% !Rnw driver = highlight::HighlightWeaveLatex()
\documentclass[11pt]{article}
%\VignetteIndexEntry{Rcpp-attributes}
%\VignetteEngine{highlight::highlight}
%\VignetteKeywords{Rcpp, attributes}
%\VignetteDepends{Rcpp}

\usepackage[USletter]{vmargin}
\setmargrb{1.25in}{1.25in}{1.25in}{1.25in}


\usepackage{textcomp}
\usepackage{color, alltt}
\usepackage[authoryear,round,longnamesfirst]{natbib}
\usepackage[colorlinks]{hyperref}
\definecolor{link}{rgb}{0,0,0.3}	%% next few lines courtesy of RJournal.sty
\hypersetup{
    colorlinks,%
    citecolor=link,%
    filecolor=link,%
    linkcolor=link,%
    urlcolor=link
}

\usepackage{microtype}                     %% cf http://www.khirevich.com/latex/microtype/
\usepackage[T1]{fontenc}		   %% cf http://www.khirevich.com/latex/font/
\usepackage[bitstream-charter]{mathdesign} %% cf http://www.khirevich.com/latex/font/

\newcommand{\proglang}[1]{\textsf{#1}}
\newcommand{\pkg}[1]{{\fontseries{b}\selectfont #1}}
\newcommand{\code}[1]{\texttt{#1}}

%% This corresponds to setting boxes=TRUE for highlight
\newsavebox{\hlbox}
\definecolor{hlBg}{rgb}{0.949019607843137,0.949019607843137,0.949019607843137}
\definecolor{hlBd}{rgb}{0.9,0.9,0.9} % border
\renewenvironment{Hchunk}{\vspace{0.5em}\noindent\begin{lrbox}{\hlbox}\begin{minipage}[b]{.98\linewidth}}%
    {\end{minipage}\end{lrbox}\fcolorbox{hlBd}{hlBg}{\usebox{\hlbox}}\vspace{0.5em}}


<<echo=FALSE,print=FALSE>>=
prettyVersion <- packageDescription("Rcpp")$Version
prettyDate <- format(Sys.Date(), "%B %e, %Y")
@

\author{J.J. Allaire \and Dirk Eddelbuettel \and Romain Fran\c{c}ois}
\title{\pkg{Rcpp} Attributes}
\date{\pkg{Rcpp} version \Sexpr{prettyVersion} as of \Sexpr{prettyDate}}

\begin{document}
\maketitle

\abstract{
  \noindent
  \textsl{Rcpp attributes} provide a high-level syntax for declaring \proglang{C++}
  functions as callable from \proglang{R} and automatically generating the code
  required to invoke them. Attributes are intended to facilitate both interactive use
  of \proglang{C++} within \proglang{R} sessions as well as to support \proglang{R}
  package development. The implementation of attributes  is based on previous
  work in the \pkg{inline} package \citep{CRAN:inline}.
}


\section{Introduction}

Attributes are a new feature of \pkg{Rcpp} version 0.10.0 \citep{CRAN:Rcpp,JSS:Rcpp}
that provide infrastructure for seamless language bindings between \proglang{R} and
\proglang{C++}. The motivation for attributes is several-fold:

\begin{enumerate}
\item
  Reduce the learning curve associated with using C++ and R together
\item
  Eliminate boilerplate conversion and marshaling code wherever
  possible
\item
  Seamless use of C++ within interactive R sessions
\item
  Unified syntax for interactive work and package development
\end{enumerate}

The core concept is to add annotations to \proglang{C++} source
files that provide the context required to automatically generate \proglang{R}
bindings to \proglang{C++} functions. Attributes and their supporting
functions include:

\begin{itemize}
\item
  \texttt{Rcpp::export} attribute to export a \proglang{C++} function
  to \proglang{R}
\item
  \texttt{sourceCpp} function to source exported functions from a file
\item
  \texttt{cppFunction} and \texttt{evalCpp} functions for inline
  declarations and execution
\item
  \texttt{Rcpp::depends} attribute for specifying additional build
  dependencies for \texttt{sourceCpp}
\end{itemize}

Attributes can also be used for package development via the
\texttt{compileAttributes} function, which automatically generates
\texttt{extern "C"} and \texttt{.Call} wrappers for \proglang{C++}
functions within packages.

\section{Using Attributes}

Attributes are annotations that are added to C++ source files to provide
additional information to the compiler. \pkg{Rcpp} supports attributes
to indicate that C++ functions should be made available as R functions,
as well as to optionally specify additional build dependencies for source files.

\proglang{C++11} specifies a standard syntax for attributes
\citep{Maurer+Wong:2008:AttributesInC++}. Since this standard isn't yet
fully supported across all compilers, \pkg{Rcpp} attributes are included in
source files using specially formatted comments.

\subsection{Exporting C++ Functions}

The \texttt{sourceCpp} function parses a \proglang{C++} file and looks for
functions marked with the \texttt{Rcpp::export} attribute. A shared
library is then built and its exported functions are made available as R
functions in the specified environment. For example, this source file
contains an implementation of convolve (note the \texttt{Rcpp::export}
attribute in the comment above the function):

<<lang=cpp>>=
#include <Rcpp.h>
using namespace Rcpp;

// [[Rcpp::export]]
NumericVector convolveCpp(NumericVector a, NumericVector b) {

    int na = a.size(), nb = b.size();
    int nab = na + nb - 1;
    NumericVector xab(nab);

    for (int i = 0; i < na; i++)
        for (int j = 0; j < nb; j++)
            xab[i + j] += a[i] * b[j];

    return xab;
}
@ 

The addition of the export attribute allows us to do this from the \proglang{R}
prompt:

<<eval=FALSE,lang=r>>=
sourceCpp("convolve.cpp")
convolveCpp(x, y)
@ 

We can now write \proglang{C++} functions using built-in \proglang{C++} types
and \pkg{Rcpp} wrapper types and then source them just as we would an
\proglang{R} script.

The \texttt{sourceCpp} function performs caching based on the last
modified date of the source file and it's local dependencies so as 
long as the source does not change the compilation will occur only 
once per R session.

\subsection{Specifying Argument Defaults}

If default argument values are provided in the C++ function definition
then these defaults are also used for the exported R function. For example,
the following C++ function:

<<lang=cpp>>=
DataFrame readData(
    CharacterVector file,
    CharacterVector colNames = CharacterVector::create(),
    std::string comment = "#",
    bool header = true)
@ 

Will be exported to R as:

<<eval=FALSE,lang=r>>=
function(file, colNames=character(), comment="#", header=TRUE)
@ 

Note that C++ rules for default arguments still apply: they must occur
consecutively at the end of the function signature and (unlike R) can't rely
on the values of other arguments.

Not all \proglang{C++} default argument values can be parsed into their
\proglang{R} equivalents, however the most common cases are supported, including:

\begin{itemize}
\item
   String literals delimited by quotes (e.g. \texttt{"foo"})
\item
   Decimal numeric values (e.g. \texttt{10} or \texttt{4.5})
\item
   Pre-defined constants including \texttt{true}, \texttt{false},
   \texttt{R\_NilValue}, \texttt{NA\_STRING}, \texttt{NA\_INTEGER},
   \texttt{NA\_REAL}, and \texttt{NA\_LOGICAL}.
\item
   Selected vector types (\texttt{CharacterVector}, \texttt{IntegerVector},
   and \texttt{NumericVector}) instantiated using the \texttt{::create}
   static member function.
\item
  \texttt{Matrix} types instantiated using the \texttt{rows},
  \texttt{cols} constructor.
\end{itemize}

\subsection{Signaling Errors}

Within \proglang{R} code the \texttt{stop} function is typically used to signal
errors. Within \proglang{R} extensions written in \proglang{C} the \texttt{Rf\_error} function is typically used. However, within \proglang{C++} code you cannot
safely use \texttt{Rf\_error} because it results in a \texttt{longjmp} over
any \proglang{C++} destructors on the stack.

The correct way to signal errors within \proglang{C++} functions is to throw an \\\texttt{Rcpp::exception}. For example:

<<lang=cpp>>=
if (unexpectedCondition)
    throw Rcpp::exception("Unexpected condition occurred");
@ 

There is also an \texttt{Rcpp::stop} function that is shorthand for throwing
an \\\texttt{Rcpp::exception}. For example:

<<lang=cpp>>=
if (unexpectedCondition)
    Rcpp::stop("Unexpected condition occurred");
@ 

In both cases the \proglang{C++} exception will be caught by \pkg{Rcpp}
prior to returning control to \proglang{R} and converted into the correct
signal to \proglang{R} that execution should stop with the specified message.

You can similarly also signal warnings with the \texttt{Rcpp::warning}
function:

<<lang=cpp>>=
if (unexpectedCondition)
    Rcpp::warning("Unexpected condition occurred");
@ 

\subsection{Supporting User Interruption}

If your function may run for an extended period of time, users will appreciate
the ability to interrupt it's processing and return to the REPL. This is
handled automatically for R code (as R checks for user interrupts periodically
during processing) however requires explicit accounting for in C and C++ 
extensions to R. To make computations interrupt-able, you should periodically
call the \texttt{Rcpp::checkUserInterrupt} function, for example:

<<lang=cpp>>=
for (int i=0; i<1000000; i++) {
    
    // check for interrupt every 1000 iterations
    if (i % 1000 == 0)
        Rcpp::checkUserInterrupt();
    
    // ...do some expensive work...
    
}
@ 

A good guideline is to call \texttt{Rcpp::checkUserInterrupt} every 1 or 2
seconds that your computation is running. In the above code, if the user
requests an interrupt then an exception is thrown and the attributes wrapper
code arranges for the user to be returned to the REPL.

Note that R provides a \proglang{C} API for the same purpose 
(\texttt{R\_CheckUserInterrupt}) however this API is not safe to use in 
\proglang{C++} code as it uses \texttt{longjmp} to exit the current scope,
bypassing any C++ destructors on the stack. The \texttt{Rcpp::checkUserInterrupt}
function is provided as a safe alternative for \proglang{C++} code.

\subsection{Embedding R Code}

Typically \proglang{C++} and \proglang{R} code are kept in their own source
files. However, it's often convenient to bundle code from both languages into
a common source file that can be executed using single call to \texttt{sourceCpp}.

To embed chunks of \proglang{R} code within a \proglang{C++}
source file you include the \proglang{R} code within a block comment that
has the prefix of \texttt{/*** R}. For example:

<<lang=cpp>>=
/*** R

# Call the fibonacci function defined in C++
fibonacci(10)

*/
@ 

Multiple \proglang{R} code chunks can be included in a \proglang{C++} file. The
\texttt{sourceCpp} function will first compile the \proglang{C++} code into a
shared library and then source the embedded \proglang{R} code.

\subsection{Modifying Function Names}

You can change the name of an exported function as it appears to \proglang{R} by
adding a name parameter to \texttt{Rcpp::export}. For example:

<<lang=cpp>>=
// [[Rcpp::export(name = ".convolveCpp")]]
NumericVector convolveCpp(NumericVector a, NumericVector b)
@ 

Note that in this case since the specified name is prefaced by a \code{.} the
exported R function will be hidden. You can also use this method to provide
implementations of S3 methods (which wouldn't otherwise be possible because
C++ functions can't contain a '.' in their name).

\subsection{Function Requirements}

Functions marked with the \texttt{Rcpp::export} attribute must meet several
requirements to be correctly handled:

\begin{itemize}
\item
   Be defined in the global namespace (i.e. not within a C++ namespace declaration)
\item
   Have a return type that is either void or compatible with \texttt{Rcpp::wrap}
   and parameter types that are compatible with \texttt{Rcpp::as} (see sections
   3.1 and 3.2 of the `\textsl{Rcpp-introduction}' vignette for more details).
 \item
   Use fully qualified type names for the return value and all parameters.
   Rcpp types may however appear without a namespace qualifier (i.e.
   \texttt{DataFrame} is okay as a type name but \texttt{std::string} must be
   specified fully).
\end{itemize}

\subsection{Random Number Generation}

\proglang{R} functions implemented in \proglang{C} or \proglang{C++} need
to be careful to surround use of internal random number generation routines
(e.g. \texttt{unif\_rand}) with calls to \texttt{GetRNGstate} and
\texttt{PutRNGstate}.

Within \pkg{Rcpp}, this is typically done using the \texttt{RNGScope} class.
However, this is not necessary for \proglang{C++} functions exported using
attributes because an \texttt{RNGScope} is established for them automatically.
Note that \pkg{Rcpp} implements \texttt{RNGScope} using a counter, so it's
still safe to execute code that may establish it's own \texttt{RNGScope} (such
as the \pkg{Rcpp} sugar functions that deal with random number generation).

The overhead associated with using \texttt{RNGScope} is negligible (only a 
couple of milliseconds) and it provides a guarantee that all C++ code
will inter-operate correctly with R's random number generation. If you are 
certain that no C++ code will make use of random number generation and the 
2ms of execution time is meaningful in your context, you can disable the
automatic injection of \texttt{RNGScope} using the \texttt{rng} parameter
of the \texttt{Rcpp::export} attribute. For example:

<<lang=cpp>>=
// [[Rcpp::export(rng = false)]]
double myFunction(double input) {
    // ...code that never uses R random number generation... 
}
@ 

\subsection{Importing Dependencies}

It's also possible to use the \texttt{Rcpp::depends} attribute to declare
dependencies on other packages. For example:

<<lang=cpp>>=
// [[Rcpp::depends(RcppArmadillo)]]

#include <RcppArmadillo.h>
using namespace Rcpp;

// [[Rcpp::export]]
List fastLm(NumericVector yr, NumericMatrix Xr) {

    int n = Xr.nrow(), k = Xr.ncol();

    arma::mat X(Xr.begin(), n, k, false);
    arma::colvec y(yr.begin(), yr.size(), false);

    arma::colvec coef = arma::solve(X, y);
    arma::colvec resid = y - X*coef;

    double sig2 = arma::as_scalar(arma::trans(resid)*resid/(n-k));
    arma::colvec stderrest = arma::sqrt(
          sig2 * arma::diagvec( arma::inv(arma::trans(X)*X)) );

    return List::create(Named("coefficients") = coef,
                        Named("stderr")       = stderrest);
}
@ 

The inclusion of the \texttt{Rcpp::depends} attribute causes \texttt{sourceCpp}
to configure the build environment to correctly compile and link against the
\pkg{RcppArmadillo} package. Source files can declare more than one dependency
either by using multiple \texttt{Rcpp::depends} attributes or with syntax like this:

<<lang=cpp>>=
// [[Rcpp::depends(Matrix, RcppArmadillo)]]
@ 

Dependencies are discovered both by scanning for package include directories
and by invoking \pkg{inline} plugins if they are available for a package.

Note that while the \texttt{Rcpp::depends} attribute establishes dependencies
for \texttt{sourceCpp}, it's important to note that if you include the same
source file in an \proglang{R} package these dependencies must still be
listed in the \texttt{Imports} and/or \texttt{LinkingTo} fields of the package
\texttt{DESCRIPTION} file.

\subsection{Sharing Code}

The core use case for \texttt{sourceCpp} is the compilation of a single
self-contained source file. Code within this file can import other C++ code
by using the \texttt{Rcpp::depends} attribute as described above.

The recommended practice for sharing C++ code across many uses of 
\texttt{sourceCpp} is therefore to create an R package to wrap the C++
code. This has many benefits not the least of which is easy distribution of
shared code. More information on creating packages that contain C++ code
is included in the Package Development section below.

\subsubsection{Shared Code in Header Files}

If you need to share a small amount of C++ code between source files 
compiled with \texttt{sourceCpp} and the option of creating a package
isn't practical, then you can also share code using local includes of C++
header files. To do this, create a header file with the definition of 
shared functions, classes, enums, etc. For example:

<<lang=cpp>>=
#ifndef __UTILITIES__
#define __UTILITIES__

inline double timesTwo(double x) {
    return x * 2;
}

#endif // __UTILITIES__
@ 

Note the use of the \texttt{\#ifndef} include guard, this is import to ensure
that code is not included more than once in a source file. You should 
use an include guard and be sure to pick a unique name for the corresponding
\texttt{\#define}. 

Also note the use of the \code{inline} keyword preceding the function. This
is important to ensure that there are not multiple definitions of
functions included from header files. Classes fully defined in header files
automatically have inline semantics so don't require this treatment.

To use this code in a source file you'd just include
it based on it's relative path (being sure to use \texttt{"} as the 
delimiter to indicate a local file reference). For example:

<<lang=cpp>>=
#include "shared/utilities.hpp"

// [[Rcpp::export]]
double transformValue(double x) {
    return timesTwo(x) * 10;
}
@ 

\subsubsection{Shared Code in C++ Files}

When scanning for locally included header files \code{sourceCpp} also checks
for a corresponding implementation file and automatically includes it in the
compilation if it exists.

This enables you to break the shared code entirely into it's own source file.
In terms of the above example, this would mean having only a function 
declaration in the header:

<<lang=cpp>>=
#ifndef __UTILITIES__
#define __UTILITIES__

double timesTwo(double x);

#endif // __UTILITIES__
@ 

Then actually defining the function in a separate source file with the 
same base name as the header file but with a .cpp extension (in the above
example this would be \code{utilities.cpp}):

<<lang=cpp>>=
#include "utilities.hpp"

double timesTwo(double x) {
    return x * 2;
}
@ 

It's also possible to use attributes to declare dependencies and exported 
functions within shared header and source files. This enables you to take
a source file that is typically used standalone and include it when compiling
another source file. 

Note that since additional source files are processed as separate translation
units the total compilation time will increase proportional to the number of
files processed. From this standpoint it's often preferable to use shared 
header files with definitions fully inlined as demonstrated above.

Note also that embedded R code is only executed for the main source file not
those referenced by local includes.

\subsection{Including C++ Inline}

Maintaining C++ code in it's own source file provides several benefits including
the ability to use \proglang{C++} aware text-editing tools and straightforward
mapping of compilation errors to lines in the source file. However, it's also
possible to do inline declaration and execution of C++ code.

There are several ways to accomplish this, including passing a code
string to \texttt{sourceCpp} or using the shorter-form \texttt{cppFunction}
or \texttt{evalCpp} functions. For example:

<<eval=FALSE,lang=r>>=
cppFunction('
    int fibonacci(const int x) {
        if (x < 2)
            return x;
        else
            return (fibonacci(x - 1)) + fibonacci(x - 2);
    }
')

evalCpp('std::numeric_limits<double>::max()')
@ 

You can also specify a depends parameter to \texttt{cppFunction} or \texttt{evalCpp}:

<<eval=FALSE,lang=r>>=
cppFunction(depends = 'RcppArmadillo', code = '...')
@ 

\section{Package Development}

One of the goals of \pkg{Rcpp} attributes is to simultaneously facilitate
ad-hoc and interactive work with \proglang{C++} while also making it very easy to
migrate that work into an \proglang{R} package. There are several benefits of
moving code from a standalone \proglang{C++} source file to a package:

\begin{enumerate}
\item
  Your code can be made available to users without \proglang{C++} development
  tools (at least on Windows or Mac OS X where binary packages are common)
\item
  Multiple source files and their dependencies are handled automatically
  by the \proglang{R} package build system
\item
  Packages provide additional infrastructure for testing, documentation
  and consistency
\end{enumerate}

\subsection{Package Creation}

To create a package that is based on \pkg{Rcpp} you should follow the
guidelines in the `\textsl{Rcpp-package}' vignette. For a new package this
is most conveniently done using  the \texttt{Rcpp.package.skeleton} function.

To generate a new package with a simple hello, world function that uses
attributes you can do the following:

<<eval=FALSE,lang=r>>=
Rcpp.package.skeleton("NewPackage", attributes = TRUE)
@ 

To generate a package based on \proglang{C++} files that you've been using
with \texttt{sourceCpp} you can use the \texttt{cpp\_files} parameter:

<<eval=FALSE,lang=r>>=
Rcpp.package.skeleton("NewPackage", example_code = FALSE,
                      cpp_files = c("convolve.cpp"))
@ 

\subsection{Specifying Dependencies}

Once you've migrated \proglang{C++} code into a package, the dependencies for
source files are derived from the \texttt{Imports} and \texttt{LinkingTo} fields
in the package \texttt{DESCRIPTION} file rather than the \texttt{Rcpp::depends}
attribute. Some packages also require the addition of an entry to the package
\texttt{NAMESPACE} file to ensure that the package's shared library is loaded
prior to callers using the package. For every package you import C++ code from
(including \pkg{Rcpp}) you need to add these entries.

Packages that provide only C++ header files (and no shared library) need only
be referred to using \texttt{LinkingTo}. You should consult the documentation
for the package you are using for the requirements particular to that package.

For example, if your package depends on \pkg{Rcpp} you'd have the following
entries in the \texttt{DESCRIPTION} file:

<<lang=bash>>=
Imports: Rcpp (>= 0.11.4)
LinkingTo: Rcpp
@ 

And the following entry in your \texttt{NAMESPACE} file:

<<lang=bash>>=
importFrom(Rcpp, evalCpp)
@ 

If your package additionally depended on the \pkg{BH} (Boost headers) package
you'd just add an entry for \pkg{BH} to the \texttt{LinkingTo} field since
\pkg{BH} is a header-only package:

<<lang=bash>>=
Imports: Rcpp (>= 0.11.4)
LinkingTo: Rcpp, BH
@ 


\subsection{Exporting R Functions}

Within interactive sessions you call the \texttt{sourceCpp} function
on individual files to export \proglang{C++} functions into the global
environment. However, for packages you call a single utility function to
export all \proglang{C++} functions within the package.

The \texttt{compileAttributes} function scans the source files within a package
for export attributes and generates code as required. For example, executing
this from within the package working directory:

<<eval=FALSE,lang=r>>=
compileAttributes()
@ 

Results in the generation of the following two source files:

\begin{itemize}
\item
  \texttt{src/RcppExports.cpp} -- The \texttt{extern "C"} wrappers required
  to call exported \proglang{C++} functions within the package.
\item
  \texttt{R/RcppExports.R} -- The \texttt{.Call} wrappers required to call
  the \texttt{extern "C"} functions defined in \texttt{RcppExports.cpp}.
\end{itemize}

You should re-run \texttt{compileAttributes} whenever functions are added,
removed, or have their signatures changed. Note that if you are using either
RStudio or \pkg{devtools} to build your package then the 
\texttt{compileAttributes} function is called automatically whenever your
package is built.

The \texttt{compileAttributes} function deals only with exporting
\proglang{C++} functions to \proglang{R}. If you want the functions to
additionally be publicly available from your package's namespace another
step may be required. Specifically, if your package \texttt{NAMESPACE} file
does not use a pattern to export functions then you should add an explicit
entry to \texttt{NAMESPACE} for each R function you want publicly available.

\subsection{Types in Generated Code}

In some cases the signatures of the C++ functions that are generated within
\texttt{RcppExports.cpp} may have additional type requirements beyond the core 
standard library and \pkg{Rcpp} types (e.g. \texttt{CharacterVector},
\texttt{NumericVector}, etc.). Examples might include convenience typedefs,
as/wrap handlers for marshaling between custom types and SEXP, or types 
wrapped by the Rcpp \texttt{XPtr} template.

In this case, you can create a header file that contains these type definitions
(either defined inline or by including other headers) and have this header
file automatically included in \texttt{RcppExports.cpp}. Headers named with
the convention \texttt{pkgname\_types} are automatically included along with
the generated C++ code. For example, if your package is named \pkg{fastcode}
then any of the following header files would be automatically included in
\texttt{RcppExports.cpp}:

<<lang=cpp>>=
src/fastcode_types.h
src/fastcode_types.hpp
inst/include/fastcode_types.h
inst/include/fastcode_types.hpp
@

There is one other mechanism for type visibility in \texttt{RcppExports.cpp}.
If your package provides a master include file for consumption by C++ clients
then this file will also be automatically included. For example, if the 
\pkg{fastcode} package had a C++ API and the following header file:

<<lang=cpp>>=
inst/include/fastcode.h
@

This header file will also automatically be included in 
\texttt{RcppExports.cpp}. Note that the convention of using \texttt{.h} for
header files containing C++ code may seem unnatural, but this comes from the
recommended practices described in `\textsl{Writing R Extensions}' 
\citep{R:Extensions}.

\subsection{Roxygen Comments}

The \pkg{roxygen2} package \citep{CRAN:roxygen2} provides a facility for
automatically generating \proglang{R} documentation files based on specially
formatted comments in \proglang{R} source code.

If you include roxygen comments in your \proglang{C++} source file with a
\texttt{//\textquotesingle} prefix then \texttt{compileAttributes} will transpose them
into R roxygen comments within \texttt{R/RcppExports.R}. For example the
following code in a \proglang{C++} source file:

<<lang=cpp>>=
//' The length of a string (in characters).
//'
//' @param str input character vector
//' @return characters in each element of the vector
// [[Rcpp::export]]
NumericVector strLength(CharacterVector str)
@ 

Results in the following code in the generated \proglang{R} source file:

<<eval=FALSE,lang=r>>=
#' The length of a string (in characters).
#'
#' @param str input character vector
#' @return characters in each element of the vector
strLength <- function(str)
@ 

\subsection{Providing a C++ Interface}

The interface exposed from \proglang{R} packages is most typically a set of
\proglang{R} functions. However, the \proglang{R} package system also provides
a mechanism to allow the exporting of \proglang{C} and \proglang{C++}
interfaces using package header files.  This is based on the
\texttt{R\_RegisterCCallable} and \texttt{R\_GetCCallable} functions described in
`\textsl{Writing R Extensions}' \citep{R:Extensions}.

\proglang{C++} interfaces to a package are published within the
top level \texttt{include} directory of the package (which within the package
source directory is located at \texttt{inst/include}). The \proglang{R} build
system automatically adds the required \texttt{include} directories for all
packages specified in the \texttt{LinkingTo} field of the package
\texttt{DESCRIPTION} file.

\subsubsection{Interfaces Attribute}

The \texttt{Rcpp::interfaces} attribute can be used to automatically
generate a header-only interface to your \proglang{C++} functions
within the \texttt{include} directory of your package.

The \texttt{Rcpp::interfaces} attribute is specified on a per-source
file basis, and indicates which interfaces (\proglang{R}, \proglang{C++},
or both) should be provided for exported functions within the file.

For example, the following specifies that both R and \proglang{C++} interfaces
should be generated for a source file:

<<lang=cpp>>=
// [[Rcpp::interfaces(r, cpp)]]
@ 

Note that the default behavior if an \texttt{Rcpp::interfaces} attribute
is not included in a source file is to generate an R interface only.

\subsubsection{Generated Code}

If you request a \texttt{cpp} interface for a source file then
\texttt{compileAttributes} generates the following header files
(substituting \emph{Package} with the name of the package code is being
generated for):

<<lang=bash>>=
inst/include/Package.h
inst/include/Package_RcppExports.h
@ 

The \texttt{Package\_RcppExports.h} file has inline definitions for all
exported \proglang{C++} functions that enable calling them using the
\texttt{R\_GetCCallable} mechanism.

The \texttt{Package.h} file does nothing other than include the
\texttt{Package\_RcppExports.h} header. This is done so
that package authors can replace the \texttt{Package.h} header with
a custom one and still be able to include the automatically generated exports
(details on doing this are provided in the next section).

The exported functions are defined within a \proglang{C++} namespace that matches
the name of the package. For example, an exported \proglang{C++} function
\texttt{bar} could be called from package \texttt{MyPackage} as follows:

<<lang=cpp>>=
// [[Rcpp::depends(MyPackage)]]

#include <MyPackage.h>

void foo() {
    MyPackage::bar();
}
@ 

\subsubsection{Including Additional Code}

You might wish to use the \texttt{Rcpp::interfaces} attribute to generate
a part of your package's \proglang{C++} interface but also provide
additional custom \proglang{C++} code. In this case you
should replace the generated \texttt{Package.h} file with one of your own.

Note that the way \pkg{Rcpp} distinguishes user verses generated files is by checking
for the presence a special token in the file (if it's present then it's known
to be generated and thus safe to overwrite). You'll see this token at the top
of the generated \texttt{Package.h} file, be sure to remove it if you want
to provide a custom header.

Once you've established a custom package header file, you need only include the
\texttt{Package\_RcppExports.h} file within your header to make available
the automatically generated code alongside your own.

If you need to include code from your custom header files within the
compilation of your package source files, you will also need to add the
following entry to \texttt{Makevars} and \texttt{Makevars.win} (both are
in the \texttt{src} directory of your package):

<<lang=bash>>=
PKG_CPPFLAGS += -I../inst/include/
@ 

Note that the R package build system does not automatically force a rebuild
when headers in \texttt{inst/include} change, so you should be sure to perform a
full rebuild of the package after making changes to these headers.

\bibliographystyle{plainnat}
\bibliography{Rcpp}

\end{document}