SQL Tuning max + group by

Question

Hi all,
I run this query
select max(prelevement_dt), mpr
from labo_demande
group by mpr;The table labo_demande contains 1.688.883 rows and an index on (MPR, PRELEVEMENT_DT) exists. (I have already rebuilt this index).
I gather stats on the table but the query still takes 10 seconds running.
The explain plan shows a TABLE ACCESS FULL on the table.
Have someone an idea on how to tune this sql statement ?
Thanks for your help

Answer 1

fcjunic wrote:
Hi all,
I run this query
select max(prelevement_dt), mpr
from labo_demande
group by mpr;The table labo_demande contains 1.688.883 rows and an index on (MPR, PRELEVEMENT_DT) exists. (I have already rebuilt this index).
I gather stats on the table but the query still takes 10 seconds running.
The explain plan shows a TABLE ACCESS FULL on the table.
Unless at least one of the two columns is declared to be non-null (or you add a predicate like (e.g.): "mpr is not null") then this is the only possible execution path for the query. If there is a NOT NULL constraint on one of the columns then an index fast full scan with "sort group by "is possible, and an index full scan with "sort group by nosort" is possible - but the cost of the the index full scan could (for valid reasons) be larger than the cost of the tablescan, and there are various bugs with costing index fast full scans that means the optimizer can make the cost of that plan too high (for invalid reasons).
If either of the paths are valid, you can hint them: /*+ index(labo_demande(mpr, prelevement_dt)) */ or /*+ index_ffs(....) */
Regards
Jonathan Lewis