1. Installing the required tools

To build an R package in Windows, you will need to install some additional software tools. These are summarized at

http://www.murdoch-sutherland.com/Rtools

1.1 Essential: Rtools

This is a collection of unix-like tools that can be run from the DOS command prompt. It also contains the MinGW compilers that are used for compiling Fortran and C code. Download using

http://www.murdoch-sutherland.com/Rtools/Rtools.exe

You should download and run it, choosing the default “Package authoring installation” to build add-on packages.

1.2 Optional: MikTeX

MikTeX is used for producing the pdf help files. You can produce an R package without it, but the package will not contain pdf help files. Most of you will have this installed anyway. Download from

http://www.miktex.org

1.3 Essential: Setting PATH variable

The PATH variable tells Windows where to find the relevant programs. To add a directory to you PATH on Windows XP select

Control Panel -> System -> Advanced -> Environment Variables

The path variable may have already been fixed in step 1.1. In any case, you should check that it looks something like this:

 C:Rtoolsbin;C:Rtoolsperlbin;C:RtoolsMinGWbin;C:Program filesRR-2.10.0bin;C:Program FilesHTML Help Workshop;<others>

• The precise directories will depend where you have installed the various tools. The above path should work if you have followed the default installation procedure. If the directory names contain spaces, put them in quotation marks.
• If you have not installed the HTML Help workshop, you will need to set WINHELP=NO in  MkRules (in the directory "C:Program filesRR-2.10.0bin").
• The htmlhelp part can be omitted if you did not install the HTML help workshop.
• I have assumed R2.10.0. For later versions, simply change the above paths to the relevant R version. It will probably then still work.
• If there are problems, please read the Rtools.txt file carefully.

2 Creating the package

Information about creating packages is provided in the document “Writing R extensions” (available under the R Help menu) or at

http://cran.r-project.org/doc/manuals/R-exts.html

The main items are summarized below to get you started, but you will almost certainly need to consult the above document if you are to successfully compile a package.

2.1 Use package.skeleton()

The simplest way to create a package is to first create an R workspace containing all the relevant functions and data sets that you want to include in the package. Delete anything from the workspace that you do not want to include in the package. Make sure the current directory is set to whereever you want create the package. Use

setwd("C:My DocumentsRpackages")

for example. Then, to create a package called "fred", use the R command

package.skeleton(name="fred")

This will generate a directory fred and several sub-directories in the required structure.

2.2 Editing the files

A package consists of a directory containing a file ‘DESCRIPTION’ and usually has the subdirectories R, data and man. The package directory should be given the same name as the package. The package.skeleton command above will have created these files for you. You now need to edit them so they contain the right information.

2.3 DESCRIPTION file

The DESCRIPTION file contains basic information about the package in the following format:

 Package: pkgname
 Version: 0.5
 Date: 2007-06-05
 Title: My first collection of functions
 Author: Joe Developer <Joe.Developer @some.domain.net>, with
 contributions from A. User <A.User @whereever.net>.
 Maintainer: Joe Developer <Joe.Developer @some.domain.net>
 Depends: R (>= 2.0.0), forecast
 Suggests: tseries
 Description: A short (one paragraph) description of what
 the package does and why it may be useful.
 License: GPL version 2 or newer
 URL: http://www.another.url

2.4 Rd files

The help files for each function and data set are given in “R documentation” (Rd) files in the man subdirectory. These are in a simple markup language closely resembling LaTeX, which can be processed into a variety of formats, including LaTeX, HTML and plain text. As an example, here is the file which documents the function seasadj in the forecast package.

name{seasadj}
alias{seasadj}
title{Seasonal adjustment}
usage{
seasadj(object)
}

arguments{
item{object}{Object created by code{link[stats]{decompose}}
or code{link[stats]{stl}}.}
}

description{Returns seasonally adjusted data constructed
by removing the seasonal component.}

value{Univariate time series.}

seealso{code{link[stats]{stl}}, code{link[stats]{decompose}}}

author{Rob J Hyndman}

examples{
plot(AirPassengers)
lines(seasadj(decompose(AirPassengers,"multiplicative")),col=4)
}

keyword{ts}

Detailed instructions for writing R documentation are at

http://cran.r-project.org/doc/manuals/R-exts.html#Writing-R-documentation-files

2.5 Including C or Fortran code

If your R code calls C or Fortran functions, the source code for these functions needs to be placed in the subdirectory src under fred.

2.6 Compiling the package for Windows

To compile the package into a zip file, go to a DOS prompt in the directory containing your package. (i.e., the directory "C:My DocumentsRpackages" in the above example. Then type

Rcmd build --binary fred

This will compile all the necessary information and create a zip file which should be ready to load in R.

2.7 Checking the package

To check that the package satisfies the requirements for a CRAN package, use

Rcmd check fred

The checks are quite strict. A package will often work ok even if it doesn’t pass these tests. But it is good practice to build packages that do satisfy these tests as it may save problems later.

2.8 Building a package for other operating systems

To build a package for something other than a Windows computer, use

Rcmd build fred

This creates a tar.gz file which can then be installed on a non-Windows computer. It can also be uploaded to CRAN provided it satisfies the above tests.