Re: [igraph] Some peculiarities

[Tamás Nepusz]

Hello,

> p <- function(y) {
> g <- graph.adjacency(y, mode = "directed");
> plot(g, layout = layout.spring);
> }
> [...] 
> p(x);
> p(x); # Output differs.
>  
> Why does the output differ between the last two calls to p()?
layout.spring starts from a random configuration and then iterates for a given 
number of iterations, so it is not guaranteed to give the same layout every 
time. Consider using layout.fruchterman.reingold or layout.kamada.kawai -- they 
are based on similar principles but allow you to specify a starting 
configuration (use the "start" parameter), so the layout will be the same if 
you use the same starting configuration.

> Second, I wonder whether there is a simple explanation why triad.census()
> may return negative values for empty subgraphs?
You are right, this is an integer overflow. You can get the correct value by 
subtracting the other values in the result vector from the number of all 
possible triads in the graph.

- T.