On 31 March 2016 at 20:18, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Also, I wonder if it'd be a good idea to provide a guard against division
> by zero --- we know rel->tuples > 0 at this point, but I'm less sure that
> reldistinct can't be zero. In the same vein, I'm worried about the first
> argument of pow() being slightly negative due to roundoff error, leading
> to a NaN result.
Yeah, that makes sense. In fact, if we only apply the adjustment when
reldistinct > 0 and rel->rows < rel->tuples, and rewrite the first
argument to pow() as (rel->tuples - rel->rows) / rel->tuples, then it
is guaranteed to be non-negative. If rel->rows >= rel->tuples (not
sure if it can be greater), then we just want the original
reldistinct.
> Maybe we should also consider clamping the final reldistinct estimate to
> an integer with clamp_row_est(). The existing code doesn't do that but
> it seems like a good idea on general principles.
OK, that seems sensible.
Regards,
Dean