MySql - 如何使用索引优化查询？

Question

我们正在尝试从数据库中获取关注者的最新 10 条通知。我们会进行一些连接，以确保我们为关注者获得正确的通知集。如果他们关注的人（他们的领导者）添加了新帖子，则关注者应该只收到在他们开始关注领导者之后添加的帖子的通知（将他们的领导者的旧帖子显示为新通知是没有意义的）。另一个加入是确保我们获得通知的read_at 时间，以便关注者知道它是否已被阅读。这是查询，但它需要 ~9 秒，这太慢了。理想情况下，它应该只需要几毫秒，特别是索引：

查询：

SELECT nf.id, nf.uuid, nf.leader_id, nf.data, nf.created_at, nfr.read_at
FROM notification_followers nf
LEFT JOIN user_follows uf ON uf.leader_id = nf.leader_id AND uf.follower_id = 14 AND uf.follow_status = 'follow'
LEFT JOIN notification_followers_read nfr ON nf.id = nfr.notification_followers_id AND nfr.follower_id = 14
WHERE (nf.created_at > uf.created_at)
ORDER BY nf.id DESC
LIMIT 10

索引：

ALTER TABLE `notification_followers` ADD INDEX `nf_lid_ca_id_idx` (`leader_id`,`created_at`,`id`);
ALTER TABLE `user_follows` ADD KEY`uf_fid_lid_fs_ca_idx` (`follower_id`,`leader_id`,`follow_status`,`created_at`)
ALTER TABLE `notification_followers_read` ADD INDEX `nfr_fid_nfid_ra_idx` (`follower_id`,`notification_followers_id`,`read_at`);

说明：

正确的结果（大约需要 9 秒）：

SQL 转储：

SQL DUMP TO REPRODUCE LOCALLY 只需在本地创建 speed_test 数据库并导入文件即可查看所有表数据的慢查询问题（~100K 行）。

我们如何优化上述内容以在几毫秒内得到正确的结果？

Answer 1

对于这个查询：

SELECT nf.id, nf.uuid, nf.leader_id, nf.data, nf.created_at, nfr.read_at
FROM notification_followers nf JOIN
     user_follows uf 
     ON uf.leader_id = nf.leader_id AND uf.follower_id = 14 AND
        uf.follow_status = 'follow' LEFT JOIN 
     notification_followers_read nfr
     ON nf.id = nfr.notification_followers_id AND nfr.follower_id = 14
WHERE nf.created_at > uf.created_at
ORDER BY nf.id DESC
LIMIT 10;

我会推荐user_follower(leader_id, follower_id, follow_status, created_at) 和notification_followers_read(notification_followers_id, follower_id, read_at) 上的索引。索引中列的顺序很重要。

请注意，我将第一个 JOIN 更改为内部联接，因为 WHERE 子句无论如何都会将其变为一个。

嗯，让我们尝试重写查询：

SELECT nf.id, nf.uuid, nf.leader_id, nf.data, nf.created_at,
       (SELECT nfr.read_at
        FROM notification_followers_read nfr
        WHERE nf.id = nfr.notification_followers_id AND nfr.follower_id = 14
       ) nfr
FROM (SELECT nf.*
      FROM notification_followers nf 
      WHERE EXISTS (SELECT 1
                    FROM user_follows uf 
                    WHERE uf.leader_id = nf.leader_id AND uf.follower_id = 14 AND
                          uf.follow_status = 'follow' AND nf.created_at > uf.created_at
                   )
      ORDER BY nf.id DESC
      LIMIT 10
     ) nf;

为此，您要确保在 notification_followers(id) 上也有索引。

根据您的数据，使用这种方法，内部子查询可能会更快：

FROM (SELECT nf.*
      FROM user_follows uf JOIN
           notification_followers nf 
           ON uf.leader_id = nf.leader_id AND nf.created_at > uf.created_at
      WHERE uf.follower_id = 14 AND uf.follow_status = 'follow' 
      ORDER BY nf.id DESC
      LIMIT 10
     ) nf

为此，索引为user_follows(follower_id, follow_status, leader_id, created_at) 和notification_followers(leader_id, created_at, id)。这可能会更快。

Answer 2

你应该试试这个。

SELECT nf.id, nf.uuid, nf.leader_id, nf.data, nf.created_at, nfr.read_at 
FROM notification_followers nf 
JOIN user_follows uf ON uf.leader_id = nf.leader_id and nf.created_at > uf.created_at AND uf.follow_status = 'follow'  AND uf.follower_id = 14 
LEFT JOIN notification_followers_read nfr ON nf.id = nfr.notification_followers_id AND nfr.follower_id = 14 
ORDER BY nf.id DESC
LIMIT 10;

创建索引。

ALTER TABLE `notification_followers` ADD INDEX `nf_lid_ca_id_idx`(`leader_id`,`created_at`,`id`);
ALTER TABLE `user_follows` ADD KEY`uf_fid_lid_fs_ca_idx`(`leader_id`,`created_at`,`follow_status`,`follower_id`)
ALTER TABLE `notification_followers_read` ADD INDEX `nfr_fid_nfid_ra_idx`(`notification_followers_id`,`follower_id`,`read_at`);