I am trying to create a POSTGRESQL function which would first INSERT
some data in a table using WHILE LOOP
and then SELECT
the results of this table.
This is an sql example:
CREATE OR REPLACE FUNCTION get_levels_test (maxlevel int) RETURNS SETOF list_of_levels AS $$
DECLARE
level int = 1;
BEGIN
TRUNCATE list_of_levels;
WHILE (level <= maxlevel) LOOP
INSERT INTO list_of_levels
SELECT level;
level = level + 1;
END LOOP;
select * from list_of_levels;
END;
$$ LANGUAGE PLPGSQL VOLATILE;
Then i try to call this function with: select get_levels_test (3)
, which shows me this error:
ERROR: query has no destination for result data
Hint: If you want to discard the results of a SELECT, use PERFORM instead.
Where: PL/pgSQL function "get_levels_test" line 12 at SQL statement
The list_of_levels
table contains just an int
column.
In case this is needed, I am using PostgreSQL 8.2.15 (Greenplum Database 4.3.3.1 build 1).
You need to use "return next" which can be used with a composite type or with a table. Please use this feature with caution as you can have performance issues with using this. Do not use this to join the results to another table. Do not return large datasets with this. Only use it for very small results such as a short report.
You also need to add exception handling into your function. I added this to my example function below.
Example table:
create table employees
(id int not null,
manager_id int null,
employee_title text not null)
distributed by (id);
Example data:
insert into employees values
(1, null, 'President'),
(2, 1, 'Vice-President'),
(3, 2, 'Manager'),
(4, 3, 'Engineer'),
(5, null, 'CMO'),
(6, 5, 'Director'),
(7, 6, 'Assistant'),
(8, 7, 'Developer');
Function:
create or replace function get_employees(p_id int) returns setof employees as
$$
declare
v_function_name text := 'get_employees';
v_location int;
v_rec record;
v_rec2 employees;
v_id employees.id%type;
begin
v_location := 1000;
create temporary table t1
(id integer) on commit drop distributed by (id);
v_location := 2000;
for v_rec in (select id from employees where id = p_id and manager_id is null order by id) loop
v_id := v_rec.id;
insert into t1 values (v_id);
while v_id is not null loop
select id into v_id from employees where manager_id = v_id;
if v_id is not null then
insert into t1 values (v_id);
end if;
end loop;
end loop;
v_location := 3000;
for v_rec2 in (select * from employees e join t1 on e.id = t1.id order by e.id) loop
return next v_rec2;
end loop;
exception
when others then
raise exception '(%:%:%)', v_function_name, v_location, sqlerrm;
end;
$$
language plpgsql;
And using the function:
select * From get_employees(1);
id | manager_id | employee_title
----+------------+----------------
1 | | President
2 | 1 | Vice-President
3 | 2 | Manager
4 | 3 | Engineer
(4 rows)
select * From get_employees(5);
id | manager_id | employee_title
----+------------+----------------
5 | | CMO
6 | 5 | Director
7 | 6 | Assistant
8 | 7 | Developer
(4 rows)