Remaining impact of broken polygon geometries

[imagico]

Although this project is no more active i wanted to share some observations here regarding the remaining impact of broken polygon geometries in the OSM database, in particular w.r.t. https://github.com/openstreetmap/osm2pgsql/pull/684.

Looking at the data actually used for the landcover and water areas layers in OSM-carto there are now remaining about 3600 landcover features and about 400 water area features that are broken in a way that prevent libosmium from handling them (that is either intersections or open rings AFAIK). This sounds like very few (about 0.01 percent - this is however not comparable to the numbers in https://github.com/openstreetmap/osm2pgsql/pull/684#issuecomment-280558187 since i count only features that are actually rendered based on the tags) but the visual impact is still significant (i would say probably about half to one third of what was before the cleanup effort).

The most frequent tags among the 3600 landcover features are:

    102 natural=scrub
    103 landuse=grass
    113 landuse=meadow
    261 amenity=parking
    286 leisure=park
    389 landuse=forest
    392 landuse=farmland
    404 landuse=residential
    733 natural=wood

This might be of interest for release and rollout of future osm2pgsql versions using libosmium and for those who routinely process polygon data with libosmium - cc @joto, @lonvia, @pnorman, @woodpeck.