I have written up a piece of C code which declares a square matrix of size 4x4. Then it samples from a sampling function called rgig
in package GeneralizedHyperbolic
in R. It inverses the matrix using a gsl library from gnu and spits out the result. This is an exercise in calling R from C.
#include <stdio.h>
#include <stdio.h>
#include <math.h>
#include <stdlib.h>
#include <stddef.h>
// for gsl
#include <gsl/gsl_machine.h>
#include <gsl/gsl_rng.h>
#include <gsl/gsl_randist.h>
#include <gsl/gsl_cdf.h>
#include <gsl/gsl_cblas.h>
#include <gsl/gsl_sf_gamma.h>
#include <gsl/gsl_vector.h>
#include <gsl/gsl_matrix.h>
#include <gsl/gsl_blas.h>
#include <gsl/gsl_linalg.h>
// for R embedding in C
#include <Rinternals.h>
#include <Rdefines.h>
#include <Rembedded.h>
#include <R_ext/Parse.h>
void gsl_square_matrix_inverse (gsl_matrix *, gsl_matrix *, int);
SEXP get_rInvGauss(void);
int main(void)
{
// Define the dimension n of the matrix
// and the signum s (for LU decomposition)
int s, i, j, n = 4;
// Define all the used matrices
gsl_matrix * m = gsl_matrix_alloc (n, n);
gsl_matrix * inverse = gsl_matrix_alloc (n, n);
// R embedding in C
char *localArgs[] = {"R", "--no-save","--silent"};
SEXP rInvGauss;
// init R embedding
Rf_initEmbeddedR(3, localArgs);
printf("\n Printing matrix m before set. size %d by %d... \n", n, n);
for(i=0; i<n; i++){
printf("\n");
for(j=0; j<n; j++){
printf(" %f ", gsl_matrix_get(m, i, j));
}
}
// set diagonal elements of matrix m from Inverse Gaussian Random samples
for(i=0; i<n; i++){
rInvGauss = get_rInvGauss();
gsl_matrix_set(m, i, i, *REAL(rInvGauss));
}
Rf_endEmbeddedR(0); // end the R embedding in C
printf("\n Printing matrix m ..... \n");
for(i=0; i<n; i++){
printf("\n");
for(j=0; j<n; j++){
printf(" %f ", gsl_matrix_get(m, i, j));
}
}
// inverse of matrix m
gsl_square_matrix_inverse (m, inverse, n);
printf("\n Printing inverse of matrix m ..... \n");
for(i=0; i<n; i++){
printf("\n");
for(j=0; j<n; j++){
printf(" %f", gsl_matrix_get(inverse, i, j));
}
}
return 0;
}
SEXP get_rInvGauss(void) {
SEXP e, s, t, tmp, result;
int errorOccurred, n=1;
double chi=5, psi=4, lambda=0.5;
// create and evaluate 'require(GeneralizedHyperbolic)'
PROTECT(e = lang2(install("require"), mkString("GeneralizedHyperbolic")));
R_tryEval(e, R_GlobalEnv, &errorOccurred);
if (errorOccurred) {
// handle error
printf("\n Error loading library GeneralizedHyperbolic:");
}
UNPROTECT(1);
// Create the R expressions using a paired list
// rgig(n = 1, chi = 5, psi = 4, lambda = 0.5) with the R API.
PROTECT(t = s = allocVector(LANGSXP, 5));
// could also be done by: PROTECT(t = s = allocList(5)); SET_TYPEOF(s, LANGSXP);
tmp = findFun(install("rgig"), R_GlobalEnv);
if(tmp == R_NilValue) {
printf("No definition for function rgig.\n");
UNPROTECT(1);
exit(1);
}
SETCAR(t, tmp); t = CDR(t);
SETCAR(t, ScalarInteger(n)); SET_TAG(t, install("n")); t= CDR(t);
SETCAR(t, ScalarReal(chi)); SET_TAG(t, install("chi")); t= CDR(t);
SETCAR(t, ScalarReal(psi)); SET_TAG(t, install("psi")); t= CDR(t);
SETCAR(t, ScalarReal(lambda)); SET_TAG(t, install("lambda")); t= CDR(t);
PROTECT(result = R_tryEval(VECTOR_ELT(s, 0), R_GlobalEnv, NULL));
UNPROTECT(2);
return(result);
}
void gsl_square_matrix_inverse (gsl_matrix *m, gsl_matrix *inverse, int n){
int s, i, j;
gsl_permutation * perm = gsl_permutation_alloc (n);
// Make LU decomposition of matrix m
gsl_linalg_LU_decomp (m, perm, &s);
// Invert the matrix m
gsl_linalg_LU_invert (m, perm, inverse);
}
I compiled the code using:
R CMD SHLIB -lgsl -lgslcblas embedR_matinv.c
with output:
gcc -arch x86_64 -std=gnu99 -I/Library/Frameworks/R.framework/Resources/include -I/Library/Frameworks/R.framework/Resources/include/x86_64 -DNDEBUG -I/usr/local/include -fPIC -g -O2 -c embedR_matinv.c -o embedR_matinv.o
gcc -arch x86_64 -std=gnu99 -dynamiclib -Wl,-headerpad_max_install_names -undefined dynamic_lookup -single_module -multiply_defined suppress -L/usr/local/lib -o embedR_matinv.so embedR_matinv.o -lgsl -lgslcblas -F/Library/Frameworks/R.framework/.. -framework R -Wl,-framework -Wl,CoreFoundation
When I submit:
R CMD embedR_matinv
its gives the error:
/Library/Frameworks/R.framework/Resources/bin/Rcmd: line 62: exec: embedR_matinv: not found
What am I doing wrong?
I also changed the main()
to test()
and made a shared object as
R CMD SHLIB -lgsl -lgslcblas embedR_matinv.c -o embedR_matinv
with the output:
gcc -arch x86_64 -std=gnu99 -I/Library/Frameworks/R.framework/Resources/include -I/Library/Frameworks/R.framework/Resources/include/x86_64 -DNDEBUG -I/usr/local/include -fPIC -g -O2 -c embedR_matinv.c -o embedR_matinv.o
If I do a dyn.load("embedR_matinv.so")
in R Studio, the code runs without any termination i.e. it hangs!
Any suggestions on what is wrong in the code?
Are you using a Mac OSX system?
try this:
sudo cp /Library/Frameworks/R.framework/Resources/library/Rserve/libs/x86_64/Rserve-bin.so\ /Library/Frameworks/R.framework/Resources/bin/Rserve