BUG: The 'where' attribute may pick up the calling environment if global is NULL

[HenrikBengtsson]

From DavisVaughan/furrr#168 (comment):

fn <- function() {
  x <- list(null = NULL, one = 1)
  globals::as.Globals(x)
}

fn()
#> $null
#> NULL
#> 
#> $one
#> [1] 1
#> 
#> attr(,"where")
#> attr(,"where")$null
#> <environment: 0x7fa5ad858e70>
#> 
#> attr(,"where")$one
#> <environment: R_EmptyEnv>
#> 
#> attr(,"class")
#> [1] "Globals" "list"

This is because:

environment(NULL)

returns the calling environment, not NULL, or emptyenv().

/cc @DavisVaughan, oh my. I, either missed that very important troubleshooting of yours, or incorrectly thought it was fixed from quickly skimming the code. So, the "fix" in futureverse/future#475, ended up with the whole thing about dropping where and the emergency back-out future.globals.keepWhere solution, which in turn gives rise to futureverse/future#608. Crossing my finger that fixing this bug here will allow me to withdraw dropping where. It's clear that I have way too much on my plate at times and missing the important details served to me.

[DavisVaughan]

will allow me to withdraw dropping where.

That would be great!

I have way too much on my plate at times

Never feel like you ever have to rush to answer any of my bug reports. I have all the time in the world to wait, and would much rather have you working at your best! It sounds to me like you need to take a break and go paragliding again 😆

[HenrikBengtsson]

Thanks. Trust me, if the all these storms that been coming in this year would stop for a while, I'd been in the sky right now ... and there's no future up there - only now