Conditional aggregation - once for each key - sql

Question

I have an aggregation problem that can probably best be described with some example data.
Below is a dataset with transports, identified by trp_no. Each such transport is loaded in a container. A container may hold multiple such transports, and in this example any transport may only be loaded in one container.
TRP_NO TRANSPORT_VOLUME COUNTRY CONTAINER_ID CONTAINER_MAX
------ ---------------- ------- ------------ -------------
1 10 SE A 80
2 20 SE A 80
3 30 SE A 80
The following keys (or functional dependencies) exists in the dataset:
trp_no -> {transport_volume, country, container_id}
container_id -> {container_max}
I want to calculate Filling Rate per Country, which is calculated as transported volume divided by the capacity. Translated into SQL, this becomes:
with sample_data as(
select 1 as trp_no, 10 as transport_volume, 'SE' as country, 'A' as container_id, 80 as container_max from dual union all
select 2 as trp_no, 20 as transport_volume, 'SE' as country, 'A' as container_id, 80 as container_max from dual union all
select 3 as trp_no, 30 as transport_volume, 'SE' as country, 'A' as container_id, 80 as container_max from dual
)
select country
,sum(transport_volume) / sum(container_max)
from sample_data
group
by country;
...which returns (10+20+30) / (80+80+80) = 25%. Which is not what I want, because all transports used the same container_id, and my query triple-counted the capacity.
The result I want is (10+20+30) / 80 = 75%.
So, I only want to sum container_max once for each container_id within the group.
Any ideas on how to fix the query?

Answer 1

This uses Rachcha's bigger sample set, which I think is necessary to really test this problem.
with sample_data as(
select 1 as trp_no, 10 as transport_volume, 'SE' as country, 'A' as container_id, 80 as container_max from dual union all
select 2 as trp_no, 20 as transport_volume, 'SE' as country, 'A' as container_id, 80 as container_max from dual union all
select 3 as trp_no, 30 as transport_volume, 'SE' as country, 'A' as container_id, 80 as container_max from dual union all
select 4 as trp_no, 10 as transport_volume, 'SE' as country, 'B' as container_id, 100 as container_max from dual union all
select 5 as trp_no, 20 as transport_volume, 'SE' as country, 'B' as container_id, 100 as container_max from dual union all
select 6 as trp_no, 30 as transport_volume, 'SE' as country, 'B' as container_id, 100 as container_max from dual union all
select 7 as trp_no, 10 as transport_volume, 'AU' as country, 'C' as container_id, 50 as container_max from dual union all
select 8 as trp_no, 15 as transport_volume, 'AU' as country, 'C' as container_id, 50 as container_max from dual union all
select 9 as trp_no, 20 as transport_volume, 'AU' as country, 'C' as container_id, 50 as container_max from dual
),
country_container_sum as
(
select country, sum(container_max) sum_container_max
from
(
select distinct country, container_id, container_max
from sample_data
)
group by country
),
country_transport_volume_sum as
(
select country, sum(transport_volume) sum_transport_volume
from sample_data
group by country
)
select country, sum_transport_volume / sum_container_max rate
from country_container_sum
join country_transport_volume_sum using (country);
Results:
COUNTRY RATE
------- ----
SE 0.666666666666667
AU 0.9

Answer 2

I added a little more sample data for illustrating a minor fix in the query that solved it-
with sample_data as(
select 1 as trp_no, 10 as transport_volume, 'SE' as country, 'A' as container_id, 80 as container_max from dual union all
select 2 as trp_no, 20 as transport_volume, 'SE' as country, 'A' as container_id, 80 as container_max from dual union all
select 3 as trp_no, 30 as transport_volume, 'SE' as country, 'A' as container_id, 80 as container_max from dual union all
select 4 as trp_no, 10 as transport_volume, 'SE' as country, 'B' as container_id, 100 as container_max from dual union all
select 5 as trp_no, 20 as transport_volume, 'SE' as country, 'B' as container_id, 100 as container_max from dual union all
select 6 as trp_no, 30 as transport_volume, 'SE' as country, 'B' as container_id, 100 as container_max from dual union all
select 7 as trp_no, 10 as transport_volume, 'AU' as country, 'C' as container_id, 50 as container_max from dual union all
select 8 as trp_no, 15 as transport_volume, 'AU' as country, 'C' as container_id, 50 as container_max from dual union all
select 9 as trp_no, 20 as transport_volume, 'AU' as country, 'C' as container_id, 50 as container_max from dual
)
select country
,sum(transport_volume / container_max) -- Note the change here
from sample_data
group
by country;
OUTPUT:
COUNTRY SUM(TRANSPORT_VOLUME/CONTAINER_MAX)
------- -----------------------------------
SE 1.35
AU .9
EDIT:
As I see your sample data, I think you need a bit of normalization in your database. The columns for a container and columns for a transport trip should reside in separate tables like this:\
TABLE CONTAINER (
container_id VARCHAR2 / INTEGER,
container_max INTEGER,
country VARCHAR2
)
TABLE trip (
trp_no INTEGER,
transport_volume INTEGER,
container_id VARCHAR2 / INTEGER REFERENCES container.container_id
)
EDIT 2:
If you want to specifically sum up the transport volumes according to the containers' capacities, you can use something like the following query (with the same sample data table sample_data from above):
select d.country,
(select sum(t.transport_volume)
from sample_data t
where t.country = d.country) /
(select sum(c.container_max)
from ( select country, container_max
from sample_data
group by container_id, country, container_max
) c
where c.country = d.country) as col1
from sample_data d
group by d.country;
OUTPUT:
COUNTRY COL1
------- -----------
SE 0.666666667
AU 0.9

Answer 3

This approach, while other ways are simpler, uses analytic functions. I only edit with this approach because, while jonearle's response gives you the correct output, you responded saying that you wanted an approach that uses analytic functions. This approach uses analytic functions.
However, you cannot use aggregate functions nor the group by clause with analytic functions (the idea itself doesn't make sense), without adding a second layer to the query. Depending on what other similar queries you want to run, this might be easier for you as a template query, however it's hard to tell without knowing what other similar queries you're running.
with sample_data as(
select 1 as trp_no, 10 as transport_volume, 'SE' as country, 'A' as container_id, 80 as container_max from dual union all
select 2 as trp_no, 20 as transport_volume, 'SE' as country, 'A' as container_id, 80 as container_max from dual union all
select 3 as trp_no, 30 as transport_volume, 'SE' as country, 'A' as container_id, 80 as container_max from dual union all
select 4 as trp_no, 10 as transport_volume, 'SE' as country, 'B' as container_id, 100 as container_max from dual union all
select 5 as trp_no, 20 as transport_volume, 'SE' as country, 'B' as container_id, 100 as container_max from dual union all
select 6 as trp_no, 30 as transport_volume, 'SE' as country, 'B' as container_id, 100 as container_max from dual union all
select 7 as trp_no, 10 as transport_volume, 'AU' as country, 'C' as container_id, 50 as container_max from dual union all
select 8 as trp_no, 15 as transport_volume, 'AU' as country, 'C' as container_id, 50 as container_max from dual union all
select 9 as trp_no, 20 as transport_volume, 'AU' as country, 'C' as container_id, 50 as container_max from dual
)
, sub as(
select x.*, sum(x.cont_mx_n) over (partition by country order by country, container_id, trp_no) as cont_mx
from(
select country
, container_id
, trp_no
, sum(transport_volume) over (partition by country order by country, container_id, trp_no) as transp_vol
, case when lead(container_id,1) over (partition by country order by country, container_id, trp_no) = container_id
then null
else container_max end as cont_mx_n
, row_number() over (partition by country order by country, container_id, trp_no) as maxchk
from sample_data
order by country, container_id, trp_no) x)
select country, transp_vol / cont_mx as rate
from sub y
where y.maxchk = (select max(x.maxchk) from sub x where x.country = y.country);
Result of the above is:
AU 0.9
SE 0.666666666666667

Answer 4

I tried this:
with sample_data as(
select 1 as trp_no, 10 as transport_volume, 'SE' as country, 'A' as container_id, 80 as container_max from dual union all
select 2 as trp_no, 20 as transport_volume, 'SE' as country, 'A' as container_id, 80 as container_max from dual union all
select 3 as trp_no, 30 as transport_volume, 'SE' as country, 'A' as container_id, 80 as container_max from dual
)
select country
,sum(transport_volume) / container_max
from sample_data
group
by country, container_max;
The result was the expected.
ps: some nice guy remembered us about also grouping container_id, which won't affect the result in this case, but might be needed in other cases :-)