Skip to content

Rpkg.org

Latest commit

 

History

History
95 lines (79 loc) · 4.69 KB

Rpkg.org

File metadata and controls

95 lines (79 loc) · 4.69 KB
Raw

R package coding project for CS599 graduate students ONLY

The goal of this project is to provide you with experience in testing/documenting/distributing your code as a free/open-source R package, as well as writing efficient C/C++ code.

To figure out how to write an R package, you can read the free online book R packages, LINK TO MY VIDEOS.

You need at least two algorithms from previous coding projects to include in your R package.

  • First project (mid-term): choose from KMEANS, GMM, HCLUST.
  • Second project (final): include BINSEG, DYNPROG, or PCA in C++ using either RcppEigen or RcppArmadillo.

For each function you need to

  • provide documentation that includes example code that shows how to use it. You can either write the Rd files yourself (recommended), or use a documentation generation system like roxygen2/inlinedocs.
  • write at least one test to make sure it returns a reasonable result (e.g. it returns an integer vector with same length as the number of rows of the input, and all values from 1 to the specified number of clusters).

One of the two functions should be implmented in C/C++ code with an Rcpp interface. For examples see the aum_sort and aum_map C/C++ functions in my aum package.

  • src/your_algo.h should define a function prototype and return/error status codes. Use different positive integer values to indicate different types of errors.
  • src/your_algo.cpp should define a standard (non-Rcpp) C/C++ function which returns an integer status code (0 for success, errors as defined in your header file). It should use standard (non-Rcpp) types as input arguments, for example double* to pass the data, and int to pass the length of the data.
  • src/interface.cpp should have an interface function with input/output Rcpp types, and which calls the function you defined in src/your_algo.cpp. This file should contain declarations, 
    #include <Rcpp.h>
#include "your_algo.h"
// [[Rcpp::export]]
  • Don’t forget to run Rcpp::compileAttributes("path/to/yourPkg") which looks for the exported functions in yourPkg/src/*.cpp and writes corresponding files: yourPkg/src/RcppExports.cpp and yourPkg/R/RcppExports.R (which are required to use your C++ code).

You need to upload your code to a GitHub repository (you need a free GitHub account so please create one if you don’t have one yet). Your repo should NOT include any files except source code/docs/tests/etc required to build the package (e.g. you will get minus points if there is a .Rhistory file, or compiled binary *.o *.so *.dll files, in your github repo).

You need to setup GitHub Actions for continuous integration checks of your R package. This will run the standard package checks, and your tests, every time you push new commits to GitHub. Make sure that your package passes checks with no WARNINGs and no ERRORs (but NOTEs are OK).

You should write a README.md or README.org file that includes at least two sections (1) Installation which explains how to install your package via remotes::install_github(“username/repo”) and (2) Usage which shows a few lines of R code about how to use your package (can be same as examples).

Please upload a PDF with

FAQ

How to get emacs working for R package compilation?

If you want to use M-x compile in emacs to compile your R package, put the following code in package/src/.dir-locals.el

((nil . ((compile-command . "R -e \"Rcpp::compileAttributes('..')\" && R CMD INSTALL .. && R --vanilla < ../tests/run_tests.R"))))

If M-x compile does not work (still runs make), and shows a message in minibuffer about irony, then the fix is M-x package-install-packages and install packages related to irony. Also if you would rather just type one key to re-compile, put the following in ~/.emacs

(global-set-key [f9] 'compile)
(add-to-list 'safe-local-variable-values '(compile-command . "R -e \"Rcpp::compileAttributes('..')\" && R CMD INSTALL .. && R --vanilla < ../tests/run_tests.R"))