Either of two SQL tables have a value - sql

Question

I have two tables, task_list_sharee and task_list_assignee. They both have a reference to a task_list table.
There's also a task table that has a reference to the task_list table since a task always exists within a task_list.
Given a task, I want to find out if either task_list_sharee OR task_list_assignee have values. Right now I'm doing it as two SQL statements, like so:
SELECT count(*)
FROM task_list_assignee a
INNER JOIN task_list l ON l.uid = a.task_list_uid
INNER JOIN task t ON t.task_list_uid = l.uid
WHERE t.uid = ?
SELECT count(*)
FROM task_list_sharee s
INNER JOIN task_list l ON l.uid = s.task_list_uid
INNER JOIN task t ON t.task_list_uid = l.uid
WHERE t.uid = ?
and if either is non-zero I punt. I'm thinking this has to be doable as just a single SQL statement but I'm a bit stumped.

Answer 1

Performance of a full count on multiple joins (even more so for LEFT JOIN) can deteriorate quickly. While all you need is proof for the existence of a single row, there is no need for this. Use EXISTS - true to its name - to allow an optimal query plan:
SELECT EXISTS (
SELECT 1
FROM task t
WHERE t.uid = ? -- provide uid here
AND (
EXISTS (
SELECT 1
FROM task_list_assignee
WHERE task_list_uid = t.task_list_uid
)
OR EXISTS (
SELECT 1
FROM task_list_sharee
WHERE task_list_uid = t.task_list_uid
)
)
);
Should be substantially faster than a full count.
I also cut out the middleman. Joining to task_list only establishes that the related row in task_list exists - which is a waste of time given that:
a task always exists within a task_list.
Ideally implemented with FK constraints to enforce referential integrity.
In the absence of actual table definitions my educated guess will have to do.
To make this fast for any table size, you need 3 btree (default) indexes on
task(uid, task_list_uid)
task_list_assignee(task_list_uid)
task_list_sharee(task_list_uid)

Answer 2

If you start with Task as your base table and do a series of left joins, you should be able to determine which tasks have values for both assignee and sharee with a coalesce:
select count(*)
from task as t
left join
task_list as l
on t.task_list_uid = l.uid
left join
task_list_assignee as a
on l.uid = a.task_list_uid
left join
task_list_sharee as s
on l.uid = s.task_list_uid
where coalesce( a.task_list_uid, s.task_list_uid ) is not null
and t.uid = ?
SQL Fiddle here