As per postgres documentation, indexSelectivity is the estimated fraction of parent table rows that will be retrieved during the index scan. This was previously hardcoded to 1, but it should instead be the fraction of parent table rows that are in our index (which can be less than 1 when we have partial indexes).
The genericcostestimate function that we call appears to compute this value for us, and so we can use that instead.
I also added tests in hnsw_cost_estimate that compute the indexSelectivity values (shown in the lantern debug logs that are outputted during this test) for partial indexes, to make sure that indexSelectivity is computed properly.
As per postgres documentation,
indexSelectivityis the estimated fraction of parent table rows that will be retrieved during the index scan. This was previously hardcoded to1, but it should instead be the fraction of parent table rows that are in our index (which can be less than1when we have partial indexes). Thegenericcostestimatefunction that we call appears to compute this value for us, and so we can use that instead.I also added tests in
hnsw_cost_estimatethat compute theindexSelectivityvalues (shown in the lantern debug logs that are outputted during this test) for partial indexes, to make sure thatindexSelectivityis computed properly.