Trying to formulate exclusive sql query from all-inclusive where clause - sql

Question

Is it possible to calculate this sales report in one query? I'm not so hot with SQL and having trouble wrapping my head around joins considering the all-inclusive where clause.
TABLES (assume 1-N for foreign keys):
TABLE contacts
id
TABLE reservations
id
created_at
contact_id
TABLE reservation_trips
id
reservation_id -- never null
amount
quantity
trip_type_id -- never null
TABLE cost_adjustments
id
reservation_trip_id -- never null
cost_type_id -- never null
TABLE trip_types
id
title
TABLE cost_types
id
title
kind
EXAMPLE WHERE:
reservations.created_at >= '2013-01-01T09:00:00+00:00' AND
reservations.created_at < '2014-01-01T09:00:00+00:00' AND
reservation_trips.trip_type_id in (1, 2, 3, 4) AND
cost_adjustments.trip_type_id in (5, 6, 7, 8) AND
contact_id in (9, 10, 11, 12)
RESULT COLUMNS:
trip_types.title as trip_type_title,
COUNT(DISTINCT reservations.id) as bookings, -- distinct reservations
SUM(reservations.commissions) as commissions, -- distinct reservations
SUM(reservation_trips.id) as guests, -- distinct reservation_trips
SUM(reservation_trips.amount * reservation_trips.quantity) as gross_sales, -- distinct reservation_trips
SUM(cost_adjustments.amount * cost_adjustments.quantity) as adjustments, -- distinct cost_adjustments where cost_types.kind != 'tax'
SUM(gross_sales + adjustments - commissions) as net_sales
The table columns would look like this:
TRIP TYPE | BOOKINGS | GUESTS | GROSS SALES | ADJUSTMENTS | COMMISSIONS | NET SALES
Thanks!

Answer 1

The main problem is to avoid counting the intermediate levels multiple times. One way to do it is do divide by the number of items at each level.
Another approach, which can be more efficient is to put each level in it's own subquery and then join them all up.
Here's an untested stab at the first method. It might need some IsNull(xxx, 0) fixup.
Select
tt.title as trip_type_title,
count(distinct r.id) as bookings,
sum(r.commissions) / count(distinct r.id) as commissions,
count(distinct rt.id) as guests, -- this seems odd, but it's in the example
sum(rt.amount * rt.quantity) / count(distinct rt.id) as gross_sales,
sum(ca.amount * ca.quantity) as adjustments,
sum(rt.amount * rt.quantity) / count(distinct rt.id) + sum(r.commissions) / count(distinct r.id) + sum(ca.amount * ca.quantity) as net_sales
From
trip_types tt
inner Join reservation_trips rt on tt.id = rt.trip_type_id
inner Join reservations r on rt.reservation_id = r.id
left outer join cost_adjustments ca on ca.reservation_trip_id = rt.id
left outer join cost_types ct on ca.cost_type_id = ct.id
Where
r.created_at >= '2013-01-01T09:00:00+00:00' and
r.created_at < '2014-01-01T09:00:00+00:00' and
rt.trip_type_id in (1, 2, 3, 4) and
ca.trip_type_id in (5, 6, 7, 8) and
contact_id in (9, 10, 11, 12) and
ct.kind != 'tax'
Group By
tt.title

Answer 2

Give this a whirl, I think it covers it
SELECT trip_types.title as trip_type_title,
COUNT(DISTINCT reservations.id) as bookings,
SUM(reservations.commissions) as commissions, distinct reservations,
SUM(reservation_trips.id) as guests, distinct reservation_trips,
SUM(reservation_trips.amount * reservation_trips.quantity) as gross_sales,
SUM(cost_adjustments.amount * cost_adjustments.quantity) as adjustments, distinct cost_adjustments where cost_types.kind != 'tax'
SUM(gross_sales + adjustments - commissions) as net_sales
FROM trip_types tt
JOIN reservation_trips on tt.id = reservation_trips.trip_type_id
JOIN reservations on reservation_trips.reservation_id = reservations.id
JOIN cost_adjustments ON cost_adjustments.reservation_trip_id = reservation_trips.id
WHERE
reservations.created_at >= '2013-01-01T09:00:00+00:00' AND
reservations.created_at < '2014-01-01T09:00:00+00:00' AND
reservation_trips.trip_type_id in (1, 2, 3, 4) AND
cost_adjustments.trip_type_id in (5, 6, 7, 8) AND
contact_id in (9, 10, 11, 12)
GROUP BY trip_types.title

Answer 3

I conferred with a friend and figured it out. The trick that I didn't explain very well in the question (sorry) was that the search should include all trips and adjustments for reservations patching the search terms in my example where clause. So you have to use a 'with' query to first find the appropriate reservations, and then filter trips and adjustments based on that. Very tricky.
Here is the answer we found with slightly different fields actually used in production, but operate to the same effect:
with reservations_search as (
select
distinct reservations.id
from
reservations
left outer join contacts on
reservations.contact_id = contacts.id
left outer join reservation_trips on
reservations.id = reservation_trips.reservation_id
left outer join cost_adjustments on
reservation_trips.id = cost_adjustments.cost_target_id
where
#{Reservation.warez_sql(params)[0].join(' AND ')}
)
select
res_level.breeze_id,
res_level.trip_category_id,
res_level.bookings,
res_level.commissions,
res_level.guests,
adjustments_level.adjustments,
res_level.gross_sales + adjustments_level.adjustments - res_level.commissions as net_sales
FROM
(select
breeze_id,
trip_category_id,
sum(trip_commission) as commissions,
count(distinct reservation_id) as bookings,
count(reservation_trip_id) as guests,
sum(gross_sales) as gross_sales
from
-- Commissions by reservation_trips.id
(select
reservations.id as reservation_id,
reservation_trips.id as reservation_trip_id,
reservation_trips.amount * reservation_trips.quantity as gross_sales,
case
when reservations.total_cost_commissionable > 0
then coalesce(reservations.commission, 0) * reservation_trips.total_cost_commissionable / reservations.total_cost_commissionable
else 0
end as trip_commission,
trip_types.breeze_id,
trip_types.trip_category_id
from
reservations
left outer join contacts on
reservations.contact_id = contacts.id
left outer join reservation_trips on
reservations.id = reservation_trips.reservation_id
left outer join trip_types on
reservation_trips.trip_type_id = trip_types.id
where
reservations.id in (select id from reservations_search) ) as trip_level
group by
breeze_id,
trip_category_id) as res_level
-- Adjustments
left outer join
(select
trip_types.breeze_id,
sum(cost_adjustments.amount * cost_adjustments.quantity) as adjustments
from
reservation_trips
left outer join reservations on
reservation_trips.reservation_id = reservations.id
left outer join cost_adjustments on
cost_target_type = 'ReservationTrip' and
reservation_trips.id=cost_adjustments.cost_target_id
left outer join cost_types on
cost_type_id = cost_types.id
left outer join trip_types on
reservation_trips.trip_type_id=trip_types.id
where
reservations.id in (select id from reservations_search) and
cost_types.commissionable IS TRUE
group by
trip_types.breeze_id) as adjustments_level on
res_level.breeze_id = adjustments_level.breeze_id