Rcpp: Install, sourceCpp and cppFunction for R and C++

Rcpp is an R package, maintained by Dirk Eddelbuettel and listed on CRAN as “Seamless R and C++ Integration”, that lets you write functions in C++ and call them from R as ordinary R functions. Researchers use it to speed up loops and simulations that are slow in plain R. After this page you should be able to install Rcpp, compile a first C++ function from R, choose between cppFunction(), sourceCpp() and a package, and record the details needed to reproduce your results. The current CRAN release at the time of reading (2026-11-08) is Rcpp 1.1.2, published 2026-07-05; it depends on R 3.5.0 or later.

What Rcpp is used for in research

The Rcpp project site describes the package as a way of extending and growing R applications through an easy-to-use, robust and performant C++ interface, mapping R data types to C++ ones. Typical research uses are numerical code that cannot be vectorised (iterative algorithms, Monte Carlo and MCMC loops, recursive calculations) and wrapping an existing C++ library so analysts can call it from R. The project site states that more than 3000 CRAN packages use Rcpp (its own figure, as read on 2026-11-08). The site also points to the Rcpp Book (Springer, useR! series), the Rcpp Gallery with 100+ examples, the Stack Overflow tag rcpp and the rcpp-devel mailing list for support.

Rcpp does not change what your statistics mean; it changes how a calculation is executed. Keep an R reference implementation of anything you move to C++ and compare outputs before trusting the faster version.

Install or access (official source only)

1. Install Rcpp from CRAN inside R: install.packages("Rcpp"), then load it with library(Rcpp). For general package-installation behaviour see How to Install R Packages (CRAN, Bioconductor, GitHub).
2. Have a C++ compiler available, because Rcpp compiles your code into a shared library. On Windows, CRAN publishes Rtools, the toolchains for building R and R packages from source. The CRAN Rtools page (read 2026-11-08) pairs versions as follows: Rtools 4.5 with R 4.5.0 and later, Rtools 4.4 with R 4.4.x, Rtools 4.3 with R 4.3.x, Rtools 4.2 with R 4.2.x, and Rtools 4.0 with R 4.0.0 to 4.1.3.
3. If you have not yet installed R itself, start with How to Install R and RStudio on Windows, Mac and Linux.

Source code is at github.com/RcppCore/Rcpp, and CRAN carries the reference manual and vignettes (introduction, attributes, extending, FAQ, modules, package, quickref, sugar and others).

Step-by-step: compile your first C++ function from R

Rcpp offers three routes. Pick by the size of the job, using what the sourceCpp reference page says about each.

Route	Use it when	What the documentation says
cppFunction()	One short function defined from an R string	Takes C++ source as a string, compiles it into a shared library and creates an R function that calls it; the cppFunction reference notes that sourceCpp is usually easier to maintain for larger projects.
sourceCpp()	A standalone .cpp file, or inline code via code=	Parses the file, finds functions marked // [[Rcpp::export]], and makes them available in the environment you give it (global by default).
An R package	More than one source file, or external dependencies	The sourceCpp page says to create an R package in that case; the Rcpp::export attribute also works inside packages through compileAttributes.

Steps

1. Install and load Rcpp: install.packages("Rcpp") then library(Rcpp).
2. Write the C++ function in a file such as fibonacci.cpp, with the export attribute above it:

   #include <Rcpp.h>
   
   // [[Rcpp::export]]
   int fibonacci(const int x) {
     if (x == 0) return(0);
     if (x == 1) return(1);
     return (fibonacci(x - 1)) + fibonacci(x - 2);
   }

3. Compile and load it from R: sourceCpp("fibonacci.cpp"). The same code can be passed as a string with sourceCpp(code = '...').
4. Call it like any R function: fibonacci(10).
5. Check it against an R version and time both, for example fib_r <- function(x) if (x < 2) x else fib_r(x - 1) + fib_r(x - 2), then identical(fibonacci(20L), fib_r(20)) and system.time(fib_r(25)) against system.time(fibonacci(25L)).

The documented rules for exported functions: mark them with [[Rcpp::export]], define them in the global namespace, use return types compatible with Rcpp::wrap and parameter types compatible with Rcpp::as. You can also embed R test code in the .cpp file inside a block comment starting /*** R; it runs after compilation, so the compiled functions are available to it.

Output and how to read it

• sourceCpp() returns, invisibly, a list with two elements: functions (names of the exported functions) and modules (names of Rcpp modules). Assign it to a variable if you want to see them.
• After a successful compile, the exported function appears in your global environment (or the env you passed) and returns ordinary R objects.
• Compiled libraries are cached. By default the cache is tempdir(), so it lasts only for the current R session; pass cacheDir to keep it across sessions. rebuild = TRUE forces recompilation, and verbose = TRUE prints details about the generated code.

Troubleshooting

• Function not found after sourceCpp(). Check the // [[Rcpp::export]] comment sits directly above a function defined in the global namespace.
• Compile error about a type. Parameter and return types must be convertible with Rcpp::as and Rcpp::wrap.
• Compiler not found on Windows. Install the Rtools release that matches your R version (table above in the install section).
• Stale behaviour after editing code. Call sourceCpp(..., rebuild = TRUE).
• Multi-file project is awkward. Move to an R package, as the documentation advises.

Reporting Rcpp in research and keeping it reproducible

For computational reproducibility, record the R version, the Rcpp version (1.1.2 as of 2026-07-05), the compiler or Rtools version, and the C++ source file itself alongside your analysis scripts. renv can lock package versions, and How to Cite Software, Code and R Packages explains how to cite Rcpp; citation("Rcpp") is the standard R call for a package’s citation details.

Free alternatives

Tool	Language	What its own documentation says
cpp11	R	A C++11 interface to R’s C API; goals and syntax similar to Rcpp, described as a ground-up rewrite with different design trade-offs; registers functions with [[cpp11::register]]; header-only (cpp11.r-lib.org).
pybind11	Python	A small header-only library that exposes C++ types in Python and Python types in C++, mainly for wrapping existing C++ code (pybind11.readthedocs.io).
Cython	Python	A compiler for Python and its Python-based extension language that produces C extensions able to call C or C++ directly (cython.org).
Numba	Python	An open-source compiler that turns a subset of Python and NumPy code into machine code at runtime via a decorator, without a separate build step or a C/C++ compiler (numba.pydata.org).

For broader tool choice see Free Statistical Software for Research.

FAQ

What is Rcpp?

An R package from CRAN that provides a C++ interface to R, so C++ functions can be compiled and called from R. The current version read on 2026-11-08 is 1.1.2.

How do I install Rcpp?

Run install.packages("Rcpp") in R, then library(Rcpp). You also need a C++ toolchain; on Windows CRAN provides Rtools matched to your R version.

When should I use sourceCpp instead of cppFunction?

Use sourceCpp() for a standalone .cpp file; use an R package once you have several source files or external dependencies.

Sources (official docs)

• CRAN: Rcpp package page, cran.r-project.org/web/packages/Rcpp/index.html (read 2026-11-08)
• Rcpp project site, www.rcpp.org (read 2026-11-08)
• Rcpp reference: sourceCpp and cppFunction, search.r-project.org/CRAN/refmans/Rcpp/html/ (read 2026-11-08)
• CRAN Rtools, cran.r-project.org/bin/windows/Rtools/ (read 2026-11-08)
• cpp11, cpp11.r-lib.org; pybind11, pybind11.readthedocs.io; Cython, cython.org; Numba, numba.pydata.org (read 2026-11-08)