Get all your packages back on R 4.0.0

3 min read

R 4.0.0 was released on 2020-04-24. Among the many news two stand out for me: First, R now uses stringsAsFactors = FALSE by default, which is especially welcome when reading in data (e.g., via read.csv) and when constructing data.frames. The second news that caught my eye was that all packages need to be reinstalled under the new version.

This can be rather cumbersome if you have collected a large number of packages on your machine while using R 3.6.x and you don’t want to spend the next weeks running into Error in library(x) : there is no package called ‘x’ errors. But there is an easy way to solve this.

After you made the update, first get your old packages:

old_packages <- installed.packages(lib.loc = "/home/johannes/R/x86_64-pc-linux-gnu-library/3.6/")
head(old_packages[, 1])
##       abind     acepack        ade4         AER   animation   anomalize 
##     "abind"   "acepack"      "ade4"       "AER" "animation" "anomalize"

lib.loc should be the location you installed the packages before updating to R 4.0.0. If unsure, you can call .libPaths(). The first path is your new lib.loc and the previous one should look the same except with 3.6 at the end.

Then you can find the packages previously installed but currently missing:

new_packages <- installed.packages()
missing_df <- as.data.frame(old_packages[
  !old_packages[, "Package"] %in% new_packages[, "Package"], 
  ])

missing_df now contains all packages you had previously installed but are not present now. In an intermediate step you might want to clean up this list a bit, in case not all of the old packages are still useful to you (I just used write.csv to export it, annotated the list and read it back in with read.csv).

Once this is done, you can install your packages back:

install.packages(missing_df$Package, Ncpus = 3)

Even on three cores at the same time, this can run fo a while…

After the installations are done, you can check the missing packages again:

missing_df <- as.data.frame(old_packages[
  !old_packages[, 1] %in% installed.packages()[, 1], 
  ])

If you’ve got all your packages back, missing_df should have zero rows. If not, you might have had some packages which are not currently on CRAN. For me those are usually packages only available on GitHub so far. I used a nice little piece of code I found in the available package to find the repositories of these packages:

library(dplyr, warn.conflicts = FALSE)
on_gh <- function(pkg) {
  repo = jsonlite::fromJSON(paste0("http://rpkg-api.gepuro.net/rpkg?q=", pkg))
  repo[basename(repo$pkg_name) == pkg,]
}
gh_pkgs <- lapply(c("quanteda.classifiers", "emo"), on_gh) %>% 
  bind_rows()
as_tibble(gh_pkgs)
## # A tibble: 2 x 3
##   pkg_name            title                           url                       
##   <chr>               <chr>                           <chr>                     
## 1 quanteda/quanteda.… quanteda textmodel extensions … https://github.com/quante…
## 2 hadley/emo          Easily insert emoji into R and… https://github.com/hadley…

Check if this grabbed the correct ones, then you can install them using remotes::install_github(gh_pkgs$pkg_name).

For me, that was it. Your mileage may vary though if some of your packages were removed from CRAN in the meantime or if you use other repos like, for example, Bioconductor.

R

Related