False positive "failed to locate global object" when using NSE

[DavisVaughan]

In the below example, I would not have expected any failures

library(globals)

fn <- function(expr) {
  expr <- substitute(expr)
  eval(expr, envir = mtcars)
}

fn(cyl)
#>  [1] 6 6 4 6 8 6 8 4 4 6 6 8 8 8 8 8 8 4 4 4 4 8 8 8 8 4 4 4 8 6 8 4

expr <- quote(fn(cyl))

globalsOf(expr)
#> Error in globalsByName(names, envir = envir, mustExist = mustExist): Identified global objects via static code inspection (fn(cyl)). Failed to locate global object in the relevant environments: 'cyl'

This is tough because you probably search for cyl in the function environment or the global environment, but it is actually evaluated in the context of mtcars through the eval()

This came up on SO as a furrr issue related to dplyr and rlang
https://stackoverflow.com/questions/64006043/using-tidy-evaluations-with-furrr

[DavisVaughan]

Actually it looks like future::getGlobalsAndPackages() is passing mustExist = FALSE down to globals, so in theory it should have "worked" already.

I think this is the true issue:

library(globals)

fn <- function(expr) {
  globalsByName("expr", envir = environment(), mustExist = FALSE)
}

fn(cyl)
#> Error in get(name, envir = env, inherits = FALSE): object 'cyl' not found

# Really `globalsByName()` does this, but `expr` shouldn't be evaluated here! It's lazy!
fn2 <- function(expr) {
  get("expr", envir = environment())
}

fn2(cyl)
#> Error in get("expr", envir = environment()): object 'cyl' not found

[DavisVaughan]

I am also now remembering that in furrr this is a "common gotcha". Each argument to the function you are going to parallelize has to be evaluated once so it can be shipped off to the worker.

So maybe this is something the user has to live with?

See my notes here https://davisvaughan.github.io/furrr/articles/articles/gotchas.html#non-standard-evaluation-of-arguments

I've added a second answer to the SO post
https://stackoverflow.com/a/64009850/7394743