Problems using the dijktstra search

graph-tool
1

Hi all.

I'm trying to use the dijktstra search But I have a exception.The property
that is used a int32_t and the GraphViews aren't empty. Any clue of what is
my error?

This is the code:

        referents_graph = GraphView(graph, vfilt= lambda v:
graph.vertex_properties["type"][v]=="referent")
        navigable_graph = GraphView(graph, reversed=True)

        referents = referents_graph.vertices()
        weight = navigable_graph.edge_properties["weight"]

        for base_referent in referents:
            # Calculate the distance of all accessible referents
            distances, predecessors =
graph_tool.search.dijkstra_search(navigable_graph,base_referent,weight)

This is the trace:

  File "/data/workspace/stanford-corenlp-python/contextgenerator.py", line
218, in resolve_corefenrence
    distances, predecessors =
graph_tool.search.dijkstra_search(navigable_graph,base_referent,weight)
  File "/usr/local/lib/python2.7/dist-packages/graph_tool/decorators.pyc",
line 2, in dijkstra_search
    Bq�Oc
  File
"/usr/local/lib/python2.7/dist-packages/graph_tool/search/__init__.py", line
126, in wrap
    ret = func(*args, **kwargs)
  File
"/usr/local/lib/python2.7/dist-packages/graph_tool/search/__init__.py", line
841, in dijkstra_search
    compare, combine, zero, infinity)
  TypeError: No registered converter was able to produce a C++ rvalue of
type int from this Python object of type float

Thanks for the time.

2

This has been fixed now in git... As a workaround, you can also do the following:

     dijkstra_search(navigable_graph,base_referent,weight, infinity=1000000)

Where the infinity value is a suitably large number.

Cheers,
Tiago

3

Thanks a lot!

I can get the correct distances normal nodes but I'm getting distances for
not reachable nodes. These distances are high compared with the others
distances (cost 140 against a mean cost of 5) but nothing that can
be programmatic filtered (like a constant inf value or negative value). If
I put infinity=-1 all distances changes to -1 (also the reachable nodes).
If is use a big infinity like 100000, the unreachable nodes get this
distance.

Is this the expected result of the function?

attachment.html (2.5 KB)

4

Yes, vertices which are unreached have a distance of "infinity". If this
is not given by the user, it is computed to be max(weight) * (N + 1),
since it is larger than any possible distance. If the user wants to
filter out these vertices, this value can be used, or the user himself
can specify the infinity value. If the user specifies the value, it
needs to be large enough, otherwise there will be problems, as you
noticed when setting it to -1.

Cheers,
Tiago

5

Thanks.

attachment.html (1.87 KB)