Sorting by geometry is slower than sorting by geohash

[komzpa]

` [local] gis@gis=# create table juno_osm_point_unclustered as (select * from juno_osm_point ORDER BY osm_id); SELECT 159446335 Time: 601714,217 ms (10:01,714)

[local] gis@gis=# create table juno_osm_point_pt1 as (select * from juno_osm_point_unclustered ORDER BY ST_GeoHash(ST_Transform(ST_Envelope(way),4326),10) COLLATE "C"); SELECT 159446335 Time: 1248105,490 ms (20:48,105) [local] gis@gis=# create table juno_osm_point_pt2 as (select * from juno_osm_point_unclustered ORDER BY way); SELECT 159446335 Time: 1287415,365 ms (21:27,415) `

This may happen as there are some other infrastructure optimizations in string sorting code, like abbreviated key comparsions.