in reply to Passing parameters to R script

Since you appear to be working under Unix, how about: 

##In a file called test_args.R
args <- commandArgs()

print (args)

[download]

which you can run from the shell 

cat test_args.R | R --slave --args all your base

[download]

which prints 

[1] "usr/lib64/R/bin/exec/R" "--slave"
[3] "--args"                 "all"
[5] "your"                   "base"

[download]

Replies are listed 'Best First'.
Re^2: Passing parameters to R script
by j1n3l0 (Friar) on Oct 17, 2007 at 19:09 UTC
    Thank you all for your input! I did eventually work it out and my solution looks like so:

    # test.R
# Parse and use command line arguments
# Invoke % R --slave --args 100 < test.R 

Args <- commandArgs();      # retrieve args
x <- c(1:as.real(Args[4])); # get the 4th argument
y <- c(x^2);                # work out square
png(filename="image.png");  # create image file
plot(x,y);                  # plot image

    [download]
    This creates the image file image.png which contains a plot of an exponential curve =)

    The variable Args contains the following:

    [1] "/usr/lib/R/bin/exec/R" 
[2] "--slave" 
[3] "--args"               
[4] "100"

    [download]
    We only require the 4th element for our graph. You can add more arguments and just pick them off accordingly =)

    I hope someone else finds this useful.

    Smoothie, smoothie, hundre prosent naturlig!
[reply]
[d/l]
[select]
      Yes, this was very useful and worked right away! Thanks a lot!
[reply]
      That works fine, but there's a slightly neater way to do it; if you provide a TRUE value to the commandArgs function, it ignores any commandline parameters before (and including) --args.
      So with Args <- commandArgs(TRUE), Args[1] is "100", and is the only element of the vector.

      It's also possible to use a simpler syntax to call the script on the command line, which implicitly assumes that any arguments are given after the script name:
      Rscript test.R 100

      Cheers,
      Julio
[reply]
[d/l]
[select]
        In Bash shell 
        # a=1
# b=10
# Rscript -e "args<-commandArgs(TRUE);x=args[1]:args[2];x;mean(x);sd(x
+)" $a $b
 [1]  1  2  3  4  5  6  7  8  9 10
[1] 5.5
[1] 3.027650

        [download]

        So you can control variables on the command line R script and you can feed these variables into R itself using commandArgs(TRUE), and manipulate the variables as args1 and args2. So if I imagine if you are doing a system call in perl, you can also construct a similar Rscript on the command line, and get the output accordingly, which you can manipulate further within perl (or any other script).

[reply]
[d/l]

        Here is simple function to parse unix way of parsing argument from R. 

        #
#
# EXAMPLE:
# R --test=mydata.csv
#> getopt('test')
#[1] "mydata.csv"

# getting numeric argument
# R --N=123
#> getopt('N',numeric=T)
#[1] 123
#

getopt <- function(argname, numeric = F)
  {
    args <- commandArgs()  
    argexp = sprintf('^--%s=',argname)
    idx <- regexpr(argexp, args);

    ret <- NA
    
    if ( any(idx > 0) )
    { # found matching argument  
    found <- args[idx > 0]
    value <- strsplit(found,"=")
    ret <- unlist(value)[2]
    }

    if ( numeric )
    ret <- as.numeric(ret)

    ret
  }

        [download]

        cheers, c.okugami

[reply]
[d/l]
      In Rscript/R 
      require(plyr)

## get and parse any commandline options/args
cmd.args <- commandArgs(trailingOnly=TRUE)

## this assumes trailingOnly in commandArgs()
mk.cmd.args.list <- function(args) {
  if (length(args)>1) {
    split.list <- strsplit( args[-1], '\\=' )  ## exclude "--args"
    values <- llply( split.list, '[[', 2 )
    names(values) <- laply( split.list, '[[', 1 )
    values } else list()  ## the possible returns 
}
cmd.args.list <- mk.cmd.args.list(cmd.args)

      [download]
[reply]
[d/l]

In Section Seekers of Perl Wisdom