Tree with recursive and default - sql

Question

Using Postgres.
I have a pricelists
CREATE TABLE pricelists(
id SERIAL PRIMARY KEY,
name TEXT,
parent_id INTEGER REFERENCES pricelists
);
and another table, prices, referencing it
CREATE TABLE prices(
pricelist_id INTEGER REFERENCES pricelists,
name TEXT,
value INTEGER NOT NULL,
PRIMARY KEY (pricelist_id, name)
);
Parent pricelist id=1 may have 10 prices.
Pricelist id=2 as a child of parent 1 may have 5 prices which override parent 1 prices of the same price name.
Child Pricelist id=3 as as a child of pricelist 2 may have 2 price which override child 2 prices of the same price name.
Thus when I ask for child 3 prices, I want to get
all prices of child 3 and
those prices of his parent (child 2) that do not exists in child 3 and
all parent 1 prices that do not exists until now.
The schema can be changed in order to be efficient.
Example:
If
SELECT pl.id AS id, pl.parent_id AS parent, p.name AS price_name, value
FROM pricelists pl
JOIN prices p ON pl.id = p.pricelist_id;
gives
| id | parent | price_name | value |
|----------|:-------------:|------------:|------------:|
| 1 | 1 | bb | 10 |
| 1 | 1 | cc | 10 |
| 2 | 1 | aa | 20 |
| 2 | 1 | bb | 20 |
| 3 | 2 | aa | 30 |
then I'm looking for a way of fetching pricelist_id = 3 prices that'd give me
| id | parent | price_name | value |
|----------|:-------------:|------------:|------------:|
| 1 | 1 | cc | 10 |
| 2 | 1 | bb | 20 |
| 3 | 2 | aa | 30 |

Answer 1

WITH RECURSIVE cte AS (
SELECT id, name, parent_id, 1 AS lvl
FROM pricelists
WHERE id = 3 -- provide your id here
UNION ALL
SELECT pl.id, pl.name, pl.parent_id, c.lvl + 1
FROM cte c
JOIN pricelists pl ON pl.id = c.parent_id
)
SELECT DISTINCT ON (p.price_name)
c.id, c.parent_id, p.price_name, p.value
FROM cte c
JOIN prices p ON p.pricelist_id = c.id
ORDER BY p.price_name, c.lvl; -- lower lvl beats higher level
Use a recursive CTE like here:
Total children values based on parent
Recursive SELECT query to return rates of arbitrary depth?
There are many related answers.
Join to prices once at the end, that's cheaper.
Use DISTINCT ON the get the "greatest per group":
Select first row in each GROUP BY group?