如何使用 JOIN 和 WHERE 优化 MySQL 查询？

Question

我有两个表：shows（横幅展示），clicks（横幅点击）。

CREATE TABLE `shows` (
`id` int(11) unsigned NOT NULL AUTO_INCREMENT,
`data` text NOT NULL,
`created_at` date NOT NULL DEFAULT current_timestamp(),
PRIMARY KEY (`id`),
KEY `created_at` (`created_at`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;

CREATE TABLE `clicks` (
`id` int(11) unsigned NOT NULL AUTO_INCREMENT,
`show_id` int(11) unsigned NOT NULL,
`data` text NOT NULL,
`created_at` date NOT NULL DEFAULT current_timestamp(),
PRIMARY KEY (`id`),
KEY `created_at` (`created_at`),
KEY `show_constaraint_idx` (`show_id`),
CONSTRAINT `show_constaraint` FOREIGN KEY (`show_id`) REFERENCES `shows` (`id`) ON DELETE CASCADE ON UPDATE CASCADE
) ENGINE=InnoDB DEFAULT CHARSET=utf8;

每个表有 40 000 000 条记录。

我在显示表中添加了 40 000 000 条记录：

drop procedure if exists doWhile;
DELIMITER //
CREATE PROCEDURE doWhile()
BEGIN
DECLARE i INT DEFAULT 1;
WHILE (i <= 40000000) DO
INSERT INTO `shows` (data, created_at) values (
    CONCAT(MD5(RAND()), MD5(RAND())),
    FROM_UNIXTIME(unix_timestamp('2021-01-01') + floor(rand() * (unix_timestamp('2021-12-31') - unix_timestamp('2021-01-01') + 1)))
);
SET i = i + 1;
END WHILE;
END;
//
CALL doWhile();

我在 clicks 表中添加了 40 000 000 条记录：

drop procedure if exists doWhile;
DELIMITER //
CREATE PROCEDURE doWhile()
BEGIN
DECLARE i INT DEFAULT 1;
WHILE (i <= 40000000) DO
INSERT INTO `clicks` (show_id, data, created_at) values (
    (FLOOR(1 + RAND() * 40000000)),
    CONCAT(MD5(RAND()), MD5(RAND())),
    FROM_UNIXTIME(unix_timestamp('2021-01-01') + floor(rand() * (unix_timestamp('2021-12-31') - unix_timestamp('2021-01-01') + 1)))
);
SET i = i + 1;
END WHILE;
END;
//
CALL doWhile();

SQL 查询执行时间为 10 秒：

SELECT shows.id,
    shows.data,
    clicks.id,
    clicks.data
FROM clicks
INNER JOIN shows ON shows.id = clicks.show_id
WHERE shows.created_at = '2021-03-03'
AND clicks.created_at >= '2021-03-03'
AND clicks.created_at <= '2021-03-06'

我尝试优化 SQL 查询，但是 SQL 查询执行时间是 10 秒：

SELECT sub_shows.id,
    sub_shows.data,
    sub_clicks.id,
    sub_clicks.data
FROM (
    SELECT clicks.id,
        clicks.show_id,
        clicks.data
    FROM clicks
    WHERE clicks.created_at >= '2021-03-03' AND clicks.created_at <= '2021-03-06'
) as sub_clicks
INNER JOIN (
    SELECT shows.id,
        shows.data
    FROM shows
    WHERE shows.created_at = '2021-03-03'
) as sub_shows ON sub_shows.id = sub_clicks.show_id

两个查询的解释相同：

如何优化查询？

回答： 将 innodb_buffer_pool_size 更改为 12G （里克·詹姆斯）

Answer 1

最好有一个复合索引涵盖连接条件的两个部分与个人。

Suggest Indexes as listed
Shows    ( created_at, id )
Clicks   ( show_id, created_at )

这两个键都有帮助的原因是索引利用了这两个部分，而无需转到原始数据页面来限定记录。只有在限定之后才需要检索结果集中的附加列。

我还会将查询调整并反转为最小粒度项目...在查询的主要 FROM 位置中的单个日期上的 SHOWS 表。

SELECT 
      shows.id,
      shows.data,
      c.id,
      c.data
   FROM
      shows s
         JOIN clicks c
            ON s.id = c.show_id
          AND c.created_at >= '2021-03-03'
          AND c.created_at <= '2021-03-06'   
   WHERE 
      s.created_at = '2021-03-03'