Question

我试图加快一个SQL查询，这个查询返回了15个最便宜的产品，这些产品具有一些零售商销售的一组特定的特征（宽度，高度，直径和负载）。 de DB中约有300k产品。总价格计算如下：

totalPrice =数量*（价格 - 折扣*价格）+ shippingCost

其中：

折扣（百分比）取决于订购数量，零售商和时间。由于同一组合可能存在许多折扣，因此选择最有利的百分比。
shippingCost 取决于订购数量，一些产品特征以及产品必须交付的目的地

我的问题是折扣和 shippingsCost 依赖于太多参数来存储每个总价格组合，以便使查询运行得更快。因此，我认为我坚持使用子查询。

以下是SQL查询的简化版本，其中产品数量设置为2。

SELECT `P`.*, `B`.`name_local` as brandName, `R`.`name` as retailerName, `D`.`amount` as discount,     `S`.`shippingCost`, ROUND(P.price * 2 + IFNULL(S.shippingCost, 0) - IFNULL(P.price * D.amount / 100 * 2, 0), 2 ) as totalPrice
FROM (`Product` P)
JOIN `Brand` B ON `B`.`id` = `P`.`idBrand`
JOIN `Retailer` R ON `R`.`id` = `P`.`idRetailer`
LEFT JOIN `Shipping` S ON `S`.`idRetailer` = `P`.`idRetailer` AND S.nbProduct = (SELECT nbProduct FROM `Shipping` WHERE nbProduct <= 2 ORDER BY nbProduct DESC LIMIT 1)
LEFT JOIN `Discount` D ON `D`.`idRetailer` = `P`.`idRetailer` AND D.amount = (SELECT MAX(amount) FROM Discount D WHERE (D.vehicle = P.vehicle OR D.vehicle = 0) AND D.idRetailer = P.idRetailer AND D.start <= 1451825895 AND D.end >=1451825895)
WHERE `width` =  '195'
AND `height` =  '65'
AND `diameter` =  '15'
AND `load` >= 0
ORDER BY `totalPrice` ASC
LIMIT 15

我使用的是mysql 14.14。在我的机器上，查询大约需要150ms才能执行。通过避免使用当前时间戳来进行折扣，可以更好地利用mysql查询缓存。但是，查询需要很长时间才能执行第一次，并且很快从查询缓存中刷新（由于许多组合）。以下是查询explain命令的结果：

+----+--------------------+----------+--------+-----------------------------------------------+------------+---------+------------------------+-------+----------------------------------------------+
| id | select_type        | table    | type   | possible_keys                                 | key        | key_len | ref                    | rows  | Extra                                        |
+----+--------------------+----------+--------+-----------------------------------------------+------------+---------+------------------------+-------+----------------------------------------------+
|  1 | PRIMARY            | P        | ref    | idBrand,idRetailer,width,height,diameter,load | width      | 12      | const,const,const      | 13268 | Using where; Using temporary; Using filesort |
|  1 | PRIMARY            | S        | ref    | idRetailer                                    | idRetailer | 4       | mydb.P.idRetailer      |     1 | Using where                                  |
|  1 | PRIMARY            | B        | eq_ref | PRIMARY                                       | PRIMARY    | 4       | mydb.P.idBrand         |     1 |                                              |
|  1 | PRIMARY            | R        | eq_ref | PRIMARY                                       | PRIMARY    | 4       | mydb.P.idRetailer      |     1 |                                              |
|  1 | PRIMARY            | D        | ref    | idRetailer                                    | idRetailer | 4       | mydb.S.idRetailer      |     1 |                                              |
|  3 | DEPENDENT SUBQUERY | D        | ref    | idRetailer,start                              | idRetailer | 4       | mydb.P.idRetailer      |     1 | Using where                                  |
|  2 | SUBQUERY           | Shipping | ALL    | NULL                                          | NULL       | NULL    | NULL                   |    48 | Using where; Using filesort                  |
+----+--------------------+----------+--------+-----------------------------------------------+------------+---------+------------------------+-------+----------------------------------------------+

是否有一种加速这种查询的优雅方法，或者我唯一的方法是改善查询缓存的效果（添加RAM，增加缓存大小等）？

Answer 1

对于第二个（更复杂的子查询）尝试使用计算折扣LEFT JOIN，让它只执行一次。喜欢这个

...
LEFT JOIN (SELECT MAX(amount) as amount, D.vehicle, D.idRetailer 
FROM Discount D 
WHERE D.start <= 1451825895 
   AND D.end >=1451825895
GROUP BY D.idRetailer, D.vehicle) D ON D.`idRetailer` = `P`.`idRetailer` 
                                 AND (D.vehicle = P.vehicle OR D.vehicle = 0)

优化包含操作和子查询的SQL查询

1 个答案: