Question

So, i have 2 tables:

Categories table

Analit  int 
PositionInMenu  int 

Item table

Analit  int 
CategoryAn  int 

So, i have about 50 categories and 2000 items. I need to take for each category count of items it includes.

1)SELECT *, (SELECT COUNT(Analit) FROM Item_table t2 WHERE t2.CategoryAn = t1.Analit) as tCount FROM Categories_table t1 ORDER BY PositionInMenu

2) Add into Categories table computed column with function call:

([dbo].[Categories_GetItemsCountInCategory]([Analit]))

and function code:

CREATE FUNCTION dbo.Categories_GetItemsCountInCategory
    (   
    @categoryId int = null
    )
RETURNS int
AS
    BEGIN
    RETURN (SELECT COUNT(Analit) 
            FROM Items 
            WHERE CategoryAn = @categoryId)
    END

And then i can simply take value of added column into my query:

SELECT *
FROM Categories_table
ORDER BY PositionInMenu

So, the question. What’s better for me?

Answer 1

You are pretty much always better off having data access inline in the query rather than in scalar UDFs.

The query optimiser does not expand out scalar (or multi statement) UDFs so you always enforce a nested loops join plan rather than allowing it to consider alternatives.

You could also consider an OUTER JOIN ... GROUP BY rather than a correlated subquery.

SELECT t1.*,
       ISNULL(tCount, 0) AS tCount
FROM   Categories_table t1
       LEFT JOIN (SELECT COUNT(Analit) AS tCount,
                         CategoryAn
                  FROM   Item_table
                  GROUP  BY CategoryAn) t2
         ON t2.CategoryAn = t1.Analit
ORDER  BY PositionInMenu