On 17.01.2013 16:01, Gerd Petermann wrote:

reg. test data: I did not yet try to understand why a short arc brakes
routing. I assume it is because you can't calculate the bearing with
an appropriate precision?
I just know that I see error messages like "road ... contains zero length arc at ... "
or "road ... contains consecutive identical nodes at ... "
when I omit the removeShortArcs call.
So, I'll try to run the first version of the patch against many tiles
and see what happens. If I don't find a problem in Europe,
it is likely that we don't need iteration.

OK?

Gerd

>

The reason As far as I can remember is that if we have the following situation

   #
   #
   #
--- O UUU ---
   #
   #
   #

Now consider both # and - U and O stand for streets. O must not be shorter than the routing resolution (which is twice the minimal distance we can put points), else O will be unroutable.
I found out about it, by having a very short stretch of a road on a bridge, where directly after the bridge (or was is actually sharing a point with the bridge?) a crossing pops up. As soon as we made sure that O is ~5.4m the routing worked again.

Im not sure if

--- UUU --- with UUU being shorter than 5.4 works though. (UUU being e.g a bridge).