如何将此 SQL 查询转换为使用 GROUP BY 和 HAVING 查询？

Question

There're products , products_filters , filters_values tables in DB.
products_filters references to products table via product_id column which is a foreign key.
Also products_filters references to filters_values table via filter_value_id column which is also a foreign key

When user selects filters, SQL query which extracts all ids of suitable products is formed.

For example, chosen filters are:
Sex: Male, Female
Brand: Brand1, Brand2, Brand3

How it should work:
It needs to select all products which have filter Sex set to Male OR Female AND filter Brand set to Brand1 OR Brand2 OR Brand3. But products having matching only in one of the chosen filter category either Sex or Brand, must not be selected. It necessiraly to have matching in all selected categories.
I think SQL should look like this:

SELECT product_id FROM products_filters WHERE 
(filter_value_id = 1 OR filter_value_id = 2) 
AND 
(filter_value_id = 3 OR filter_value_id = 4 OR filter_value_id = 5)

Where 1 is Male, 2 is Female, 3 is Brand1, 4 is Brand2, 5 is Brand3.

But this query doesn't work.

In my previous question I was answered that GROUP BY and HAVING may help.

Q: How can I transform SQL above with GROUP BY and HAVING?

Answer 1

Given

    drop table if exists t;
    create table t(id int ,gender varchar(1), brand varchar(1));
    
    insert into t values
    (1,'m',1),(1,'f',2),(1,'m',3),(2,'f',1),(2,'f',2),(2,'f',3),
    (3,'m',1),(4,'f',2);

Correlated sub queries with distinct

    select distinct id
    from t
    where
             (select count(distinct gender) from t t1 where gender in('m','f') and t1.id = t.id) = 2 and
             (select count(distinct brand)  from t t1 where brand in(1,2,3) and t1.id = t.id) = 3
    ;

+------+
| id   |
+------+
|    1 |
+------+
1 row in set (0.002 sec)

Answer 2

SELECT product_id
    FROM products_filters AS f1
    JOIN products_filters AS f2  USING(product_id)
    WHERE f1.filter_value_id IN (1,2)
      AND f2.filter_value_id IN (3,4,5)

(I don't think GROUP BY...HAVING COUNT(*) = 2 is reasonable for this case.)

(The EAV schema design is a pain to deal with.)