C structures in R

Isn't this a question clearly in R-devel's domain?

  R-devel is intended for questions and discussion about code development
  in R. Questions likely to prompt discussion unintelligible to
  non-programmers should go to to R-devel.

The short answer is that quite a bit of code, e.g pwilcox and RODBC, does 
things like this.  You don't need to pass the pointer back to R, but if 
you do external pointers are designed for this job.  Nevertheless, 
graphics recording on Windows makes use of integers AFAIR to store C-level 
structures (but it predates raw and external pointers).

Is it safe to call C code from R that mallocs memory for a structure,
returning a pointer to that structure via some 'raw()' parameter. Then,
pass that pointer to another C routine, and finally free the malloced
memory by passing the raw() data to another C routine?

I've written some code that does this, I'm just not sure if anything
could go wrong. The worst I can come up with is a memory leak if the
structure's memory isn't freed - possibly because the R is interrupted.

R isn't going to stomp on memory that's been malloced by an included C
routine between .C calls though, is it?

Barry

[[
gory details:

  R code calls a C routine with .C passing a 4-byte (because 4 is
sizeof(char*) in my architecture) 'raw' object, the C code then mallocs
the structure and copies the address of the structure into the 4 bytes
that the raw object (which appears as a char* in the C routine argument
list) points to. This gets returned back to R. Another option would be
to create a raw object of the right size to store my structure, pass
that to C and not malloc anything in C. But that would mean altering the
C code which I want to do as little as possible.

gory code available on request.

]]

______________________________________________
R-help at stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.

Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

The short answer is that quite a bit of code, e.g pwilcox and RODBC, 
does things like this.  You don't need to pass the pointer back to R, 
but if you do external pointers are designed for this job.  

[reads a bit more of 'Writing R Extensions'...]

  Right yes, this does look like the tool for the job. I'll try and come 
up with a minimal example that duplicates what I'm doing with raw and 
pointers, it might be a useful illustration in the documentation.

Barry

"BaRow" == Barry Rowlingson <B.Rowlingson at lancaster.ac.uk>
    on Fri, 01 Dec 2006 14:09:13 +0000 writes:

>> The short answer is that quite a bit of code, e.g pwilcox
    >> and RODBC, does things like this.  You don't need to pass
    >> the pointer back to R, but if you do external pointers
    >> are designed for this job.

    BaRow>   [reads a bit more of 'Writing R Extensions'...]

    BaRow>   Right yes, this does look like the tool for the
    BaRow> job. I'll try and come up with a minimal example that
    BaRow> duplicates what I'm doing with raw and pointers, it
    BaRow> might be a useful illustration in the documentation.

Yes, indeed.  I've wanted to have a small well documented
example on using external pointers more than once.

And if it would end up to be too large for "Writing R Ext..", 
you could put it at least on the R-Wiki.

Thanks in advance!
Martin