Hi Andres,
> Yea, it's something we should improve. Have you checked if this has
> performance impact for heap? Should we also consider planning costs?
Since the visibility map is very small, all pages of the visibility map will
usually reside in memory. The IO cost of accessing the visibility map can
be ignored. We should add the CPU cost of accessing visibility map. The
CPU cost of accessing visibility map is usually smaller than cpu_tuple_cost.
But Postgres does not have a Macro to describe such a small cost. Should
We add one?
> As far as I can tell you have not acually attached the patch.
Ah, forgot to upload the patch. Attach it below.