>
> > Is the caluculation "rows of Nested loop = rows of OuterPlan * rows of
> > InnerPlan" wrong ?
>
> Careful --- rows produced and cost are quite different things. The
> cost estimate for a nestloop is "cost of outerplan + rows of outerplan *
> cost of innerplan", but we don't necessarily expect to get as many rows
> out as the product of the row counts. Typically, it'd be lower due to
> join selectivity. Above you see only 3 rows out, which is not too bad
> a guess, certainly better than 2*15905 would be.
>
I see. rows of Join = rows of outerplan * rows of innerplan * "join
selectity".
and "join selectivity" is calcutated by eqjoinsel() etc.
Thanks.
Hiroshi Inoue
Inoue@tpf.co.jp