Hi all, >Yeah, in my last email, I said I tried this way. But I worried that >rel->tuples may be zero for an empty relation. In my previous email， I worried rel->tuples may be zero for an empty relation. But here it's safe, because an empty relation has no tuples in pg_statistic. So it will not enter if (HeapTupleIsValid(vardata.statsTuple)). Sorry for the noise.
Tom Lane <[email protected]> 于2026年3月1日周日 08:08写道： > Hmm ... doesn't this contradict your argument that avgfreq and > mcv_freq need to be calculated on the same basis? Admittedly > that was just a heuristic, but I'm not seeing why it's wrong. > Agree > > The reason for this is that estfract is calculated as: > > estfract = 1.0 / ndistinct; > > where ndistinct has been adjusted to account for restriction clauses. > > Therefore, we must also use the adjusted avgfreq when adjusting > > estfract here: > > It feels like that might end up double-counting the effects of > the restriction clauses. > > Anyway, we all seem to agree that s/rel->rows/rel->tuples/ is the > correct fix for a newly-introduced bug. I'm inclined to proceed > by committing that fix (along with any regression test fallout) > and then investigating the avgfreq change as an independent matter. +1 -- Thanks, Tender Wang
