[forgot to attach the second patch in the first mail, sorry.]
Hello,
Appending information to the profiler's output seems to generate
problems. Here is a small example of code :
<code r>
require(boot)
Rprof( memory.profiling = TRUE)
Rprof(NULL)
for(i in 1:2){
Rprof( memory.profiling = TRUE, append = TRUE)
example(boot)
Rprof(NULL)
}
</code>
The problem is that the file Rprof.out contains more than once the
header information:
$ grep "sample.interval=" Rprof.out
memory profiling: sample.interval=20000
memory profiling: sample.interval=20000
memory profiling: sample.interval=20000
and `summaryRprof` or `R CMD Rprof` are not dealing with it
> idx <- grep( "sample", rownames( smp <- summaryRprof()[[1]] ) );
smp[idx, ]
self.time self.pct total.time total.pct
sample.interval=20000 0 0 0.04 0.1
`sample.interval=20000` is incorrectly considered as a function.
This is not too much of a big deal, but then if I ask for memory
profiling information as well, I get nothing:
> summaryRprof( mem = "stats")
Error in tapply(1:1L, list(index = c("sample.interval=20000:profiling:", :
arguments must have same length
> summaryRprof( mem = "tseries")
Error in data.frame(..., check.names = FALSE) :
arguments imply differing number of rows: 1, 1490
> summaryRprof( mem = "both")
Error in apply(sapply(strsplit(memstuff, ":"), as.numeric), 1, diff) :
dim(X) must have a positive length
$ R CMD Rprof Rprof.out
Each sample represents seconds.
Total run time: 0 seconds.
Total seconds: time spent in function and callees.
Self seconds: time spent in function alone.
% total % self
total seconds self seconds name
Illegal division by zero at /usr/local/lib/R/bin/Rprof line 91, <> line
1491.
A quick fix could be to ignore all the lines containing sample.interval=
except the first one, but then if someone wants to actually change the
interval argument in Rprof, it would not be correct. I attach a patch
against R-devel/src/library/utils/R/summRprof.R to do that anyway, but I
will look at a better solution.
I am not fluent enough in Perl to do the same in the Rprof script, it
looks like it does not handle the memory profiling information anyway. I
am about to learn Perl, so I guess it could be a useful way to do it. I
also attach a patch against R-devel/src/scripts/Rprof that would at
least get rid of the memory profiling information (if any). This is not
as good as dealing with it, but ... this is the first time I ever touch
a Perl script.
Cheers,
Romain
> version
_
platform
i686-pc-linux-gnu
arch
i686
os
linux-gnu
system i686,
linux-gnu
status Under development
(unstable)
major
2
minor 6.0
year
2007
month
03
day
30
svn rev
40983
language
R
version.string R version 2.6.0 Under development (unstable) (2007-03-30
r40983)
--
Mango Solutions
data analysis that delivers
Tel: +44(0) 1249 467 467
Fax: +44(0) 1249 467 468
Mob: +44(0) 7813 526 123
--- summRprof.R 2007-03-31 21:10:31.000000000 +0100
+++ /tmp/R-devel/src/library/utils/R/summRprof.R 2006-10-05 07:54:38.000000000 +0100
@@ -32,10 +32,8 @@
repeat({
chunk<-readLines(filename,n=chunksize)
- chunk <- chunk[ -grep("sample\\.interval=", chunk) ]
if (length(chunk)==0)
break
-
if (memory.profiling){
memprefix<-attr(regexpr(":[0-9]+:[0-9]+:[0-9]+:[0-9]+:",chunk),"match.length")
if (memory=="both"){
@@ -121,7 +119,6 @@
repeat({
chunk<-readLines(filename,n=chunksize)
- chunk <- chunk[ -grep("sample\\.interval=", chunk) ]
if (length(chunk)==0)
break
memprefix<-attr(regexpr(":[0-9]+:[0-9]+:[0-9]+:[0-9]+:",chunk),
--- Rprof 2007-03-31 21:23:16.000000000 +0100
+++ /tmp/R-devel/src/scripts/Rprof 2007-03-31 19:58:29.000000000 +0100
@@ -57,12 +57,11 @@
%totalcounts = ();
while (<>) {
- if (/.*sample\.interval=/) {
- s/.*sample\.interval=//;
+ if (/^sample\.interval=/) {
+ s/sample\.interval=//;
$sample = $_ / 1e6;
} else {
chomp;
- s/[^\"]*\"/"/;
@line = split(/ /);
%names = ();
$leaf = @line[0];
______________________________________________
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
