Edge weight not changing node position in neato?

Help
1

I’m trying to use the [weight] attribute in a neato graph but changing the values does not appear to do anything?

Example file:

graph G {
  a -- b [weight="10.0"]
  a -- c [weight="1.0" ]
  a -- d [weight="1000.1"]
  b -- c [weight="1"]
}

test-graph

I was expecting a -- d to be much closer together than the others, but it doesn’t seem to matter what values I put, the graph does not change at all.

I am on windows if that matters.

2

If you want to get d closer to a, try the len attribute. Weight will just help enforce len

… a larger weight encourages the layout to make the edge length closer to that specified by the len attribute.

3

I believe the default len attribute is 1.0 if not specified? Should I make len higher? Or make it bigger for the ones that should be farther apart?

This is just an example, my actual data is a social graph with like 100 nodes where I’ve calculated an edge-weight between nodes based on amount of interaction between the nodes. Should I be trying to programmatically generate different len values in addition or instead?

I was hoping that I could just specify a weight attribute and then the force direction algorithm would just “figure it out” based on that attribute since the only initial information I have are the edges that exist between nodes and their weight I’m trying to use graphviz to draw and thereby better understand what the shape of the social graph is.

4 
graph G {
  a -- b [weight="1.0" len=2 ]
  a -- c [weight="1.0" len=2 ]
  a -- d [weight="1" len=1 ]
  b -- c [weight="1" len=2 ]
}

If I set different len values, that seems to change the graph:
test-graph

But setting a weight value still doesn’t seem to do anything. Whether the weight is 1 or 1000 the graph remains the same. Is that because it only matters when there are more nodes to try to lay out?

BTW, the CLI command I’m running is neato -Tpng neato-test.gv -o test-graph.png

5

I was able to generate a graph using real data that looked more reasonable by skipping the weight and just setting len inversely to weight (higher weight = smaller len) and that seemed to do the trick.