Thanks, no problem. My assumption is if the accessed rows on each steps are different, then the execution should be different, so I will tune the statement or play with the statistics. But in this case, I am confused that execution plan is same, but accessed rows are big difference in return. The sql statement should be no problem, since one ran without any problem. if the statistics is the problem, then wouldn't it be different execution plan?
Alvaro Jose Fernandez wrote:
I would recommend - that's adventurous, but... - have the PeopleSoft people or the person in charge of this "new" SQL to maybe consider otheralternatives writing the query/report.I don't work on 10.2.0.3, but had some headaches with merge join cartesians on 9i . They work wery well in many cases, but some others not. I got rid of that ones 10076 event which activate costing for MJC(on 9i the CBO don't computed cost for MJC by default).Sorry I cannot be of help on 10.2.0.3 which is your case. regards.
-- //www.freelists.org/webpage/oracle-l