On Thu, Dec 15, 2022 at 1:03 AM Sandro Santilli <s...@kbt.io> wrote: > > On Tue, Dec 13, 2022 at 11:31:05AM -0800, Paul Ramsey wrote: > > Just to put a nail in this one: the last time our selectivity estimates for > > Contains/Within were good was version 2.1, which used the && operator > > instead of hte ~/@ operators in the SQL wrapper functions. > > I think I was the one chaing from && to ~/@ as they are more correct. > If we're going to change the selectivity estimator we should consider > using a different name to estimate the different filters, just in case > we'll use different implementation for one or the other. For points > they are really equal but for polygons estimates could be very > different for intersect and full containment (polygons could only > touch or have small overlap but very never be nested).
The reality of estimating off of the gridded histogram is there's no really any way to distinguish between intersects/contains strategies. You don't know the extents of the samples that fill in the grid, just their presence/absence in a given grid cell. I'd rather not build foundations for a house I will never erect. P > > --strk; > _______________________________________________ > postgis-users mailing list > postgis-users@lists.osgeo.org > https://lists.osgeo.org/mailman/listinfo/postgis-users _______________________________________________ postgis-users mailing list postgis-users@lists.osgeo.org https://lists.osgeo.org/mailman/listinfo/postgis-users