R keeps crashing when executing 'rlogspline'

On Thu, 31 May 2007, Jacques Wagnor wrote:

Dear List,

I have a simple model as follows:

x <- rnorm(500)
library(logspline)
fit <- logspline(x)
n <- 1000000
y <- replicate(n, sum(rlogspline(rpois(1,10), fit))) # last line

What you do not see in RGui is this:


  *** caught segfault ***
address (nil), cause 'memory not mapped'

Traceback:
  1: .C("rpqlsd", as.double(c(fit$coef.pol, fit$coef.kts)), 
as.double(fit$knots),     as.double(fit$bound), as.integer(0), qq = 
as.double(p), as.integer(length(fit$knots)),     as.integer(length(p)), 
PACKAGE = "logspline")
  2: qlogspline(pp, fit)
[rest deleted]

I think you are getting zero from rpois(1,10) with larger values of n.

rlogspline does not check for n==0

This results in calling qlogspline( numeric(0), fit ) and generates the 
segfault, I guess.

So figure out what should happen when rpois(1,10) returns zero and avoid 
calling rlogspline in those cases.

If 'zero.value' is a suitable value, then this will work:

library(logspline)
x <- rnorm(500)
fit <- logspline(x)
res <- sapply(1:50, function(x) sum(rlogspline(x, fit)))
n <- 1000000
rp <- rpois(n,10)
y <- c( zero.value, res )[ rp+1 ] ## table lookup is faster!

Chuck

The problem I keep getting is Rcrashes when doing the last line.  It
seems to be fine if n is small, but not if n is 1000000.  The message
I keep getting is:

"R for Windows GUI front-end has encountered a problem and needs to
close.  We are sorry for the inconvenience.  If you were in the middle
of something, the information you were working on might be lost."

Any insights would be appreciated,

Jacques

platform       i386-pc-mingw32
arch           i386
os             mingw32
system         i386, mingw32
status
major          2
minor          5.0
year           2007
month          04
day            23
svn rev        41293
language       R
version.string R version 2.5.0 (2007-04-23)

Charles C. Berry                        (858) 534-2098
                                          Dept of Family/Preventive Medicine
E mailto:cberry at tajo.ucsd.edu	         UC San Diego
http://biostat.ucsd.edu/~cberry/         La Jolla, San Diego 92093-0901