[Routing] Gosmore routing server

[Nic Roets]

> After reading your post I decided to give a routing website a try. My
> first working result is: <http://tile.openstreetmap.nl/~lambertus/routing>.

> A few questions popped up while trying some routes:
> - Can Gosmore be used as a cgi-bin namefinder? If so, how?

I can implement if you can give me an example of how it is better than
Dave Earl's http://gazetteer.openstreetmap.org

> - Relatively often Gosmore is not able to find a route while looking at
> the OSM data does not reveal any obvious problems. It would be

There is a bug in SVN : Gosmore will start and end at the nearest
segment to the start and end points respectively, even if that segment
is an administrative boundary. Here is a patch :
--- gosmoreBadRoute.cpp 2008-05-25 17:50:52.000000000 +0200
+++ gosmore.cpp 2008-05-28 13:41:56.000000000 +0200
@@ -392,6 +392,7 @@
       // We don't do for (int dir = 0; dir < 1; dir++) {
       // because if our search box is large enough, it will also give us
       // the other node.
+      if (!(((wayType*)(data + itr.nd[0]->wayPtr))->bits & (1<<car))) continue;
       if (itr.nd[0]->other[0] < 0) continue;
       __int64 lon0 = lon - itr.nd[0]->lon, lat0 = lat - itr.nd[0]->lat,
               lon1 = lon - (ndBase + itr.nd[0]->other[0])->lon,


> interesting in such circumstances to be able to somehow 'debug' this.

The easiest would be to return the "most promising" incomplete route.
Then it will start with jump. I'll code it later today.

> - In one instance Gosmore ran for more than 30 minutes to calculate a
> route (which is an exception) so it would be nice for the routing server
> to be able to set some sort of maximum run time for Gosmore.

Gosmore limits the amount of nodes (and therefore time) it will spend
according to the direct distance.

You can try to impose an additional limit on it with 'ulimit -t'
command. If that doesn't work I'll write some more code.