Rcpp Evaluation error: subscript out of bounds

Question

I want to use the "glasso" package in Rcpp code, and the cpp code is as follow:

#include <RcppArmadillo.h>
 // [[Rcpp::depends(RcppArmadillo)]]
using namespace Rcpp;
// [[Rcpp::export]]
List sec(arma::mat x,double lam){
Environment gla("package:glasso");
Function gl=gla["glasso"];
double thr=1e-2;bool approx=0;
bool diag=1; bool nu=0;
List bc(5);
bc=gl(x,lam,nu,thr,thr,approx,approx,nu,nu,diag);
return(bc);}

But When I source the code in R and I get follow error.

set.seed(100)
x<-matrix(rnorm(50*10),ncol=10)
s<- var(x)
library(glasso)
a<-sec(s, 0.01)
"Error in sec(s, 0.01) : Evaluation error: subscript out of bounds."

I check document of "glasso" package, the result list contain 5 values, so I'm confused where is the problem.

I assume you do understand that calling R code from C++ still runs at R speed. — Dirk Eddelbuettel, May 14 '18 at 15:47

score 4 · Accepted Answer · answered May 14 '18 at 03:35

There are a couple problems in your code. The latest version of glasso() function returns a list of 7 (not 5) In my answer I will use this version of glasso(). To pass NULL value to glasso() function within Rcpp call, use R_NilValue. In my example I omitted a few values at the end of argument list glasso() function since I was not sure what values you want to pass there (it looks like you are using an older version of this package):

#include <RcppArmadillo.h>
 // [[Rcpp::depends(RcppArmadillo)]]
using namespace Rcpp;
// [[Rcpp::export]]
List sec(arma::mat x,double lam){
Environment gla("package:glasso");
Function gl=gla["glasso"];
double thr=1e-2;bool approx=0;
bool diag=1; bool nu=0;
List bc(7);
bc = gl(x, lam, R_NilValue,thr,thr,approx,diag);
return(bc);}

Here is my R code:

library(Rcpp)

sourceCpp("test.cpp")

library(glasso)
set.seed(100)
x<-matrix(rnorm(50*10),ncol=10)
s<- var(x)
a<-sec(s, 0.01)


   #output
   a
    # $w
    # [,1]        [,2]         [,3]         [,4]        [,5]        [,6]
    # [1,]  0.680779447  0.33966152 -0.009663187 -0.108946142  0.07360211  0.21757108
    # [2,]  0.339661517  1.43157518 -0.042659411 -0.190859910  0.11234565  0.15097851
    #...
    # 
    # $wi
    # [,1]        [,2]        [,3]         [,4]         [,5]        [,6]
    # [1,]  2.26552621 -0.57002634  0.00000000  0.114629054 -0.159489421 -0.46727499
    # [2,] -0.58556329  1.01623488 -0.04678184  0.236972034  0.000000000 -0.03616841
    #...
    # 
    # $loglik
    # [1] -43.31512
    # 
    # $errflag
    # [1] 0
    # 
    # $approx
    # [1] FALSE
    # 
    # $del
    # [1] 0.008940439
    # 
    # $niter
    # [1] 1

Thanks for your answer, I think it works for my problem. For this issue I have 2 more questions.First, in some R function, when parameter has default value which I don't want to change, in this sitution, does it still necessary to set the parameter value? Second, in my example if I want set the parameter start="warm",which kind of type I need set for start? "char" or "int"? — Xia.Song, May 14 '18 at 04:05
@Xiaol.Song If the default values are at the end - yes, you do not have to define them. As for start, the following works: char start[]="cold"; bc = gl(x, lam, R_NilValue,thr,thr,approx,diag,start,R_NilValue); — Katia, May 14 '18 at 04:33

Rcpp Evaluation error: subscript out of bounds

1 Answers1