Re: [igraph] building R package from development branch igraph--main--0.4

[Tamas Nepusz]

Dear Vincent,

> I ran a few community algorithms, walktrap, edge.betweenness., fastgreedy on
> a ~150k collaboration network, fastgreedy successfully completed on a
> simplified connected graph but failed on the nonsimplified version with what
> appeared to be a memory access violation.
Yes, that's not a surprise: you should always simplify the network
before doing the actual community analysis. Of course one could add an
explicit igraph_simplify call right in the beginning of the actual
community detection implementation, but that's a waste of time if the
graph is already simple. You should also note that I've completely
re-implemented the fast greedy community detection algorithm during
the weekend - the current implementation is much faster. I think it
would also be feasible to issue a warning if the algorithm detects a
loop edge or multiple edges between the same nodes.

--
Tamas