I am using the multicore package in R for parallelizing my code. However, if the tcltk package is loaded, forking processes with the multicore package will cause R to hang i
Another way to avoid loading a particular package as a dependency is, based on the assumption that none of the functions you require depend on that package, would be to reference the functions you need using their namespace:
lattice::xyplot(1~1)
This way, you don't need to load the package with your function, and you don't inadvertently load the problem package.
If immediately detaching the package after it's been attached is a good enough solution, then try something like the following:
setHook(hookName = packageEvent("tcltk", "attach"),
value = function(...) detach(package:tcltk))
# Try it out
library(tcltk)
# Loading Tcl/Tk interface ... done
# Error in as.environment(pos) : invalid 'pos' argument
search()
# [1] ".GlobalEnv" "package:graphics" "package:grDevices"
# [4] "package:utils" "package:datasets" "package:methods"
# [7] "Autoloads" "package:base"
If (as seems likely) the very act of loading & attaching the package is causing the problem, you might also pursue a strategy like the one sketched out in the comments to your question. Namely:
"C:/R/Library/dummy/"
..libPaths
by executing .libPaths(c("C:/R/Library/dummy/", .libPaths()))
.Then, if any package attempts to load tcltk, it will first look for packages in "C:/R/Library/dummy/"
, and, finding one of that name, will load it for a moment (before it's immediately detached by the hook described above).