Question

Alright, let’s say I have these two tables:

• items with columns id, stuff
• item_properties with columns item_id, prop_id

Now I want to execute a query like

SELECT stuff FROM items WHERE
      EXISTS(SELECT * FROM item_properties WHERE prop_id = 123 AND item_id = items.id)
  AND EXISTS(SELECT * FROM item_properties WHERE prop_id = 456 AND item_id = items.id)
  AND NOT EXISTS(SELECT * FROM item_properties WHERE prop_id = 789 AND item_id = items.id)
  AND NOT EXISTS(SELECT * FROM item_properties WHERE prop_id = 101 AND item_id = items.id)

Which works, but looks ugly and is slow. Can anyone think of a smarter way to do this? I can also get the 123,456 and 789,101 lists via a subquery from a third table, if necessary. I am open to suggestions to change my table design as well.

The number of property IDs I need to check the properties of an item against can vary.

Thanks!

Answer 1

if you have a table that contains prop_id to include and to exclude itemsIU (item_id, prop_id, include)

select distinct stuff 
  from items i
  join item_properties ip on i.id = ip.item_id
  join itemsIU iiu on ip.prop_id = iiu.prop_id
group by i.id
having sum(include) = (select count(1) 
                         from itemsIU iiu2 
                        where i.id = iiu2.item_id
                          and iiu2.include = 1)

for you particular example you can use:

select distinct stuff 
  from items i
  join item_properties ip on i.id = ip.item_id
  join (          select 123 prop_id, 1 include
        union all select 456, 1
        union all select 789, 0
        union all select 101, 0) iiu on ip.prop_id = iiu.prop_id
group by i.id
having sum(include) = 2