One of the nice features of this library is that it provides a set of data structures that are sufficient to work with most of the output from functions inside R. There are two ways to evaluate code. The first is with evalR. It runs the code you’ve specified and returns the output as an Robj struct (which is how ALL data in R is stored). You could do the following:
Robj x = evalR("rnorm(15)");x would hold a Robj struct with a pointer to the output of that command. There are two problems with using this primitive approach:
rnorm(15) was allocated by R. Since R has no way to know what you’re doing with the data inside your D program, it could be reclaimed by the garbage collector at any time. You could wrap the evalR command like this: Rf_protect(evalR("rnorm(15)")). That would prevent R from ever collecting the data, so to prevent a memory leak, you’d need a corresponding call Rf_unprotect_ptr(x) or Rf_unprotect(1). See the R extensions manual for further information. The bottom line is that this is messy. You’re having to manually manage memory on top of a garbage collector.Vector struct. In other words, you can do thisauto x = Vector("rnorm(15)");The Vector struct uses reference counting to handle the memory management for you, and it allows convenient access to the data, for instance:
double x1 = x[1];
x[2] = 4.9;The other way to evaluate R code is with evalRQ. This is much simpler, because it does not return anything. One example is printing something to the screen:
evalRQ(`print("This was printed by R")`);It’s also useful for intermediate results. Maybe you are reading in a big dataset and you only need one variable.
evalRQ(`data.raw <- read.csv("file.csv")
auto x = Vector("data.raw[,2]");evalRQ was used to read in the data, then Vector was used to acquire a pointer to the variable you want to work with in D.