On Sun, Jul 25, 2021 at 12:15 AM Steve Haroz <steve.ha...@gmail.com> wrote: > > Hello, > > I'd like to propose moving the default library install location on Windows > from: > %USERPROFILE%/Documents/R > to some other location such as: > %USERPROFILE%/R > > For many users the Documents folder is backed up or synchronized. > Installing libraries thrashes Documents, and it causes synchronization > issues with Dropbox (I confirm this one), OneDrive, and users with > Network IT policies. > > The vast majority of R users won't touch that folder and don't need it > backed up. And, its contents are not really "documents".
I very much support this request. R's behavior of defaulting HOME to an unusual (legacy) directory is also a source of bugs and interoperability problems: R packages that interface c/c++ libraries that require a user config file (libssh, libgit2, etc) sometimes cannot find the user config because it gets stored on a different location than expected. A good example is 'git'. The 'git for windows' command line utility stores my user config in C:\users\Jeroen\.gitconfig. However, for R packages that link to libgit2 (e.g. git2r, gert) the user config gets loaded/stored from C:\users\Jeroen\Documents\.gitconfig because libgit2 inherits the R home directory. Having 2 different git configurations is obviously confusing for users. Another problem is that when you start R from "windows bash" or the msys2 shell (including "git for windows" or rtools40), the "~" directory is different than in RGui. This is because these shells do set the HOME variable (to the usual value), and R inherits that. For example: when I start R in RGui for Windows I see: > normalizePath("~") [1] "C:\\Users\\jeroen\\Documents" But when I start R or Rterm from the "git for windows" shell we get: > normalizePath("~") [1] "C:\\Users\\jeroen" An ambiguous home directory is obviously going to break R code that refers to paths in your home. Because these problems affect all Windows users, I think the benefits of fixing the default behavior easily outweigh the backward-compatibility and "there is a documented workaround" arguments. ______________________________________________ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel