Mysql 查询很慢，其中一个条件为 Exists

Question

I got 3 tables and the relations between them is many to many.

This is the Image of my tables with its columns

I need to get the hashtag names that has files with specific category_id.

the problem is that when i use the below query without specifying the category, it performs well like 0.05s.

select `hashtags`.`slug` 
from `hashtags` 
where EXISTS (
   select * from `files` 
   inner join `file_hashtags` on `files`.`id` = `file_hashtags`.`file_id` 
   where `hashtags`.`id` = `file_hashtags`.`hashtag_id` 
);

but when i perform the below query with specified category it goes like 3s to perform.

select `hashtags`.`slug` 
from `hashtags` 
where EXISTS (
   select * from `files` 
   inner join `file_hashtags` on `files`.`id` = `file_hashtags`.`file_id` 
   where `hashtags`.`id` = `file_hashtags`.`hashtag_id` 
   and `files`.`category_id`=2 
);

what can i do to improve this to get the better query time? also i did this query using IN instead of Exists, but the result is the same with a little like 0.1s better perform in time.

about the indexes:

• files table has ID as primary key, and category_id as BTREE index (need this for when i need to perform easy queries like get files with specific category), and slug as Unique index.

• hashtags table has ID as primary key, and slug as Unique index.

• file_hashtags table has two foregin keys for their tables, also (file_id, hashtag_id) is Primary.

also there are about 150k rows in files table, 75 in hashtags table and 260k in the pivot table.

Answer 1

You can use inner join without exist:

select distinct `hashtags`.`slug` 
from `hashtags`
inner join `file_hashtags` on `hashtags`.`id` = `file_hashtags`.`hashtag_id`
inner join `files` on `files`.`id` = `file_hashtags`.`file_id`
where `files`.`category_id`=2
group by `hashtags`.`slug`;

Answer 2

Try like this.

SELECT hashtags.slug
FROM hashtags
WHERE EXISTS
(SELECT * FROM (SELECT * 
FROM files
WHERE category_id = 2) A
INNER JOIN file_hashtags ON A.id = file_hashtags.file_id
WHERE hashtags.id = file_hashtags.hashtag_id)

If the number of rows in the files table is large, the number of join records can be reduced by performing a filter before proceeding with the join operation.

Answer 3

Please use SHOW CREATE TABLE . It shows what indexes you have

I assume that file_hashtags is a many-to-many mapping? Then it needs

PRIMARY KEY(file_id, hashtag_id)
INDEX(hashtag_id, file_id)

More discussion: http://mysql.rjweb.org/doc.php/index_cookbook_mysql#many_to_many_mapping_table

Because of and files.category_id=2 , files needs

INDEX(category_id)

So, have those ids, and simply the query to:

SELECT  h.slug
    FROM  files AS f
    JOIN  file_hashtags AS fh  ON fh.file_id = f.id
    JOIN  hashtags AS h  ON h.id = fh.hashtag_id
    WHERE  f.category_id = 2;

(I assume id is the PRIMARY KEY of each table, though id is not needed in file_hashtags .)

I do not believe EXISTS helps with performance for this query.

If you are not using ENGINE=InnoDB , my answer is inadequate. You should be using InnoDB.