Capturing output from new threads in R

Question

I am running a multithreaded R script but I am having trouble with generating output from the cluster.

outFun <- function()
{
  cat(sample(0:9,1));
}
require(snow)
clust <- makeCluster(4)
clusterExport(clust,"outFun")
clustFun <- function(i){outFun()}
clusterApplyLB(clust,1:8,clustFun)

I understand that I don't see any of the output from outFun() because it's in a new R thread, but I was hoping that there would be some way to forward this output back to the master thread so it was visible when printed.

EDIT: This question answers this for a Linux machine, but the solution does not work for Windows. The workaround given is to simply use file output, but I am curious if anyone knows a solution to be able to actually send output back to the master thread in Windows.

@DWin as I have demonstrated in my code above, there are packages to make it multi-threaded... — jpd527, Oct 07 '13 at 14:42

score 2 · Accepted Answer · answered Oct 14 '13 at 13:29

The makeCluster outfile="" option doesn't work with Rgui on Windows because Rgui doesn't send output from child processes to the display window. However, outfile="" does work with the Rterm program because it's a simple console program:

C:\Program Files\R\R-3.0.2\bin\i386> rterm -q
> library(parallel)
> clust <- makeCluster(4, outfile="")
starting worker pid=1596 on localhost:11862 at 09:13:30.005
starting worker pid=1192 on localhost:11862 at 09:13:30.342
starting worker pid=1616 on localhost:11862 at 09:13:30.679

The "starting worker" messages are coming from the worker processes just before they execute the slaveLoop function. You should also see output from the workers from executing tasks:

> clusterEvalQ(clust, message("hello"))
hello
hello
hello
hello
[[1]]
NULL

[[2]]
NULL

[[3]]
NULL

[[4]]
NULL

Interesting solution, would not have thought the problem lied in the IDE. — jpd527, Oct 14 '13 at 15:12

Capturing output from new threads in R

1 Answers1