MySQL 查询优化 - 子查询 & 使用 where；使用临时的；使用文件排序

Posted 2023-04-14

【中文标题】MySQL 查询优化 - 子查询 & 使用 where；使用临时的；使用文件排序

【英文标题】：MySQL Query Optimisation - Sub Queries & Using where; Using temporary; Using filesort

【发布时间】：2013-02-16 00:26:04

【问题描述】：

我正在寻找优化以下 mysql 查询：

SELECT 
    `l`.`id` AS `l__id`, `l`.`latitude` AS `l__latitude`, `l`.`longitude` AS `l__longitude`, `l`.`name` AS `l__name`, `f`.`id` AS `f__id`, `f`.`day` AS `f__day`, `f`.`temperature_low` AS `f__temperature_low`, `f`.`temperature_high` AS `f__temperature_high`, `c`.`id` AS `c__id`, `c`.`name` AS `c__name` 
FROM `location` `l` 
INNER JOIN `forecast` `f` ON `l`.`id` = `f`.`location_id` 
INNER JOIN `condition` `c` ON `f`.`condition_id` = `c`.`id` 
WHERE (
    `f`.`day` IN ('2012-12-28', '2012-12-29') AND 
    EXISTS (
        SELECT 
            `f2`.`id` AS `f2__id` 
        FROM `forecast` `f2` 
        WHERE (
            `f2`.`location_id` = `l`.`id` AND `f2`.`day` = "2012-12-28" AND `f2`.`condition_id` IN (6, 7, 9, 10, 11, 12, 13, 14, 18) AND `f2`.`temperature_high` <= 30 AND `f2`.`temperature_low` >= 0
        )
    ) AND 
    EXISTS (
        SELECT 
            `f3`.`id` AS `f3__id` 
        FROM `forecast` `f3` 
        WHERE (
            `f3`.`location_id` = `l`.`id` AND `f3`.`day` = "2012-12-29" AND `f3`.`condition_id` IN (6, 7, 9, 10, 11, 12, 13, 14, 18) AND `f3`.`temperature_high` <= 30 AND `f3`.`temperature_low` >= 0
        )
    ) 
    AND `l`.`latitude` IS NOT NULL AND `l`.`longitude` IS NOT NULL
);

运行 EXPLAIN 会给出以下输出：

+----+--------------------+-------+--------+--------------------------------------+-----------------+---------+------------------------+------+----------------------------------------------+
| id |    select_type     | table |  type  |            possible_keys             |       key       | key_len |          ref           | rows |                    Extra                     |
+----+--------------------+-------+--------+--------------------------------------+-----------------+---------+------------------------+------+----------------------------------------------+
|  1 | PRIMARY            | f     | range  | location_id_idx,condition_id_idx,day | day             |       3 | NULL                   | 1298 | Using where; Using temporary; Using filesort |
|  1 | PRIMARY            | l     | eq_ref | PRIMARY                              | PRIMARY         |       8 | weather.f.location_id  |    1 | Using where                                  |
|  1 | PRIMARY            | c     | eq_ref | PRIMARY                              | PRIMARY         |       8 | weather.f.condition_id |    1 |                                              |
|  3 | DEPENDENT SUBQUERY | f3    | ref    | location_id_idx,condition_id_idx,day | location_id_idx |       9 | weather.l.id           |  276 | Using where                                  |
|  2 | DEPENDENT SUBQUERY | f2    | ref    | location_id_idx,condition_id_idx,day | location_id_idx |       9 | weather.l.id           |  276 | Using where                                  |
+----+--------------------+-------+--------+--------------------------------------+-----------------+---------+------------------------+------+----------------------------------------------+

我知道问题是“使用 where；使用临时；使用文件排序”和子查询，但我不确定如何重写它以提高性能。

提前感谢任何可以帮助我解决此类问题的建议。

皮特

【参考方案1】：

在您的查询中，您提供了谓词来限制 forecast 集，但正在无序地加入您的表。 mysql 优化器做了一些工作并在预测中使用day 键，并在加入其他表之前使用f.day IN ('2012-12-28', '2012-12-29') 限制预测集。优化器无法在加入之前使用您的依赖子查询（看起来没有必要）来进一步限制condition_id、temperature_high 和temperature_low 设置的预测，因此这是在所有连接完成后完成的。为了改进查询，您应该删除依赖子查询并限制预测集开始。

SELECT *
FROM forecast f 
INNER JOIN location l 
  ON l.id = f.location_id AND l.latitude IS NOT NULL AND l.longitude IS NOT NULL
INNER JOIN condition c
  ON f.condition_id = c.id
WHERE f.day IN ('2012-12-28', '2012-12-29') 
  AND f.condition_id IN (6, 7, 9, 10, 11, 12, 13, 14, 18) 
  AND f.temperature_high <= 30 AND f.temperature_low >= 0

【讨论】：

感谢这双令人神清气爽的眼睛。你完全正确，我不需要每天循环创建子查询，因为每天的条件和温度都是相同的。我认为在不同日子选择不同条件的能力是我留下的。干杯！