Converting PostgreSQL Subqueries into Joins - sql

Question

Below is an example schema with 3 tables. I'm trying to run a query that returns all Jobs where all child Shifts are of status 6. If a Job has a child Shift with a status of 5, the Job should not be returned. The proper response for a query from the sample data inserted below is no rows returned.
There is a working query below with the comment "Works". I am trying to refactor the "works" query to use joins instead of subqueries. The query with the comment "Does not work" is my attempt.
-- begin setup and table creation: only run this section once.
CREATE EXTENSION "uuid-ossp";
CREATE TABLE jobs
(
id uuid NOT NULL DEFAULT uuid_generate_v4(),
CONSTRAINT jobs_pkey PRIMARY KEY (id)
);
CREATE TABLE bookings
(
id uuid NOT NULL DEFAULT uuid_generate_v4(),
job_id uuid,
CONSTRAINT bookings_pkey PRIMARY KEY (id)
);
CREATE TABLE shifts
(
id uuid NOT NULL DEFAULT uuid_generate_v4(),
booking_id uuid,
status integer,
CONSTRAINT shifts_pkey PRIMARY KEY (id)
);
insert into jobs (id) values ('e857c86c-bc31-11e6-9aae-57793f585d49');
insert into bookings (id, job_id) values ('736da82c-bc32-11e6-b9b8-f36753d321ac', 'e857c86c-bc31-11e6-9aae-57793f585d49');
insert into bookings (id, job_id) values ('7d839e5c-bc32-11e6-8bb3-4fa95be86a74', 'e857c86c-bc31-11e6-9aae-57793f585d49');
insert into shifts (booking_id, status) values ('736da82c-bc32-11e6-b9b8-f36753d321ac', 6);
insert into shifts (booking_id, status) values ('7d839e5c-bc32-11e6-8bb3-4fa95be86a74', 5);
-- end setup and table creation
We want all jobs where all child shifts are of status 6. If a job has a child shift with a status of 5, the job should not be returned. The proper response for a query from the sample data inserted above is no rows returned.
Does not work :(
SELECT "jobs".*
FROM "jobs"
inner join bookings b1 on jobs.id = b1.job_id
inner join shifts s1 on b1.id = s1.booking_id
left outer join bookings b2 on jobs.id = b2.job_id
left outer join shifts s2 on b2.id = s2.booking_id and s2.status IN (2,3,4,5)
WHERE s1.status = 6
AND s2.id IS NULL
GROUP BY "jobs"."id";
Works
SELECT "jobs".*
FROM "jobs"
WHERE jobs.id IN (
SELECT job_id
FROM bookings
WHERE bookings.id IN (
SELECT booking_id FROM shifts WHERE status = 6
)
) AND jobs.id NOT IN (
SELECT job_id FROM bookings WHERE bookings.id IN (
SELECT booking_id FROM shifts WHERE status IN (2,3,4,5)
)
)
GROUP BY "jobs"."id";
How can I refactor the "works" query to use joins instead of subqueries? The "does not work" query is my attempt.

Answer 1

Try this (haven't tested so there may be typos):
with prohibited_jobs as (
select distinct jobs.id
from jobs
join bookings on jobs.id == bookings.job_id
join shifts on shifts.booking_id = booking.job_id
where shift.status != 6
)
select jobs.*
from jobs
left outer join prohibited_jobs p on p.id = jobs.id
where
p.id IS NULL
It's not completely free from subqueries (doing everything with joins would most certainly be less efficient), but it removes some unnecessary checks, so may be a little bit faster (which I suspect is your goal).
There is a small difference to your working query, in that it returns all jobs where all shifts have status 6 (as you said you want), whereas your query also ensures that the job has at least one shift (of status 6).