[PATCH] reverse dead-end ways (to avoid routing problem?)
Perhaps, the attached patch would help the problem of routing to the wrong end of the dead-end way?
Perhaps, the attached patch would help the problem of routing to the wrong end of the dead-end way?
It will solve some, but not all problems. What should happen in the case, where a street has two dead ends and a crossing in the middle? I feel that your patch solves the sympthoms, not the unknown base problem.
It will solve some, but not all problems. What should happen in the case, where a street has two dead ends and a crossing in the middle?
In that case, it won't be trying to leave the way at either end. Perhaps, it will work OK - only one way to find out.
I feel that your patch solves the sympthoms, not the unknown base problem.
But we don't know what the problem is, yet. Why not try the patch? If it doesn't work for the situation you mentioned, we can try something else. If it breaks other things, we can try something else. Adding a node to the end of dead-end ways may be what's required but that will bloat the map files when it may not be necessary. Ha, that's just jogged a memory, this is what Steve wrote the other day:
However, the NOD section is much larger on the MP version; it contains fewer roads but more routing nodes.
I assume that the different number of roads is differences in how they are joined together. I can't see or think why the number of nodes would be different.
I think we know where those extra nodes are being used, now. Cheers, Mark
Interesting... Looking at: http://www.cgpsmapper.com/tool.htm#1 It says of the MapRoute program: # Fix several potential problems in routable maps * setting NodID point always at the beginning and end of the road (important for dead-end roads) Perhaps, adding a node at the dead-ends is a must. Something else has just occurred to me. Hasn't it been reported previously that people were seeing the pink line go to the wrong end of motorways? I wonder if that's related to our current issue and was caused by the fact that the motorway was incomplete (or was clipped!) and so one or more of its ends were not terminated with nodes. You can't reverse the direction of the motorway, so making its ends into nodes would be the only option. So, the best approach may be making all dead-ends into nodes, after all. I shall make a patch to do that. Mark
On Feb 14, 2009, at 17:46, Mark Burton wrote:
It says of the MapRoute program:
# Fix several potential problems in routable maps
* setting NodID point always at the beginning and end of the road (important for dead-end roads)
Perhaps, adding a node at the dead-ends is a must.
Ok then. I looked at some cGPSmapper-created maps that had dead ends without nodes, so I thought this was something legal, but probably those maps were just broken. I didn't ever try those out. We could simplify the code a little if we assume that all road ends have nodes (see the startsWithNode-stuff in RoadDef).
Something else has just occurred to me. Hasn't it been reported previously that people were seeing the pink line go to the wrong end of motorways?
I'd assumed that these problems were fixed by your earlier patch, the one with the incorrect offsets. Cheers Robert
We could simplify the code a little if we assume that all road ends have nodes (see the startsWithNode-stuff in RoadDef).
Yes, we could probably remove some code there.
Something else has just occurred to me. Hasn't it been reported previously that people were seeing the pink line go to the wrong end of motorways?
I'd assumed that these problems were fixed by your earlier patch, the one with the incorrect offsets.
I don't think so, I remember at least one person reporting seeing the routing line go to the wrong end of a motorway. Not to some random place, but to the end of the way it should have gone to. I attach a simplified version of my last patch - it now always increments the highways count for the end nodes. By doing that, we know that the resulting highway counts for the end nodes will always be one more than the real number of ways that join those nodes (if we ever need to know that).
participants (3)
-
Johann Gail -
Mark Burton -
Robert Vollmert