Re: [igraph] closeness and average path length in terms of the harmonic distance between vertices

[Tamás Nepusz]

Hi Bob,

> Sorry I only come back to you only now.  shortest.paths does not seem to 
> indicate the vertices that have been traversed in the shortest paths - just a 
> matrix of those shortest paths.
Yup, but you can use get.shortest.paths() to get _one_ shortest path per vertex 
pair, or get.all.shortest.paths() to get _all_ possible shortest paths between 
vertices. Of course the output of the latter function can be huge, especially 
if the graph has a regular structure (like a lattice) and there are many 
shortest paths between the same pair of vertices.

> Having looked at closeness(), is seems that closeness is currently done as:
> res <- .Call("R_igraph_closeness", graph, vids - 1, mode,  weights, PACKAGE = 
> "igraph")
> 
> How do I learn more of what is being done by this function? Many thanks.
It goes down to the C core of igraph, where it eventually ends up in a function 
named igraph_closeness:

http://igraph.sourceforge.net/doc/html/ch13s06.html#igraph_closeness

The source code is here:

https://github.com/igraph/igraph/blob/master/src/centrality.c#L2565

However, this function requires extensive knowledge of the C layer of igraph, 
so I don't know how useful it will be to you in the end. It boils down to doing 
a breadth first search on the graph and recording the distances in a vector.

-- 
T.