Checking condition in IF vs WHERE clause - sql

Question

I have a problem and I have 2 solutions for that problem. I have to write a procedure which will return rows falling into required categories. The required categories will be passed as boolean (or bit) values. The first solution is:-
DECLARE #IsRowType1Req bit;
DECLARE #IsRowType2Req bit;
DECLARE #tbl1 table (ID int, RowType varchar(50));
DECLARE #tmpTbl table (ID int, RowType varchar(50));
IF #IsRowType1Req = 1
INSERT INTO
#tmpTbl
SELECT
*
FROM
#tbl1
WHERE
RowType = 'RowType1';
IF #IsRowType2Req = 1
INSERT INTO
#tmpTbl
SELECT
*
FROM
#tbl1
WHERE
RowType = 'RowType2';
SELECT * FROM #tmpTbl;
Now, this solution uses IF clause to select only required type of rows. The other one is:-
DECLARE #IsRowType1Req bit;
DECLARE #IsRowType2Req bit;
DECLARE #tbl1 table (ID int, RowType varchar(50));
SELECT
*
FROM
#tbl1
WHERE
RowType = 'RowType1' AND
#IsRowType1Req = 1
UNION
SELECT
*
FROM
#tbl1
WHERE
RowType = 'RowType2' AND
#IsRowType2Req = 1
This solution utilizes WHERE clause to restrict unwanted type of rows. Can anybody please tell me which one will be faster and why?

Answer 1

With the information you have given there seems no need to use a UNIONof IF.
SELECT *
FROM #tbl1
WHERE
(RowType = 'RowType2' AND #IsRowType2Req = 1)
OR (RowType = 'RowType1' AND #IsRowType1Req = 1);

Answer 2

The answer would likely depend on the size of the table as well as the size of each subset returned.
The IF solution has to insert rows into another table, which doesn't come cheap. The WHERE solution, on the other hand, uses UNION which implies sorting to remove duplicate entries. If you replaced UNION with UNION ALL (which you could safely do because your two subsets would never have duplicates), the WHERE solution would become the better one of the two hands down.
However, since rows in #tbl1 cannot be more than one type, you could solve the problem differently. You could match #tbl1 against a dynamically built table consisting of 0, 1 or 2 rows depending on the values of #IsRowType1Req and #IsRowType2Req. You would build that table like this:
SELECT 'RowType1' WHERE #IsRowType1Req = 1
UNION ALL
SELECT 'RowType2' WHERE #IsRowType2Req = 1
and then join it to #tbl1:
SELECT
*
FROM
#tbl1 AS t
INNER JOIN
(
SELECT 'RowType1' WHERE #IsRowType1Req = 1
UNION ALL
SELECT 'RowType2' WHERE #IsRowType2Req = 1
) AS f (RowType)
ON
t.RowType = f.RowType
;
In a way, this might mean that WHERE still "wins", but you could rewrite the virtual table without using WHERE:
SELECT CASE #IsRowType1Req WHEN 1 THEN 'RowType1' END
UNION ALL
SELECT CASE #IsRowType2Req WHEN 1 THEN 'RowType2' END
or like this, using the VALUES constructor introduced in SQL Server 2008:
VALUES
(CASE #IsRowType1Req WHEN 1 THEN 'RowType1' END),
(CASE #IsRowType2Req WHEN 1 THEN 'RowType2' END)
That way the table would always consist of 2 rows, each row containing either a requested type or NULL. The result of the join with that table would still be the same and match your desired result.

Answer 3

In this case, I would go with the IF, but only because you are using the UNION clause and that means your are going to run two single queries and then combine the results.
If fact, when you use UNION, SQL Server needs to make sure the combined result set does not contain any duplicated rows, so it needs to run both queries and then compare the result set to check for coincidences.
Here you have a detailed explanation http://webbtechsolutions.com/2009/08/06/the-effects-union-in-a-sql-query/