Chapters - One Page R

Hands-On Data Science with R
Doing it with Style
[email protected]
8th December 2014
Visit http://HandsOnDataScience.com/ for more Chapters.
T
Data Scientists program over data—they write programs to manage, manipulate, and model data
in a variety of ways. Our programs will need to read and understood by others. Here we present
guidelines for programming in R (R Core Team, 2014) that assists with the transparency of our
programs.
DR
AF
The aims of any programming style include ensuring consistency and ease of understanding.
When we write programs we should write for others to easily read and to learn from
and to build upon. Think of writing a story to keep the reader (including ourselves) informed
and engaged. If a reviewer can’t follow the program then we have not succeeded. Keep it
simple.
Of course, we generally write programs to have it run on a computer which cares little about
programming style—a compiler or interpreter will translate our program into machine code that
the computer understands. Very ugly code will also work and achieve that goal but why not
make it attractive and accessible to other humans?
As we work through this chapter, new R commands will be introduced. Be sure to review the
command’s documentation and understand what the command does. You can ask for help using
the ? command as in:
?read.csv
We can obtain documentation on a particular package using the help= option of library():
library(help=rattle)
This chapter is intended to be hands on. To learn effectively, you are encouraged to have R
running (e.g., RStudio) and to run all the commands as they appear here. Check that you get
the same output, and you understand the output. Try some variations. Explore.
©
Copyright
2013-2014 Graham Williams. You can freely copy, distribute,
or adapt this material, as long as the attribution is retained and derivative
work is provided under the same license.
Data Science with R
1
Hands-On
Doing it with Style
Naming Files
1. Filenames use the R extension. This aligns with the fact that the language is unambiguously
called “R” and not “r.”
Preferred
generatePlots.R
Discouraged
generatePlots.r
2. The file name should match the name of the main function defined within the file. For
example, if the function defined in the file is fancyPlot() then name the file as:
Preferred
fancyPlot.R
fancy_plot.R
fancy.plot.R
fancy_plot.r
T
Discouraged
DR
AF
3. R binary data filenames end in ”.RData”. I have no strong motivation for this except that
it conforms to the capitalised naming scheme.
Preferred
weather.RData
Discouraged
weather.rdata
weather.Rdata
weather.rData
4. Other standard file names use lowercase where there is a choice.
Preferred
weather.csv
Discouraged
weather.CSV
Copyright
© 2013-2014 [email protected]
Module: StyleO
Page: 1 of 8
Data Science with R
2
Hands-On
Doing it with Style
Naming Objects
5. Function names use capitalised verbs, beginning with lowercase.
Preferred
displayPlotAgain()
Discouraged
DisplayPlotAgain()
displayplotagain()
display_plot_again()
6. Variable names and function argument names use dot separated words.
Preferred
list.of.frames
lib.cmd
list_of_frames
T
Discouraged
DR
AF
7. Constants are all capitals. This makes them stand out and makes it clear that they should
not be changed.
Preferred
MAX.LINES
Discouraged
const.max.lines
8. Variables within a dataset (i.e., data frame) are lowercase and use underscore to separate the
words. This has the advantage that underscore is acceptable in SQL databases for columns,
whereas a period is often used to identify the server/database/table/schema names. We
often load/save data in data frames from/to databases. We can use normVarNames() from
rattle (Williams, 2014) to normalise variables names in this way.
Preferred
min_temp
wind_gust_speed
Discouraged
max.pressure
wind.dir
WindSpeed
Copyright
© 2013-2014 [email protected]
Module: StyleO
Page: 2 of 8
Data Science with R
3
Hands-On
Doing it with Style
Layout
9. Named arguments in parameter lists do not have a space around the =. I prefer this as
visually it ties the named arguments strongly together. This is the only situation where
I tightly couple a binary operator. In all other situations there should always be a space
around the operator. Another motivation is that it avoids splitting the line between the
argument name and the argument value.
Preferred
read.csv(file="data.csv", sep=";", na.strings=".")
Discouraged
read.csv(file = "data.csv", sep =
";", na.strings
= ".")
T
10. Use an indentation of 2. Some argue this is not enough to show the structure when using
smaller fonts. If it is an issue for you then 4 is okay. But I would choose a different font
instead.
DR
AF
11. Align curly braces. Thus an opening curly brace is on a line by itself. This is a particular
difference with many other programming styles. My motivation is that the open and close
curly braces are then aligned visually and this provides an added visual check of syntax
correctness and visually gives a very strong code block view. The placement of the open
curly bracket at the end of the previous line is endemic and in my opinion is bad practise,
hiding the opening of a block of code simply to save on having some additional lines (which
was only important in my much younger days where we used punched cards or terminals
limited to 24 lines). The preferred style also makes it easier to comment out, for example,
just the line containing the “while” and still have valid syntax. Don’t be afraid of the extra
white space—for the human reader, white space is good, and the computer does not care.
Preferred
while (blueSky())
{
openTheWindows()
doSomeResearch()
}
retireForTheDay()
Discouraged
while (blueSky()) {
openTheWindows()
doSomeResearch()
}
retireForTheDay()
Copyright
© 2013-2014 [email protected]
Module: StyleO
Page: 3 of 8
Data Science with R
4
Hands-On
Doing it with Style
Function Definition Layout
12. Align function arguments by comma. This is a controversial style, but it works to emphasize
the arguments and makes it easier to comment out some arguments with little fuss.
Interesting Option
dialPlot <- function(label="UseR!"
, value=78
, dial.radius=1
, value.cex=3
, value.color="black"
, label.cex=3
, label.color="black")
{
...
}
Traditional
DR
AF
T
dialPlot <- function(label="UseR!", value=78, dial.radius=1, value.cex=3,
value.color="black", label.cex=3, label.color="black")
{
...
}
dialPlot <- function(label="UseR!",
value=78,
dial.radius=1,
value.cex=3,
value.color="black",
label.cex=3,
label.color="black")
{
...
}
Copyright
© 2013-2014 [email protected]
Module: StyleO
Page: 4 of 8
Data Science with R
5
Hands-On
Doing it with Style
Function Call Layout
13. Similarly when we call the function. Once again, the prefix comma on the line is quite
convenient in allowing us to quickly comment out the whole line and retain correct syntax.
Interesting Option
dialPlot(label="UseR!"
, value=78
, dial.radius=1
, value.cex=3
, value.color="black"
, label.cex=3
, label.color="black")
Traditional
DR
AF
dialPlot(label="UseR!",
value=78,
dial.radius=1,
value.cex=3,
value.color="black",
label.cex=3,
label.color="black")
T
dialPlot(label="UseR!", value=78, dial.radius=1, value.cex=3,
value.color="black", label.cex=3, label.color="black")
Copyright
© 2013-2014 [email protected]
Module: StyleO
Page: 5 of 8
Data Science with R
6
Hands-On
Doing it with Style
Kuhn Checklist
14. Max Kuhn, author of caret (Kuhn et al., 2014) developed a checklist and posted it to the
R developers mailing list in January 2012. I have paraphrased some of the points here and
embellished it a little with my views, but they are quite in sync with Kuhn’s views.
(a) Extend the work of others and avoid redundancy. Reuse others functions, with due
credit, to add any missing features.
(b) For a categorical model builder ensure the target is a factor (like Yes/No) rather than
integers (like 1/0). The factor levels should be identified in the resulting model object
and the predict() function should return predicted classes as factors with the same
levels and ordering of levels. Support a type= to switch between predicted classes and
class probabilities. Use type="prob" for probabilities.
(c) Implement a separate predict(), using predict.class() where class is the class
of the object returned by the model builder. Do not use special functions like
modelPredict().
DR
AF
T
(d) Provide both a formula interface as in foo(y~x, data=ds) and non-formula interface
as in foo(x, y) to the function. “Formula methods are really inefficient at this time
for large dimensional data but are fantastically convenient. There are some good
reasons to not use formulas, such as functions that do not use a design matrix (e.g.,
cforest()) or need factors to be handled in a non-standard way (e.g., cubist()).”
(e) “Don’t require a test set when model building.”
(f) If not all variables are used in the resulting model, allow the required subset of variables to be provided for predict() and not all the original variables, and avoid referencing variables by position rather than name.
Copyright
© 2013-2014 Gr[email protected]
Module: StyleO
Page: 6 of 8
Data Science with R
7
Hands-On
Doing it with Style
Further Reading
The Rattle Book, published by Springer, provides a comprehensive
introduction to data mining and analytics using Rattle and R.
It is available from Amazon. Other documentation on a broader
selection of R topics of relevance to the data scientist is freely
available from http://datamining.togaware.com, including the
Datamining Desktop Survival Guide.
This chapter is one of many chapters available from http://
HandsOnDataScience.com. In particular follow the links on the
website with a * which indicates the generally more developed chapters.
I like the guidelines at Google but I have my own idiosyncrasies. The style decisions I have made
I motivate above, based on over 30 years of programming in very many different languages. Also
see Wikipedia for an excellent summary of many styles.
DR
AF
T
Rasmus B˚
a˚
ath, in The State of Naming Conventions in R, reviews naming conventions used in
R, finding that the initial lower case capitalised word scheme for functions was the most popular,
and dot separated names for arguments similarly. This is the style I prefer.
Copyright
© 2013-2014 [email protected]
Module: StyleO
Page: 7 of 8
Data Science with R
8
Hands-On
Doing it with Style
References
Kuhn M, Wing J, Weston S, Williams A, Keefer C, Engelhardt A, Cooper T, Mayer Z, Kenkel
B, the R Core Team, Benesty M (2014). caret: Classification and Regression Training. R
package version 6.0-37, URL http://CRAN.R-project.org/package=caret.
R Core Team (2014). R: A Language and Environment for Statistical Computing. R Foundation
for Statistical Computing, Vienna, Austria. URL http://www.R-project.org/.
Williams GJ (2009). “Rattle: A Data Mining GUI for R.” The R Journal, 1(2), 45–55. URL
http://journal.r-project.org/archive/2009-2/RJournal_2009-2_Williams.pdf.
Williams GJ (2011). Data Mining with Rattle and R: The art of excavating data for knowledge discovery. Use R! Springer, New York. URL http://www.amazon.com/gp/product/
1441998896/ref=as_li_qf_sp_asin_tl?ie=UTF8&tag=togaware-20&linkCode=as2&camp=
217145&creative=399373&creativeASIN=1441998896.
DR
AF
T
Williams GJ (2014). rattle: Graphical user interface for data mining in R. R package version
3.3.1, URL http://rattle.togaware.com/.
This document, sourced from StyleO.Rnw revision 544, was processed by KnitR version 1.8 of
2014-11-11 and took 7.7 seconds to process. It was generated by gjw on nyx running Ubuntu
14.04.1 LTS with Intel(R) Xeon(R) CPU W3520 @ 2.67GHz having 8 cores and 12.3GB of
RAM. It completed the processing 2014-12-08 20:38:37.
Copyright
© 2013-2014 [email protected]
Module: StyleO
Page: 8 of 8