Charles Galpin wrote:
Hi Steve
On May 17, 2010, at 11:03 AM, Stephen Woodbridge wrote:
I do not think you should be passing -1 as the reverse cost, what is that supposed to mean? Edge weights should not be negative. If You are just trying to avoid it from taking a reverse edge for testing then set the value to a very large number instead.
I simply want to force the routing to not support traversing a link in reverse (in other words force the routing to follow the only real world path). My links are directed (one way).
According to the following wiki page, a negative cost prevents an edge from being inserted into the graph.
OK, it has been a while since I read the docs, so maybe I should go read them again 
I can see having a large number working by making the cost high, but this doesn't seem right.
Typically, you should be passing the column that contains the reverse cost and not a constant. In you example above "length" is a column that contains the cost, if your cost and reverse cost are the same and you just want to test a bidirectional graph then you could set you reverse cost to "length" also, like:
SELECT gid AS id, source::integer,target::integer, length::double
precision AS cost, length::double precision as reverse_cost FROM linksDoesn't this just say my links are bidirectional and the cost is the same?
Yes
This is exactly what I don't want
Thanks,
charles