Re: [igraph] Importing very large graphs in igraph-R

[Gábor Csárdi]

Eric,

graph.data.frame() is much slower than graph(), because it uses
symbolic vertex names. Using the low-level vertex ids and the graph()
function should take a couple of minutes for your graph.

Of course, if you have symbolic vertex names, then you will need to
convert them to numeric ids first, but maybe this can be done faster
by hand than graph.data.frame() is doing it.

Best,
Gabor

On Thu, Sep 3, 2009 at 6:54 PM, Eric Sun<address@hidden> wrote:
> Hi,
>
> First of all, thank you guys for your excellent work.  I searched the
> mailing lists for this issue but didn’t find anything:
>
> I’m trying to import a very large graph (around 1 million nodes and 50
> million edges) using 64-bit R (2.10.0 unstable) and igraph 0.5.2-2, MacBook
> Pro 2.66Ghz Core 2 Duo, 4GB RAM.
> I’m calling    graph.data.frame(mydata, directed = FALSE)
> where mydata is a data frame with 2 columns (containing the edge
> connections).  Since it’s an undirected graph I only store A, B  (but not B,
> A  to save space).
>
> The command has been running for about 38 hours.  It seems to use only 1 CPU
> core since the rest of my computer is still snappy.  top reports the
> following:
>   PID COMMAND      %CPU   TIME   #TH #PRTS #MREGS RPRVT  RSHRD  RSIZE  VSIZE
>  1853 R           94.5% 38:19:14   1    17     94 1213M  5200K  1197M  2740M
>
> Does anyone have any idea how long I should expect the command to take?  If
> it’s another day or two I will wait, but if this is going to take months I’m
> interested in seeing if there is a more scalable option I can try.
>
> Thanks!
> Eric
> _______________________________________________
> igraph-help mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/igraph-help
>
>



-- 
Gabor Csardi <address@hidden>     UNIL DGM