> Bruce Momjian writes:
>
> > No, we have no ability to randomly pick rows to use for estimating
> > statistics. Should we have this ability?
>
> How's reading a sufficiently large fraction of random rows going to be
> significantly faster than reading all rows? If you're just going to read
> the first n rows then that isn't really random, is it?
Ingres did this too, I thought. You could specify a certain number of
random rows, perhaps 10%. On a large table, that is often good enough
and much faster. Often 2% is enough.
--
Bruce Momjian | http://candle.pha.pa.us
pgman@candle.pha.pa.us | (610) 853-3000
+ If your life is a hard drive, | 830 Blythe Avenue
+ Christ can be your backup. | Drexel Hill, Pennsylvania 19026