Skip to content

parallel error message extraction (in mclapply)?

2 messages · ivo welch, Neal H. Walfield

ivo welch
# 
dear R experts---I am looking at a fairly uninformative error in my program:

Error in mclapply(1:nrow(opts), solveme) :
  (converted from warning) all scheduled cores encountered errors in user code

the doc on ?mclapply tells me that

  In addition, each process is running the job inside try(...,
silent=TRUE) so if error occur they will be stored as try-error
objects in the list.

I looked up "try", but this manpage creates as many new riddles as it
solves.  so, it isn't obvious to me how I get to the try-error object
list, or how to find out what triggered the abort.  of course, I can
use lapply to debug this, but this could be slow for some programs.
(I know the answer for my problem is that a uniroot deep inside my
code complained that my endpoints were not opposite.)

in general, is there a way to get some more information from mclapply
failures?  could this be added to the docs for mclapply?

/iaw


----
Ivo Welch (ivo.welch at gmail.com)
http://www.ivo-welch.info/
Neal H. Walfield
# 
Hi, Ivo,

At Fri, 28 Dec 2012 16:34:03 -0800,
ivo welch wrote:
A try object is an object that captures the execution context at the
time of an error.  Consider the following code:

  > library(multicore)
  > result = mclapply(1:10, function (i) { if (i == 4) foo() else i })
  Warning message:
  In mclapply(1:10, function(i) { :
    scheduled core 4 encountered error in user code, all values of the job will be affected

To investigate the error, we can use the try object that has been
saved in position 4:

  > print(result[[4]])
  [1] "Error in FUN(4L[[1L]], ...) : could not find function \"foo\"\n"
  attr(,"class")
  [1] "try-error"
  attr(,"condition")
  <simpleError in FUN(4L[[1L]], ...): could not find function "foo"
  > traceback(result[[4]])
  1: Error in FUN(4L[[1L]], ...) : could not find function "foo"

Neal