[R] line profiling

Florent D. flodel at gmail.com
Fri Apr 5 12:39:12 CEST 2013


This is about the new "line profiling" feature in R 3.0.0. As I was
testing it, I find the results somewhat disappointing so I'd like to
get your opinion.

I put some poorly written code in a test.R file, here are the contents:

double <- function(x) {
  out <- c()
  for (i in x) {
    out <- c(out, 2*i)   # line 4

Then this how I source the file and run a profile:

source("test.R", keep.source = TRUE)
Rprof("test.profile", line.profiling=TRUE)
y <- double(1:20000)
summaryRprof("test.profile", lines = "both")

The relevant part of the output is:

         total.time total.pct self.time self.pct
"double"       0.98    100.00      0.00     0.00
"c"            0.92     93.88      0.92    93.88
test.R#4       0.06      6.12      0.06     6.12

My problem is I was expecting test.R#4 (line 4 of my file) to show up
somewhere between "c" and "double", because "c" is called on line #4
of my file, and line #4 is inside the body of the "double" function.
If I look at the "test.profile" where I told Rprof() to record, there
are a lot of

"c" "double"

records while I was hoping to find

"c" 1#4 "double"

Can someone please explain why, when the profiler records "c", it does
not record the extra information that "c" was called on line 4? Is it
not possible to know? Is it a choice the developers made? An

Thank you in advance for your help.

More information about the R-help mailing list