Dynamic table created from CTE (parent/child) - sql

Question

If I have a very simple table called tree
create table if not exists tree (id int primary key, parent int, name text);
And a few rows of data
insert into tree values (1, null, 'A');
insert into tree values (2, 1, 'B');
insert into tree values (3, 1, 'C');
insert into tree values (4, 2, 'D');
insert into tree values (5, 2, 'E');
insert into tree values (6, 3, 'F');
insert into tree values (7, 3, 'G');
I can easily run CTEs on it, and produce an output giving me path like this
with recursive R(id, level, path, name) as (
select id,1,name,name from tree where parent is null
union select tree.id, level + 1, path || '.' || tree.name, tree.name from tree join R on R.id=tree.parent
) select level,path,name from R;
Which gives the output
level | path | name
-------+-------+------
1 | A | A
2 | A.B | B
2 | A.C | C
3 | A.B.D | D
3 | A.B.E | E
3 | A.C.F | F
3 | A.C.G | G
What I'm wondering, is it possible to somehow project this output into another table, dynamically creating columns based on level (level1, level2, level3 etc), giving me something like this in return
id | level1 | level2 | level3
---+--------+--------+-------
1 | A | |
2 | A | B |
3 | A | C |
4 | A | B | D
5 | A | B | E
6 | A | C | F
7 | A | C | G
Any help would be appreciated.

Answer 1

If you know the maximum depth of your tree, I'd keep your approach and simplify it using array concatenation to produce the desired output.
So for a 5 level tree, that would look like this :
WITH RECURSIVE R(id, path) AS (
SELECT id, ARRAY[name::text] FROM tree WHERE parent IS NULL
UNION SELECT tree.id, path || tree.name FROM tree JOIN R ON R.id=tree.parent
)
SELECT id,
path[1] AS l1,
path[2] AS l2,
path[3] AS l3,
path[4] AS l4,
path[5] AS l5
FROM R;
PS : sorry for not commenting on Ziggy's answer which is very close, but I don't have enough reputation to do so. I don't see why you would need a windowing function here ?

Answer 2

PostgreSQL requires to always define the type of the output, so you can't have the columns levelX produced dynamically. However, you can do the following:
with recursive
R(id, path) as (
select id,ARRAY[name::text] from tree where parent is null
union
select tree.id, path || tree.name::text from tree join R on R.id=tree.parent
)
select row_number() over (order by cardinality(path), path), id,
path[1] as level1, path[2] as level2, path[3] as level3
from R
order by 1
In the example above, the column row_number happens to match id, but probably that wouldn't happen with your real data.