Re: [igraph] problem with large graph

[Tamas Nepusz]

Dear Uri,

I'm not an R expert (I mostly use igraph from C or Python), so that's  
why I didn't answer your message so far. The size of the graph you are  
working with (215K nodes, 3M edges) should be no problem for igraph's  
core, at least when it has to store the edges only (without any  
additional metadata). E.g., doing the following works for me in R  
(version 2.8.1 on openSuSE 11.1 -- I know that this R version is  
outdated, but this is the default in openSuSE):

> xs <- floor(runif(3000000)*215000)
> ys <- floor(runif(3000000)*215000)
> df <- data.frame(xs=xs, ys=ys)
> g <- graph.data.frame(df)
> ecount(g)
[1] 3e+06
> vcount(g)
[1] 215000
> dd <- degree.distribution(g)
 [1] 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00
 [6] 0.000000e+00 0.000000e+00 0.000000e+00 9.302326e-06 2.325581e-05
[...lines omitted here...]
[56] 4.651163e-06 0.000000e+00 0.000000e+00 4.651163e-06

However, it could have happened that the edge data together with the  
metadata attached to each edge is too large for R or igraph itself.  
(Note that there is a point when the whole graph is stored twice in  
the memory, once as a data frame, once as an igraph graph, which  
contains a copy for each edge attribute). So, I would first try to  
construct an igraph graph without the attached attributes, i.e. create  
a data frame with only profile_id_a and profile_id_b and check if it  
works or not. If it works, try adding the attributes back to the data  
frame one by one and find where the whole process breaks down. I'm not  
familiar with R's internal details, so it could also have happened  
that you are hitting an internal limit somewhere; maybe Gabor can tell  
more about this.

--
Tamas