Hi Stan, > I'm back from vacation and did some measurements this morning. > I also find that #3 performs better than #4 with a more complicated query. > In this experiment, #3 needs more buffer reads than #4 (#4 uses 20% less > buffer reads), > but performs much better (#4 needs 60% more time than #3).
Great, so let's include your addition. In the mean time, I double-checked with the legacy parts of our E-Banking application. Most of it uses #3 as well. > So unless an oracle guru advises against it, I would implement #3. I've had some DBA advise against #3 in favour of #4, but my measurements could not confirm that. Besides, others advised the other way round. Can you create a pull request from this: https://github.com/stanislas/jOOQ/commit/cea35b05390cd4ccebccf19621ecfafacf2da148 Cheers Lukas 2012/5/24 Stanislas Nanchen <[email protected]>: > Hi Lukas, > > I'm back from vacation and did some measurements this morning. > I also find that #3 performs better than #4 with a more complicated query. > In this experiment, #3 needs more buffer reads than #4 (#4 uses 20% less > buffer reads), > but performs much better (#4 needs 60% more time than #3). > > So unless an oracle guru advises against it, I would implement #3. > > cu. stan.
