Passing arguments to R via the command line

We have seen how to run a “static” program – i.e., one that runs the same way each time. In the earlier example, the code will generate different p-values each time (because it will be run with a different random seed each time), but we can’t run multiple versions simultaneously (well, we can, but we would overwrite sim.RData each time).

To fix this problem, we need to be able to pass an argument from the command line to R. This is accomplished via the --args option, as in a command like:

R CMD BATCH --no-save --no-restore "--args 3" sim.R .sim.Rout

From within R, the argument (3) is accessible via the commandArgs() function. So, let us rewrite sim.R as follows:

N <- 10000
p <- numeric(N)
n <- 10
for (i in 1:N) {
  x <- rnorm(n)
  y <- rnorm(n, sd=3)
  p[i] <- t.test(x, y, var.equal=TRUE)$p.value
}
id <- commandArgs(TRUE)
filename <- paste("sim", id[1], ".RData", sep="")
save(p, file=filename)

Here, the TRUE option means that we only want the stuff in the --args section and we don’t care about the --no-save and --no-restore options. Now, we can run

R CMD BATCH --no-save --no-restore "--args 1" sim.R .sim.Rout

which runs our simulation and saves the resulting p-values in a file called sim1.RData; likewise,

R CMD BATCH --no-save --no-restore "--args 2" sim.R .sim.Rout

runs another simulation and saves those results as sim2.RData, and so on.

The result of commandArgs is a character vector, so if you want to use it for numeric purposes, you have to convert it with id <- as.numeric(commandArgs(TRUE)). Then you could use it to, say, set the simulation seed with set.seed(id) or change the simulation settings with y <- rnorm(n, sd=id).