Studying CRAN package names

[This article was first published on R and Finance, and kindly contributed to R-bloggers]. (You can report issue about the content on this page here)
Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.

Setting a name for a CRAN package is an intimate process. Out of an
infinite range of possibilities, an idea comes for a package and you
spend at least a couple of days writing up and testing your code before
submitting to CRAN. Once you set the name of the package, you cannot
change it. You choice index your effort and, it shouldn’t be a surprise
that the name of the package can improve its impact.

Looking at package
one strategy that I commonly observe is to use small words, a verb or
noun, and add the letter R to it. A good example is dplyr. Letter d
stands for dataframe, ply is just a tool, and R is, well, you know. In a
conventional sense, the name of this popular tool is informative and
easy to remember. As always, the extremes are never good. A couple of
bad examples of package naming are A3, AF, BB and so on. Googling
the package name is definitely not helpful. On the other end, package
samplesizelogisticcasecontrol provides a lot of information but it is
plain unattractive!

Another strategy that I also find interesting is developers using names
that, on first sight, are completely unrelated to the purpose of the
package. But, there is a not so obvious link. One example is package
sandwich. At first sight, I challenge anyone to figure out what it
does. This is an econometric package that computes robust standard
errors in a regression model. These robust estimates are also called
sandwich estimators because the formula looks like a
But, you only know that if you studied a bit of econometric theory. This
strategy works because it is easier to remember things that surprise us.
Another great example is package janitor. I’m sure you already
suspect that it has something do to with data cleaning. And you are
right! The message of the name is effortless and it works! The author
even got the privilege of using letter R in the name.

While I can always hand pick good and bad examples, let’s dig deeper. In
this post, we will study the names of packages available in CRAN by
comparing them to the whole English vocabulary. We are going use the
following datasets:

  • List of CRAN package, available with function
  • List of English words, available at WordNet
    This is a comprehensive database of English words that I once used
    in a
    It contains several tables, including all possible words from the
    English language.

First, let’s have a look at the distribution of size (number of
characters) for all packages available in CRAN:


# get data
df.pkgs <- = '')) %>%
  mutate(Package = as.character(Package),
         n.char = nchar(Package)) %>% 
  rename(pkg = Package) %>%
  select(pkg, n.char)

# plot it!
p <- ggplot(df.pkgs, aes(x=n.char)) +
  geom_histogram(binwidth = 1)

As I suspected, the names of CRAN packages are usually small, with an
average of 5-6 characters. We have a couple of packages with more than
25 characters. Let’s see their names:


## [1] "AnglerCreelSurveySimulation"   "FractalParameterEstimation"   
## [3] "ig.vancouver.2014.topcolour"   "RoughSetKnowledgeReduction"   
## [5] "samplesizelogisticcasecontrol"

I am sorry for the authors, but, in my opinion, I’m sure we could find
better names. I am also sorry for those who are using these packages but
do not use the autocomplete

of RStudio and need to type the loooooooooong names.

As for my hypothesis that CRAN package have short names, let’s compare
the distribution of package names against all words in the English
language. For that, let’s load the WordNet database and do some


# get data
conn <- dbConnect(drv = SQLite(), 'WordNet/sqlite-31.db')
words <- dbReadTable(conn, 'wordsXsensesXsynsets') %>%

# some are duplicate (same word, different types)
words <- unique(words)
words$nchar <- nchar(words$lemma)

# set df to plot <- data.frame(n.char = c(df.pkgs$n.char, words$nchar), 
                         source.char = c(rep('CRAN pkgs', nrow(df.pkgs)),
                                         rep('English Vocabulary', nrow(words))))

p <- ggplot(, aes(x=n.char, color=source.char )) +
  geom_density(size=1) + coord_cartesian(xlim=c(0, 40))


As I suspected, the distributions are very different. There is no need
to apply a formal test as the visual evidence is clear: CRAN package
have a tendency for shorter names.

Now, let’s look at the distribution of used letters in relative terms:


temp <- str_split(str_to_upper(df.pkgs$pkg), '')
all.chars <- = c,args = temp)
char.counts.pkg <- table(all.chars)

temp <- str_split(str_to_upper(words$lemma), '')
all.chars <- = c,args = temp)
char.counts.words <- table(all.chars) <- data.frame(perc.count = c(char.counts.pkg/sum(char.counts.pkg), 
                         char = c(names(char.counts.pkg),
                         source.char = c(rep('CRAN pkgs', length(char.counts.pkg)),
                                         rep('WordNet', length(char.counts.words))))

# only keep LETTERS
idx <-$char %in% LETTERS <-[idx, ]

p <- ggplot(, aes(x=char, y = perc.count, color=source.char,width=.5)) +
  geom_col(position = 'dodge') + scale_y_continuous(labels = percent_format())  


The result is really interesting! I was expecting far more differences
in the relative use of characters. Not surprisingly, letter R is more
used in package naming than in the English vocabulary. Still, the
difference is not that large. Given that R is the name of the
programming language, I was expecting a much greater proportion of R
characters. My intuition was clearly wrong! In comparison, letters P and
M have more difference in relative terms than letter R. I’m really not
sure why that is. Overall, it is pretty clear the use of characters in
the names of packages follow the distribution of words in the English

While the distribution of letter is similar, we find just a few package
with names exactly as in the English language. For all 10524 packages
found in CRAN, only 698 are an exact match of all 147478 unique words in
the English vocabulary. If we can’t match them all, let’s see how far
they are from the English dictionary. For that, we use package
stringdist to compute the minimum editing distance that we can find
for all package names with respect to the English vocabulary. In a
nutshell, the editing distance measures how many string modifications we
need in order for two strings to match each other. By computing the
minimum editing distance of package’s names against the English
vocabulary, we have a measure of equality. Here I’m using method='lv',
which seems to be the most appropriate in this study.

my.fct <- function(,possible.names ){
  #my.dist<- possible.names[which.min(stringdist(, possible.names ))]
  my.dist<- min(stringdist(, possible.names, method='lv'))
  #my.dist<- min(adist(, possible.names ))

char.distances <- pbapply::pbsapply(df.pkgs$pkg, FUN = my.fct, 

## Loading required package: stringdist

Let’s look at the results:

p <- ggplot(data.frame(char.distances), aes(x=char.distances))+
  geom_histogram(binwidth = 1) 


As we can see, most of packages names are just three or four edits away.
This shows how similar the choice of packages is to the English

Summing up, our data analysis shows that the names of packages are
usually shorter than the words in the English language. However, when
looking at distribution of used characters and editing distances, it is
pretty clear that the names are based on the English language, usually
with a few modifications of a base word.

I hope you enjoyed this post. In the next one I will explore the
package’s authors and the use of comments in R code.

To leave a comment for the author, please follow the link and comment on their blog: R and Finance. offers daily e-mail updates about R news and tutorials about learning R and many other topics. Click here if you're looking to post or find an R/data-science job.
Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.

Never miss an update!
Subscribe to R-bloggers to receive
e-mails with the latest R posts.
(You will not see this message again.)

Click here to close (This popup will not appear again)