If you have read my previous post about this (if not, you can find it here), you know that I found a cardinality issue with 12c top-N queries (or row limiting as Oracle call it).
I didn’t write it, but before I even published the post, I opened an SR to Oracle about this, and after supplying all the information they needed, I’ve been told that this is probably how it works, Oracle transforms the SQL to use analytic functions and it behaves as expected.
Then came OOW15.
I love going to Oracle Open World, as this is the place to meet the right people, so I went to talk to one of the directors or development of the optimizer. As I explained the problem, he realized that this is probably a real issue, and when the transformation takes place, nothing updates the optimizer about the cardinality. He asked for the link to my post and I sent him all the information.
Besides him, I had the opportunity to talk to some support people (including one of the support VPs). As part of the interesting conversation we had, I mentioned the SR I opened and the way it got closed. They asked for the SR details and I provided them. Today, the SR got updated. It seems that this was escalated, and analyzed again by one of the engineers. He opened a bug to the development as something does seem wrong.
I’ll keep updating about this issue, stay tuned.