@scy clever use of a range query here, whatever "compressed Hilbert ordering" is it must be cool. I wonder what other datasets could benefit from that treatment.