Question

I’m looking for a more efficient way to build this query (double subqueries make me cringe):

SELECT contact_id FROM (
  SELECT * FROM (
    SELECT mr.contact_id, di.district
    FROM recipients mr
    JOIN address a ON mr.contact_id = a.contact_id
    JOIN district_values di ON a.id = di.entity_id
    WHERE mr.mid = 29
    ORDER BY di.district DESC ) addrSingle
  GROUP BY mr.contact_id ) addrNull
WHERE di.district IS NULL

Let me explain what’s going on here.

Recipients holds a list of contacts. Each contact may have multiple addresses. Each address has a related district_values table. I need to retrieve contacts where the district_values.district column is null for ALL addresses.

For example:

Contact A
Address 1.district = 4
Address 2.district = null
= don't include

Contact B
Address 1.district = null
= include

Contact C
Address 1.district = null
Address 2.district = 3
= don't include

The logic of my existing query is as follows:

1. retrieve contacts with related addresses and districts, order so that any addresses with a non null value are ordered first
2. apply group by so i reduce to a single contact record and if addresses with a district are retained
3. apply where clause to remove addresses with at least one district value

It works — it’s just a bit ugly.

Answer 1

You could try this, use LEFT JOIN and count the related record which is zero.

SELECT mr.contact_id 
    FROM recipients mr 
LEFT JOIN address a ON mr.contact_id = a.contact_id  
LEFT JOIN district_values di ON a.id = di.entity_id 
    WHERE mr.mid = 29 
GROUP BY mr.contact_id
    HAVING COUNT(a.*) = 0 AND COUNT(di.*) = 0