The R Project SVN R

% File src/library/utils/vignettes/Sweave.Rnw
% Part of the R package, https://www.R-project.org
% Copyright 2002-2026 Friedrich Leisch and the R Core Team
% Distributed under GPL 2 or later

\documentclass[a4paper]{article}

%\VignetteIndexEntry{Sweave User Manual}
%\VignettePackage{utils}
%\VignetteDepends{tools, datasets, stats, graphics}

\title{Sweave User Manual}
\author{Friedrich Leisch, Vincent Goulet and R Core Team}

\usepackage[round]{natbib}
\usepackage{graphicx, Rd}
\usepackage{listings}

%% colors inspired by classicthesis.sty
\usepackage{color}
\definecolor{link}{rgb}{0,0.4,0.6}        % liens internes
\definecolor{url}{rgb}{0.6,0,0}           % liens externes
\definecolor{citation}{rgb}{0,0.5,0}      % citations
\definecolor{delim}{rgb}{0.0,0.5,0.5}

%% display noweb markers in color in listings; 'literate' key used to
%% leave references to chunk 'boxp' alone
\lstset{language=[LaTeX]TeX,
  basicstyle=\small\ttfamily,
  commentstyle=\slshape,
  keywordstyle=\mdseries,
  moredelim=*[s][\color{delim}]{<<}{>>=},
  moredelim=[l][\color{delim}]{@},
  literate={<<boxp>>}{<<boxp>>}8,
  frame=trbl,
}

\usepackage{hyperref}
\hypersetup{%
  colorlinks,%
  plainpages=true,%
  linkcolor=link,%
  citecolor=citation,%
  urlcolor=url,%
  % pdfstartview=FitH,% or Fit
  pdfstartview={XYZ null null 1},%
  pdfview={XYZ null null null},%
  pdfpagemode=UseNone,% for no outline
  pdfauthor={Friedrich Leisch and R Core Team},%
  pdftitle={Sweave User Manual},%
  pdfsubject={R vignette documentation system}%
}
\def\subsectionautorefname{Section}

\newcommand{\I}[1]{#1}

\sloppy

\begin{document}

\maketitle

\section{Introduction}
\label{sec:intro}

Sweave provides a flexible framework for mixing text and \R{} code for
automatic document generation. A single source file contains both
documentation text and \R{} code. The file may then either be
\emph{woven} into a final document containing the documentation text
together with the \R{} code and its output (text, graphs), or
\emph{tangled} to extract the \R{} code to a script.

This literate programming system \citep{fla:Knuth:1984} allows the
re-generation of a report if the input data change while also
documenting the code to reproduce the analysis in the same file that
contains the report. The \R{} code of the complete analysis is
embedded into a \LaTeX{} document\footnote{\url{https://www.ctan.org}}
using the \code{noweb} syntax \citep{flm:Ramsey:1998}. Hence, the full
power of \LaTeX{} (for high-quality typesetting) and \R{} (for data
analysis) can be used simultaneously. See
\cite{e1071-papers:Leisch:2002} and references therein for more
general thoughts on dynamic report generation and pointers to other
systems.

Sweave uses a modular concept using different drivers for the actual
translations. Obviously different drivers are needed for different
text markup languages (\LaTeX{}, HTML, \ldots). Several packages on
CRAN provide support for other word processing systems (see
\autoref{sec:faq}).


\section{\I{Noweb} files}
\label{sec:noweb}

\code{noweb} \citep{flm:Ramsey:1998} is a simple
literate-programming tool which allows combining program source code
and the corresponding documentation into a single file.  A
\code{noweb} file is a simple text file which consists of a sequence
of code and documentation segments, called \emph{chunks}:
\begin{description}
\item[Documentation chunks] start with a line that has an at sign
  (\verb|@|) as first character, followed by a space or newline
  character. The rest of this line is a comment and ignored.
  Typically documentation chunks will contain text in a markup
  language like \LaTeX{}.  The chunk continues until a new code or
  documentation chunk is started: note that Sweave does not interpret
  the contents of documentation chunks, and so will identify chunk
  indicators even inside \LaTeX{} verbatim environments.

\item[Code chunks] start with \verb|<<|\code{\var{options}}\verb|>>=| at the beginning of
  a line; again the rest of the line is a comment and ignored.
\end{description}
The default for the first chunk is documentation.

In the simplest usage of \command{noweb}, the options (if present) of
the code chunks give the names of source code files, and the tool
\command{notangle} can be used to extract the code chunks from the
\code{noweb} file. Multiple code chunks can have the same name, the
corresponding code chunks are then concatenated when the source code is
extracted. \command{noweb} has some additional mechanisms to
cross-reference code chunks (the \verb|[[...]]| operator, etc.),
Sweave does currently not use nor support these features, so they are
not described here.


\section{Sweave files}
\label{sec:sweavefile}

Sweave source files are \code{noweb} files with some additional
syntax that allows some additional control over the final
output. Traditional \code{noweb} files have the extension
\file{.nw}, which is also fine for Sweave files (and fully supported
by the software). Additionally, Sweave currently recognizes files with
extensions \file{.rnw}, \file{.Rnw}, \file{.snw} and
\file{.Snw} to directly indicate a \code{noweb} file with Sweave
extensions. We will use \file{.Rnw} throughout this document.

\subsection{A simple example}

A minimal Sweave file is shown in \autoref{fig:ex1.Rnw}. The file
contains three code chunks embedded in a simple \LaTeX{}
document. Running the following expressions creates the \LaTeX{}
document shown in \autoref{fig:ex1.tex}.
<<>>=
rnwfile <- system.file("Sweave", "example-1.Rnw", package = "utils")
Sweave(rnwfile)
@
<<echo=FALSE>>=
rnwfileb <- basename(rnwfile)
texfileb <- sub("Rnw$", "tex", rnwfileb)
pdffileb <- sub("Rnw$", "pdf", rnwfileb)
plotfileb <- sub("\\.tex$", "-003.pdf", texfileb)
@
Compiling this \LaTeX{} document results in the PDF in
\autoref{fig:ex1.pdf}. The latter can also be created directly from
within \R{} using
<<>>=
tools::texi2pdf("example-1.tex")
@
\begin{figure}[htbp]
  \centering
  \begin{minipage}{0.9\textwidth}
    \lstinputlisting{\Sexpr{rnwfile}}
  \end{minipage}
  \caption{A minimal Sweave file: \file{\Sexpr{rnwfileb}}.}
  \label{fig:ex1.Rnw}
\end{figure}

\begin{figure}[htbp]
  \centering
  \begin{minipage}{0.9\textwidth}
    \lstinputlisting{\Sexpr{texfileb}}
  \end{minipage}
  \caption{Running \code{Sweave("\Sexpr{rnwfileb}")} produces the file
    \file{\Sexpr{texfileb}}.}
  \label{fig:ex1.tex}
\end{figure}

\begin{figure}[htbp]
  \centering
  \fbox{\begin{minipage}{0.8\textwidth}
    \includegraphics[width=\textwidth]{\Sexpr{pdffileb}}
  \end{minipage}}
  \caption{The final document \file{\Sexpr{pdffileb}} created by
    compiling \file{\Sexpr{texfileb}}.}
  \label{fig:ex1.pdf}
\end{figure}

The first difference between \file{\Sexpr{rnwfileb}} and
\file{\Sexpr{texfileb}} is that the \LaTeX{} style file
\file{Sweave.sty} is automatically loaded, which provides environments
for typesetting \R{} input and output (the \LaTeX{} environments
\code{Sinput} and \code{Soutput}). Otherwise, the documentation chunks
are copied without any modification from
\file{\Sexpr{rnwfileb}} to \file{\Sexpr{texfileb}}.

The real work of Sweave is done on the code chunks: The first code
chunk has no name, hence the default behavior of Sweave is used, which
transfers both the \R{} commands and their respective output to the
\LaTeX{} file, embedded in \code{Sinput} and \code{Soutput}
environments, respectively.

The second and third code chunks show one of the Sweave extensions to the
\code{noweb} syntax: code chunk names can be used to pass options to
Sweave which control the final output.
\begin{itemize}
\item The option \code{eval=FALSE} for the second chunk indicates that
  the code should not be evaluated (yet). The code is reused in the third
  chunk (details follow in \autoref{sec:chunkrefs}).
\item The third chunk is marked as a figure chunk (\code{fig=TRUE})
  such that Sweave creates (by default) a PDF file of the plot created
  by the commands in the chunk. Furthermore, the statement
  \verb|\includegraphics{\Sexpr{sub(".pdf", "", plotfileb, fixed=TRUE)}}|
  is inserted into the \LaTeX{} file (details on the choice of file
  names for figures follow later in this manual).
\item Option \code{echo=FALSE} indicates that the \R{} input should not
  be included in the final document (no \code{Sinput} environment).
\end{itemize}

\subsection{Sweave options}

Options control how code chunks and their output (text, figures) are
transferred from the \file{.Rnw} file to the \file{.tex} file. All
options have the form \code{\var{key}=\var{value}}, where \code{\var{value}} can be
a number, string or logical value.  Several options can be specified at
once (separated by commas), all options must take a value (which must
not contain a comma or equal sign). Logical options can take the
values \code{TRUE}, \code{FALSE}, \code{T}, \code{F} as well
as lower-case and capitalized versions of the first two.

In the \file{.Rnw} file options can be specified either
\begin{enumerate}
 \item inside the angle brackets at the beginning of a code chunk,
  modifying the behaviour \emph{only for this chunk}, or
 \item anywhere in a documentation chunk using the command %
  \begin{quote}
    \verb|\SweaveOpts{|\code{\var{opt1}=\var{value1}, \var{opt2}=\var{value2}, ..., \var{optN}=\var{valueN}}\verb|}|
  \end{quote}
  which modifies the defaults for the rest of the document, i.e.,
  \emph{all code chunks after the statement}. Hence, an
  \verb|\SweaveOpts| statement in the preamble of the document sets
  defaults for all code chunks.
\end{enumerate}
Further, global options can be specified (as a comma-separated list of
\code{\var{key}=\var{value}} items) in the call to the function
\code{Sweave}, via the \option{--options=} flag of \command{R CMD
  Sweave}, and in the environment variable \env{SWEAVE\_OPTIONS}.

Which options are supported depends on the driver in use. All drivers
should at least support the following options (all options appear
together with their default value, if any):
\begin{description}
  \item[\code{split=FALSE}] a logical value. If \code{TRUE}, then the output is
   distributed over several files, if \code{FALSE} all output is
   written to a single file.  Details depend on the driver.

 \item[\code{label}] a text label for the code chunk, which is used
   for file-name creation when \code{split=TRUE}. It is also used in
   the comments added above the chunk when the file is processed by
   \command{Rtangle} (provided the \code{annotate} option is true, as
   it is by default), and as part of the file names for files
   generated by figure chunks. Because labels can be part of file
   names, they should contain only alphanumeric characters and
   \code{\#+-\_}. (Including the period \code{.} can cause
   confusion with file extensions.)
\end{description}

The first (and only the first) option in a code chunk name can be
optionally without a name, then it is taken to be a label. Therefore,
starting a code chunk with
\begin{quote}
  \verb|<<hello, split=FALSE>>|
\end{quote}
is the same as
\begin{quote}
  \verb|<<split=FALSE, label=hello>>|
\end{quote}
but
\begin{quote}
  \verb|<<split=FALSE, hello>>|
\end{quote}
gives a syntax error. Having an unnamed first argument for labels is
needed for \code{noweb} compatibility. If only %
\code{{\textbackslash}SweaveOpts} % <- protect from Sweave catching \SweaveOpts
is used for setting options, then Sweave files can be written to be
fully compatible with \code{noweb} (as only file names appear in code
chunk names).

Note that \code{split=TRUE} should \emph{not} be used in package
vignettes.

The help pages for \code{RweaveLatex} and \code{Rtangle} list the
options supported by the default \code{Sweave} and \code{Stangle} drivers.

Now for the promised details on the file names corresponding to
figures. These are of the form \code{\var{prefix}-\var{label}.\var{ext}}.
\begin{itemize}
\item \code{\var{prefix}} can be set by the option
  \code{prefix.string}, otherwise is the basename of the output file
  (which unless specified otherwise is the basename of the input
  file).
\item \code{\var{label}} is the label of the code chunk if it has
  one, otherwise the number of the code chunk in the range
  \texttt{001} to \texttt{999}.
\item \code{\var{ext}} is the appropriate extension for the type of
  graphics, e.g.{} \code{pdf}, \code{eps}, \ldots{}.
\end{itemize}
So for the \file{\Sexpr{rnwfileb}} file, the PDF version of the
boxplot is \file{\Sexpr{plotfileb}}, since it is the third code chunk
in the document. Note that if \code{prefix.string} is used it can
specify a directory as part of the prefix. For example, using
\verb|\SweaveOpts{prefix.string=figures/fig}| will result in the
auto-generated figures to be placed in the subdirectory \verb|figures|
(the Sweave document should arrange to create this directory before
use).

\subsection{Using multiple input files}

\LaTeX{} files can include others via \verb|\input{}| commands.  These
can also be used in Sweave files, but the included files will be
included by \LaTeX{} and are not processed by Sweave.  The equivalent
if you want the included files to be processed is the
\verb|\SweaveInput{}| command.

Included files should use the same Sweave syntax (see below) and
encoding as the main file.

\subsection{Using scalars in text}

There is limited support for using the values of \R{} objects in text
chunks. Any occurrence of
\verb|\Sexpr|\verb|{|\code{\var{expr}}\verb|}| is replaced by the
string resulting from coercing the value of the expression \code{expr}
to a character vector; only the first element of which being used.
For example, \verb|\Sexpr{sqrt(9)}| will be replaced by the string
\code{3} in the \LaTeX{} document.

The expression in \verb|\Sexpr| is evaluated in the same environment
as the code chunks. Therefore, one can access all objects defined in the
code chunks which have appeared before the expression and were
evaluated. The expression may contain any valid \R{} code, only braces
(\verb|{ }|) are not allowed. (This is not really a limitation because
more complicated computations can be easily done in a hidden code
chunk and the result then be used inside \verb|\Sexpr|.)

\subsection{Code chunk reuse}
\label{sec:chunkrefs}

Named code chunks can be reused in other code chunks following later
in the document. Consider the simple example
\begin{quote}% NB: this is indented to avoid interpretation by Sweave.
\begin{verbatim}
 <<a>>=
 x <- 10
 @

 <<b>>=
 x + y
 @

 <<c>>=
 <<a>>
 y <- 20
 <<b>>
 @
\end{verbatim}
\end{quote}
which is equivalent to defining the last code chunk as
\begin{quote}
\begin{verbatim}
 <<c>>=
 x <- 10
 y <- 20
 x + y
 @
\end{verbatim}
\end{quote}

The chunk reference operator \verb|<<>>| takes only the name of the
chunk as its argument: no additional Sweave options are allowed.
It has to occur at the beginning of a line.

References to unknown chunk names are omitted, with a warning.

\subsection{Syntax definition}

So far we have only talked about Sweave files using \code{noweb} syntax
(which is the default). However, Sweave allows the user to redefine
the syntax marking documentation and code chunks, using scalars in
text or reuse code chunks.

\autoref{fig:ex1.Stex} shows the example from \autoref{fig:ex1.Rnw}
using the \code{SweaveSyntaxLatex} definition. It can be created using
<<results=hide>>=
SweaveSyntConv(rnwfile, SweaveSyntaxLatex)
@

\begin{figure}[htbp]
  \centering
  \begin{minipage}{0.9\textwidth}
    \lstinputlisting{\Sexpr{sub("Rnw$", "Stex", rnwfileb)}}
  \end{minipage}
  \caption{An Sweave file using \LaTeX{} syntax:
    \file{\Sexpr{sub("Rnw$", "Stex", rnwfileb)}}.}
  \label{fig:ex1.Stex}
\end{figure}

Code chunks are now enclosed in \code{Scode}
environments, code chunk reuse is performed using
\verb|\Scoderef{|\code{\var{chunkname}}\verb|}|. All other operators are the same as in
the \code{noweb}-style syntax.

Which syntax is used for a document is determined by the extension of
the input file, files with extension\footnote{%
  The lowercase versions are also supported for convenience on
  case-insensitive file systems.} %
\file{.Rtex} or \file{.Stex} are assumed to follow the \LaTeX-style
syntax. Alternatively the syntax can be changed at any point within
the document using the commands
\begin{quote}
  \verb|\SweaveSyntax{SweaveSyntaxLatex}|
\end{quote}
or
\begin{quote}
  \verb|\SweaveSyntax{SweaveSyntaxNoweb}|
\end{quote}
at the beginning of a line within a documentation chunk. Syntax
definitions are simply lists of regular expression for several Sweave
commands: see the two default definitions mentioned above for
examples.

\subsection{Encoding}

\LaTeX{} documents written in most languages other than English
usually require a special input encoding to support accented
characters and various symbols beyond those of ASCII. The modern way
to deal with this situation is to use the UTF-8 input encoding and
compile with Lua\LaTeX{}.

As from \R{} 3.1.0, the UTF-8 encoding is handled preferentially to
other encodings. Sweave will leave UTF-8 code in that encoding, and
output in that encoding as well. However, it remains required to tell
Sweave that a file uses this encoding. This can be done through
either the argument \code{encoding} of the function \code{Sweave} or
the flag \option{--encoding} of \command{R CMD Sweave}, or by
inserting the comment
\begin{quote}
  \verb|%\SweaveUTF8|
\end{quote}
in the file to process.

If you are still using pdf\LaTeX{} with non-ASCII documents, you
should be using the \LaTeX{} package \pkg{inputenc}, with a statement
such as
\begin{quote}
  \verb|\usepackage[utf8]{inputenc}|
\end{quote}
in the document's preamble. There is a wide range of input encodings
which are supported, at least partially, and the more recent \LaTeX{}
package \pkg{inputenx} supports more. Other encodings commonly
encountered in Sweave documents are \code{latin1}, \code{latin9} and
\code{ansinew} (the CP1252 codepage on Windows). Greater
coverage\footnote{Including Eastern European, Greek and Hebrew
  letters.} of UTF-8 can be obtained by
\begin{quote}
  \verb|\usepackage[utf8]{inputenx}| \\
  \verb|\input{ix-utf8enc.dfu}|
\end{quote}

The previous paragraphs covered the documentation sections, but what of
the code sections where \LaTeX{} escapes cannot be used?  The first
piece of advice is to use only ASCII code sections as anything else
will reduce the portability of your document. But \R{} code can be
included if in the input encoding declared in the preamble.

The next complication is inclusions: the final vignette may well
include \R{} output, \LaTeX{} or other Sweave files \emph{via}
\verb|\input{}| and \verb|\SweaveInput{}| lines, a bibliography and
figures. It is the user's responsibility that the text inclusions are
covered by the declared input encoding.  \LaTeX{} allows the input
encoding to be changed by
\begin{quote}
  \verb|\inputencoding{|\code{\var{something}}\verb|}|
\end{quote}
statements: these may not work well in Sweave processing. Since
\LaTeX{} package loading is typically not legal \LaTeX{} code in an
included file, it is easiest to declare the encoding to Sweave using
the \verb|%\SweaveUTF8| comment.

It is all too easy for B\textsc{ib}\TeX{} to pick up UTF-8-encoded
references for a Latin-1 document, or \emph{vice versa}.

R output is again to a large extent under the user's control.  If a
Latin-1 Sweave document is processed by \R{} running a Latin-1 locale or
a UTF-8 document is processed in a UTF-8 locale, the only problems
which are likely to arise are from handling data in another encoding,
but it may be necessary to declare the document's encoding to cover
the \R{} output which is generated even if the document is itself ASCII.
One common problem is the quotes produced by \code{sQuote()} and
\code{dQuote()}: these will be in UTF-8 when \R{} is run in a UTF-8
locale, and will be in CP1252 when Sweave is run from \command{Rgui.exe}
on Windows.  Two possible solutions are to suppress fancy quotes by
\code{options(useFancyQuotes=FALSE)} or to force UTF-8 by
\code{options(useFancyQuotes="UTF-8")}.

The encoding of figures is not usually an issue as they are either
bitmaps or include encoding information: it may be necessary to use
the \code{pdf.encoding} Sweave option to set the \code{pdf()}
device up appropriately.


\section{Weaving and tangling}

The user front-ends of the Sweave system are the two \R{} functions
\code{Sweave} and \code{Stangle}, both contained in
package~\pkg{utils}. \code{Sweave} runs the code chunks through \R{}
and replaces them with any combination of their input and output.
\code{Stangle} extracts only the code chunks from an \file{.Rnw} file
and writes them out to one or several files. \code{Stangle} is actually
just a wrapper function for \code{Sweave}, which uses a tangling
instead of a weaving driver by default. See
<<eval=FALSE>>=
help("Sweave")
@
for more details and arguments of the functions.

\subsection{The \texorpdfstring{\code{RweaveLatex}}{RweaveLatex} driver}

This driver transforms \file{.Rnw} files with \LaTeX{} documentation
chunks and \R{} code chunks to proper \LaTeX{} files (for typesetting both
with standard \command{latex} or \command{pdflatex}), see
<<eval=FALSE>>=
help("RweaveLatex")
@
for details, including the supported options.

\subsubsection{Using objects as chunk options}

Starting with \R{} 4.6.0, the \code{RweaveLatex} driver allows the
\var{value} of logical and numerical chunk options (only) to be the
name of an object defined in earlier, evaluated code chunks. This is
useful to pass computed values to options. The Sweave file in
\autoref{fig:ex2.Rnw} provides two examples of use. Processing the
file with
<<results=hide>>=
rnwfile <- system.file("Sweave", "example-2.Rnw", package = "utils")
Sweave(rnwfile)
tools::texi2pdf("example-2.tex")
@
<<echo=FALSE>>=
rnwfileb <- basename(rnwfile)
pdffileb <- sub("Rnw$", "pdf", rnwfileb)
@
creates the PDF in \autoref{fig:ex2.pdf}. Note that the second code
chunk was not evaluated, and how the proportions of the plot differ
from the otherwise same one in \autoref{fig:ex1.pdf}.

\begin{figure}[htbp]
  \centering
  \begin{minipage}{0.9\textwidth}
    \lstinputlisting{\Sexpr{rnwfile}}
  \end{minipage}
  \caption{Sweave file \file{\Sexpr{rnwfileb}} with options taking
    their values from objects in the second and third code chunks.}
  \label{fig:ex2.Rnw}
\end{figure}

\begin{figure}[htbp]
  \centering
  \fbox{\begin{minipage}{0.8\textwidth}
    \includegraphics[width=\textwidth]{\Sexpr{pdffileb}}
  \end{minipage}}
  \caption{The final document \file{\Sexpr{pdffileb}} created from
    \file{\Sexpr{rnwfileb}}.}
  \label{fig:ex2.pdf}
\end{figure}

\subsubsection{Unparsable code in code chunks}

In traditional Sweave usage, the expressions in code chunks are always
parsed on weaving, whether option \code{eval} is \code{TRUE} or
\code{FALSE}. Furthermore, code chunks with \code{eval=FALSE} are
automatically and irrevocably commented out on tangling. These
conventions may prove limiting in uses of literate programming beyond
pure statistical analysis. For example, they make impossible to
include in a code chunk, say for didactic purposes, invalid R code
that should appear uncommented in the tangled script.

To ease with this sort of scenario, \R{} 4.6.0 introduced options for
the standard drivers to completely ignore code chunks on weaving,
tangling, or both. With the \code{RweaveLatex} driver, chunks using
the option \code{ignore.on.weave=TRUE} are not parsed on weaving, but
remain written out verbatim on tangling. A shorter alias \code{weave}
for \code{!ignore.on.weave} is also supported. An option \code{ignore}
also sets at once \code{ignore.on.weave} and the equivalent option for
\code{Rtangle}.

The Sweave file in \autoref{fig:ex3.Rnw} provides examples of use of
the option \code{ignore.on.weave} (and of \code{Rtangle} options
that are covered later in \autoref{sec:weaving:rtangle}).
<<results=hide>>=
rnwfile <- system.file("Sweave", "example-3.Rnw", package = "utils")
Sweave(rnwfile)
tools::texi2pdf("example-3.tex")
@
<<echo=FALSE>>=
rnwfileb <- basename(rnwfile)
pdffileb <- sub("Rnw$", "pdf", rnwfileb)
@
Note the ``missing'' code chunks in the PDF output shown in
\autoref{fig:ex3.pdf}.

\begin{figure}[htbp]
  \centering
  \begin{minipage}{0.9\textwidth}
    \lstinputlisting{\Sexpr{rnwfile}}
  \end{minipage}
  \caption{Sweave file \file{\Sexpr{rnwfileb}} with finer control of
    weaving and tangling.}
  \label{fig:ex3.Rnw}
\end{figure}

\begin{figure}[htbp]
  \centering
  \fbox{\begin{minipage}{0.8\textwidth}
    \includegraphics[width=\textwidth]{\Sexpr{pdffileb}}
  \end{minipage}}
  \caption{The final document \file{\Sexpr{pdffileb}} created from
    \file{\Sexpr{rnwfileb}}.}
  \label{fig:ex3.pdf}
\end{figure}

\subsubsection{Writing to separate files}

If \code{split} is set to \code{TRUE}, then all text corresponding
to code chunks (the \code{Sinput} and \code{Soutput} environments)
is written to separate files. The file names are of form
\file{prefix.string-label.tex}. If several code chunks have the same
label, their outputs are concatenated. If a code chunk has no label,
then the number of the chunk is used instead. The same naming scheme
applies to figures.  You do need to ensure that the file names
generated are valid and not so long that there are not regarded as the
same by your OS (some file systems only recognize 13 characters).

\subsubsection{\LaTeX{} style file}

The driver automatically inserts the statement
\verb|\usepackage{Sweave}| as the last line before
\verb|\begin{document}| in the final \LaTeX{} file if the statement
is not found in the Sweave source file. The style file
\file{Sweave.sty} defines the environments \code{Sinput} and
\code{Soutput} for typesetting code chunks. If you do not want to
include the standard style file, perhaps because you have your own
definitions for \code{Sinput} and \code{Soutput} environments in a
different place, simply insert a comment like
\begin{quote}
  \verb|% \usepackage{Sweave}|
\end{quote}
in the \LaTeX{} preamble of to file to prevent automatic insertion of
the line.

\subsubsection{Figure sizes}

The style file \file{Sweave.sty} sets the default \LaTeX{} figure
width (which is independent of the size of the generated EPS or PDF
files). The current default is 80\% of the current text width as set
with:
\begin{quote}
  \verb|\setkeys{Gin}{width=0.8\textwidth}|
\end{quote}
If you want to use another width for the figures that are
automatically generated and included by Sweave, simply add a line
similar to the one above \emph{after} \verb|\begin{document}|. If you
want no default width for figures, load the style manually with the
option \code{nogin} by inserting the statement
\begin{quote}
  \verb|\usepackage[nogin]{Sweave}|
\end{quote}
in the preamble.

Note that a new graphics device is opened for each figure chunk (one
with option \code{fig=TRUE}), hence all graphical parameters of the
\code{par()} command must be set in each single figure chunk and are
forgotten after the respective chunk (because the device is closed
when leaving the chunk).

One thing that gets easily confused are the width and height
parameters of the \R{} graphics devices and the corresponding
arguments to the \LaTeX{} command \verb|\includegraphics|. The Sweave
options \code{width} and \code{height} are passed to the \R{} graphics
devices, and hence affect the default size of the produced EPS and PDF
files. They do not affect the size of figures in the document, which
is the default mentioned above. If you want to modify the size of
included figures, use \verb|\setkeys{Gin}| or explicit
\verb|\includegraphics| commands in combination with the Sweave option
\code{include=FALSE}.

\subsubsection{Prompts and text width}

By default the driver gets the prompts used for input lines
and continuation lines from \R{}'s \code{options()} settings. To set new
prompts use something like
<<eval=FALSE>>=
options(prompt = "MyR> ", continue = "...")
@
See \code{help(options)} for details. Similarly the width of the
output text is controlled by the \R{} option \code{"width"}.

\subsubsection{Graphics devices}

The default graphics device for figure chunks is
\code{pdf()}: the standard options \code{pdf}, \code{eps},
\code{png} and \code{jpg} allow one or more of these to be
selected for a particular chunk or (via \verb|\SweaveOpts|) for the
whole document.  It can be convenient to select PNG for a particular
figure chunk by something like
\begin{verbatim}
<<figN, fig=TRUE, pdf=FALSE, png=TRUE>>
\end{verbatim}
pdf\LaTeX{} and Lua\LaTeX{} should then automatically include the
\abbr{PNG} file for that chunk.

You can define your own graphics devices and select one of them by the
option \code{grdevice}. The value of the option is a function device
defined in a hidden code chunk. For example, we could make use of the
\code{cairo\_pdf} device by using \code{grdevice=my.Swd} with
\code{my.Swd} defined as:
\begin{verbatim}
my.Swd <- function(name, width, height, ...)
    grDevices::cairo_pdf(filename = paste(name, "pdf", sep = "."),
                         width = width, height = height)
\end{verbatim}

Specialized custom graphics devices may need a customized way to shut
them down in place of \verb|graphics.off()|: this can be supplied
\emph{via} a companion function \verb|my.Swd.off|, which is called
with no arguments.

For a complete example see the file
\file{src/library/utils/tests/customgraphics.Rnw} in the \R{} sources.

\subsection{The \texorpdfstring{\code{Rtangle}}{Rtangle} driver}
\label{sec:weaving:rtangle}

The driver \code{Rtangle} can be used to extract \R{} code chunks from
a \file{.Rnw} file. Code chunks can either be written to one large
file, or to separate files (one for each code chunk). The options
\code{split}, \code{prefix}, and \code{prefix.string} have the same
defaults and interpretation as for the \code{RweaveLatex} driver. See
<<eval=FALSE>>=
help("Rtangle")
@
for further details.

Note that when using \code{split=TRUE} with labeled code chunks, the
resulting script files sorted alphabetically will most likely not
preserve the logical order of the code.

As of \R{} 4.6.0, a number of options give users finer control on the
tangling process. First, the option \code{chunk.sep} allows to specify
the separator between code chunks when they are collected in a single
file (two blank lines by default). One may also use
\code{chunk.sep=FALSE} to completely omit the separator (between two
chunks when used as a chunk option).

Second, \code{Rtangle} also gains an option \code{ignore.on.tangle} to
completely omit a code chunk on tangling, irrespective of the value of
option \code{eval} (the option \code{drop.evalFALSE=TRUE} only omits
unevaluated chunks). The driver also supports the shorter alias
\code{tangle} for \code{!ignore.on.weave}, and the option
\code{ignore} to set both \code{ignore.on.weave} and
\code{ignore.on.tangle} at once.

Finally, \code{Rtangle} now provides an option \code{extension} to
specify the extension, without the leading dot, for the file name of a
tangled code chunk when splitting is selected. (Using an extension
other than \code{.R} when \code{split=FALSE} does not appear as a
feature worth supporting.) If used as \code{extension=TRUE}, the
default extension (usually \code{R}) is used. If the option is
\code{FALSE}, no extension is added to the file name.

The option \code{extension} is specially useful with
\code{ignore.on.weave=TRUE} of \code{RweaveLatex} to include code or
text that the engine would not be able to parse on weaving, yet that
needs to be tangled to a file with a specific extension, such as
\code{.sh} or \code{.txt}.

The Sweave file in \autoref{fig:ex3.Rnw} takes advantage of the
\code{Rtangle} options above. Tangling the file with
<<>>=
Stangle(rnwfile, split = TRUE, annotate = FALSE, chunk.sep = "\n")
@
<<echo=FALSE>>=
rfile <- sub("\\.Rnw$", "-main.R", rnwfileb)
@
creates the files in \autoref{fig:ex3.R}.

\begin{figure}[htbp]
  \centering
  \begin{minipage}{0.9\textwidth}
    \lstinputlisting[title=\file{\Sexpr{rfile}}]{\Sexpr{rfile}}
  \end{minipage}
  \begin{minipage}{0.9\textwidth}
    \lstinputlisting[title=\file{hello.sh}]{hello.sh}
  \end{minipage}
  \begin{minipage}{0.9\textwidth}
    \lstinputlisting[title=\file{README}]{README}
  \end{minipage}
  \caption{Files created by tangling \file{\Sexpr{rnwfileb}}.}
  \label{fig:ex3.R}
\end{figure}


\section{Obtaining the name of the processed file}

So far, we presented weaving and tangling as separate, somewhat
orthogonal procedures. However, one may devise various schemes where
both weaving \emph{and} tangling are needed to create a document. Here
are a few examples:
\begin{itemize}
\item maintaining documentation and code together even though the code
  is not directly used to create the text;
\item code that needs to be included verbatim in a document, but not
  necessarily as or where it appears in the sources;
\item pieces of code that rely on other pieces being saved as files.
\end{itemize}
This manual is an example of the third item: the \file{.tex} and PDF
files in the figures need to be created from the \file{.Rnw} files
before they can be included in this document.

A nice trick to achieve this consists of embedding a tangling
procedure \emph{inside} a weaving procedure by calling \code{Stangle}
in a code chunk. Now, this requires the name of the file to
process\ldots{} which is the file being processed. To avoid hard
coding the file name in the \code{Stangle} call, package~\pkg{utils}
provides, since \R{} 4.6.0, a function \code{SweaveGetSourceName} that
returns the name of the file being woven by an \code{Sweave} process
launched from the command line. This allows to start \texttt{Stangle}
inside \texttt{Sweave} with something like:
\begin{quote}% NB: this is indented to avoid interpretation by Sweave.
\begin{verbatim}
 <<echo=FALSE, results=hide>>=
 FILE <- SweaveGetSourceName()
 Stangle(FILE)
 @
\end{verbatim}
\end{quote}

The function supports weaving processes started by means of either
\command{R~CMD~Sweave}, \command{R~-e}, or \command{Rscript~-e} (see
\autoref{sec:faq}).


\section{Adding Sweave Drivers}

Adding drivers is relatively simple by modelling them on the existing
\code{RweaveLatex} and \code{Rtangle} drivers.\footnote{%
  If you copy the standard drivers, \textbf{do} be careful not to
  infringe R-core's copyright: the copyright notices in the \R{}
  sources \textbf{must} also be copied.}

An Sweave driver is a function of no arguments which should return a
named list of five functions:
\begin{description}
\item{\code{setup(file, syntax, \dots)}} Set up the driver, e.g.{}
  open the output file. Its return value is an object which is passed
  to the next three functions, and can be updated by the next two. The
  value should be a list, and contain a component \code{options}, a
  named list of the default settings for the options needed by
  \code{runcode}.
\item{\code{runcode(object, chunk, options)}} Process a code chunk.
  Returns a possibly updated \code{object}. Argument \code{chunk} is a
  character vector, and \code{options} is a options list for this
  chunk.
\item{\code{writedoc(object, chunk)}} Write out a documentation
    chunk.  Returns a possibly updated \code{object}.
\item{\code{finish(object, error)}} Finish up, or clean up if
    \code{error} is true.
\item{\code{checkopts(options)}} Convert or validate \code{options}
    given as a named list of character strings.
\end{description}

Note that the \code{setup} function should have a \code{\dots}
argument.  It will be passed additional arguments from a
\code{Sweave()} call, but in future \code{Sweave} may itself set
options and pass them on the setup function: one such may be the
encoding used for the text to be processed.

\bibliographystyle{jss}
\bibliography{Sweave}

\newpage
\appendix

\section{Frequently Asked Questions}
\label{sec:faq}

  \subsection{How can I get Emacs to automatically recognize files
    in Sweave format?}

  \abbr{ESS}\footnote{\url{https://ESS.R-project.org/}} automatically
  recognize files with extension \file{.Rnw} as Sweave files and turn
  on the correct modes. Please follow the instructions on the
  \abbr{ESS} homepage on how to install \abbr{ESS} on your computer.

  \subsection{Can I run Sweave directly from a shell?}

  You may run Sweave and Stangle directly from a shell, rather than
  manually starting \R{} and then running calling \code{Sweave} or
  \code{Stangle}, using \command{R~CMD~Sweave} and
  \command{R~CMD~Stangle} (these commands take the name of the file to
  process in argument). For \code{Sweave} and \code{Stangle} calls
  requiring many arguments, it may be simpler to pass the call to \R{}
  by means of \command{R~-e} or \command{Rscript~-e}.

  \subsection{Why does \LaTeX{} not find my EPS and PDF graphic files when
     the file name contains a dot?}

   Sweave uses the standard \LaTeX{} package \pkg{graphicx} to handle
   graphic files, which automatically chooses the type of file,
   provided the basename given in \verb|\includegraphics| has no
   extension. Hence, you may run into trouble with graphics handling
   if the basename of your Sweave file contains dots: \file{foo.Rnw}
   is OK, while \file{foo.bar.Rnw} is not.

   \subsection{Empty figure chunks give \LaTeX{} errors.}

   When a code chunk with \code{fig=TRUE} does not call any plotting
   functions, invalid PDF (or EPS) files may be created. Sweave cannot
   know if the code in a figure chunk actually plotted something or
   not, so it will try to include the graphics, which is bound to
   fail.

   \subsection{Why do R lattice graphics not work?}

   In recent versions of Sweave they do if they would when run at the
   command line: some calls (e.g.{} those inside loops) need to be
   explicitly \code{print()}-ed.

   \subsection{How can I get Black \& White lattice graphics?}

   What is the most elegant way to specify that strip panels are to have
   transparent backgrounds and graphs are to be in black and white when
   lattice is being used with Sweave?  I would prefer a global option that
   stays in effect for multiple plots.

   Answer by \I{Deepayan Sarkar}: I'd do something like this as part of
   the initialization:
\begin{verbatim}
 <<...>>=
 library(lattice)
 ltheme <- canonical.theme(color = FALSE)     ## in-built B&W theme
 ltheme$strip.background$col <- "transparent" ## change strip bg
 lattice.options(default.theme = ltheme)      ## set as default
 @
\end{verbatim}

  \subsection{Creating several figures from one figure chunk does
     not work}

   Consider that you want to create several graphs in a loop similar
   to
\begin{verbatim}
 <<fig=TRUE>>=
 for (i in 1:4) plot(rnorm(100)+i)
 @
\end{verbatim}
   This will currently \emph{not} work, because Sweave allows
   \emph{only one graph} per figure chunk. The simple reason is that
   Sweave opens a device before executing the code and closes it
   afterwards. If you need to plot in a loop, you have to program it
   along the lines of
\begin{verbatim}
 <<results=tex, echo=FALSE>>=
 for(i in 1:4) {
    fname <- paste("myfile", i, ".pdf", sep = "")
    pdf(file = fname, width = 6, height = 6)
    plot(rnorm(100)+i)
    dev.off()
    cat("\\includegraphics{", fname, "}\n\n", sep = "")
 }
 @
\end{verbatim}


 \subsection{How can I set default
   \texorpdfstring{\code{par()}}{par()} settings for figure  chunks?}

  Because Sweave opens a new device for each graphics file in each
  figure chunk, using \code{par()} has only an effect if it is used
  inside a figure chunk. If you want to use the same settings for a
  series of figures, it is easier to use a hook function than
  repeating the same \code{par()} statement in each figure chunk.

  The effect of
\begin{verbatim}
  options(SweaveHooks = list(fig = function() par(bg = "red", fg = "blue")))
\end{verbatim}
  should be easy to spot.

  \subsection{How can I change the formatting of R input and output
     chunks?}

   Sweave uses the \pkg{fancyvrb} package for formatting all \R{} code
   and text output. \pkg{fancyvrb} is a very powerful and flexible
   package that allows fine control for displaying text in verbatim
   environments. If you want to change the default layout, simply read
   the \pkg{fancyvrb} documentation and modify the definitions of
   the \code{Sinput} and \code{Soutput} environments in
   \file{Sweave.sty}, respectively.


  \subsection{How can I change the line length of R input and
     output?}

   Sweave respects the desired line length of R, namely
   \code{options("width")}. Therefore, after e.g.\ \code{options(width
     = 40)} lines will be formatted to have at most 40 characters (if
   possible).


   \subsection{Can I use Sweave for Word files?}

   Not directly, but package~\CRANpkg{officer} provides
   functionality to manipulate Microsoft Word documents from R.

   \subsection{Can I use Sweave for \I{OpenDocument} files?}

   Not directly. Package \CRANpkg{odfWeave} (archived in 2018) provided functions for using Sweave in
   combination with OpenOffice Writer rather than \LaTeX.
   See the \href{https://CRAN.R-project.org/view=ReproducibleResearch}{CRAN
     Task View: Reproducible Research} for an up-to-date list of related
   developments.

   \subsection{Can I use Sweave for HTML files?}

   Yes, package \CRANpkg{R2HTML} provides a driver for using Sweave in
   combination with HTML rather than \LaTeX.

   \subsection{After loading package \pkg{R2HTML} Sweave doesn't
     work properly!}

   Package \CRANpkg{R2HTML} registers an Sweave driver for HTML files
   using the same file extensions as the default \code{noweb}
   syntax, and after that its syntax is on the search list before the
   default syntax.  Using
\begin{verbatim}
  options(SweaveSyntax = "SweaveSyntaxNoweb")
\end{verbatim}
or calling Sweave like
\begin{verbatim}
  Sweave(..., syntax = "SweaveSyntaxNoweb")
\end{verbatim}
ensures the default syntax even after loading package \CRANpkg{R2HTML}.

\end{document}

%%% Local Variables:
%%% mode: latex
%%% TeX-master: t
%%% End: