Torge <torgato(at)posteo(dot)de> writes:
> Now I would like to humbly propose a feature that gives an easy way to
> get a quick count estimate for any condition - index based or not -
> based on a random sample of rows, that does not require a custom
> function creation or complex SQL statement
Can't you do that already using TABLESAMPLE? For example, to
use a 1% sample:
select count(*) * 100 from mytab tablesample system(1) where <condition>;
You do have to remember to multiply by the factor corresponding
to your sample rate, but aside from that annoyance this (a)
already exists, (b) is SQL-standard, and (c) can be adapted to
a lot of other kinds of analysis besides plain count(*).
regards, tom lane