Invalid flagging of faulty ... usage (part 2)

[DavisVaughan]

Follow up of #62, after receiving this comment #62 (comment) reproduced below:

library(purrr)
library(future)

catcher <- function(x) {
  map(x, ~list(...))
}

catcher(list(1:2, 1:3))
#> [[1]]
#> [[1]][[1]]
#> [1] 1 2
#> 
#> 
#> [[2]]
#> [[2]][[1]]
#> [1] 1 2 3

future({
  catcher(list(1:2, 1:3))
})
#> Error in getGlobalsAndPackages(expr, envir = envir, tweak = tweakExpression, : Did you mean to create the future within a function?  Invalid future expression tries to use global '...' variables that do not exist: {; catcher(list(1:2, 1:3)); }

^{Created on 2021-01-12 by the reprex package (v0.3.0.9001)}

It looks like the dots are being returned as an NA that gets wrapped with the DotDotDotList class.
a05c813#diff-7b1d4e2796df08171808d5d5341fadf8c78a22cc2ee1c2874cc4d3ebb2504dcbR233

But then in future::getGlobalsAndPackages(), it checks to make sure that the dots are a list, and errors otherwise (as we see above) https://github.com/HenrikBengtsson/future/blob/0755d6f30219f596f27f948a1e1df1a6ddc71ed0/R/globals.R#L174

Maybe instead of NA, it should be setting ddd to list() in some cases like this one? That is my initial guess.

[HenrikBengtsson]

Thxs. Finally got around to this one. I decided to fix this for now in future, cf. HenrikBengtsson/future@89d28d8.

I think the original reason to return an structure(NA, class = "DotDotDotList") was to be conservative and catch cases like this that I couldn't foresee at the time. I think the proper, long-term fix is to not return a globals[["..."]] element at all, because it's not a global. I'll create a separate issue for that.