R development

Creating R Packages,
Using CRAN, R-Forge,
And Local R Archive Networks
And Subversion (SVN) Repositories
Spencer Graves Sundar Dorai-Raj
PDF Solutions Google
San José CA Mountain View CA
spencer.graves@prodsyse.com sdorairaj@google.com

© 2008 PDF Solutions, Inc. PDF Solutions has made these materials available for public noncommercial use, and
they may be reproduced, in part or in whole, consistent with the following requirements without charge or further
permission from PDF Solutions: (1) users exercise due diligence in ensuring the accuracy of the materials
reproduced; (2) “PDF Solutions, Inc.” be clearly and predominantly identified as the source; and, (3) the reproduction
is not represented as an official version of the materials reproduced, nor as having been made, in affiliation with or
with the endorsement of PDF Solutions.

Motivation
R is the language of choice for a large and growing
proportion of people developing new statistical algorithms
Comprehensive R Archive Network (CRAN) makes it easy
to benefit from others’ work and to share your work and get
feedback on potential improvements
Creating R packages
• Provides a system for creating software with documentation
including unit tests, and thereby
• Increases software quality & development productivity
Local R Archive Networks can increase your productivity in
developing new code and sharing it with coworkers
R-Forge and local Subversion (SVN) repositories make
collaboration on joint software development easy &
productive
2 / PDF Solutions Inc.
Yield, Performance, Profitability

Outline

Installing R and R Packages
• From CRAN
• From a local package
• From alternative repositories
• Getting help
Obtaining source code
Creating R packages
Establishing and Maintaining Local R Archive
Networks
Using Subversion (SVN)


Installing R And R Packages

Installing R from CRAN
Installing R contributed packages from
• CRAN
• local package
• alternative repositories


Installing R
www.r-project.org
CRAN
(select a local repository)
Download an appropriate precompiled version or
package source to suit your operating system
Configure ...
• R Installation and Administration manual http
://cran.r-project.org/doc/manuals/R-admin.pdf
• modify default options in “~R/etc/Rprofile.site”:
• default repositories (including local?)
• max.print
options(repos = c(CRAN = "http://cran.cnr.berkeley.edu",
• ... CRANextra = "http://www.stats.ox.ac.uk/pub/RWin"),
max.print=222)

system.file()

Installing R Packages From CRAN
install.packages(‘packageName’)
OR in Rgui:

• select a local repository (if needed)
• select package(s) from list


Installing R Packages From Local Zip Files (Windows)
in Rgui:

find packageName.zip


Or From R Command Prompt (Any OS)

Windows binary
• install.packages(“packageName.zip”, repos = NULL)

Any OS provided appropriate tools for compiling
source are available
• install.packages(“packageName.tar.gz”, repos =
NULL)
• Windows requires “Rtools”
• http://www.murdoch-sutherland.com/Rtools/
• Mac requires Xtools
• For most Linux/UNIX systems the required toolsets
are available

Getting Help
?functionName
• help pages for packages in the search path
Fuzzy search
• “help.search” function
• www.r-project.org → search or ‘RSiteSearch’ function
Other R search engines and R Wiki
Google
r-help listserve
• PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html and provide
commented, minimal, self-contained, reproducible code.
• Reading “r-help”, “r-devel”, “r-sig-___” is like attending a
professional meeting a few minutes a day

Outline

Creating R packages
Networks


Obtaining Source Code For R
www.r-project.org CRAN (select a
repository)
For R:


Obtaining Source Code For A Package

Load CRAN in
browser
• Click “Packages”
link
• 1700 objects
including
packages (as of
2009-03-11)
• Find the
package of
interest by first
letter
• click name

“lme4” Package
Package pages contain
links to:
• Package dependencies
• Package source
• Package binaries
• Reference manual
• Archives for old source
tarballs
• Maintainer contact info
• And, if applicable,
• Project URL
• Task Views
• Vignettes

Using An Installed Package
help(package = fortunes) or library(help =
fortunes)
• to get an overview of package capabilities
library(fortunes)
• to attach it as the second in the search path
?fortune
• to get ‘help’ on the function ‘fortune’

> fortune('RTFM')

This is all documented in TFM. Those who WTFM don't want
to have to WTFM again on the mailing list. RTFM.
-- Barry Rowlingson
R-help (October 2003)

“DierckxSpline” Package

Click
• Download to your hard
drive
• Unzip


“DierckxSpline” Package Contents
data sets
files not checked by ‘R CMD check’
Help files
R function definition files
source code in Fortran, C, C++, ...
Package description
Names to be exported
Not all packages have all of these
Some packages have others
Ultimate documentation = source code
“debug” function: walk through R code line by line until
we understand what it does; “browser” for check points

Outline

Creating R packages
• Why?
• How to create?
• How to check?
• How to share?
Networks


Why Create R Packages
Productivity
• Tripled my software development productivity overnight
• Help file with examples first; code to these examples
• “R CMD check” finds when new changes break previous
tests
• Version control
Quality:
• Examples = unit testing
• http://en.wikipedia.org/wiki/Unit_test
• Chambers’ “Prime Directive”: Trustworthy software
• (2008) Software for Data Analysis (Springer)

• as well as documentation
Easy to share results
Easy to understand what I did a couple of years ago

How to Create an R Package
Copy existing package(s)
“package.skeleton” function
“Writing R Extensions” manual
• http://cran.r-project.org/doc/manuals/R-exts.pdf
Other references:
• Rossi, Peter (2006) Making R Packages under Windows
http://faculty.chicagogsb.edu/peter.rossi/research/bayes%20book/bayesm/Making%20R%20Packages%20Under%20W
, accessed 2008.11.02

• Leisch, Friedrich (2008) Creating R Packages: A Tutorial
http://epub.ub.uni-muenchen.de/6175/

R-devel listserve (r-devel@stat.math.ethz.ch)
Rolf Turner: In the middle of a Saturday morning (in my Time Zone!) I send out a plea for help,
and in just over 20 minutes my problem is solved! I don't think you get service like that
anywhere else. This R-help list is BLOODY AMAZING!
Spencer Graves: 'The sun never sets on the (former) British Empire.' Today, it never sets on
R-Help.
-- Rolf Turner and Spencer Graves
R-help (May 2005)

See section 1.1 in “Writing R
Extensions”
Package Directory Structure

packageName

• DESCRIPTION – describes the package contents Required
• man – Rd help files
• R – R code files

Optional
• NAMESPACE – defines the package name space
• data – contains files with data (txt, csv, rda)
• inst – contents are copied to installed package
• src – C, Fortran code to compile with the package
• tests – R code for testing package functions


Building Packages On Windows

Requires Rtools
• Contains all compilers and Unix tools
• http://www.murdoch-sutherland.com/Rtools
LaTeX: http://www.miktex.org
For additional help, see:
• Google
• r-devel mailing list
• FAQ: http://cran.cnr.berkeley.edu/bin/windows/base/rw-FAQ.html
• http://faculty.chicagogsb.edu/peter.rossi/research/bayes%20book/b
, accessed 2008.11.02


Building Packages On Mac

Mac tools are usually not loaded “out-of-the-box”
• Required tools maybe downloaded or installed from
the OSX installation CDs
• http://developer.apple.com/tools/xcode/
Latex: http://www.tug.org/mactex/
Building packages on PPC and Intel Macs slightly
different
• See the FAQ 5.4 on link below
Help
• http://cran.cnr.berkeley.edu/bin/macosx/RMacOSX-FA
• R-SIG-Mac mailing list


Typical Package Check And Install Sequence
R CMD build packageName
• (or “R CMD build pkg” with an R-Forge package)
• Windows: in a “Command Prompt” window with
‘packageName’ in the local directory current package
• Creates “packageName_x.y-z.tar.gz” version number
R CMD check packageName_x.y-z.tar.gz
R CMD install packageName_x.y-z.tar.gz
• Installs it in your local installation of R
R CMD install --build packageName_x.y-z.tar.gz
• Creates “packageName_x.y-z.zip”, which can be
used to install “packageName” on other Windows
computers All “R CMD” commands are executed in a Windows
CMD terminal (or analogous terminal for other OSes)

Cryptic Error Message?
invisible(lapply(list.files(“~packagepath/R", full =
TRUE, pattern=".R$"), source))
• This call individually sources every R file in a
directory
• Identifies particular functions and lines with syntax
errors
Google
RSiteSearch
• www.r-project.org Search
• Function in R (i.e. RSiteSearch(restrict = “functions”))
• R-devel mailing list
Undo recent changes and try again from the last
working version

Submitting A Package To CRAN
www.r-project.org -> CRAN -> (select a local mirror)

Build packageName_x.y-z with the current version of R
Upload to ftp://cran.r-project.org/incoming
• packageName_x.y-z.tar.gz
• (With firewall problems, can you use a different computer?)
Email “cran@r-project.org”
• subj: packageName_x.y-z.tar.gz now on CRAN
• text: “uploaded to CRANincoming. GPL (>= 2)”


Outline

Creating R packages
Networks


Local R Archive Networks
Why:
• Share work with others that
you may not want to share
with the world
How:
• Requires access to a web server
• Then setting up a very specific directory structure to hold
both source and binary packages
• bin directory contains compiled packages for Windows
(*.zip) or Mac (*.tgz)
• Must contain a subdirectory for every supported version of
R
• src directory contains package source (*.tar.gz)


Repository Directory Structure

/www (directory that is visible from web)
• bin
• windows
• contrib
• 2.7 package1_x.y-z.zip
package2_x.y-z.zip
• 2.8 PACKAGES

• macosx
• contrib
package1_x.y-z.tgz
• 2.7 package2_x.y-z.tgz
• 2.8 PACKAGES

• src
package1_x.y-z.tar.gz
• contrib package2_x.y-z.tar.gz
PACKAGES

Accessing The Repository Via install.packages
The PACKAGES file
identifies which version to
install
• Contents of PACKAGES
equal DESCRIPTION file
from each package

Installing a package
• install.packages(“packageName”, repos =
“http://my.Rrepos.com”)
• Or add to Rprofile.site (in $RHOME/etc)
options(repos = c(CRAN = "http://cran.cnr.berkeley.edu",
myCRAN = "http://my.Rrepos.com",
CRANextra = "http://www.stats.ox.ac.uk/pub/RWin"),
max.print=222)
R.home() # R installation directory

Outline

Creating R packages
Networks
• Why?
• Installing and Using Subversion
• R-Forge
• a local Subversion (SVN) repository
• How to use
• How to establish and maintain

Why Use A Subversion Repository?
Easy to collaborate on package development
Help learn R
• Find an R package that interests you
• Make suggestions to the package maintainer
• A maintainer may ask if you’d like do make those
changes in their subversion repository
Audit trail on all changes
• Relatively easy to identify and reverse changes
selectively

Creating an SVN repository (e.g. R-Forge)
typically requires help from Information
Technology

Installing and Using Subversion (SVN) Client

SVN – http://subversion.tigris.org
Windows client – TortoiseSVN
http://tortoisesvn.tigris.org
Mac client – Finder plugin
http://scplugin.tigris.org
Symbols
• Green check: No local changes since
“Commit”
• Red exclamation point: local change not
in the repository
• Yellow exclamation point: an “SVN
Update” conflicted with local changes

SVN Checkout, Update, Commit

SVN Checkout
• Creates a local copy of a package on an SVN
repository
SVN Update
• Updates local copies to newer versions on the
repository
• Identifies conflicts between recent changes made
locally and elsewhere
SVN Commit
• Uploads recent changes from the local copy to the
repository


Two Subversion Repositories For R: RForge & R-Forge
RForge: www.rforge.net
• 37 projects as of 2009-03-11

R-Forge: r-forge.r-project.org
• 340 projects as of 2009-03-11
• including DierckxSpline, FinTS, maxLik, fda, Rmetrics, ...

Both are free
Installation of Packages in R: If an R-Forge package
passed the quality check it can be installed directly
via:
• install.packages(“DierckxSpline",repos="http://r-forge.r-
project.org")

Anonymous Subversion Access From R-Forge
svn checkout svn://svn.r-forge.r-
project.org/svnroot/dierckxspline
Windows: right-click on a new folder & select SVN
Checkout


Developer Subversion Access Via SSH

Only project developers can access the SVN tree
via this method. SSH must be installed on your
client machine. Substitute developername with
the proper values. Enter your site password when
prompted.
svn checkout svn+ssh://developername@svn.r-
forge.r-project.org/svnroot/dierckxspline


A Local Subversion Repository

Why?
• Facilitate collaboration on software development
How?
• Different people typically work on different functions
• “SVN Update” downloads recent changes made by
others
• “R CMD check” makes sure everything passes the
programmed unit tests
• “SVN Commit” uploads recent local changes


How To Establish/Maintain An SVN Repository

Creating a repository server typically requires
help from your local IT department
• We won’t discuss that here.
Once established, TortoiseSVN can be used to
create projects.
To add a new project to the repository:
• “Import” to the repository
• “Checkout” an official local copy
• which contains the bookkeeping SVN requires that is
NOT included in your “Import”


“Import” To The Repository
Click on the
folder containing
the package
(DESCRIPTION,
MAN, R, ...)
Tortoise SVN
Import
Enter “URL of
repository”
• with the name of
your package


“Checkout”
Your original does NOT contain the bookkeeping
information required by SVN
Therefore, you need to “Checkout” an official copy properly
configured for SVN
To do that
• Create a new folder
to contain this
version
• Right-click:
TortoiseSVN
• Checkout
• Enter “URL of
Repository” and
“Checkout Directory”

Outline

Creating R packages
Networks


Annotated Bibliography

Writing R Extensions
• http://cran.r-project.org/doc/manuals/R-exts.pdf
• THE official reference manual for R package
development
• BUT: It IS a reference manual, NOT a tutorial
Rossi, Peter (2006) Making R Packages under
Windows: A Tutorial
• http://faculty.chicagogsb.edu/peter.rossi/research/bay
, accessed 2008.11.02
• Excellent overview


Annotated Bibliography – 2

Falcon, Seth (2006) Modeling package
dependencies using graphs. R News, 6(5):8-12,
December 2006.
• “pkgDepTools” package for viewing dependencies
between packages
Gilbert, Paul, R (2004) package maintenance. R
News, 4(2):21-24, September 2004.
• Reviews the “Make” capabilities described more fully
in “Writing R Extensions”
Ligges, Uwe (2003) R help desk: Package
management. R News, 3(3):37-39, December 2003.
• Managing packages in multiple ‘libraries’



Leisch, Friedrich (2008): Creating R Packages: A
Tutorial. In: Brito, Paula (ed.) , Compstat 2008 -
Proceedings in Computational Statistics. Physica
Verlag: Heidelberg, Germany.
• http://epub.ub.uni-muenchen.de/6175/
Ripley, Brian D. (2005) Packages and their
management in R 2.1.0. R News, 5(1):8-11, May
2005.
• Updates Ligges (2003) to R 2.1.0



Rougier, Jonathan (2005) Literate programming
for creating and maintaining packages. R News,
5(1):35-39, May 2005.
• “The basic idea of literate programming is ... to keep
the code and the documentatation ... together, in one
file” using the “noweb” literate programming tool.


R development

More Related Content

Similar to R development (20)

R development