How to identify and list Circular Reference elements in Postgres - postgresql

Question

I have an employee , manager hierarchy which could end up being circular.
Ex:
28397468N>88518119N>87606705N>28397468N
Create Table emp_manager ( Emp_id varchar(30), Manager_id varchar(30));
Insert into emp_manager values ('28397468N','88518119N');
Insert into emp_manager values ('88518119N','87606705N');
Insert into emp_manager values ('87606705N','28397468N');
My requirement is:
When my proc is called and there are circular hierarchies in the emp_manager table, we should return an error listing the employees in the hierarchy.
The below link contains some useful info:
https://mccalljt.io/blog/2017/01/postgres-circular-references/
I have modified it as below:
select * from (
WITH RECURSIVE circular_managers(Emp_id, Manager_id, depth, path, cycle) AS (
SELECT u.Emp_id, u.Manager_id, 1,
ARRAY[u.Emp_id],
false
FROM emp_manager u
UNION ALL
SELECT u.Emp_id, u.Manager_id, cm.depth + 1,
(path || u.Emp_id)::character varying(32)[],
u.Emp_id = ANY(path)
FROM emp_manager u, circular_managers cm
WHERE u.Emp_id = cm.Manager_id AND NOT cycle
)
select
distinct (path) d
FROM circular_managers
WHERE cycle
AND path[1] = path[array_upper(path, 1)]) cm
BUT, the problem is, it is returning all combinations of the hierarchy:
{28397468N,88518119N,87606705N,28397468N}
{87606705N,28397468N,88518119N,87606705N}
{88518119N,87606705N,28397468N,88518119N}
I need a simple answer like this:
28397468N>88518119N>87606705N>28397468N
even this will do:
28397468N>88518119N>87606705N
Please help!

Answer 1

So all references:
{28397468N,88518119N,87606705N,28397468N}
{87606705N,28397468N,88518119N,87606705N}
{88518119N,87606705N,28397468N,88518119N}
are correct but just start from different element.
I need a simple answer like this: 28397468N>88518119N>87606705N>28397468N
So what's needed is a filter for the same circle refs.
Let's do that in a way:
sort distinct items in arrays
aggregate them back - so for all references it will be '{28397468N,87606705N,88518119N}'
use produced value for DISTINCT FIRST_VALUE
WITH D (circle_ref ) AS (
VALUES
('{28397468N,88518119N,87606705N,28397468N}'::text[]),
('{87606705N,28397468N,88518119N,87606705N}'::text[]),
('{88518119N,87606705N,28397468N,88518119N}'::text[])
), ordered AS (
SELECT
D.circle_ref,
(SELECT ARRAY_AGG(DISTINCT el ORDER BY el) FROM UNNEST(D.circle_ref) AS el ) AS ordered_circle
FROM
D
)
SELECT DISTINCT
FIRST_VALUE (circle_ref) OVER (PARTITION BY ordered_circle ORDER BY circle_ref) AS circle_ref
FROM
ordered;
circle_ref
{28397468N,88518119N,87606705N,28397468N}
DB Fiddle: https://www.db-fiddle.com/f/6ytb2v11s8T95PPLoTZZed/0

Answer 2

To prevent circular references, you can use a closure table and a trigger - as explained in https://stackoverflow.com/a/38701519/5962802
The closure table will also allow you to easily get all subordinates for a given supervisor (no matter how deep in the hierarchy) - or all direct bosses of a given employee (up to the root).
Before using the rebuild_tree stored procedure you will have to remove all circular references from the hierarchy.