NYCPHP Meetup

[Steve Manes]

Ophir Prusak wrote:
> thanx for the explanations everyone!
> 
> fyi, i did the following two queries which use the same columns but
> return a different number of results. Indeed the %30 rule seems to be
> the case.

If you recall, we had the same situation on Postgresql at Trafficmac 
with some of those mondo adserver queries which had gobs of duped 
indices.  That's why I built those summary tables.  I believe that most 
SQL engines are smart enough to see that a flat table scan may be faster 
than diving through an overpopulated index tree.