range graph removing edges about pandana

[fscottfoti]

Good question. Actually this is not an indicator of anything being wrong per se. It actually means that there are street edges that aren't useful for taking shortest paths between different places (certain streets aren't on the way to anywhere I guess). It's just a byproduct of the algorithm and is really just means things are working as designed (see http://en.wikipedia.org/wiki/Contraction_hierarchies for more info)

[bridwell]

That makes sense. Looks like where I have edges being removed tends to be for 2-way networks with duplicates or 1-way network where the edges have the same weight in both directions.

OK one more question: when it states 'removed x edges of n' the n is always double the number of input edges. This makes sense for the 2-way case, but for the 1-way case are extra edges being added for the opposite direction? I guess I'm confused because I would think that the bi-directional edges needed for the dijkstra shortest path search are being provided already. Or is the n reported to the console just a simple approximation and not representative of the true edge count?

[fscottfoti]

For the one way case extra edges are not being added for the opposite direction. It's a bit disconcerting that the n is twice the edges even in the 1-way case. Unfortunately I don't have an answer for that since this is coming from an underlying library and not code we wrote. I do think it works though in the tests that I ran but let me know if you see troublesome results.