Group by is slow with Order by

Can you please post the explain plan of your SQL with Order by and without Order by and the Database version?

HOW TO: Post a SQL statement tuning request - template posting

What about the structure of those tables?

What does the explain plan show? I mean what does the explain-plan look in with order by and without order by? Maybe order by determines the use of some index that is not so good. If that be the case, an alter index rebuild may help. There must be a sort group by in the select, then another sort for the order. Still, does the access path for the joined table change in case you add the order by? In case it does change, then try what Nimish suggested - I'd try that anyway.

3154057 wrote:Hi All,I am running following SQL, which it taking almost 4-5 seconds and returning only 18 records.SELECT  SUM(TOTAL_AMOUNT) C1, a.LEVEL2_ENAME AS c2, c.CURR_YEARMO AS c3 FROM TERRITORY_PRESET_MSNP a, CV_RESTATED_MSNP b LEFT OUTER JOIN MONTH_D c ON b.YEARMO = c.CURR_YEARMO, PRODUCT_OFFERING dWHERE ( b.PO_ID = d.ROW_ID AND b.DATASOURCE_ID = 10 AND b.YEARMO = 201704 AND b.OWNER_TERR_ID = a.TERR_ID AND c.CURR_YEARMO = 201704 AND a.YEARMO = 201706GROUP BY c.CURR_YEARMO, a.LEVEL2_ENAMEORDER BY C3, C2;If I remove the ORDER BY clause it is returning results in 1 second. Can you guide me what is the issue?~Shalini

Post the EXACT SQL you are running (The above won't even compile due to an unclosed parenthesis).

Post the explain plan with and without the ORDER BY.

It looks like you are stroing dates as numbers - please don't do that. Dates should be in DATE columns.

It is not too surprising that an ORDER BY increases the time taken - you are asking it to sort, but we will

know more when you post the plans.

Nimish Garg has picked up an important general point. A query with "group by X,Y order by X,Y" will probably do a SORT GROUP BY operation to eliminate a sort for the "order by", while a query with just "group by X,Y" will probably do "HASH GROUP BY" which returns the data in an arbitrary order.

Given the optimizer can find lots of ways of getting the arithmetic wrong it's then possible for a switch from SORT to HASH to result in other parts of the plan changing, and this may affect the performance dramatically even when it's obvious to you that it shouldn't.  In a case like yours, if there's no other "pure" code alternative than the suggested code rewrite to force a separate HASH GROUP BY followed by SORT GROUP BY is a suitable alternative. (You could force a HASH GROUP BY with the use_hash_aggregation() hint, though that's potentially harder for other people to manage later if the code has to change.)

THere are a couple of problems with your SQL, though:

a) It's not executable - there's a missing ")".

b) The mix of older (Oracle) and newer (ANSI) syntax is highly undesirable

c) The left outer join should disappear because the subsequent "AND c.CURR_YEARMO = 201704" will eliminate the preserved rows.

You may want to review the code; and, as others have said, the first step to understanding the difference in performance is looking at the execution plans.

Regards

Jonathan Lewis