logo separator

[mkgmap-dev] routing problems when style creates multiple routable ways for one OSM way

From Gerd Petermann GPetermann_muenchen at hotmail.com on Sat Feb 24 08:52:19 GMT 2018

Hi all,

in [1] Felix suggested to add a filter so that a style would not add too many routable ways for one OSM way.

For those who are not familiar with that idea: Felix' style adds e.g. additional routable ways with other routing
attributes when a way is very steep, e.g. one saying something like "nice to cycle"  and the other with oneway=-1 and "not nice to cycle".
The effect on NOD is that at some "RouteNodes" we find a lot more so called "RouteArcs",
many of them with equal angles and lengths. 

Reg. the filter I see no problem to change mkgmap so that it detects the number of routable ways N which were added for one 
OSM way and if N > max_N (I think Felix says max_N is 4)  ignore the last N - max_N ways. 
It might also be possible to add a style function which gives access to the number of previously added routable ways, if that is needed,
but I see no need for that. I think the style rules could just add routable ways in order of importance. Right?

Anyway, the question is if this will always help, as I think that the problem is probably the number of RouteArcs with equal 
angles in one RouteNode. Maybe we just have to make sure that we don't have exactly equal RouteArcs, 
maybe the overall number is the problem. The angles are stored with a rather poor resolution (max. 8 bits), but maybe we can manipulate
the other attributes like length or curviness of an arc to avoid the crash.
Maybe the problem is not the number of arcs in one RouteNode but the overall number of arcs that have to be checked  when a route 
is calculated.

@Felix: Maybe you have already found out more details? You wrote about  "bsod", I know the meaning for Windows (Blue screen of Death),
but what exactly happens with a device? Is it dangerous the play with this on my Oregon?  What exactly do I have to do to to reproduce 
such a crash? 


[1] http://www.mkgmap.org.uk/pipermail/mkgmap-dev/2018q1/028256.html

More information about the mkgmap-dev mailing list