Learn how to build R packages

• Local: Y station:/home/wangjl/data/project/learn_R_pkg
• github: https://github.com/BioMooc/learn_R_pkg

About R pkg

• https://cran.r-project.org/doc/manuals/R-exts.html

• https://r-pkgs.org/

• https://github.com/hadley/r-pkgs

• https://r-pkgs.org/src.html

About Rcpp

• https://dirk.eddelbuettel.com/code/rcpp.html

• http://www.rcpp.org/

• http://gallery.rcpp.org

• http://adv-r.had.co.nz/Rcpp.html

• https://adv-r.hadley.nz/rcpp.html

About Matrix

• https://github.com/RcppCore/RcppEigen
• https://github.com/RcppCore/RcppArmadillo

About C++

• Effective C++, More Effective C++ and Effective STL.
• https://teuder.github.io/rcpp4everyone_en/

Examples

On Ubuntu 20.04

$ which R
/usr/local/bin/R
$ ls -lth /usr/local/bin/R
lrwxrwxrwx 1 root root 26 Sep  3  2021 /usr/local/bin/R -> /data/public/R-4.1.1/bin/R

devtools 2.4.2
roxygen2 7.1.2
Rcpp 1.0.7

ggplot2 3.3.5
ggrepel 0.9.1

pkgS1 The basic R package building process

pkgS2 How to depend on functions in other pkgs?

pkgS3 How to use C/C++ in my R pkg? .C()

• http://dept.stat.lsa.umich.edu/~yizwang/software/maxLinear/noteonR.html
• http://users.stat.umn.edu/~geyer/rc/
• https://github.com/cjgeyer/foo

1. compile the C/C++ codes into a shared library (.dll in Windows and .so (.o) in linux),
2. load the shared library by dyn.load("LIBNAME"), and
3. call the compiled C function by .C("FUNNAME",...). Here LIBNAME is the name of the shared library file and FUNNAME is the name of the C function.

Do settings when loading the pkg, use Hooks for Namespace Events like .onLoad(). In pkgS3/R/zzz.R

$ cat C_funcs.c #函数必须返回 void 类型; 只能通过指针传递计算结果;
void useC(int *i) {
  i[0] = 11;
}
$ R CMD SHLIB C_funcs.c

> dyn.load("C_funcs.so")
> a <- 1:10
> out <- .C("useC", b = as.integer(a)) #第一个参数是C函数名，接着是其他R参数
> out$b
[1] 11 2 3 4 5 6 7 8 9 10


use #' @useDynLib pkgS3 on one function comment.

pkgS4 How to use C/C++ in my R pkg? .Call()

$ vim src/helloC1.c
#include <R.h>
#include <Rdefines.h>
#include <string.h>

SEXP add2(SEXP a, SEXP b) {
  SEXP result = PROTECT(allocVector(REALSXP, 1));
  REAL(result)[0] = asReal(a) + asReal(b);
  UNPROTECT(1);

  return result;
}

$ R CMD SHLIB src/helloC1.c

> dyn.load("src/helloC1.so")
> .Call("add2", 10, 20)
[1] 30

pkgS5 How to use C/C++ in my R pkg? Rcpp pkg.

Step1: put this in a R/xx.R file, to load the compiled .so file
#' @useDynLib pkgS5
#' @import Rcpp
#'
NULL

Step2: put the Rcpp pkg you use in DESCRIPTION file, like
LinkingTo: Rcpp (>= 0.11.0), RcppArmadillo, RcppEigen

Step3: write your C++ file in src/xx.cpp
#include <Rcpp.h>
using namespace Rcpp;

//' @export
// [[Rcpp::export]]
NumericVector timesTwo(NumericVector x) {
  return x * 2;
}


(optional) put `//' @export` before your C++ function to expose it outside this R pkg

Step4: test
> devtools::document()
> unlink("NAMESPACE")
> devtools::document()
> devtools::load_all()
> timesTwo(3:5)
[1]  6  8 10

Step5: build and test
> devtools::build()
> install.packages("~/data/project/learn_R_pkg/pkgS6_0.1.0.tar.gz", repos = NULL, type = "source")
Restart R session: shift+ctrl+F10
> library(pkgS6)
> timesTwo(1:3)
[1] 2 4 6

Install

> #devtools::install_github("userName/my_rpackage/rmytools")
> devtools::install_github("BioMooc/learn_R_pkg/pkgS2")