[BioC] slow graph methods [was: 'slow insertions in to graphNEL object (24 hours for 16k nodes)']

[Wolfgang Huber]

Dear Paul,

> Your second suggestion sounds like it might work.  Is there documentation
> somewhere that shows how to convert an environment into an S4 graph object?
> And perhaps the complementary conversion, from a graph into an environment?

Seth's and Martin' answers pretty much cover what I was thinking. Beyond 
Seth's, I know of no other existing solutions. (I.e. you could consider 
this an opportunity to write and contribute your own 'efficient' graph 
class based on the virtual "graph" class in the "graph" package, and 
with similar functionality as "graphNEL" but different internal 
implementation.

The nice thing about Bioconductors "graph" class is that it allows to do 
that.

I and others have dabbled in using one of the sparse matrix 
implementations for R for storing the adjacency matrix - but none of 
these attempts are mature, afaIk (?).

Best wishes
   Wolfgang

------------------------------------------------------------------
Wolfgang Huber  EBI/EMBL  Cambridge UK  http://www.ebi.ac.uk/huber