Optimise a SQL query

Hi,

normally you would use a FROM T1 INNER JOIN T2 ON T1.PK1 = T2.K1 instead of using a CROSS JOIN and a WHERE. But modern query machines should find that out by themselves and create an identical execution plan in both cases.

Instead of creating the WHERE at the end you could reduce the number of rows before joining like this:

SELECT COUNT(distinct Q1.PK1)
FROM ( SELECT T1.PK1 FROM T1 WHERE T1.C1 = 3) AS Q1
INNER JOIN (SELECT T2.K1 FROM T2 WHERE T2.C1 > 0) AS Q2
ON Q1.PK1 = Q2.K1

Select allOpen in new window

But I think also this is something the query engine will find out by itself and optimize on it's own. Just create an execution plan for both and look at the result to see if there's any difference. You can also activate the execution time statistics (both in case of SQL Server, don't know which database you use).

Cheers,

Christian

Thanks
Will SQL stop looking in T2 as soon as it finds one T2.C1 > 0 for each PK1 candidate ?

Depending on the cardinalities involved, you need the follwoing two indices: T1(C1) and T2(C1) to support the predicates for T1.C1=3 and T2.C1>0

No problem with the predicates
But my question remains : will SQL stop looking in T2 as soon as it sees a T2.C1 > 0 for each PK involved ?

For info, the table T1 is very small and the number of rows in T2 for each T1.PK is large (2.000 per PK)
The primary key in T2 is (K1,C1)

Sure.

After rereading your post again, the index on T2 should be T2(K1, C1). Cause it should give you an index seek instead of a table scan.

When T2(K1) is indexed, a common practice, then you should get a index seek on this index with an key lookup for C1, which is slower, cause it requires the additional lookup.