oracle sql left join and count, sum, group by
Solution 1
i ended up doing something ghastly like below by left joining two subqueries. if someone has a better way of doing this, please post.
select t1.sp_question_id, t2.projectCount, t2.amount, t2.percentTotal
from (select
q.sp_question_id
from questions q
where q.fiscal_year = 2014) t1
left join
(select
q.sp_question_id,
count(p.project_id) as projectCount,
sum(p.funding) as amount,
round(sum(p.funding)/sum(sum(p.funding)) over() *100) as percentTotal
from questions q
left join projects p on p.fiscal_year = q.fiscal_year
join objectives o on o.sp_objective_id = p.sp_objective_id
and o.sp_question_id = q.sp_question_id
and o.fiscal_year = p.fiscal_year
join funders f on p.funder_id = f.funder_id
where f.funder_short_name ='foo'
and q.fiscal_year = 2014
group by q.sp_question_id
order by q.sp_question_id) t2
on t1.sp_question_id = t2.sp_question_id
Solution 2
The problem you're having relates to one of the most unknown problems of sql: the transitivity of relationships.
You're doing an optional (outer) join between projects and questions, but you're asking for a mandatory (inner join) relationship between projects and objectives (and funders):
The transitivity system which calculates the intermediary resultsets gives prevalence to the inner join, which as a result basically means the outer join in the middle is ignored. To be exact, it is not ignored, but it becomes an inner join instead.
What you get is the result of an inner join all the way down, while you're actually expecting the left join to behave as is and return null rows for the questions not related to any projects... But the engine does not work like this, as described above.
Out of my head, I just think you could use left joins all the way down. But then you probably would face more null rows than you expect. It really depends what you're doing with the data.
Also, it seems there's a broad join in your query that should be modified. Please try this:
select
q.sp_question_id,
count(p.project_id) as projectCount,
sum(p.funding) as amount,
round(sum(p.funding)/sum(sum(p.funding)) over() *100) as percentTotal
from questions q
left join objectives o on
o.sp_question_id = q.sp_question_id
and o.fiscal_year = 2014
left join projects p on o.fiscal_year = p.fiscal_year and o.sp_objective_id = p.sp_objective_id
left join funders f on p.funder_id = f.funder_id and f.funder_short_name ='foo'
where q.fiscal_year = 2014
group by q.sp_question_id
order by q.sp_question_id;
milesmiles55
Updated on June 04, 2022Comments
-
milesmiles55 almost 2 years
i want to calculate the project count, amount, and percent total for each question restricted to a given funder and fiscal year. for some reason when i run the query below, i'm not seeing the null values from the left join.
the relationship between projects and questions is that a question can have many projects. the questions table is linked by the objectives table.
UPDATED
select q.sp_question_id, count(p.project_id) as projectCount, sum(p.funding) as amount, round(sum(p.funding)/sum(sum(p.funding)) over() *100) as percentTotal from questions q left join projects p on p.fiscal_year = q.fiscal_year join objectives o on o.sp_objective_id = p.sp_objective_id and o.sp_question_id = q.sp_question_id and o.fiscal_year = p.fiscal_year and o.fiscal_year = 2014 join funders f on p.funder_id = f.funder_id where f.funder_short_name ='foo' and q.fiscal_year = 2014 group by q.sp_question_id order by q.sp_question_id;
questionId projectCount amount percentTotal q1 14 54510 4 q2 29 1083598.72 76 q3 1 19900 1 q4 5 145631 10 q5 1 124999 9 q6 1 0 0
expected result
questionId projectCount amount percentTotal q1 14 54510 4 q2 29 1083598.72 76 q3 1 19900 1 q4 5 145631 10 q5 1 124999 9 q6 1 0 0 q7 <null> <null> <null>
query with left joins all the way down
select q.sp_question_id, count(p.project_id) as projectCount, sum(p.funding) as amount, round(sum(p.funding)/sum(sum(p.funding)) over() *100) as percentTotal from questions q left join projects p on p.fiscal_year = q.fiscal_year left join objectives o on o.sp_objective_id = p.sp_objective_id and o.sp_question_id = q.sp_question_id and o.fiscal_year = p.fiscal_year and o.fiscal_year = 2014 left join funders f on p.funder_id = f.funder_id where f.funder_short_name ='foo' and q.fiscal_year = 2014 group by q.sp_question_id order by q.sp_question_id;
result
questionId projectCount amount percentTotal na 51 1428638.72 11 q1 51 1428638.72 11 q2 51 1428638.72 11 q3 51 1428638.72 11 q4 51 1428638.72 11 q5 51 1428638.72 11 q6 51 1428638.72 11 q7 51 1428638.72 11 qother 51 1428638.72 11
query with left joins all the way down and where clauses adjusted
select q.sp_question_id, count(p.project_id) as projectCount, sum(p.funding) as amount, round(sum(p.funding)/sum(sum(p.funding)) over() *100) as percentTotal from questions q left join projects p on p.fiscal_year = q.fiscal_year left join objectives o on o.sp_objective_id = p.sp_objective_id and o.sp_question_id = q.sp_question_id and o.fiscal_year = p.fiscal_year and o.fiscal_year = 2014 left join funders f on p.funder_id = f.funder_id and f.funder_short_name ='foo' where q.fiscal_year = 2014 group by q.sp_question_id order by q.sp_question_id;
result
questionId projectCount amount percentTotal na 1225 299628985.01 11 q1 1225 299628985.01 11 q2 1225 299628985.01 11 q3 1225 299628985.01 11 q4 1225 299628985.01 11 q5 1225 299628985.01 11 q6 1225 299628985.01 11 q7 1225 299628985.01 11 qother 1225 299628985.01 11
-
milesmiles55 over 9 yearsthis makes good sense. i tried using left joins all the way down but, it's not working @Sebas
-
milesmiles55 over 9 yearsthe post has been updated. all of the questionIds are coming back on the left side but, the aggregate functions for each row is off. the amounts given are the totals for the funder. @Sebas
-
Sebas over 9 years@milesmiles55 you need to add the conditions of the where clause to the left joins they belong. Otherwise the outer joins become inner join again, due to the fact that the where clause is considered mandatory.
-
milesmiles55 over 9 yearsupdated @Sebas seems like where clauses are missing, can't put my finger on it
-
Sebas over 9 years@milesmiles55 Comment out the windowed function line (
percentTotal
). It might be the reason why you get the rolled up results. After that if it's the problem we can figure another way to do it -
milesmiles55 over 9 yearsLet us continue this discussion in chat.