SQL query for statistic info - sql

Question

So let's assume I have two tables, cars and engines. Every car has an id column and sold_date column which points to a date when the car was sold. Every engine has car_id and type columns, where the first one is a foreign key to the cars table and the second is an engine's type name, which can be anything from V1 to V999.
So what I want to get is a list of dates from let's say August 1st to August 3rd with every type of engine and a number of sold cars like this:
sold_date engine_type number_of_sold_cards
08.01.2015 V8 6
08.01.2015 V6 8
08.01.2015 V4 9
08.02.2015 V8 15
08.02.2015 V6 0
08.02.2015 V4 5
08.03.2015 V8 4
08.03.2015 V6 6
08.03.2015 V4 0
The example assumes that for these 3 days were sold only cars with engines' types of V8, V6 and V4. What it means is that if there had been sold 5 types of engines for the period (V8, V6, V4, V2, V0) instead, I'd need 5 rows for every date

Answer 1

Partition outer join to the rescue!
with cars as (select 1 id, to_date('01/08/2015', 'dd/mm/yyyy') sold_date from dual union all
select 2 id, to_date('01/08/2015', 'dd/mm/yyyy') sold_date from dual union all
select 3 id, to_date('02/08/2015', 'dd/mm/yyyy') sold_date from dual union all
select 4 id, to_date('03/08/2015', 'dd/mm/yyyy') sold_date from dual union all
select 5 id, to_date('01/08/2015', 'dd/mm/yyyy') sold_date from dual union all
select 6 id, to_date('01/08/2015', 'dd/mm/yyyy') sold_date from dual union all
select 7 id, to_date('03/08/2015', 'dd/mm/yyyy') sold_date from dual union all
select 8 id, to_date('03/08/2015', 'dd/mm/yyyy') sold_date from dual union all
select 9 id, to_date('02/08/2015', 'dd/mm/yyyy') sold_date from dual union all
select 10 id, to_date('02/08/2015', 'dd/mm/yyyy') sold_date from dual union all
select 11 id, to_date('02/08/2015', 'dd/mm/yyyy') sold_date from dual union all
select 12 id, to_date('04/08/2015', 'dd/mm/yyyy') sold_date from dual union all
select 13 id, to_date('04/08/2015', 'dd/mm/yyyy') sold_date from dual union all
select 14 id, to_date('04/08/2015', 'dd/mm/yyyy') sold_date from dual union all
select 14 id, to_date('04/08/2015', 'dd/mm/yyyy') sold_date from dual),
engines as (select 1 car_id, 'V8' engine_type from dual union all
select 2 car_id, 'V8' engine_type from dual union all
select 3 car_id, 'V8' engine_type from dual union all
select 4 car_id, 'V8' engine_type from dual union all
select 5 car_id, 'V4' engine_type from dual union all
select 6 car_id, 'V4' engine_type from dual union all
select 7 car_id, 'V4' engine_type from dual union all
select 8 car_id, 'V4' engine_type from dual union all
select 9 car_id, 'V6' engine_type from dual union all
select 10 car_id, 'V6' engine_type from dual union all
select 11 car_id, 'V6' engine_type from dual union all
select 12 car_id, 'V6' engine_type from dual union all
select 13 car_id, 'V2' engine_type from dual union all
select 14 car_id, 'V2' engine_type from dual union all
select 15 car_id, 'V0' engine_type from dual),
engine_types as (select distinct engine_type from engines),
res as (select c.id car_id,
e.engine_type,
c.sold_date
from engines e
inner join cars c on (e.car_id = c.id)),
final_res as (select et.engine_type,
res.sold_date
from engine_types et
left outer join res on (et.engine_type = res.engine_type)),
dates as (select to_date('01/08/2015', 'dd/mm/yyyy') - 1 + level dt -- paramaterise the start date
from dual
connect by level <= to_date('03/08/2015', 'dd/mm/yyyy') - to_date('01/08/2015', 'dd/mm/yyyy') + 1 -- paramaterise the start and end dates
)
select dts.dt,
fr.engine_type,
count(fr.sold_date) cnt
from dates dts
left outer join final_res fr partition by (fr.engine_type) on (dts.dt = fr.sold_date)
group by dts.dt,
fr.engine_type
order by dts.dt,
fr.engine_type;
DT ENGINE_TYPE CNT
---------- ----------- ----------
01/08/2015 V0 0
01/08/2015 V2 0
01/08/2015 V4 2
01/08/2015 V6 0
01/08/2015 V8 2
02/08/2015 V0 0
02/08/2015 V2 0
02/08/2015 V4 0
02/08/2015 V6 3
02/08/2015 V8 1
03/08/2015 V0 0
03/08/2015 V2 0
03/08/2015 V4 2
03/08/2015 V6 0
03/08/2015 V8 1
The first and second subqueries ("cars" and "engines") are just mimicking your tables; you would not need to include them in your query.
The "engine_types" subquery is just getting the distinct list of engine_types used in the engines table. If you have some other table that lists the available engine_types, then use that instead.
The "dates" subquery is just generating a list of dates between a given date range - in an ideal world, the start and end dates would be parameterised (assuming this is being run in PL/SQL or some such).
The "res" subquery does the join on the cars and engines table, to get the type of engine sold on each date.
The "final_res" subquery outer joins the res and engine_types subqueries, so that every engine type is listed, along with the sold_date, if it was sold.
Once you have that, then it's easy to do a partition outer join on the final_res subquery to the dates subquery.

Answer 2

You need to join cars and engine tables. Need to count the sold cars and group by sold_date and engine_type:
select c.sold_date
,e.engine_type
,count(*) as number_of_sold_cars
from cars c
inner join engines e on c.id = e.car_id
where c.sold_date between date '2015-08-01' and date '2015-08-03'
group by c.sold_date,e.engine_type
Update:
With this query you will have all engine types sold in the selected period. You will have 0 for number_of_sold_cars if this engine type is not sold on the particular date:
with engine_type as (
select distinct e.engine_type
from cars c
inner join engines e on c.id = e.car_id
where c.sold_date between date '2015-08-01' and date '2015-08-03'
)
select c.sold_date
,t.engine_type
,count(*) as number_of_sold_cars
from cars c
inner join engines e on c.id = e.car_id
left join enginte_type t on t.engine_type = e.engine_type
where c.sold_date between date '2015-08-01' and date '2015-08-03'
group by c.sold_date,t.engine_type