Folks,
Nope, it's definitely proportional. As a simple example, a sample of
500 rows in a table of 1000 rows should yeild stats estimates with 90%+
accuracy. But a sample of 500 rows in a 600,000,000 row table is so
small as to be nearly useless; it's quite possible to get all the same
value in a random sample of < 0.1% even on a column with a D/N of 0.001.
I meant "a D/N of 0.1". Sorry.
--Josh
---------------------------(end of broadcast)---------------------------
TIP 9: In versions below 8.0, the planner will ignore your desire to
choose an index scan if your joining column's datatypes do not
match
