script works in Rstudio but not with Rscript

OK, this suggestion brings up some interesting results.  No solution,
however.  But it's interesting ... and maybe some helpful leads.

The basic/short answer to your question is no, it doesn't work with "R
-f script".

The longer answer: Note that the script is currently set up to be
called with a command line argument, "./affinity_propagation.R 3".  So
first I tried "R -f affinity_propagation.R 3".  That failed, but
because it wasn't using the "3" as an argument.  So I modified the
code to hard-code the value 3 in there and not read the command line
argument using commandArgs(TRUE).  Then I ran the script using
"./affinity_propagation.R" -- and it worked!  "R -f
affinity_propagation.R" also worked.

Then I noticed the --args command line option for R, and I ran "R -f
affinity_propagation.R --args 3" on the original script, and it
failed, the same error as I described below.

Also, when you say "Try loading methods", do you mean just a line like
the following?:

library(methods)

I tried that as well, and it didn't help, same error.

So, given all that, any new ideas?  Is the call to commandArgs()
screwing something up?  Can that be fixed somehow?  Is there another
library/function I could/should use to read command line arguments?

Hmmm, playing around with things some more, it looks like if I do:

num <- as.integer(args[1])

rather than:

num <- args[1]

things work.  So maybe this is an issue with types, scalar vs. array,
number vs. string?

Milt Epstein
Programmer in Computational Genomics
Institute for Genomic Biology (IGB)
University of Illinois at Urbana-Champaign (UIUC)
mepstein at illinois.edu


On Wed, 8 Apr 2015, Henrik Bengtsson wrote:

Does it work with R -f script?  If so, then it's because Rscript does not
attaching methods package by default, but R does. Try loading methods at
the top of your script.

My $.02

Henrik
On Apr 8, 2015 07:41, "Milt Epstein" <mepstein at illinois.edu> wrote:

Greetings.  I am new to R, but have quite a bit of experience
programming with other languages (e.g., Perl, Java, Python, shell
scripting).  I'm now working on a project where I need to use R.  A
colleague wrote a number of small scripts that work fine in Rstudio,
but a couple of them don't work when run using Rscript (which we're
planning on doing).  The behavior is basically the same on a few
different machines (two of them are Linux, one I think is a Mac).  To
run the scripts using Rscript, we put one of the following lines at
the top of the script:

#!/usr/bin/Rscript
#!/usr/bin/env Rscript

or called the script using Rscript:

Rscript scriptname

In all cases, the behavior is the same.

Here's one of the scripts:

#!/usr/bin/env Rscript

library(apcluster)
options(stringsAsFactors = FALSE)

args <- commandArgs(TRUE)
num <- args[1]

numClusters <- num
mydata <- read.csv("input_data.csv")
xData <- mydata[, 1]
yData <- mydata[, 2]

fit <- apclusterK(negDistMat(r=2), mydata, K = numClusters)
#Leave the rest commented for now
#output <- data.frame(xData, yData, c(as.data.frame(fit[1]),
as.data.frame(fit[2]), as.data.frame(fit[3])))
#write.csv <- write.table(output, file = "output_AP.csv", sep = ",",
row.names = FALSE, col.names = FALSE)

Here's a call to the script:

$ ./affinity_propagation.R 3

Attaching

The following object is masked package::

    heatmap

Trying p = -15.41969
   Number of clusters: 17
Error in tmpk - K : non-numeric argument to binary operator
Calls: apclusterK ... apclusterK -> .local -> apclusterK -> apclusterK ->
.local
Execution halted

The line with the expression "tmpk - K" is from the code for the
apclusterK() function (from the apcluster library).  The
definition/value of tmpk is set using a call to the function length().
I mention this because the failure we're getting with the other script
also seems to involve an expression with call to length().  Here's the
error from calling that script:

$ ./spectral_clustering.R 3
Loading required package: methods
Error in length(tmpsig) * nc : non-numeric argument to binary operator
Calls: specc -> specc -> .local -> matrix
Execution halted

The script is calling the specc() function in the kernlab library.

Any ideas what's going on here, why it's not working and what we can
do to get it to work?  Is there something that needs to be set or run
in .Renviron or .Rprofile, say?

Thanks.

Milt Epstein
Programmer in Computational Genomics
Institute for Genomic Biology (IGB)
University of Illinois at Urbana-Champaign (UIUC)
mepstein at illinois.edu