No, I'm merely pointing out that your observation of "skip scan is like an
index fast full scan" can appear to be true in some cases thanks to recent
optimizer changes.
That doesn't affect the way you have to decide column ordering, which
should be driven by patterns, frequency, and precision of predicate usage.
Regards
Jonathan Lewis
On Wed, 18 Aug 2021 at 19:35, Mladen Gogala <gogala.mladen@xxxxxxxxx> wrote:
Replies in-line
On 8/18/21 8:02 AM, Jonathan Lewis wrote:
Not true.
But thanks to (relatively) recent changes in the costing of index skip
scans it will appear this way if the number of index entries per value
of the skipped column(s) is so small that that take less than two
blocks (which often means you need 3 or 4 hundred rows per leading
values)
Thanks Jonathan! If I understand you correctly, you are advocating
putting the most selective column in order to avoid an expensive
operation, not unlike FFS?
--
Mladen Gogala
Database Consultant
Tel: (347) 321-1217
https://dbwhisperer.wordpress.com
