Ahoj,
while cycle.travel snaps to and calculates routes along the edges of area highways (even to `highway=footway` and `area=yes`), it fails to do so if the area highway is defined as a relation.
See for example this route.
The plaza in the centre is this OSM entity, and it is not possible to snap at it's edge (I tried at the north edge). It also does not route along it. Instead, it snaps to this OSM entity (an area highway not defined as relation).
This also results in a detour, see attached screenshot.
GraphHopper, OSRM, BRouter and OpenRouteService all have the same issue:
* https://github.com/graphhopper/graphhopper/issues/3104
* https://github.com/Project-OSRM/osrm-backend/issues/7146
* https://github.com/abrensch/brouter/issues/778
* https://github.com/GIScience/openrouteservice/issues/1944
Regards!
Comments
Indeed, yes. cycle.travel runs on a fork of OSRM so will have the same issue here. Unfortunately routing across areas is exceptionally complex, and multipolygons even more so; so although it’s something I’d like to fix I can’t see any likelihood of that in the near future.
(Of course, it would be better if OSM actually had proper areas rather than the horrid kludge that is multipolygon relations, which stuff geometry information into a metadata object!)