How can I `print` or `cat` when using parallel

Question

If I call a function using parSapply then print, message or cat statements inside that function don't seem to output to the console.

My process takes a very long time, so I need some way of seeing the progress and getting the results output as they are done. Are there any special commands that would allow me to print to the console from a parallel process?

Example:

library(parallel)

oneloop = function(x) {
  for(i in 1:50) {
    a = rnorm(100000)
    a = sort(a)
  }
  print(x)
  message(x)
  cat(x)
}

cl <- makeCluster(5)
output = parSapply(cl, 1:10, oneloop)
stopCluster(cl)

Solution 1

Using outfile param in makeCluster you can redirect the output to a file and then check that file to see how your program progresses.

Interestingly on a Linux machine setting it to "" outputs to the console, but that doesn't work for me on a Windows machine. File output works on both.

Solution 2

I am using makeCluster with outfile on Windows with doParallel package version 1.0.8. I'm able to tail outfile using a simple Powershell command powershell Get-Content c:/path/to/log/mylog.txt -wait. Works for me.

Comments

[Ben]

A slightly different approach to getting a progress log file when using %dopar% is documented in detail here: viksalgorithms.blogspot.com/2012/02/…

[Steve Weston]

outfile="" doesn't work on Windows with Rgui, but it does with Rterm.

[MichaelChirico]

It's strange how poorly-documented this is. I had to pore through source code for a good 15 minutes before I found exactly what outfile is doing (basically being added to a command being sent to terminal as OUT= outfile). Sending to terminal with "" is consistent with, e.g., write.table, which takes the connection "" to mean stdout

[JMT2080AD]

Using the setting outfile = "" to print to the terminal works for me using emacs and ESS in Windows.