Bioconductor downloads and dependencies

downloadsbydeps.R

library(igraph)
library(BiocPkgTools)

d <- biocDownloadStats()

pkgs <- biocPkgList()

whmaintainerpkgs <- grep("maintainer", pkgs$Maintainer)
coreteam <- c("interdonato", "michafla", "jmacdon",
              "morgan", "andrzej", "pages",
              "ramos", "shepherd", "grimbough",
              "turaga", "VanTwisk", "jwang96")
whcoreteampkgs <- grep(paste(coreteam, collapse="|"), pkgs$Maintainer)
corepkgs1 <- pkgs$Package[unique(c(whmaintainerpkgs, whcoreteampkgs))]

biocviews <- sapply(pkgs$biocViews, paste, collapse="_")
whcorepkgs2 <- grep("DataRepresentation", biocviews)

corepkgs <- unique(c(corepkgs1, corepkgs2))

dep_df <- buildPkgDependencyDataFrame(repo="BioCsoft",
                                      dependencies=c("Depends", "Imports"))
g <- buildPkgDependencyIgraph(dep_df)

## add vertices for Bioconductor packages without dependencies
g <- g + vertices(pkgs$Package[!pkgs$Package %in% names(V(g))])

## exclude packages outside Bioconductor
excludedpkgs <- names(V(g))[!names(V(g)) %in% pkgs$Package]
g <- induced_subgraph(g, setdiff(names(V(g)), excludedpkgs))

## define non-core Bioconductor packages
noncorepkgs <- setdiff(pkgs$Package, corepkgs)

## calculate for non-core Bioconductor packages their
## average monthly download in the last 12 months, the total
## number of dependences and the number dependences to
## "core infrastructure packages"
res <- data.frame(Downloads=integer(length(noncorepkgs)),
                  Ndeps=integer(length(noncorepkgs)),
                  Ncoredeps=integer(length(noncorepkgs)),
                  row.names=noncorepkgs, check.names=FALSE)
for (p in noncorepkgs) {

  ## fetch dependences
  deps <- setdiff(names(subcomponent(g, p, mode="out")), p)

  ## fetch number of dependences to "core infrastructure packages"
  ncoredeps <- sum(deps %in% corepkgs)

  ## fetch median number of downloads through the last 12 months
  d.pkg <- d[d$Package %in% p & d$Month != "all", ]
  lastfullmonth <- Sys.Date() - as.POSIXlt(Sys.Date())$mday
  mask <- d.pkg$repo == "Software" &
          d.pkg$Date >= (lastfullmonth-365) & d.pkg$Date <= lastfullmonth
  d.pkg <- d.pkg$Nb_of_distinct_IPs[mask]
  d.pkg <- d.pkg[d.pkg > 0]
  if (length(d.pkg) == 12)
    res[p, ] <- c(median(d.pkg), length(deps), ncoredeps)
  else ## if a package has no download data for the last 12 months set NA
    res[p, ] <- c(NA, length(deps), ncoredeps)
}

## discard new packages with no download data for the last 12 months
res <- res[!is.na(res$Downloads), ]

saveRDS(res, file="downloadsbydeps.rds")

ncd <- cut(res$Ncoredeps, breaks=c(0, 1, max(res$Ncoredeps)),
           right=FALSE, include.lowest=TRUE)
plot(log10(res$Downloads) ~ ncd, xlab="Number of core dependences",
     ylab="log10 Monthly downloads", las=1)
points(1:2+0.1, tapply(log10(res$Downloads), ncd, mean), pch=23, bg="black")