SQL: Get records having min/max value per group

Here’s one possible solution for a case when you need to fetch a set of records unique on one (or several) column with a min/max value in the other column (usually a timestamp). The solution doesn’t require any changes to the DB scheme (denormalization or creation of views) and can be performed in the scope of a single SQL query.

Scenario

Imagine you have a table named customer_subscriptions with the following fields:

  • user_name (we’ll use name instead of user_id for simplicity here consider user_name to be unique for this example)
  • plan_name (we’ll use name instead of plan_id here as well)
  • created_at (the timestamp when the subscription was created)

For illustration, consider the following data in customer_subscriptions:

Sample customer_subscriptions table

Let’s say we want to show a table showing what was the first subscription plan for each user:

Desired query result

Subquery JOIN solution

  1. First, we must find the minimum created_at value for each user subscription entry separately in a subquery.
  2. Then, we join the subquery on user_name and created_at to filter down to the first subscription record for each user.
SELECT cs.user_name, cs.plan_name, cs.created_at
FROM customer_subscriptions cs
JOIN (
    SELECT user_name, MIN(created_at) AS min_created_at
    FROM customer_subscriptions
    GROUP BY user_name
) AS sub
ON cs.user_name = sub.user_name AND cs.created_at = sub.min_created_at;

This solution would work perfectly for both MySQL and PostgreSQL.

PostgreSQL-specific solution

In PostgreSQL, we can use DISTINCT ON to filter the resulting set to be unique on specific columns, such as user_name. We won’t need to join an additional subquery.

  1. DISTINCT ON (user_name): Ensures that each user_name appears only once in the result set.
  2. ORDER BY user_name, created_at: Orders the records by user_name and then by created_at within each user_name, so the first record (with the smallest created_at value) is selected.
SELECT DISTINCT ON (user_name) user_name, plan_name, created_at
FROM customer_subscriptions
ORDER BY user_name, created_at;

Conclusion

Fetching the first subscription record for each user based on the earliest created_at value can be efficiently achieved using different SQL constructs available in MySQL and PostgreSQL. While MySQL relies on subqueries and joins, PostgreSQL’s DISTINCT ON clause provides a straightforward and elegant solution.