Various C++ syntax errors resulting from expansion of macro ‘RcppExport’ within R Markdown

Question

I have the following .cpp file, which I want to use within R:

#include <Rcpp.h>
#include <RcppEigen.h>

using Eigen::Map;
using Eigen::EigenBase;
using Eigen::HouseholderQR;
using Eigen::MatrixXd;
using Eigen::SparseMatrix;
using Eigen::SparseQR;
using Eigen::COLAMDOrdering;

// [[Rcpp::depends(RcppEigen)]]

// [[Rcpp::export]]
SEXP qr_dense_residop(Map<MatrixXd> X, Map<MatrixXd> Y){
  const HouseholderQR<MatrixXd> QR(X);
  return Rcpp::wrap(Y - (X * QR.solve(Y)));
}

// [[Rcpp::export]]
SEXP qr_sparse_residop(Map<SparseMatrix<double>> X, Map<SparseMatrix<double>> Y){
  const SparseQR<SparseMatrix<double>, COLAMDOrdering<int>> QR(X);
  return Rcpp::wrap(Y - (X * QR.solve(Y)));
}

If I compile it with Rcpp::sourceCpp(), it compiles correctly. However, I want to include the source code in an Rmarkdown document, and so I instead am including it like this:

```{Rcpp, file="resid.cpp"}
```

Then, when I knit the document, I get the following error:

using C++ compiler: ‘g++ (GCC) 11.3.0’
/stornext/System/data/apps/gcc/gcc-11.3.0/bin/g++ -std=gnu++17 -I"/stornext/System/data/apps/R/R-4.3.2/lib64/R/include" -DNDEBUG   -I"/home/users/allstaff/milton.m/R/x86_64-pc-linux-gnu-library/4.3/Rcpp/include" -I"/home/users/allstaff/milton.m/R/x86_64-pc-linux-gnu-library/4.3/RcppEigen/include" -I"/vast/scratch/users/milton.m/tmp/RtmpABumCC/sourceCpp-x86_64-pc-linux-gnu-1.0.12" -I/usr/local/include    -fpic  -g -O2  -w -c file153c34131d6db.cpp -o file153c34131d6db.o
In file included from /home/users/allstaff/milton.m/R/x86_64-pc-linux-gnu-library/4.3/RcppEigen/include/RcppEigenForward.h:26,
                 from /home/users/allstaff/milton.m/R/x86_64-pc-linux-gnu-library/4.3/RcppEigen/include/RcppEigen.h:25,
                 from file153c34131d6db.cpp:1:
file153c34131d6db.cpp: In function ‘SEXPREC* qr_sparse_residop(Eigen::Map<Eigen::SparseMatrix<double, 0, int> >, Eigen::Map<Eigen::SparseMatrix<double, 0, int> >)’:
/home/users/allstaff/milton.m/R/x86_64-pc-linux-gnu-library/4.3/Rcpp/include/RcppCommon.h:140:27: error: expected unqualified-id before string constant
  140 | #define RcppExport extern "C" attribute_visible
      |                           ^~~
file153c34131d6db.cpp:33:1: note: in expansion of macro ‘RcppExport’
   33 | RcppExport SEXP sourceCpp_1_qr_dense_residop(SEXP XSEXP, SEXP YSEXP) {
      | ^~~~~~~~~~
/home/users/allstaff/milton.m/R/x86_64-pc-linux-gnu-library/4.3/Rcpp/include/RcppCommon.h:140:27: error: expected unqualified-id before string constant
  140 | #define RcppExport extern "C" attribute_visible
      |                           ^~~
file153c34131d6db.cpp:45:1: note: in expansion of macro ‘RcppExport’
   45 | RcppExport SEXP sourceCpp_1_qr_sparse_residop(SEXP XSEXP, SEXP YSEXP) {
      | ^~~~~~~~~~
file153c34131d6db.cpp:54:2: error: expected ‘}’ at end of input
   54 | }
      |  ^
file153c34131d6db.cpp:20:81: note: to match this ‘{’
   20 | SEXP qr_sparse_residop(Map<SparseMatrix<double>> X, Map<SparseMatrix<double>> Y){
      |                                                                                 ^
make: *** [file153c34131d6db.o] Error 1

What is happening here, and how can I avoid this error?

Answer 1

You are doing something wrong but it is not clear what. If you look at the existing documentation for knitr and its Rcpp engine, it is clear that having the // [[Rcpp::depends(RcppEigen)]] that is used by sourceCpp() should be enough ... as that is what knitr uses!

Here is a complete Rmd file setting up something simpler: the getEigenValues() from the README of the RcppEigen repo. It works. Maybe you can build up from this.

---
title: "RcppEigen in RMarkdown"
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```

Standard `getEigenValue` example from README


```{Rcpp, eigenValues}
#include <RcppEigen.h>

// [[Rcpp::depends(RcppEigen)]]

using Eigen::Map;                       // 'maps' rather than copies
using Eigen::MatrixXd;                  // variable size matrix, double precision
using Eigen::VectorXd;                  // variable size vector, double precision
using Eigen::SelfAdjointEigenSolver;    // one of the eigenvalue solvers

// [[Rcpp::export]]
VectorXd getEigenValues(Map<MatrixXd> M) {
    SelfAdjointEigenSolver<MatrixXd> es(M);
    return es.eigenvalues();
}
```

From `?eigen`:


```{r}
M <- cbind(c(1,-1), c(-1,1))
eigen(M)   # base R
getEigenValues(M)
```

It does the trick. (Eigen values are invariant to ordering so c(2,0) is equivalent to c(0,2).)