Stored aggregate function in HAVING: ERROR 1054, or a wrong result when the column is also selected

A stored aggregate function (CREATE AGGREGATE FUNCTION) cannot be used directly in HAVING. Built-in aggregates like SUM work there.

• If the column passed to the function is not in the select list: ERROR 1054 Unknown column 'v' in 'HAVING'.
• If the column is also in the select list: no error, but a wrong result. The function fetches once per row of the group, but gets the value of the group's first row every time.

  CREATE TABLE t (g INT, v INT);

  INSERT INTO t VALUES (1, 10), (1, 20), (2, 5);

  DELIMITER //

  CREATE AGGREGATE FUNCTION my_sum(x INT) RETURNS INT

  BEGIN

  DECLARE s INT DEFAULT 0;

  DECLARE CONTINUE HANDLER FOR NOT FOUND RETURN s;

  LOOP

  FETCH GROUP NEXT ROW;

  SET s = s + x;

  END LOOP;

  END //

  DELIMITER ;

  SELECT g, my_sum(v) FROM t GROUP BY g; -- ok: 1 -> 30, 2 -> 5

  SELECT g FROM t GROUP BY g HAVING SUM(v) > 25; -- ok: 1

  SELECT g FROM t GROUP BY g HAVING my_sum(v) > 25; -- ERROR 1054 (42S22): Unknown column 'v' in 'HAVING'

  SELECT g, v FROM t GROUP BY g HAVING my_sum(v) > 25; -- empty set, expected g = 1

  SELECT g, v FROM t GROUP BY g HAVING my_sum(v) = 20; -- returns g = 1, so my_sum(v) is 10 + 10, not 10 + 20

  Expected: HAVING my_sum(v) > 25 returns g = 1, the same as HAVING SUM(v) > 25.
  What the function sees, using a copy of my_sum that appends each fetched x to a user variable @seen:

  in the select list: 10 20 = 30; 5 = 5;

  in HAVING: 10 10 = 20; 5 = 5;

  Also:

• Same with a qualified column, HAVING my_sum(t.v) > 25, and without GROUP BY, SELECT 1 FROM t HAVING my_sum(v) > 10.
• A column that is in GROUP BY works: HAVING my_sum(g) > 1.
• ORDER BY my_sum(v) works.
  Workarounds: an alias (SELECT g, my_sum(v) AS s FROM t GROUP BY g HAVING s > 25) or a derived table.

I searched Jira and found no existing issue. MDEV-14213 is about a plain stored function in HAVING, not an aggregate.

Found while testing stored aggregates for a university project: https://github.com/ayoublabrinssi/Capstone-Starter-MariaDB-custom-aggregates