Useful R CMD envvars? about callr HOT 6 CLOSED

Hmmm. You can already pass env vars, and .libPaths() is also transferred,.

There is also rcmd_safe(), which does set most (all?) of these env vars:

❯ callr::rcmd_safe
function (cmd, cmdargs = character(), libpath = .libPaths(),
    repos = c(getOption("repos"), c(CRAN = "https://cran.rstudio.com")),
    system_profile = FALSE, user_profile = FALSE, env = c(CYGWIN = "nodosfilewarning",
        R_TESTS = "", R_BROWSER = "false", R_PDFVIEWER = "false"), ...)

from callr.

I guess the only thing that's missing is the NOT_CRAN env var. How would you feel about having callr set that? (I think it would be best in rcmd).

It still might be worthwhile to pull out c(CYGWIN = "nodosfilewarning", R_TESTS = "", R_BROWSER = "false", R_PDFVIEWER = "false") into its own function so that you can easily add to the env vars set by rcmd_safe()

from callr.

Yeah, I can add NOT_CRAN. I would add it to rcmd_safe, though.

rcmd is "run this in a replica environment"

rcmd_safe is "run this in a safe environment"

Pulling it out is a good idea, too.

from callr.

It might be worth documenting rcmd and rcmd_safe in the same file to make the connection more clear.

from callr.

Also, why isn't safety the default? i.e. why isn't it (e.g.) rcmd() and rcmd_unsafe() (or maybe rcmd_mimic())

from callr.

Just question. Could be.

At the time I wrote it, I needed the mimic behavior more than the safe one, especially for r().
Not that there are huge differences between rcmd() and rcmd_safe() currently....

But this is already on CRAN, so.... although I don't think many people use it.

from callr.