Guidance on getting desired SQL results

Question

Thanks to the help on my last question , I now have one final step to fully grasp this method of querying for my problem.

Given these two tables:

airlines_detail :

airplane_id	airline_id	total_seats
2187	425	260
4361	747	290
3478	747	270
7292	425	250
5833	425	310
3472	425	300
4472	747	290
2624	425	320

Bookings :

airplane_id	booked
2187	40
4361	30
4361	10
5833	30
4361	30
2624	30
4472	40
4472	40
2624	10
7292	20
2187	20
4472	30
3478	20
3472	40
4472	30
4472	10
4361	20
3478	30
2187	30
2187	10

I need to get this result:

airline_id	airplanes
425	7292
747	4361

The result is the airplane_id whose SUM of empty seats is closest to the AVG number of empty seats per each airline_id (in this case, there's only 425 and 747 to worry about).

So, we would need to find the AVG of SUM(empty_seats) GROUP BY airline_id , then compare the SUM(empty_seats) GROUP BY airplane_id for each airline, and return a table that shows the airplanes that are closest to their respective airline's average of empty seats.

For airline with id 425 the average number of empty seats are 242. Therefore the airplane which is closest to this average is 7292.

Equipped with the answer from the previous question in getting the sum of empty seats for each plane, I'm hoping this next step is a simple one for a guru.

Answer 1

You would start with the query you have, then use that as an intermediate result to get the averages per airline, and then rank the airplanes by how much their empty seat count deviates from their airline's average. For that ranking you can use ROW_NUMBER() and a windowing clause ( OVER ). Finally, select the rows in that result that are ranked first.

For this chain of building upon previous results, it is handy to use the WITH syntax:

WITH base as (
    SELECT
        a.airplane_id,
        a.airline_id,
        a.total_seats - SUM(b.booked) AS empty_seats
    FROM
        airlines_detail a
    LEFT JOIN bookings b
        ON
            a.airplane_id = b.airplane_id
    GROUP BY  
        a.airplane_id,
        a.airline_id
), average as (
    SELECT
        airline_id,
        AVG(empty_seats) avg_empty_seats
    FROM
        base
    GROUP BY
        airline_id
), ranked as (
    SELECT
        b.airplane_id,
        b.airline_id,
        ROW_NUMBER() OVER (
            PARTITION BY b.airline_id 
            ORDER BY ABS(b.empty_seats - a.avg_empty_seats)
        ) rn
    FROM
        base b
    INNER JOIN average a
        ON
            a.airline_id = b.airline_id
)
SELECT
    airline_id,
    airplane_id
FROM
    ranked
WHERE
    rn = 1;