On Sun, 26 Apr 2020 at 00:40, Tomas Vondra <tomas(dot)vondra(at)2ndquadrant(dot)com> wrote:
> This reminds me our attempts to add bloom filters to hash joins, which I
> think ran into mostly the same challenge of deciding when the bloom
> filter can be useful and is worth the extra work.
Speaking of that, it would be interesting to see how a test where you
write the query as IN(VALUES(...)) instead of IN() compares. It would
be interesting to know if the planner is able to make a more suitable
choice and also to see how all the work over the years to improve Hash
Joins compares to the bsearch with and without the bloom filter.
David