Passing arguments to R script from command lines

Note: The program just loads a text file containing data, filters out non numeric variables and writes a text file with remaining numeric variables only.

R style

The most natural way to pass arguments from the command line is to use the function commandArgs. This function scans the arguments which have been supplied when the current R session was invoked. So creating a script named sillyScript.R which starts with

#!/usr/bin/env Rscriptargs = commandArgs(trailingOnly=TRUE)

and running the following command line

Rscript --vanilla sillyScript.R iris.txt out.txt

will create a string vector args which contains the entries iris.txt and out.txt. Missing and default arguments can be handled this way:

# test if there is at least one argument: if not, return an error
if (length(args)==0) {
  stop("At least one argument must be supplied (input file).n", call.=FALSE)} else if (length(args)==1) {
  # default output file
  args[2] = "out.txt"}

The simple use case described in the introduction thus gives

## program...df = read.table(args[1], header=TRUE)
num_vars = which(sapply(df, class)=="numeric")
df_out = df[ ,num_vars]write.table(df_out, file=args[2], row.names=FALSE)

Finally, the command lines

Rscript --vanilla sillyScript.R iris.txt out.txt

or

Rscript --vanilla sillyScript.R iris.txt

will both load the file iris.txt, filter out the non numeric variables and write the resulting data in out.txt. Whereas running

Rscript --vanilla sillyScript.R

Error: At least one argument must be supplied (input file).
Execution halted

python style

One package allows to obtain the same result in a python-like style: the package optparse. The package can be used to perform a similar task. Basically, the package contains the functions

• make_option to declare options, their flags, types, default values and help messages;
• OptionParser to read the arguments passed to the R script and parse_args to parse them according to what has been declared thanks to make_option.

The functions are used as follows:

#!/usr/bin/env Rscript
library("optparse")
 
option_list = list(
  make_option(c("-f", "--file"), type="character", default=NULL, 
              help="dataset file name", metavar="character"),
	make_option(c("-o", "--out"), type="character", default="out.txt", 
              help="output file name [default= %default]", metavar="character")); 
 
opt_parser = OptionParser(option_list=option_list);
opt = parse_args(opt_parser);

which produces a list opt that contains all the arguments sorted by order of appearance in option_list and which can be called by their names as declared in this object: opt$file and opt$out. Then, managing null arguments is performed as follows:

if (is.null(opt$file)){
  print_help(opt_parser)
  stop("At least one argument must be supplied (input file).n", call.=FALSE)}

in which the function print_help print the help page of the option list as declared in the object option_list.

The remaining of the function is almost unchanged:

## program...df = read.table(opt$file, header=TRUE)
num_vars = which(sapply(df, class)=="numeric")
df_out = df[ ,num_vars]write.table(df_out, file=opt$out, row.names=FALSE)

If the entire script is saved in a file called yasrs.R,

Rscript --vanilla yasrs.R

will give

Usage: testScript.R [options]


Options:
  -f CHARACTER, --file=CHARACTER
		dataset file name

	-o CHARACTER, --out=CHARACTER
		output file name [default= out.txt]

	-h, --help
		Show this help message and exit


Error: At least one argument must be supplied (input file).
Execution halted

and

Rscript --vanilla yasrs.R --help

will print the help

Usage: testScript.R [options]


Options:
  -f CHARACTER, --file=CHARACTER
		dataset file name

	-o CHARACTER, --out=CHARACTER
		output file name [default= out.txt]

	-h, --help
		Show this help message and exit

and finally

Rscript --vanilla yasrs.R -f iris.txt

or

Rscript --vanilla yasrs.R -f iris.txt -o out.txt

will both create the file out.txt as described in the introduction of this post.

My dearest Céline, I hope that this post is explicit enough so that you can make your choice. Now you know that everything can be done with R, even this kind of map.