On Mon, Mar 28, 2022 at 7:07 PM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Right, that's the case I had in mind --- apologies if my terminology
> was faulty. btree can actually handle such a case now, but what it
> fails to do is re-descend from the tree root instead of plowing
> forward in the index to find the next matching entry.
KNNGIST seems vaguely related to what we'd build for nbtree skip scan,
though. GiST index scans are "inherently loose", though. KNNGIST uses
a pairing heap/priority queue, which seems like the kind of thing
nbtree skip scan can avoid.
> +1. We at least need to be sure we all are using these terms
> the same way.
Yeah, there are *endless* opportunities for confusion here.
--
Peter Geoghegan