Question

I’m storing a tree in Postgres linked via a parent_id column. I want to find the leaves of the tree, as in, those elements that have no children. Here’s the query I have so far:

where("id NOT IN ( SELECT parent_id FROM elements WHERE parent_id IS NOT NULL )")

Is there a better way to write this query? I was trying to think of a way to handle it with a join, as I’ve heard that’s more efficient, but wasn’t able to come up with anything. Also, is there a way to do this without raw SQL and instead use ActiveRecord?

UPDATE: Here’s the schema:

   Column   |            Type             |     Modifiers      
------------+-----------------------------+--------------------
 id         | character varying(36)       | not null
 user_id    | character varying(36)       | not null
 parent_id  | character varying(36)       | 
 title      | character varying(255)      | 
 created_at | timestamp without time zone | not null
 updated_at | timestamp without time zone | not null
Indexes:
    "index_elements_on_id" UNIQUE, btree (id)
    "index_elements_on_parent_id" btree (parent_id)
    "index_elements_on_user_id" btree (user_id)

Answer 1

The technique is called “exclusion join”.

select e.* from elements e
left outer join elements leafs
on e.id = leafs.parent_id
where leafs.parent_id is null