[This article was first published on Rcpp Gallery, and kindly contributed to R-bloggers]. (You can report issue about the content on this page here)
Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.

We have seen the use of the STL transform functions in the posts STL transform and Transforming a matrix. We use the same logic in conjuction with a logical (ie boolean) vector in order subset an initial vector.

```#include <Rcpp.h>

using namespace Rcpp;
using namespace std;

const double flagval = __DBL_MIN__; // works
//const double flagval = NA_REAL;   // does not

// simple double value 'flagging' function
inline double flag(double a, bool b) { return b ? a : flagval; }

// [[Rcpp::export]]
NumericVector subsetter(NumericVector a, LogicalVector b) {
// We use the flag() function to mark values of 'a'
// for which 'b' is false with the 'flagval'
transform(a.begin(), a.end(), b.begin(), a.begin(), flag);

// We use sugar's sum to compute how many true values to expect
NumericVector res = NumericVector(sum(b));

// And then copy the ones different from flagval from a into
// res using the remove_copy function from the STL
remove_copy(a.begin(), a.end(), res.begin(), flagval);
return res;
}
```

We can illustrate this on a simple example or two:

```a <- 1:5
subsetter(a, a %% 2 == 0)

 2 4

subsetter(a, a > 2)

 3 4 5
```

Casual benchmarking (not shown) shows this to be comparable to and even slightly faster than basic indexing in `R` itself.