------------+-----------+-------------+----------+--------+------+------------
| cache_deals | 0 | PRIMARY | 1 | deal_id_replication | A | 663532 | NULL | NULL | | BTREE
| | |
| cache_deals | 1 | handset_numeric_id | 1 | handset_numeric_id | A | 759 | NULL | NULL | YES | BTREE
| | |
| cache_deals | 1 | network_numeric_id | 1 | network_numeric_id | A | 8 | NULL | NULL | YES | BTREE
| | |
| cache_deals | 1 | tariff_numeric_id | 1 | tariff_numeric_id | A | 1091 | NULL | NULL | YES | BTREE
| | |
| cache_deals | 1 | manufacturer_numeric_id | 1 | manufacturer_numeric_id | A | 23 | NULL | NULL | YES | BTREE
| | |
| cache_deals | 1 | operating_system | 1 | operating_system | A | 42 | NULL | NULL | YES | BTREE
| | |
| cache_deals | 1 | image_url | 1 | image_url | A | 755 | NULL | NULL | YES | BTREE
| | |
| cache_deals | 1 | tariff_id | 1 | tariff_id | A | 1091 | NULL | NULL | YES | BTREE
| | |
| cache_deals | 1 | term | 1 | term | A | 7 | NULL | NULL | YES | BTREE
| | |
| cache_deals | 1 | minutes | 1 | minutes | A | 26 | NULL | NULL | YES | BTREE
| | |
| cache_deals | 1 | texts | 1 | texts | A | 14 | NULL | NULL | YES | BTREE
| | |
| cache_deals | 1 | data | 1 | data | A | 14 | NULL | NULL | YES | BTREE
| | |
| cache_deals | 1 | org_line_rental | 1 | org_line_rental | A | 128 | NULL | NULL | YES | BTREE
| | |
| cache_deals | 1 | effective_monthly_cost | 1 | effective_monthly_cost | A | 2147 | NULL | NULL | YES | BTREE
| | |
| cache_deals | 1 | free_gift_id | 1 | free_gift_id | A | 105 | NULL | NULL | YES | BTREE
| | |
| cache_deals | 1 | cashback | 1 | cashback | A | 2 | NULL | NULL | YES | BTREE
| | |
| cache_deals | 1 | free_lr | 1 | free_lr | A | 2 | NULL | NULL | YES | BTREE
| | |
| cache_deals | 1 | half_lr | 1 | half_lr | A | 2 | NULL | NULL | YES | BTREE
| | |
| cache_deals | 1 | clearance_flag | 1 | clearance_flag | A | 2 | NULL | NULL | YES | BTREE
| | |
| cache_deals | 1 | handset_colour | 1 | handset_colour | A | 17 | NULL | NULL | YES | BTREE
| | |
| cache_deals | 1 | popularity | 1 | popularity | A | 718 | NULL | NULL | YES | BTREE
| | |
+-------------+------------+-------------------------+--------------+-------------------------+-----------+-------------+----------+--------+------+------------
我是否正确索引此表？
谢谢你的帮助。
答案 0 :(得分：3)
REGEXP正在杀死潜在的表现，因为它无法使用索引。查询的那一部分可以用其他方式编写，还是可以规范化为另一个表？如果可以，那可能会有所帮助。除此之外，phone_cost和handset_numeric_id上的综合索引应该有所帮助。将popularity添加到该索引可能有所帮助，但MySQL已经对查询不满意了，因为您并不真的应该能够使用GROUP BY中不包含的列子句。
作为旁注，列的数据类型可能更有效。由于int(5)的限制，MEDIUMINT与INT的工作方式相同，只是MEDIUMINT小一个字节。此外，int(20)没有意义，因为INT类型不会那么大。
答案 1 :(得分：0)
select [name the columns you actually want returned]
from puresen_mv_shop.cache_deals
where feature_ids = 'i,t,d'
and phone_cost > 100.00
order
by popularity
LIMIT 0,10;
索引features_id或（features_id，phone_cost）