Using build tools to run tar_make

[jmbuhr]

I noticed that in the targets tutorial you make heavy use of roxygen comments for your functions and the folder contains a DESCRIPTION file much like a package. So I was wondering if we could take this a step further (and maybe this is a neat trick to add to the documentation):
If you include a function:

.onLoad <- function(libname, pkgname) {
  targets::tar_make()
}

RStudio's build panel and the shortcut Ctrl+Shift+B not only generate the documentation for the functions (which can be viewed in RStudios help panel), it also runs the targets pipeline!

[wlandau]

Good thoughts, but the trouble is that tar_vistnetwork(), tar_watch(), etc. need to load the project's functions without running tar_make(). Also, devtools::load_all() has gotchas when it comes to targets workflows. For one, tar_option_get("envir") (the environment where targets looks for global functions and other objects) is the global environment by default, but devtools::load_all() puts these objects in a package environment. In addition, if you call devtools::load_all() in _targets.R, then parallel workers in tar_make_clustermq() and tar_make_future() cannot access the package contents. If you implement a workflow as an R package, it is best to fully install the package and supply it to the packages and imports options as described at https://books.ropensci.org/targets/practices.html#packages-based-invalidation.

[wlandau]

Related: #374, #375