On Mon, Aug 9, 2021 at 11:05 AM Bruce Momjian <br...@momjian.us> wrote:
> > > selectivity = (1 - null_frac1) * (1 - null_frac2) * min(1/ > > num_distinct1, > > 1/num_distinct2) > > = (1 - 0) * (1 - 0) / max(10000, 10000) > > = 0.0001 > > Nice, can you provide a patch please? > > Change the line: selectivity = (1 - null_frac1) * (1 - null_frac2) * min(1/num_distinct1, 1/num_distinct2) to be: selectivity = (1 - null_frac1) * (1 - null_frac2) / max(num_distinct1, num_distinct2) The wording already talks about "divide by max". Though: "so we use an algorithm that relies only on the number of distinct values for both relations together with their null fractions:" maybe adds a parenthetical note: "so we use an algorithm that relies only on the number of distinct values (the row count estimate for the whole table, not the -1 in the column statistics) for both relations together with their null fractions:" Just note I haven't tried to absorb that whole page, let alone the implementation, and am not all that familiar with this part of PostgreSQL. Its seems right, though, in isolation. David J.
