simpler tips, tricks and tools - rob j. hyndman · goodhabits indentingandcommenting....

Simple.

......tips, tricks & tools

A brief history of R & R1976: S language developed at Bell Laboratories.

1980: S first used outside Bell.1987: I first used an implementation of S (calledAce) distributed by CSIRO.1988: S-PLUS produced, and I start using it.1996: I heard Ross Ihaka give a talk about R at astatistics conference.1997: CRAN began with 12 packages.2000: R 1.0.0 released.2001: I stopped using S-PLUS and switch to R.2004: I contributed my first function to R(quantile).2006: I contributed my first package to CRAN(forecast).

2

A brief history of R & R1976: S language developed at Bell Laboratories.1980: S first used outside Bell.

1987: I first used an implementation of S (calledAce) distributed by CSIRO.1988: S-PLUS produced, and I start using it.1996: I heard Ross Ihaka give a talk about R at astatistics conference.1997: CRAN began with 12 packages.2000: R 1.0.0 released.2001: I stopped using S-PLUS and switch to R.2004: I contributed my first function to R(quantile).2006: I contributed my first package to CRAN(forecast).

2

A brief history of R & R1976: S language developed at Bell Laboratories.1980: S first used outside Bell.1987: I first used an implementation of S (calledAce) distributed by CSIRO.

1988: S-PLUS produced, and I start using it.1996: I heard Ross Ihaka give a talk about R at astatistics conference.1997: CRAN began with 12 packages.2000: R 1.0.0 released.2001: I stopped using S-PLUS and switch to R.2004: I contributed my first function to R(quantile).2006: I contributed my first package to CRAN(forecast).

2

A brief history of R & R1976: S language developed at Bell Laboratories.1980: S first used outside Bell.1987: I first used an implementation of S (calledAce) distributed by CSIRO.1988: S-PLUS produced, and I start using it.

1996: I heard Ross Ihaka give a talk about R at astatistics conference.1997: CRAN began with 12 packages.2000: R 1.0.0 released.2001: I stopped using S-PLUS and switch to R.2004: I contributed my first function to R(quantile).2006: I contributed my first package to CRAN(forecast).

2

A brief history of R & R1976: S language developed at Bell Laboratories.1980: S first used outside Bell.1987: I first used an implementation of S (calledAce) distributed by CSIRO.1988: S-PLUS produced, and I start using it.1996: I heard Ross Ihaka give a talk about R at astatistics conference.

1997: CRAN began with 12 packages.2000: R 1.0.0 released.2001: I stopped using S-PLUS and switch to R.2004: I contributed my first function to R(quantile).2006: I contributed my first package to CRAN(forecast).

2

A brief history of R & R1976: S language developed at Bell Laboratories.1980: S first used outside Bell.1987: I first used an implementation of S (calledAce) distributed by CSIRO.1988: S-PLUS produced, and I start using it.1996: I heard Ross Ihaka give a talk about R at astatistics conference.1997: CRAN began with 12 packages.

2000: R 1.0.0 released.2001: I stopped using S-PLUS and switch to R.2004: I contributed my first function to R(quantile).2006: I contributed my first package to CRAN(forecast).

2

A brief history of R & R1976: S language developed at Bell Laboratories.1980: S first used outside Bell.1987: I first used an implementation of S (calledAce) distributed by CSIRO.1988: S-PLUS produced, and I start using it.1996: I heard Ross Ihaka give a talk about R at astatistics conference.1997: CRAN began with 12 packages.2000: R 1.0.0 released.

2001: I stopped using S-PLUS and switch to R.2004: I contributed my first function to R(quantile).2006: I contributed my first package to CRAN(forecast).

2

A brief history of R & R1976: S language developed at Bell Laboratories.1980: S first used outside Bell.1987: I first used an implementation of S (calledAce) distributed by CSIRO.1988: S-PLUS produced, and I start using it.1996: I heard Ross Ihaka give a talk about R at astatistics conference.1997: CRAN began with 12 packages.2000: R 1.0.0 released.2001: I stopped using S-PLUS and switch to R.

2004: I contributed my first function to R(quantile).2006: I contributed my first package to CRAN(forecast).

2

A brief history of R & R1976: S language developed at Bell Laboratories.1980: S first used outside Bell.1987: I first used an implementation of S (calledAce) distributed by CSIRO.1988: S-PLUS produced, and I start using it.1996: I heard Ross Ihaka give a talk about R at astatistics conference.1997: CRAN began with 12 packages.2000: R 1.0.0 released.2001: I stopped using S-PLUS and switch to R.2004: I contributed my first function to R(quantile).

2006: I contributed my first package to CRAN(forecast).

2

A brief history of R & R1976: S language developed at Bell Laboratories.1980: S first used outside Bell.1987: I first used an implementation of S (calledAce) distributed by CSIRO.1988: S-PLUS produced, and I start using it.1996: I heard Ross Ihaka give a talk about R at astatistics conference.1997: CRAN began with 12 packages.2000: R 1.0.0 released.2001: I stopped using S-PLUS and switch to R.2004: I contributed my first function to R(quantile).2006: I contributed my first package to CRAN(forecast).

2

Outline

1 Getting help

2 Finding functions

3 Digging into functions

4 Writing functions

5 Debugging

6 Version control

7 My R workflow

3

Getting helpStackOverflow.comFor programmingquestions.

4


CrossValidated.com

For statisticalquestions.

4


CrossValidated.com

For statisticalquestions.

R-help mailing listsstat.ethz.ch/mailman/listinfo/r-helpOnly when all-elsefails.

4

stat.ethz.ch/mailman/listinfo/r-help



Outline

1 Getting help

2 Finding functions


4 Writing functions

5 Debugging

6 Version control

7 My R workflow

5

How to find the right functionFunctions in installed packages.

......

help.search("neural"). Equivalently: ??neuralAlso built into RStudio help.

Functions in other CRAN packages.

......

library(sos)findFn("neural")RSiteSearch("neural")

findFn only searches functions.RSiteSearch searches more widely.

rseek.org

Google customized search on R-related sites.

6


......



......


findFn only searches functions.

RSiteSearch searches more widely.rseek.org


6


......



......



rseek.org


6


......



......



rseek.orgGoogle customized search on R-related sites.

6

CRAN Task Views.

......cran.r-project.org/web/views/

Curated reviews ofpackages by subject

Use install.views()and update.views()in the ctv package.

7

cran.r-project.org/web/views/


CRAN Task Views.

......cran.r-project.org/web/views/

Curated reviews ofpackages by subjectUse install.views()and update.views()in the ctv package.

7



Outline

1 Getting help

2 Finding functions


4 Writing functions

5 Debugging

6 Version control

7 My R workflow

8

Digging into functions.Example: How does forecast for ets work?..

......

forecastforecast.etsforecast:::pegelsfcast.C

Typing the name of a function gives itsdefinition.Be aware of classes and methods.Type package:::function for hiddenfunctions.Download the tar.gz file from CRAN ifyou want to see any underlying C orFortran code.

9


......


Typing the name of a function gives itsdefinition.

Be aware of classes and methods.Type package:::function for hiddenfunctions.Download the tar.gz file from CRAN ifyou want to see any underlying C orFortran code.

9


......


Typing the name of a function gives itsdefinition.Be aware of classes and methods.

Type package:::function for hiddenfunctions.Download the tar.gz file from CRAN ifyou want to see any underlying C orFortran code.

9


......


Typing the name of a function gives itsdefinition.Be aware of classes and methods.Type package:::function for hiddenfunctions.

Download the tar.gz file from CRAN ifyou want to see any underlying C orFortran code.

9


......


Typing the name of a function gives itsdefinition.Be aware of classes and methods.Type package:::function for hiddenfunctions.Download the tar.gz file from CRAN ifyou want to see any underlying C orFortran code.

9

Outline

1 Getting help

2 Finding functions


4 Writing functions

5 Debugging

6 Version control

7 My R workflow

10

Good habitsindenting and commenting.

Reindenting using RStudio.formatR package: I run tidy.dir() beforereading student code!Google R style guide:Hadley’s R style guide:

11

Good habitsindenting and commenting.Reindenting using RStudio.

formatR package: I run tidy.dir() beforereading student code!Google R style guide:Hadley’s R style guide:

11

Good habitsindenting and commenting.Reindenting using RStudio.formatR package: I run tidy.dir() beforereading student code!

Google R style guide:Hadley’s R style guide:

11

Good habitsindenting and commenting.Reindenting using RStudio.formatR package: I run tidy.dir() beforereading student code!Google R style guide:

Hadley’s R style guide:

11

Good habitsindenting and commenting.Reindenting using RStudio.formatR package: I run tidy.dir() beforereading student code!Google R style guide:Hadley’s R style guide:

11

Outline

1 Getting help

2 Finding functions


4 Writing functions

5 Debugging

6 Version control

7 My R workflow

12

Simple debugging tools.When something goes wrong:..

......

traceback()

options(error=recover)debug()browser()trace()

.More extensive debugging tools discussed at..

......

www.stats.uwo.ca/faculty/murdoch/software/debuggingR/.

......RStudio plans to have debugging tools in afuture release.

13

www.stats.uwo.ca/faculty/murdoch/software/debuggingR/



......

traceback()options(error=recover)

debug()browser()trace()


......



13




......

traceback()options(error=recover)debug()

browser()trace()


......



13




......

traceback()options(error=recover)debug()browser()

trace()


......



13




......

traceback()options(error=recover)debug()browser()trace()


......



13




......



......


.


13




......



......



13



Outline

1 Getting help

2 Finding functions


4 Writing functions

5 Debugging

6 Version control

7 My R workflow

14

Version control.

......In any non-trivial project, it is important to keeptrack of versions of R code and documents.

Dropbox

Every version of your files for the last 30 days(or longer if you buy Packrat).

But no built-in diff facilities or changelog.

Git

Free version control system.

Easy to fork or roll back.

Install git software on your local computer.

Manage via RStudio.

15

Outline

1 Getting help

2 Finding functions


4 Writing functions

5 Debugging

6 Version control

7 My R workflow

16

R projects

Basic idea

Every paper, book or consulting report is a“project”.

Every project has its own folder and Rworkspace.

Every project is entirely scripted. That is, allanalysis, graphs and tables must be able to begenerated by running one R script. The finalreport must be able to be generated byprocessing one LATEX file.

17

R projectsfunctions.R contains all non-packagedfunctions used in the project.

main.R sources all other R files in the correctorder.Data files provided by the client (ordownloaded from a website) are never edited.All data manipulation is scripted in R (orsome other language).Tables generated via xtable or texregpackages.Graphics in pdf format.Report or paper in LATEX pulls in the tables andgraphics.

18

R projectsfunctions.R contains all non-packagedfunctions used in the project.main.R sources all other R files in the correctorder.

Data files provided by the client (ordownloaded from a website) are never edited.All data manipulation is scripted in R (orsome other language).Tables generated via xtable or texregpackages.Graphics in pdf format.Report or paper in LATEX pulls in the tables andgraphics.

18

R projectsfunctions.R contains all non-packagedfunctions used in the project.main.R sources all other R files in the correctorder.Data files provided by the client (ordownloaded from a website) are never edited.All data manipulation is scripted in R (orsome other language).

Tables generated via xtable or texregpackages.Graphics in pdf format.Report or paper in LATEX pulls in the tables andgraphics.

18

R projectsfunctions.R contains all non-packagedfunctions used in the project.main.R sources all other R files in the correctorder.Data files provided by the client (ordownloaded from a website) are never edited.All data manipulation is scripted in R (orsome other language).Tables generated via xtable or texregpackages.

Graphics in pdf format.Report or paper in LATEX pulls in the tables andgraphics.

18

R projectsfunctions.R contains all non-packagedfunctions used in the project.main.R sources all other R files in the correctorder.Data files provided by the client (ordownloaded from a website) are never edited.All data manipulation is scripted in R (orsome other language).Tables generated via xtable or texregpackages.Graphics in pdf format.

Report or paper in LATEX pulls in the tables andgraphics.

18

R projectsfunctions.R contains all non-packagedfunctions used in the project.main.R sources all other R files in the correctorder.Data files provided by the client (ordownloaded from a website) are never edited.All data manipulation is scripted in R (orsome other language).Tables generated via xtable or texregpackages.Graphics in pdf format.Report or paper in LATEX pulls in the tables andgraphics.

18

R projects

.Advantages over Sweave and knitR..

......

å Keeps R and LATEX files separate.

å Multiple R files that can be runseparately.

å Easier to rebuild sections (e.g., onlysome R files).

å Easier to collaborate.

19

Figures without whitespaceR graphics have too much surrounding whitespace for inclusion in reports.

The following function fixes the problem.

.

......

savepdf <- function(file, width=16, height=10){fname <- paste("figs/",file,".pdf",sep="")pdf(fname, width=width/2.54, height=height/2.54,

pointsize=10)par(mgp=c(2.2,0.45,0), tcl=-0.4, mar=c(3.3,3.6,1.1,1.1))

}

savepdf("fig1")plot(x,y)dev.off()

20

R with Makefiles.

......

A Makefile provides instructions about how tocompile a project..Makefile..

......

# list R filesRFILES := $(wildcard *.R)# pdf figures created by RPDFFIGS := $(wildcard figs/*.pdf)# Indicator files to show R file has runOUT_FILES:= $(RFILES:.R=.Rdone)

all: $(OUT_FILES)

# RUN EVERY R FILE%.Rdone: %.R functions.RRscript $< && touch $@

21

Makefiles for R & LATEX.Makefile..

......

# Usually, only these lines need changingTEXFILE= paperRDIR= ./figsFIGDIR= ./figs

# list R filesRFILES := $(wildcard $(RDIR)/*.R)# pdf figures created by RPDFFIGS := $(wildcard $(FIGDIR)/*.pdf)# Indicator files to show R file has runOUT_FILES:= $(RFILES:.R=.Rdone)# Indicator files to show pdfcrop has runCROP_FILES:= $(PDFFIGS:.pdf=.pdfcrop)

all: $(TEXFILE).pdf $(OUT_FILES) $(CROP_FILES)

22

Makefiles for R & LATEX.Makefile continued …..

......

# May need to add something here if some R files# depend on others.

# RUN EVERY R FILE$(RDIR)/%.Rdone: $(RDIR)/%.R $(RDIR)/functions.RRscript $< && touch $@

# CROP EVERY PDF FIG FILE$(FIGDIR)/%.pdfcrop: $(FIGDIR)/%.pdfpdfcrop $< $< && touch $@

# Compile main tex file$(TEXFILE).pdf: $(TEXFILE).tex $(OUT_FILES) $(CROP_FILES)latexmk -pdf -quiet $(TEXFILE)

23

The magic of RStudioAn RStudio projectcan have anassociated Makefile.Then building andcleaning can be donefrom within RStudio..

...... rstudio.com

24

rstudio.com

Keep up-to-date

RStudio blog:

.

...... blog.rstudio.org

I blog about R (and other things):

.

...... robjhyndman.com/researchtips

R-bloggers:

.

...... www.r-bloggers.com

25

blog.rstudio.org

robjhyndman.com/researchtips

www.r-bloggers.com

Keep up-to-date

RStudio blog:....... blog.rstudio.org

I blog about R (and other things):

.

...... robjhyndman.com/researchtips

R-bloggers:

.


25

blog.rstudio.org


www.r-bloggers.com

Keep up-to-date


I blog about R (and other things):....... robjhyndman.com/researchtips

R-bloggers:

.


25

blog.rstudio.org


www.r-bloggers.com

Keep up-to-date


I blog about R (and other things):....... robjhyndman.com/researchtips

R-bloggers:....... www.r-bloggers.com

25

blog.rstudio.org


www.r-bloggers.com

simpler tips, tricks and tools - rob j. hyndman · goodhabits indentingandcommenting....

Documents