Question

I’d like to know the most efficient SQL query for achieving this problem:

Say we have a table with two columns, one storing entry ids (entry_id) and one storing category ids (cat_id):

 entry_id       cat_id
 3              1
 3              2
 3              3
 3              20
 4              1
 4              2
 4              21

I’d like to count how many distinct entry_id‘s there are in the categories 1, 2 OR 3 but that also must be in cat_id 20.

For example, categories 1, 2 and 3 might represent music genres (Country, Pop etc.), while category 20 might be recording formats (CD, Vinyl etc.). So another way of putting it verbally could be: “How many products are there that are on Vinyl and in either the Pop or Country category?”

I could achieve this with a nested loop in code (PHP) or possibly with a nested SQL subquery, but neither feels that efficient. I feel there must be an obvious answer to this staring me in the face…

EDIT TO ADD:
I would also like to do this without modifying the database design, as it’s a third party system.

FURTHER EXAMPLE TO CLARIFY:
Another real-world example of why I’d need this data:

Let’s say the category ids instead represent either:

• Accommodation Types (Camping = 20, Holiday Cottage = 21)

OR

• Continents and their sub-regions (i.e. Europe = 1, UK = 2, England = 3)

Let’s say someone has selected that they are interested in camping (cat_id = 1). Now we need to count how many camping products there are in the Europe. A product might be tagged as both Europe (parent), UK (child) AND England (grand-child), giving us an array of category ids 1, 2 or 3. So we now need to count how many distinct products there are in both those categories AND the original accommodation category of 1 (camping).

So having selected Camping, the end result might look something like:

• Europe: 4 camping products
  • UK: 2 camping products
    • England : 1 camping product
    • Wales : 1 camping product
  • France: 2 camping products
    etc.

Hope that helps…

Answer 1

I believe you want GROUP BY, COUNT() and EXISTS()

declare @t table(entry_id int, cat_id int)

insert @t select 1, 1
insert @t select 2, 1
insert @t select 1, 2
insert @t select 2, 2
insert @t select 3, 1
insert @t select 1, 20

select t1.cat_id, COUNT(*)
from @t as t1
where exists(
    select * from @t
    where t1.entry_id = entry_id 
    and cat_id = 20)
group by t1.cat_id

V2 using join instead of EXISTS()

declare @t table(entry_id int, cat_id int)

insert @t select 1, 1
insert @t select 2, 1
insert @t select 1, 2
insert @t select 2, 2
insert @t select 3, 1
insert @t select 1, 20

select t1.cat_id, COUNT(*)
from @t as t1
join @t as t2 on t1.entry_id = t2.entry_id and t2.cat_id = 20
group by t1.cat_id