Reverse aggregation inside of common table expression

Question

I would have expected that the following query returns all people with their respective children.

WITH RECURSIVE nested_people (id, name, children) AS (
   SELECT id, name, NULL::JSON AS children
   FROM people
   WHERE parent_id IS NULL
 UNION ALL
   SELECT people.id, people.name, ROW_TO_JSON(nested_people.*) AS children
   FROM people
   JOIN nested_people ON people.parent_id = nested_people.id
)
SELECT * FROM nested_people;

But actually it does the exact reverse. I can't think of a way to do correct nesting without having to an additional CTE. Is there a way?

Example data

+----+-------+-----------+
| id | name  | parent_id |
+----+-------+-----------+
|  1 | Adam  | null      |
|  2 | Abel  | 1         |
|  3 | Cain  | 1         |
|  4 | Enoch | 3         |
+----+-------+-----------+

Result

+----+-------+--------------------------------------------------------------------------+
| id | name  |                                children                                  |
+----+-------+--------------------------------------------------------------------------+
|  1 | Adam  | null                                                                     |
|  2 | Abel  | {"id":1,"name":"Adam","children":null}                                   |
|  3 | Cain  | {"id":1,"name":"Adam","children":null}                                   |
|  4 | Enoch | {"id":3,"name":"Cain","children":{"id":1,"name":"Adam","children":null}} |
+----+-------+--------------------------------------------------------------------------+

Expected Result

+----+-------+----------------------------------------------------------------------------------------------------------------------+
| id | name  |                                                       children                                                       |
+----+-------+----------------------------------------------------------------------------------------------------------------------+
|  1 | Adam  | [{"id":2, "name":"Abel", "children":null},{"id":3,"name":"Cain","children":[{"id":4,"name":"Enoch","children":null}] |
|  2 | Abel  | null                                                                                                                 |
|  3 | Cain  | [{"id":4,"name":"Enoch","children":null}]                                                                            |
|  4 | Enoch | null                                                                                                                 |
+----+-------+----------------------------------------------------------------------------------------------------------------------+

Can you add some data sample and from this sample what would be the desired result? — Jorge Campos, Dec 12 '15 at 23:51

score 2 · Accepted Answer · edited May 23 '17 at 11:45

This rCTE traverses the tree from the other side:

WITH RECURSIVE cte AS (
   SELECT id, parent_id, name, NULL::JSON AS children
   FROM   people p
   WHERE  NOT EXISTS (  -- only leaf nodes; see link below
      SELECT 1 FROM people
      WHERE  parent_id = p.id
      )
   UNION ALL
   SELECT p.id, p.parent_id, p.name, row_to_json(c) AS children
   FROM   cte c
   JOIN   people p ON p.id = c.parent_id
   )
SELECT id, name, json_agg(children) AS children
FROM   cte
GROUP  BY 1, 2;

SQL Fiddle.

Use json_agg() to aggregate multiple branches per node in the outer SELECT.
Minor differences to your desired result:

This includes the parent_id in the children column.
A single node is not wrapped into an array.

Either can be adapted, but I'd expect the result is OK for you as is.

How to identify leaf nodes:

Select rows which are not present in other table

Reverse aggregation inside of common table expression

1 Answers1

Linked