Re: [jgrapht-developers] dumping adjacency and laplacian matrices

[John V. S. <js...@gm...>]

Another possibility is a parameter telling the exporter whether to
ignore/collapse vs. assert on detecting a loop or multi-edge.

It would be nice if there were a way to query a graph for these
properties.  In the case where they were not being enforced by the graph
implementation itself, the query would actually inspect the graph
content; in case the graph already knows this inspection could be
short-circuited.  Arranging this communication in a generic way so that
we don't end up with an explosion of interface combinations is the
tricky part.  Yet another enhancement request?

JVS

On Tue, 2005-12-13 at 19:04 -0500, Charles Fry wrote:
> Now that I'm getting a little deeper into this, I realized that a
> Laplacian matrix is only defined for "an undirected, unweighted graph
> without graph loops (i,i) or multiple edges from one node to another [1]."
> 
>    1. http://mathworld.wolfram.com/LaplacianMatrix.html
> 
> Currently there are interfaces for Directed and Undirected graphs, but
> no interfaces are used to mark Multigraphs or Pseudographs. It is thus
> impossible for me to tell beforehand whether or not the graph provided
> contains loops or multiple edges.
> 
> The possible solutions that I see to this are to:
> 
>  - Define interfaces that indicate when multiple edges or loops are
>    permitted in a graph, and explicitely check those, possibly issuing a
>    warning.
> 
>  - Ignore loops, and collapse multiple edges when creating the Laplacian
>    matrix.
> 
> I don't see any reason not to do the latter, although the former may
> still be beneficial for other reasons.
> 
> Thoughts?
> 
> Charles
>