SQL Join and count relations - sql

Question

Having a PostgreSQL query problem, and wondering if there's an efficient way to get this in a single query. Let's take the following simple table structure. Think of it as the traditional many to many relationship.
users <-> user_collections <-> collections
Given a users id, I'd like to first get all of their collections. This is the simple part for which I have a query:
SELECT c.id, c.name, c.description, c.created_at, c.updated_at
FROM collections c
JOIN user_collections uc ON c.id = uc.collection_id
WHERE uc.user_id = $1
ORDER BY created_at DESC
So for example:
users
id | email
1 | user1#example.com
2 | user2#example.com
3 | user3#example.com
user_collections
id | user_id | collection_id
1 | 1 | 1
2 | 2 | 1
3 | 3 | 1
collections
id | name | description
1 | Example | Demo collection
In the above case, querying for collections for user one would yield the first collection. However I'd also like to get a count of how many users are associated with each collection. In this case, a total count of 3, since all three members share this collection. A member count if you will. Is there a sensible way to do this in one query, or is two probably better?

Answer 1

An alternative approach to using correlated subquery is to pre-calculated the number of users in each collection, then join it to your existing query.
i.e.
with collection_counts as (
select
collection_id
, count(1) as collection_count
from user_collections
group by collection_id
)
SELECT
c.id
, c.name
, c.description
, c.created_at
, c.updated_at
, cc.collecion_count
FROM collections c
JOIN user_collections uc ON c.id = uc.collection_id
join collection_counts as cc on c.id = cc.collection_id
WHERE uc.user_id = $1
ORDER BY created_at DESC

Answer 2

You need to use here a correlated sub query to get the corresponding total user of that collection [Correlated sub query help https://www.geeksforgeeks.org/sql-correlated-subqueries/]
SELECT c.* , (select count(user_id) from user_collections sc where sc.collection_id=uc.collection_id) as GroupCount
FROM collections c
JOIN user_collections uc ON c.id = uc.collection_id
WHERE uc.user_id = $1
ORDER BY created_at DESC

Answer 3

One method is conditional aggregation:
SELECT c.id, c.name, c.description, c.created_at, c.updated_at,
COUNT(*) as num_users
FROM collections c JOIN
user_collections uc
ON c.id = uc.collection_id
GROUP BY c.id
HAVING COUNT(*) FILTER (WHERE uc.user_id = $1) > 0
ORDER BY created_at DESC;
That said, it might be faster to do:
SELECT c.id, c.name, c.description, c.created_at, c.updated_at,
(SELECT COUNT(*)
FROM user_collections uc
WHERE c.id = uc.collection_id
) as num_users
FROM collections c
WHERE EXISTS (SELECT 1
FROM user_collections uc
WHERE c.id = uc.collection_id AND
uc.user_id = $1
)
ORDER BY created_at DESC;
This would be faster for two reasons:
It avoids the outer aggregation. Aggregations on larger data are generally more expensive.
It calculates the count only for the collections that are in the result set.
This can also make use of indexes on the table -- which if you care about performance, you should have.