Re: [igraph] community_walktrap memory leak

[Tamas Nepusz]

> When executing g.community_walktrap(), memory just rises and rises
> until this message appears:
> 
> terminate called after throwing an instance of 'std::bad_alloc'
> [...]
> Has anyone experienced this before with this community algorithm?
No, this is the first time we heard about this.
I tried running the algorithm on a generated Barabasi-Albert graph with
20000 vertices and 40000 edges. The memory usage peaks at 1390 MB, but
all of it is given back to the OS after the algorithm terminates (and
before the caller program terminates), so I guess this is not a memory
leak. With 80000 vertices and 160000 edges, the memory usage rockets up
to 4 GB at which point I simply terminated the program. So I guess the
community detection algorithm simply uses a lot of memory. I'm not
familiar with the internals of the algorithm as we used the original
source code of the authors almost intact.

You might try setting the steps= parameter of the algorithm to 3 instead
of 4. This way I was able to finish the analysis of my random network
with 80000 vertices; the memory usage peaked at 2.8 GB, but this was
directly from C, not from Python.

-- 
Tamas