Hannu Krosing wrote:
> the per-page clustering would make sure that all the tuples would be on
> 1 (or on a few) pages.
I understand that You can cluster on one column, but how do you do it for
indexes on other columns?
BTW, lossy variants also lose count(), group by only from index
> and what comes to updating the index, you have to do it only once per
> 100 pages ;)
Sorry, how does that work, if I update foo = 'bar'->'baz' - I can flip
the 'baz' bit
on right away but I have to check every other row to see
if I can turn the 'bar' bit off
Andre