Why select NOT IN returns 0 row while there are actually many results

Question

If I run:

select ANY_VALUE(id)
from my_table
group by title

it will return 5563 rows which are all id(s) of rows with distinct 'title'.

Table my_table has total 6497 rows. (If I run: Select * from my_table I got 6497 rows)

Now I expect this query will give me the list of the rest (934 rows) of the id(s) in my_table :

Select * from my_table where id NOT IN (
select ANY_VALUE(id)
from my_table
group by title)

but it gives me 0 row.

I also tried:

  Select * from my_table where id NOT IN (
    select ANY_VALUE(id)
    from my_table
    group by title) AND id IS NOT NULL

or

Select * from my_table  where not exists (
select ANY_VALUE(id) AS value
from my_table
group by title) and id is not NULL;

All of them return 0 row. What I am doing wrong ?

Investigate: [fiddle](https://dbfiddle.uk/?rdbms=mysql_8.0&fiddle=5ee7e59c11f85a3c1d756ef24063d43c). — Akina, Mar 04 '20 at 12:27
You **must remember** that `ANY_VALUE()` is non-deterministic function. So it cannot be used except output expressions list. — Akina, Mar 04 '20 at 12:35
Why not just use `MAX` or `MIN` functions? Will do the same job in that case. — Michał Turczyn, Mar 04 '20 at 12:44

fancyPants · Accepted Answer · 2020-03-04T12:45:25.600

I'd say this is a bug.

To reproduce:

create table foo(id int auto_increment primary key, a int);
insert into foo(a) values(1), (1), (2);

When you do an explain and a show warnings (in MySQL <= 5.6 you have to do explain extended) you can see what MySQL is actually executing after the optimizer did his job:

mysql > explain select * from foo where id not in (select id from foo group by a);
+----+--------------------+-------+------------+-----------------+---------------+---------+---------+------+------+----------+-------------+
| id | select_type        | table | partitions | type            | possible_keys | key     | key_len | ref  | rows | filtered | Extra       |
+----+--------------------+-------+------------+-----------------+---------------+---------+---------+------+------+----------+-------------+
|  1 | PRIMARY            | foo   | NULL       | ALL             | NULL          | NULL    | NULL    | NULL |    3 |   100.00 | Using where |
|  2 | DEPENDENT SUBQUERY | foo   | NULL       | unique_subquery | PRIMARY       | PRIMARY | 4       | func |    1 |   100.00 | NULL        |
+----+--------------------+-------+------------+-----------------+---------------+---------+---------+------+------+----------+-------------+
2 rows in set, 1 warning (0.00 sec)

mysql > show warnings;
+-------+------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Level | Code | Message                                                                                                                                                                                                                                               |
+-------+------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Note  | 1003 | /* select#1 */ select `playground`.`foo`.`id` AS `id`,`playground`.`foo`.`a` AS `a` from `playground`.`foo` where (not(<in_optimizer>(`playground`.`foo`.`id`,<exists>(<primary_index_lookup>(<cache>(`playground`.`foo`.`id`) in foo on PRIMARY))))) |
+-------+------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+

As you can see this is a totally different query. No more grouping. To get the result you'd expect, you have to nest the query one more time like this:

select * from foo where id not in (
    select id from (
        select any_value(id) as id from foo group by a
    ) sq
);

Thank you, your last query does give the expected results. Although I do not understand clearly your explanation but it is the solution for me. — user1314404, Mar 04 '20 at 14:16

Why select NOT IN returns 0 row while there are actually many results

1 Answers1