[Rcpp-devel] conflict between R_ext/Applic.h and RcppArmadillo.h?

Hi Richard,

| Hi,
| 
| I am trying to call Rdqags (from R_ext/Applic.h) in a C++ program that
| uses RcppArmadillo. Everything seems to work fine except the compiler
| warns about conflicts. Is there a way to avoid these conflicts? Could
| they cause any unforeseen problems? Below is a self-contained example
| illustrating the problem. Note that RcppArmadillo is not actually
| required in the example, but it is in my program. See below for
| sessionInfo().

Hm, that's interesting.   I haven't used R_ext/Applic.h much; it says

    /* This header file contains routines which are in the R API and ones which
   are not.

   Those which are not can be used only at the user's risk and may change
   or disappear in a future release of R.
   */

so in a way that gives access from C to the R API.  Not surprising that it
bites with what we have.

Rcpp and RcppArmadillo have a moderately involved hierarchy of headers, we
generally do not recommend mixing or changing.

So the simplest suggestion I have ... is to 'split' things. Have your
RcppArmadillo code, have it call a C function ... defined over in another
source file and header which just includes R_ext/Applic.h and communicates
with the other side via standard C (or C++ / STL types).  Makes sense?

Let us know how that goes.  

Dirk


| Thanks,
| Richard
| 
| 
| library(RcppArmadillo)
| library(inline)
| 
| # Integrand
| inc <- "
| #include <R_ext/Applic.h>
| void f(double *x, int n, void *ex) {
|   double *v;
|   v = (double*)ex;
|   double sigma = v[0];
|   for(int i=0; i<n; i++) {
|     x[i] = exp(-(x[i]*x[i]) / (2*sigma*sigma));
|   }
| }
| "
| 
| # Compute integral
| f <- cxxfunction(signature(sigma_="numeric",
|                            lower_="numeric", upper_="numeric"),
|   "
|     double sigma = as<double>(sigma_);
|     double lower = as<double>(lower_);
|     double upper = as<double>(upper_);
|     void *ex;
|     ex = &sigma;
|     double epsabs = 0.001;
|     double epsrel = 0.001;
|     double result = 0.0;
|     double abserr = 0.0;
|     int neval = 0;
|     int ier = 0;
|     int limit = 50;
|     int lenw = 200;
|     int last = 0;
|     int iwork = 50;
|     double work = 200.0;
|     Rdqags(f, ex, &lower, &upper, &epsabs, &epsrel, &result, &abserr,
|            &neval, &ier, &limit, &lenw, &last, &iwork, &work);
|     return Rcpp::List::create(result, abserr, last, ier);
|   ", plugin="RcppArmadillo", includes=inc)
| 
| fr <- function(x, sigma=1) exp(-x^2 / (2*sigma*sigma))
| 
| sig <- 10
| f(sig, 0, 2)
| str(integrate(fr, 0, 2, sigma=sig))
| 
| 
| 
| 
| 
| > sessionInfo()
| R version 2.14.1 (2011-12-22)
| Platform: i386-pc-mingw32/i386 (32-bit)
| 
| locale:
| [1] LC_COLLATE=English_United States.1252
| [2] LC_CTYPE=English_United States.1252
| [3] LC_MONETARY=English_United States.1252
| [4] LC_NUMERIC=C
| [5] LC_TIME=English_United States.1252
| 
| attached base packages:
| [1] stats     graphics  grDevices utils     datasets  methods   base
| 
| other attached packages:
| [1] inline_0.3.8         unmarked_0.9-6       RcppArmadillo_0.2.34
| [4] Rcpp_0.9.9           lattice_0.20-0       reshape_0.8.4
| [7] plyr_1.7.1
| 
| loaded via a namespace (and not attached):
| [1] grid_2.14.1  tools_2.14.1
| _______________________________________________
| Rcpp-devel mailing list
| Rcpp-devel at lists.r-forge.r-project.org
| https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel

"Outside of a dog, a book is a man's best friend. Inside of a dog, it is too
dark to read." -- Groucho Marx

Richard,

| 
| Hi Richard,
|

| | Hi,
| | 
| | I am trying to call Rdqags (from R_ext/Applic.h) in a C++ program that
| | uses RcppArmadillo. Everything seems to work fine except the compiler
| | warns about conflicts. Is there a way to avoid these conflicts? Could
| | they cause any unforeseen problems? Below is a self-contained example
| | illustrating the problem. Note that RcppArmadillo is not actually
| | required in the example, but it is in my program. See below for
| | sessionInfo().

Another idea would be to dig up some old posts on the about how Rcpp lets you
get R API access points directly.  Davor used that to get to loess.  That way
you would get to use Rdqags but without needing the conflicting header.  Just
another thought.

Dirk

"Outside of a dog, a book is a man's best friend. Inside of a dog, it is too
dark to read." -- Groucho Marx

Hi Dirk,

Thanks for those suggestions. I am trying them both but without
success so far. I will keep at it, but first, I wonder how serious
these warnings are? I can still pass R CMD check, and my program seems
to work very well. Could something disastrous happen if I ignore the
warnings? Do you think CRAN would accept the package?

Richard

Richard,

On 1 March 2012 at 15:36, Dirk Eddelbuettel wrote:
|
| Hi Richard,
|
| On 1 March 2012 at 16:25, Richard Chandler wrote:
| | Hi,
| |
| | I am trying to call Rdqags (from R_ext/Applic.h) in a C++ program that
| | uses RcppArmadillo. Everything seems to work fine except the compiler
| | warns about conflicts. Is there a way to avoid these conflicts? Could
| | they cause any unforeseen problems? Below is a self-contained example
| | illustrating the problem. Note that RcppArmadillo is not actually
| | required in the example, but it is in my program. See below for
| | sessionInfo().

Another idea would be to dig up some old posts on the about how Rcpp lets you
get R API access points directly. ?Davor used that to get to loess. ?That way
you would get to use Rdqags but without needing the conflicting header. ?Just
another thought.

Dirk

--
"Outside of a dog, a book is a man's best friend. Inside of a dog, it is too
dark to read." -- Groucho Marx

Hi Richard.

| Hi Dirk,
| 
| Thanks for those suggestions. I am trying them both but without
| success so far. I will keep at it, but first, I wonder how serious
| these warnings are? I can still pass R CMD check, and my program seems

You never showed them to us, so we can't tell.

| to work very well. Could something disastrous happen if I ignore the
| warnings? Do you think CRAN would accept the package?

For CRAN you probably want a package framework anyway rather than an inline
call.  And what R CMD check says matters -- which is a superset of compiler
warnings plus a boatload or two of other tests.  Which are sometimes a pain,
but beneficial to all of us in the long run.

Dirk
 
| Richard
|

| >
| > Richard,
| >

| > |
| > | Hi Richard,
| > |

| > | | Hi,
| > | |
| > | | I am trying to call Rdqags (from R_ext/Applic.h) in a C++ program that
| > | | uses RcppArmadillo. Everything seems to work fine except the compiler
| > | | warns about conflicts. Is there a way to avoid these conflicts? Could
| > | | they cause any unforeseen problems? Below is a self-contained example
| > | | illustrating the problem. Note that RcppArmadillo is not actually
| > | | required in the example, but it is in my program. See below for
| > | | sessionInfo().
| >
| > Another idea would be to dig up some old posts on the about how Rcpp lets you
| > get R API access points directly. ?Davor used that to get to loess. ?That way
| > you would get to use Rdqags but without needing the conflicting header. ?Just
| > another thought.
| >
| > Dirk
| >
| > --
| > "Outside of a dog, a book is a man's best friend. Inside of a dog, it is too
| > dark to read." -- Groucho Marx

"Outside of a dog, a book is a man's best friend. Inside of a dog, it is too
dark to read." -- Groucho Marx