Order by objects relation (PostgreSQL) - postgresql

Question

Have 2 tables for example:
In 1st: object & parent columns
object | parent
-------+---------
object1| null
object2| object1
object3| null
2nd has: object & reference columns
object | reference
-------+---------
object1| null
object2| null
object3| object1
Need to query tables to order like following: parent is first, then - child(s), objects which have reference(s) to parent.
object1
object2
object3
Is it possible to do in one SQL query or need to sort manually in an array? Seems it is a classical task, probably solution already exists somewhere?

Answer 1

Is this what you're looking for?
CREATE TABLE oparen (object varchar(10), parent varchar(10));
CREATE TABLE oref (object varchar(10), ref varchar(10));
INSERT INTO oparen VALUES
('object1',null),('object2','object1'),
('object3',null),('object4','object2');
INSERT INTO oref VALUES
('object1',null),('object2',null),('object3','object1'),
('object5','object6'),('object6','object1'),('object7','object4');
WITH hier AS (
SELECT parent AS obj, 1 AS rank FROM oparen
WHERE parent IS NOT NULL
UNION
SELECT object, 2 FROM oparen
WHERE parent IS NOT NULL
UNION
SELECT object, 3 FROM oref
WHERE ref IS NOT NULL),
allobj AS (
SELECT object AS obj FROM oparen
UNION
SELECT object FROM oref)
SELECT a.obj, coalesce(h.rank, 4) AS rank
FROM allobj a LEFT JOIN hier h ON a.obj = h.obj
ORDER BY coalesce(h.rank, 4), a.obj;
EDIT: After the improved example in the answer below, the following query should do the trick:
WITH parents AS (
SELECT parent AS obj, 1 AS rank FROM oparen
WHERE parent IS NOT NULL
),
family AS (
SELECT * FROM parents
UNION ALL
SELECT object, 2 FROM oparen op
WHERE parent IS NOT NULL
AND NOT EXISTS (SELECT obj FROM parents WHERE obj = op.object)
),
hier AS (
SELECT * FROM family
UNION ALL
SELECT object AS obj, coalesce(f.rank + 2, 5) AS rank
FROM oref LEFT JOIN family f ON oref.ref = f.obj
WHERE ref IS NOT NULL
),
allobj AS (
SELECT object AS obj FROM oparen
UNION
SELECT object FROM oref)
SELECT a.obj, h.rank AS rank
FROM allobj a LEFT JOIN hier h ON a.obj = h.obj
ORDER BY h.rank, a.obj;
Testbed creation in the top is updated according to the new requirements.

Answer 2

I inserted following data:
INSERT INTO oparen VALUES
('object1',null),('object2','object1'),('object3',null),('object4','object2');
INSERT INTO oref VALUES
('object1',null),('object2',null),('object3','object1'),('object5','object6'),('object6','object1');
Order is incorrect and object2 listed twice. DISTINCT on obj breaks the order also. Should go 6 then 5.

Answer 3

No, does not work: checked for another data and simplified to use and only by oref table content:
INSERT INTO oref VALUES
('object1',null),('object2',null),('object3','object1'),
('object5','object6'),('object6','object1'),('object7','object4'), ('object4','object5');
WITH family AS (
SELECT object AS obj, 1 AS rank FROM oref
WHERE ref IS NULL
),
hier AS (
SELECT * FROM family
UNION ALL
SELECT object AS obj, coalesce(f.rank + 2, 5) AS rank
FROM oref LEFT JOIN family f ON oref.ref = f.obj
WHERE ref IS NOT NULL
),
allobj AS (
SELECT object AS obj FROM oref)
SELECT a.obj, h.rank AS rank
FROM allobj a
LEFT JOIN hier h ON a.obj = h.obj
ORDER BY h.rank, a.obj;
Think need to use recursive queries here. Will write and post here.

Answer 4

Following recursive query works:
WITH RECURSIVE tables(object, rank) AS (
SELECT DISTINCT o.object, 1 AS rank FROM oref o
WHERE o.ref IS NULL
UNION
SELECT o.object, t.rank + 1 AS rank
FROM (SELECT DISTINCT o.object, o.ref FROM oref o
WHERE ref IS NOT NULL) o, tables t
WHERE o.ref = t.object AND rank <= t.rank
),
ordered AS (
SELECT * FROM tables
)
SELECT * FROM tables
WHERE tables.rank = (SELECT MAX(rank) FROM ordered WHERE ordered.object = tables.object)
ORDER BY rank;
Any comments, questions, objections, propositions? ;)