What R package for phylogenetics is the most popular?

While writing my first R package and its associated manuscript, I needed to talk about some other R packages in the phylogenetics research community. The most obvious choice would be to just cite the ones that I actually use, but that doesn’t necessarily mean that other practicing phylogeneticists do the same. I needed to get some stats on which phylogenetics packages were actually popular, but luckily the R ecosystem has the tools to make this easy. Click here to skip straight to the popularity table.

Instructions

First, let’s install the CRAN Task Views package and the CRAN-logs API package. We’ll use the development version of cranlogs since it hasn’t been updated on CRAN in a while and some stuff has changed.

install.packages("ctv")
devtools::install_github("metacran/cranlogs")
library(ctv)
library(cranlogs)

We can get a list of all the CRAN task views using the available.views function. Annoyingly, there’s no way to filter and extract JUST the Phylogenetics task view, so we’ll have to write a short filter to extract it.¹

available.views()
phylo_ctv <- Filter(function(x) x$name == “Phylogenetics”, available.views())[[1]]

Now we can extract the list of packages that are associated with the “Phylogenetics” task view, and using that list of packages, query the CRAN-logs server to figure out the most popular phylogenetics packages in the last year.

phylo_ctv$packagelist
phylo_packages <- phylo_ctv$packagelist$name
output <- cran_downloads(to = “2018-10-01”, from = “2017-10-01”, packages = phylo_packages)
head(output)

Of course, what about the phylogenetics packages that aren’t in the Phylogenetics task view? Another way to view it is to consider if a package depends on the ape package. Pretty much every phylogenetics package will use ape in some form or another, so it might also be a good proxy of what a “phylo” package is.

Get a list of all the reverse dependencies of ape using devtools.²

revdep_packages <- devtools::revdep("ape")
all_phylo_packages <- unique(c(phylo_packages, revdep_packages))
output2 <- cran_downloads(to = "2018-10-01", from = "2017-10-01", packages = all_phylo_packages)

By default, the output from cranlogs is the number of downloads for a given package on a given date. We want to sum up all of these counts so we have a total number of downloads per package.

Aggregate these data using dplyr.

library(dplyr)
output2 %>% group_by(package) %>% summarise(downloads = sum(count)) %>% arrange(-downloads)

Exercises

1. Filter the table to only include packages in both the CRAN task view and reverse dependencies list. (This will exclude e.g., ggplot2 and other arguably-peripheral packages.)

2. Use the lubridate package to find the ten most popular packages by year. (The CRAN logs go back to October 2012.)

Table

Here’s the full version of the table. There’s some packages in here that are only peripherally associated with phylogenetics, but it gives a good picture of what the state of the field looks like. I’ve also annotated each package with which list it came from, the CRAN Task View list or the reverse dependencies list.

1. Note that we can’t use the typical filtering mechanism using the single square bracket [ because of the way lists work. In particular, there’s no good destructuring syntax for lists-of-lists as there are for simple vectors. See ?Extract for more details. ↩

2. The builtin package tools also has it, but it only returns packages that you have currently installed. ↩