The dependency network of all CRAN packages
In crandep: Network Analysis of Dependencies of CRAN Packages

In the introduction we have see that a dependency network can be built using get_dep(). While it is theoretically possible to use get_dep() iteratively to obtain all dependencies of all packages available on CRAN, it is not practical to do so. This package provides two functions get_dep_all_packages() and get_graph_all_packges() for obtaining the dependencies of all CRAN packages directly, as well as an example dataset.

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

library(crandep)
library(dplyr)
library(ggplot2)
library(igraph)
library(visNetwork)

All types of dependencies, in a data frame

The example dataset cran_dependencies contains all dependencies as of 2020-05-09.

data(cran_dependencies)
cran_dependencies
dplyr::count(cran_dependencies, type, reverse)

This is essentially a snapshot of CRAN. We can obtain all the current dependencies using get_dep_all_packages(), which requires no arguments:

df0.cran <- get_dep_all_packages()
head(df0.cran)
dplyr::count(df0.cran, type, reverse) # numbers in general larger than above

df1.cran <- df0.cran |> dplyr::count(from, type, reverse) |> dplyr::count(from)
v9.all <- dplyr::filter(df1.cran, n == 9L)$from
v0.all <- dplyr::filter(df1.cran, n == 10L)$from

As of r Sys.Date(), there are r length(v0.all) packages that have all 10 types of dependencies, and r length(v9.all) packages that have 9 types of dependencies: r paste(v9.all, collapse = ", ").

Network of one type of dependencies, as an igraph object

We can build dependency network using get_graph_all_packages(). Furthermore, we can verify that the forward and reverse dependency networks are (almost) the same, by looking at their size (number of edges) and order (number of nodes).

g0.depends <- get_graph_all_packages(type = "depends")
g0.depends

We could obtain essentially the same graph, but with the direction of the edges reversed, by using the argument reverse:

# Not run
g0.rev_depends <- get_graph_all_packages(type = "depends", reverse = TRUE)
g0.rev_depends

The dependency words accepted by the argument type is the same as in get_dep(). The two networks' size and order should be very close if not identical to each other. Because of the dependency direction, their edge lists should be the same but with the column names from and to swapped.

For verification, the exact same graphs can be obtained by filtering the data frame for the required dependency and applying df_to_graph():

g1.depends <- df0.cran |>
  dplyr::filter(type == "depends" & !reverse) |>
  df_to_graph(nodelist = dplyr::rename(df0.cran, name = from))
g1.depends # same as g0.depends

If we extract the equivalent graph of reverse dependencies, we should obtain the same graph as before (had it been extracted above):

# Not run
g1.rev_depends <- df0.cran |>
  dplyr::filter(type == "depends" & reverse) |>
  df_to_graph(nodelist = dplyr::rename(df0.cran, name = from))
g1.rev_depends # should be same as g0.rev_depends

The networks obtained above should all be directed acyclic graphs:

igraph::is_dag(g0.depends)
igraph::is_dag(g1.depends)

External reverse dependencies & defunct packages

One may notice that there are external reverse dependencies which won't be appear in the forward dependencies if the scraping is limited to CRAN packages. We can find these external reverse dependencies by nodelist = NULL in df_to_graph():

df1.rev_depends <- df0.cran |>
  dplyr::filter(type == "depends" & reverse) |>
  df_to_graph(nodelist = NULL, gc = FALSE) |>
  igraph::as_data_frame() # to obtain the edge list
df1.depends <- df0.cran |>
  dplyr::filter(type == "depends" & !reverse) |>
  df_to_graph(nodelist = NULL, gc = FALSE) |>
  igraph::as_data_frame()
dfa.diff.depends <- dplyr::anti_join(
  df1.rev_depends,
  df1.depends,
  c("from" = "to", "to" = "from")
)
head(dfa.diff.depends)

This means we are extracting the reverse dependencies of which the forward equivalents are not listed. The column to shows the packages external to CRAN. On the other hand, if we apply dplyr::anti_join() by switching the order of two edge lists,

dfb.diff.depends <- dplyr::anti_join(
  df1.depends,
  df1.rev_depends,
  c("from" = "to", "to" = "from")
)
head(dfb.diff.depends)

the column to lists those which are not on the page of available packages on CRAN (anymore). These are either defunct or core packages.

Summary statistics

Using the data frame df0.cran, we can also obtain the degree for each package and each type:

df0.summary <- dplyr::count(df0.cran, from, type, reverse)
head(df0.summary)

We can look at the "winner" in each of the reverse dependencies:

df0.summary |>
  dplyr::filter(reverse) |>
  dplyr::group_by(type) |>
  dplyr::top_n(1, n)

This is not surprising given the nature of each package. To take the summarisation one step further, we can obtain the frequencies of the degrees, and visualise the empirical degree distribution neatly on the log-log scale:

df1.summary <- df0.summary |>
  dplyr::count(type, reverse, n)
gg0.summary <- df1.summary |>
  dplyr::mutate(reverse = ifelse(reverse, "reverse", "forward")) |>
  ggplot2::ggplot() +
  ggplot2::geom_point(ggplot2::aes(n, nn)) +
  ggplot2::facet_grid(type ~ reverse) +
  ggplot2::scale_x_log10() +
  ggplot2::scale_y_log10() +
  ggplot2::labs(x = "Degree", y = "Number of packages") +
  ggplot2::theme_bw(20)
gg0.summary

This shows the reverse dependencies, in particular Reverse_depends and Reverse_imports, follow the power law, which is empirically observed in various academic fields.

Visualisation

We can now visualise (the giant component of) the CRAN network of Depends, using functions in the package visNetwork. To do this, we will need to convert the igraph object g0.depends to the node list and edge list as data frames.

prefix <- "http://CRAN.R-project.org/package=" # canonical form
degrees <- igraph::degree(g0.depends)
df0.nodes <- data.frame(id = names(degrees), value = degrees) |>
  dplyr::mutate(title = paste0('<a href=\"', prefix, id, '\">', id, '</a>'))
df0.edges <- igraph::as_data_frame(g0.depends, what = "edges")

We could use igraph::membership() & igraph::cluster_*() for community detection and visualisation of the clusters using different colours, which however will take too much computing time and therefore not shown here.

By adding the column title in df0.nodes, we enable clicking the nodes and being directed to their CRAN pages, in the interactive visualisation below:

set.seed(2345L)
vis0 <- visNetwork::visNetwork(df0.nodes, df0.edges, width = "100%", height = "720px") |>
  visNetwork::visOptions(highlightNearest = TRUE) |>
  visNetwork::visEdges(arrows = "to", color = list(opacity = 0.5)) |>
  visNetwork::visNodes(fixed = TRUE) |>
  visNetwork::visIgraphLayout(layout = "layout_with_drl")
vis0

Going forward

Methods in social network analysis, such as stochastic block models, can be applied to study the properties of the dependency network. Ideally, by analysing the dependencies of all CRAN packages, we can obtain a bird's-eye view of the ecosystem. The number of reverse dependencies is modelled in this other vignette.