Comments on: Retrieving Records with Max Values using TOP http://www.rabidgadfly.com/index.php/2007/08/15/retrieving-records-with-max-values-using-top/ Glenn Gervais's methods, tips, and musings on Flash, ColdFusion, XML, and other web technology. Wed, 01 Sep 2010 12:20:23 +0000 http://wordpress.org/?v=2.9.2 hourly 1 By: rabidGadfly http://www.rabidgadfly.com/index.php/2007/08/15/retrieving-records-with-max-values-using-top/comment-page-1/#comment-42766 rabidGadfly Wed, 15 Aug 2007 17:44:31 +0000 http://www.rabidgadfly.com/?p=45#comment-42766 Brian, Thanks for the improved solution and the concise explanation. Makes perfect sense. Brian,

Thanks for the improved solution and the concise explanation. Makes perfect sense.

]]> By: Brian Kotek http://www.rabidgadfly.com/index.php/2007/08/15/retrieving-records-with-max-values-using-top/comment-page-1/#comment-42760 Brian Kotek Wed, 15 Aug 2007 16:48:48 +0000 http://www.rabidgadfly.com/?p=45#comment-42760 Understand though that using correlated subqueries like that can be quite inefficient, depending on the database platform, schema, and the amount of data. Correlated subqueries are evaluated once for every matching record. Which means that the subquery can potentially run for every cartesian product created in the join between GameScores and Games. Different databases have different approaches to trying to optimize this, but it is definitely something to be aware of. Consider this alternative: select gs.scoreID, v.maxScore, g.gameName from GameScores gs inner join ( select max(score) maxScore, gameID from GameScores group by gameID ) v on v.maxScore = gs.score and v.gameID = gs.gameID inner join Games g on gs.gameID = g.gameID This should give the same results, but it uses an inline view that has no correlation to the rest of the query, which means that it will be more efficient than a correlated subquery. Note that the original query you posted also suffers from another flaw: what happens if there are two records that tie for the high score for a game? Using Top 1 will essentially drop one of the two records. Probably not a big deal for high scores for a game, but if you were calculating which sales rep had the highest sales for the quarter this might be a much bigger issue. ;-) Understand though that using correlated subqueries like that can be quite inefficient, depending on the database platform, schema, and the amount of data. Correlated subqueries are evaluated once for every matching record. Which means that the subquery can potentially run for every cartesian product created in the join between GameScores and Games. Different databases have different approaches to trying to optimize this, but it is definitely something to be aware of.

Consider this alternative:

select gs.scoreID, v.maxScore, g.gameName
from
GameScores gs
inner join
(
select
max(score) maxScore,
gameID
from
GameScores
group by
gameID
) v
on v.maxScore = gs.score and v.gameID = gs.gameID
inner join
Games g
on gs.gameID = g.gameID

This should give the same results, but it uses an inline view that has no correlation to the rest of the query, which means that it will be more efficient than a correlated subquery.

Note that the original query you posted also suffers from another flaw: what happens if there are two records that tie for the high score for a game? Using Top 1 will essentially drop one of the two records. Probably not a big deal for high scores for a game, but if you were calculating which sales rep had the highest sales for the quarter this might be a much bigger issue. ;-)

]]>