Thanks to your question I’ve finally got around to doing two upgrades – then shifted everything on to a new laptop so that now I can run the four most important versions of Oracle simultaneously.

Nothing sinister or exciting; I just forgot that I ought to do the upgrade. I need two have two versions of 10.2 because of the significant histogram change across the 3/4 break – but I’m less worried about changes from 4 to 5.

I think you’re asking whether the performance benefit could come from the fact that we rerun the same subquery against the same data because we have pre-sorted the inputs and therefore get a caching benefit. The answer is no (although I have pre-sorted a partial result to get this benefit with a nested loop join in the past).

You need to read the article that I linked to in my answer to comment 5 above.

It’s an fantastic video.. !!
I doubt, During the demonstration – where you have said that “order by (sub-query with sort)” while using sub-query remembers the previous result came from the main query. Finally, when have seen that demo, I observed that, as the blocks are referenced previously and might be exists in buffer due to the query would be fetched most recently used buffer from hot of the LRU ?

Other wise i might be thinking way out different or totally in wrong path.

Further all the best for the upcoming book and very much waiting from long time… !!

thanking you in advance.. !!

Six: 8.1.7.4, 9.2.0.8, 10.2.0.3, 10.2.0.4, 11.1.0.6 (ought to upgrade to .7), 11.2.0.2.