Use one select query instead of query + subquery - sql

Question

SQL Server 2014, data has been changed to protect the data. I hope the below makes sense.
I have to search a table for all categories with a Moderate risk. Then I need to go into another table (test) and retrieve those Moderate categories where only the Toby test has failed. Categories are unique in that first table.
So in the sample data below, categories 4 and 5 both are Moderate risk, and both also have Toby with a result of Fail in the test table.
However I want to exclude category 4 from my final output because that Category also has the Bill test that failed.
My goal is only to show Category 5 as an output. I can do this with a query and sub-query, where the sub-query returns categories 4 and 5, and the main query filters on that. But can I achieve the same thing with a single query somehow?
Update:
My current query is below. I've had to munge it a bit for this post, I hope it's sufficient. Basically the sub-query pulls in all categories that have any failed test for a Moderate category, and the main query filters out any categories with other failures.
select tt.category, tt.[name]
from test_table mt
where tt.category in (select mt.category
from main_table mt
inner join test_table tt on tt.category = mt.category
where mt.risk= 'Moderate' and tt.result = 'Fail')
and tt.[name] <> 'Toby'
and tt.[result] = 'Fail'
Output:
category risk
----------------------
1 Minimal
2 Critical
3 Elevated
4 Moderate
5 Moderate
category name result
-------------------------------
1 Mark Pass
1 Bill No Result
1 John Pass
1 Toby Pass
2 Mark Pass
2 Bill No Result
2 John Fail
2 Toby Pass
3 Mark Pass
3 Bill No Result
3 John Pass
3 Toby Pass
4 Mark Pass
4 Bill Fail
4 John Pass
4 Toby Fail
5 Mark Pass
5 Bill Pass
5 John Pass
5 Toby Fail

Answer 1

Join the tables, group by category and set the conditions in the HAVING clause:
select c.category
from categories c inner join test t
on c.category = t.category
where c.risk = 'Moderate'
group by c.category
having
sum(case when t.name = 'Toby' and t.result = 'Fail' then 1 else 0 end) > 0
and
sum(case when t.name <> 'Toby' and t.result = 'Fail' then 1 else 0 end) = 0
or:
select c.category
from categories c inner join test t
on c.category = t.category
where c.risk = 'Moderate' and t.result = 'Fail'
group by c.category
having count(distinct t.name) = 1 and max(t.name) = 'Toby'
See the demo.
Results:
> | category |
> | -------: |
> | 5 |

Answer 2

A pretty direct reading of your question suggests exists/not exists:
select c.*
from categories c
where c.risk = 'Moderate' and
exists (select 1
from tests t
where t.category = c.category and
t.name = 'Toby' and
t.result = 'Fail'
) and
not exists (select 1
from tests t
where t.category = c.category and
t.name <> 'Toby' and
t.result = 'Fail'
);
Conditional aggregation is also a viable solution -- and a solution that I prefer when the filter is on only one table. However, the filtering here is on two tables, and this will use of an index on test(category, result, name) and avoid the outer aggregation.