Question

I have a table which contains the item and category ids:

create table SomeTable (
    ItemId int,
    CategoryId int
)

Given some category ids (Set X) I would like to determine all item ids that share at least one category id and some stats for each of these item ids:

A – Number of category ids of item id that are not in set x
B – Number of category ids shared between item id and set x
C – Number of category ids in set x but which are not associated with item id

I have written some tsql code which involves a cross join and several ctes plus left joins. It works but is fairly slow.

I am sure someone must have encountered a similar problem. I would provide the code but the above description is simplified. Thanks.

Answer 1

Here’s a couple of ideas. (I don’t know how they’ll compare performance wise with what you have already. Left for you to benchmark.)

set nocount on;

-- create a sample table
declare @T table ( ItemId int identity(1,1), CategoryId int );
insert @T values ( 100 );
insert @T values ( 100 );
insert @T values ( 100 );
insert @T values ( 100 );
insert @T values ( 100 );
insert @T values ( 200 );
insert @T values ( 200 );
insert @T values ( 300 );
insert @T values ( 300 );
insert @T values ( 300 );
insert @T values ( 300 );
insert @T values ( 500 );
insert @T values ( 500 );
insert @T values ( 500 );
insert @T values ( 600 );
insert @T values ( 700 );
insert @T values ( 800 );
insert @T values ( 800 );
insert @T values ( 800 );
insert @T values ( 900 );


-- grab some CategoryIDs to work with
declare @X table ( CategoryId int );
insert @X
    select CategoryID=200 union
    select CategoryID=400 union
    select CategoryID=600 union
    select CategoryID=800 


-- A. Number of category ids of item id that are not in set x
select distinct t.CategoryID from @T t
where not exists(select 1 from @X x where t.CategoryID = x.CategoryID)

-- or, using the set difference operator
select CategoryID from @T
except
select CategoryID from @X


-- B. Number of category ids shared between item id and set x
select distinct x.CategoryID from @X x
join @T t on t.CategoryID = x.CategoryID;

-- or, using set intersection
select CategoryID from @T
intersect
select CategoryID from @X


-- C. Number of category ids in set x but which are not associated with item id 
select distinct x.CategoryID from @X x
where not exists(select 1 from @T t where t.CategoryID = x.CategoryID)

-- or, using the set difference operator
select CategoryID from @X
except
select CategoryID from @T