Question

I’m wondering if there’s any way to optimize the following SELECT query. (Note: I typed this when writing my question for nonexistent tables and I might not have the correct syntax.)

The goal is, if Table2 contains any related rows I want to set the value of the third column to the number of related rows in Table2. Otherwise, if Table3 contains any related rows I want to set the column to the number of related rows in Table3. Otherwise, I want to set the column value to 0.

SELECT Id, Title,
    CASE
        WHEN EXISTS (SELECT * FROM Table2 t2 WHERE t2.RelatedId = Table1.Id) THEN
            (SELECT COUNT(1) FROM Table2 t2 WHERE t2.RelatedId = Table1.Id)
        WHEN EXISTS (SELECT * FROM Table3 t3 WHERE t3.RelatedId = Table1.Id) THEN
            (SELECT COUNT(1) FROM Table3 t3 WHERE t3.RelatedId = Table1.Id)
        ELSE 0
    END AS RelatedCount
    FROM Table1

I don’t like the fact that I’m basically performing the same query twice (in two cases). Is there any way to do what I want while only performing the query once?

Note that this is part of a much larger query with multiple JOINs and UNIONs so it’s not easy to take a completely different approach.

Answer 1

This query should perform much better. You are not just performing the same query twice; since they are correlated subqueries, they will run once per row.

SELECT Id, Title,
    coalesce(t2.Count, t3.Count, 0) AS RelatedCount
    FROM Table1 t
left outer join (
    SELECT RelatedId, count(*) as Count
    FROM Table2
    group by RelatedId
) t2 on t1.Id = t2.RelatedId
left outer join (
    SELECT RelatedId, count(*) as Count
    FROM Table3
    group by RelatedId
) t3 on t1.Id = t3.RelatedId