Rcpp function crashing R when using a “long” input

Question

I have a C++ function for linear interpolation that I'm running using Rcpp the function used to work fine when I was doing some interpolation with initial vectors that are not so "long", but when I triple the size of the initial vector the function works fine out of a loop but when i decide to loop the function with same inputs Rstudio crashes. I had a look at r aborted when using rcpp and R crashes when calling a Rcpp function in a loop but I just did not find any of the solutions helpfull to my problem.

This is the C++ function:

double  InterpolacionRAWCpp(NumericVector plazos,NumericVector tasas,double plazoRequerido){

  double interpolacionRAW,a,p1,p2;
  int numero_plazo;
  double ayuda_plazo = plazoRequerido;
  NumericVector tn(tasas.size()+1);
  NumericVector pn(tasas.size()+1);
  tn[0] = tasas[0];
  pn[0] = 0;
  tn[Rcpp::Range(1,tn.size())] = tasas;
  pn[Rcpp::Range(1,pn.size())] = plazos;
  interpolacionRAW = 0;
  numero_plazo = tasas.size();
  if(ayuda_plazo==0){
    interpolacionRAW=tn[0];
  } else {
    for(int i =0; i<numero_plazo;++i){
      if((plazoRequerido > pn[i]) & (plazoRequerido<=pn[i+1])){
        a = pn[i+1]-pn[i];
        p1 = (pn[i+1]*tn[i+1]-pn[i]*tn[i])/a;
        p2 = (pn[i+1]*pn[i]*(tn[i]-tn[i+1]))/(a*plazoRequerido);
        interpolacionRAW=p2+p1;  
        break;
      } else if(plazoRequerido>pn[(numero_plazo-1)]){
        // Cuando el plazo es mayor al maximo plazo del vector de plazos
        // entonces se extrapola
        a=pn[(numero_plazo)]-pn[numero_plazo-1];
        p1=(pn[(numero_plazo)]*tn[(numero_plazo)]-pn[numero_plazo-1]*tn[numero_plazo-1])/a;
        interpolacionRAW=(p1*(plazoRequerido-pn[(numero_plazo)])+pn[(numero_plazo)]*tn[(numero_plazo)])/plazoRequerido;
        break;
      } else if(plazoRequerido<pn[0]){
        // Cuando el plazoRequerido es menor que el plazo minimo del vector de plazos
        // se extrapola, para atraz
        a=pn[1]-pn[0];
        p1=(pn[1]*tn[1]-pn[0]*tn[0])/a;
        interpolacionRAW=(p1*(plazoRequerido-pn[1])+pn[1]*tn[1])/plazoRequerido;
        break;
      }

    }

  }

  return interpolacionRAW;
}

With this inputs I won't get any problem looping the function:

plazo2 = c(0.25,0.50,1.00,3.00,5.00,7.00,10.00,15.00)
dato2 = c(4.147594,4.197599 ,4.403012,5.281392,6.169297,7.124895,8.699570,11.574581)
x = NULL
for(i in 1:10000){
  x = rbind(x,InterpolacionRAWCpp(plazo2,dato2,10))
}

but if I change plazo2 and dato2 for this:

plazo1 = c(0.08333,0.16667,0.25000,0.50000,1.01389,2.02778,3.04167,4.05556,5.06944,
           6.08333,7.09722,8.11111,9.12500,10.13889,11.15278,12.16667,13.18056,14.19444,15.20833)
dato1 = c(4.11798,4.13244,4.14501,4.19365,4.39560,4.80608,5.27216,5.71510,6.16038,
         6.62284,7.11100,7.67899,8.17816,8.68633,9.20055,9.71490,10.32713,10.94182,11.55650)

Rstudio will crash

Answer 1

This is an indexing error. C++ indexes range from 0 to n-1 instead of R 's 1 to n system.

To an extent, you consider this here:

// Note vectors are n+1
NumericVector tn(tasas.size()+1);
NumericVector pn(tasas.size()+1);

// Assign into 0
tn[0] = tasas[0];
pn[0] = 0;

However, the next portion is problematic:

// Here we are using n+1 but need just n. 
tn[Rcpp::Range(1,tn.size())] = tasas;
pn[Rcpp::Range(1,pn.size())] = plazos;

Switch the last part to:

// Retrieves n+1 and by subtracting 1 we stay in range.
tn[Rcpp::Range(1,tn.size() - 1)] = tasas;
pn[Rcpp::Range(1,pn.size() - 1)] = plazos;