Select associated records based on comparison with array column - sql

Question

Suppose a hypothetical scenario with following tables
Vehicles
+----+--------+
| id | number |
+----+--------+
| 1 | v1 |
| 2 | v2 |
| 3 | v3 |
+----+--------+
Users
+-----+------+-------------+
| id | name | vehicle_ids |
+-----+------+-------------+
| 100 | u1 | {1} |
| 200 | u2 | {1,2} |
| 300 | u3 | {2,3} |
| 400 | u4 | {} |
+-----+------+-------------+
Given number of a vehicle I am looking for a query that returns all the users associated with that vehicle,
where vehicle_ids contains the id of vehicle
where vehicle_ids only has the id of vehicle (i.e single element)

Answer 1

We can use ANY and ALL constructs (more details here Row and Array Comparisons)
USING "ANY"
http://sqlfiddle.com/#!17/ec886/21
For first case where we need to return users that are associated with vehicle represented by search query (along with others) we can use ANY. So, following query will return user rows with name u2 and u3 when supplied with v2 as vehicle name
SELECT "users".*
FROM "users"
WHERE (
(SELECT "vehicles"."id" FROM "vehicles" WHERE "vehicles"."number" = 'v2') =
ANY("users"."vehicle_ids")
)
You can replace "number" = 'v2' with any other vehicle name to get all the users which are associated with this vehicle number.
USING "ALL"
http://sqlfiddle.com/#!17/ec886/27
For second case where we need exact match i.e. only one element in the array, we will use All. So following query will return only user row with name u1 when v1
SELECT "users".*
FROM "users"
WHERE (
"users"."vehicle_ids" != '{}' and
(SELECT "vehicles"."id" FROM "vehicles" WHERE "vehicles"."number" = 'v1') = ALL("users"."vehicle_ids")
)
If we don't use "users"."vehicle_ids" != '{}' then the result will also contain user u4
Note:
This subquery method will only work for single vehicle i.e. we can't do WHERE "vehicles"."number" in ('v2', v3) as the subquery will return multiple rows and Postgres will throw an error. We might have to explore unnest if it can serve the same purpose as original question and also if it is generic enough to support aforementioned use case.

Answer 2

You can use unnest() as well:
select u.*
from users u cross join lateral
unnest(vehicle_ids) x(vehicle_id) join
vehicles v
on v.id = x.vehicle_id
where v.number = 'v2';
Here is a db<>fiddle.

Answer 3

Question #1: "where vehicle_ids contains the id of vehicle"
You can join the two tables with a condition on the array
select u.*
from users u
join vehicles v on v.id = any(u.vehicle_ids)
where v.number = 'v1';
Question #2: "where vehicle_ids only has the id of vehicle (i.e single element)"
This can be achieved by create a single element array from a SELECT statement
select u.*
from users u
where u.vehicle_ids = array(select v.id from vehicle v where v.number = 'v1');