Hi Hervé,
I'm not sure I can be more help that Daniel was. I merged the GSoC code into the release and wrote the test cases, but beyond that you would have to check with the original authors. If you have more information that would useful in debugging this issue, I have reopened #106 also note #105.
Is upstream cost negative cost values?
I suspect that negative costs are not support and just cause the edge to be removed for that direction.
-Steve
On 11/7/2013 12:14 PM, Hervé Pella wrote:
I saw in "https://github.com/pgRouting/pgrouting/issues/107" the
discussion about the shortest path pairs method :"It seems that APSP Floyd-Warshall does not return all shortest path
pairs, at least not with the documentation example:
http://docs.pgrouting.org/dev/src/apsp_warshall/doc/index.htmlIt also returns a different number of results than APSP Johnson
algorithm, which has a similar issue #106."I work on hydrographic network and more specifically on connectivity
between different fish habitat. I try to calculate the distance
between arcs through the network (both downstream direction but also
upstream). I tested the two methods for calculate this metrics : APSP
Johnson APSP Floyd-Warshall (also with the reverse_cost) from
pgrouting 2.0 , postgis 2.1 and postgres 9.3.They give me the same results and do not go back in networks. They
only calculate distance to downstream without going upstream in the
network. Do you have more information about this problem ? Do you
know someone in the PgRouting community who would work on this issue
?Best regards
Hervé Pella Irstea Lyon - France