Custom table join - sql

Question

Suppose I have 4 tables in Postgresql DB:
users {
id: int
}
cars {
id: int
}
usage_items {
id: int,
user_id: int,
car_id: int,
start: date,
end: date
}
prices {
id: int,
car_id: int,
price: int
}
When a user is renting a car I am creating a usage_item record to track the rental time. At the end of the month, I am sending him an invoice with calculated costs. The SQL is pretty simple here:
SELECT usage_items.start, usage_items.end, prices.price
FROM usage_items
JOIN prices ON prices.car_id = usage_items.car_id
(I omitted here WHERE clause with dates comparison, the rest of calculations I do in my Ruby code)
The problem I struggle with now is that some of my users have custom contracts with me ensuring lower prices for them. I am looking for a way to express this logic in my DB.
I came up with an idea to add user_id column to the prices table but this way I would need to create prices for every single user. So I decided to implement the following logic: if car_id in prices row is null it means that it is the default price for all the users. Otherwise, it is specific to a user. But I have no idea how to write SQL for this case, because:
SELECT usage_items.start, usage_items.end, prices.price
FROM usage_items
JOIN prices ON prices.car_id = usage_items.car_id
WHERE prices.user_id IS NULL OR prices.user_id = usage_items.user_id
returns rows for both prices. And I need only the one with an associated group or if it does not exist the one with null group_id.
Can you help me fix this SQL? Or maybe my design is bad and I should change it somehow?

Answer 1

Given your existing schema, this is one way to accomplish what you want:
Setup:
CREATE TABLE usage_items (user_id INTEGER, car_id INTEGER);
CREATE TABLE prices (user_id INTEGER, car_id INTEGER, price INTEGER);
INSERT INTO usage_items VALUES (1, 10), (2, 11), (3, 12);
INSERT INTO prices VALUES
(1, 10, 101),
(2, 11, 102),
(4, 12, 104),
(NULL, 10, 201),
(NULL, 11, 202),
(NULL, 12, 304);
Query (I'm not using start/end but it's the same thing):
SELECT DISTINCT ON (u.user_id, u.car_id) u.user_id, u.car_id, p.price
FROM usage_items u
LEFT JOIN prices p
ON u.car_id = p.car_id
AND (u.user_id = p.user_id OR p.user_id IS NULL)
ORDER BY u.user_id, u.car_id, CASE WHEN p.user_id IS NOT NULL THEN 1 ELSE 2 END
Result:
| user_id | car_id | price |
| ------- | ------ | ----- |
| 1 | 10 | 101 |
| 2 | 11 | 102 |
| 3 | 12 | 304 |
As you can see, the records in usage_items with a corresponding record for their car AND user_id in prices get their custom price rather than the NULL version; user 3 who does not have a custom price gets the NULL version (and not the custom price for a different customer).
Test here https://www.db-fiddle.com/f/wPVWEY3r22n22iKpDMrcMC/0