Use of C++ in Packages

[Peter_Griffin]

I have read this article several times, but still feel confused about the conclusion of it, which is don’t use C++ to interface with R. There have been so many R packages relying on C++, but now one of the R Core Team members is telling us that we shouldn't use C++ to interface with R. I used to think C++ is the solution for computation intensive tasks in R, since writing C or Fortran code is more challenging for most users.
https://developer.r-project.org/Blog/public/2019/03/28/use-of-c---in-packages/index.html

[technocrat]

Hi,

The article is talking about writing C++ programs that make calls to R programs and explains the problems that can cause with the C++ code. Going the other way is not a problem (although it is a challenge), with RCpp and other packages, an R package can bring in a C++ library and tons and tons of them do.

[hinkelman]

There was some discussion of this blog post on r-devel: https://stat.ethz.ch/pipermail/r-devel/2019-March/077560.html

[JohnMount]

In my opinion (formed after re-reads, research, and even asking around)- the original negative r-project.org article is wrong on some of its key technical points (for example: there is no C++ interface to R, article fails to emphasize Rcpp properly uses the C interface, article doesn't mention important history of .C versus .Call, and so on) and does not support a clear conclusion (let alone a conclusion of "don’t use C++ to interface with R"). I respect that the article is coming from a knowledgeable R authority, but feel the article has not been a good thing for the R community.

If the article had been carefully labeled as being only about "writing C++ programs that make calls to R" that would have helped. But it is hard to support that position as the original article starts with "About 20% packages from CRAN and BIOC repositories include some native code and more than a half of those include some code in C++". That commonly means R using C++ for extensions.

Perhaps the article should have been on a personal blog, or r-project.org should become a formally refereed source (if one is going to treat it as authoritative).