Scoping issue with future.callr and tar_make_future(): variables declared in _targets.r can conflict with functions from imported packages

[mkoohafkan]

Originally posted in ropensci/targets#761

In this example, variables defined in _targets.r conflict with identically-named functions imported by packages specified via tar_option_set(). The following example reproduces the error with plan(callr) and tar_make_future(). The issue does not occur when calling tar_make() or using plan(sequential). In this example, a variable year declared in _targets.r conflicts with the function lubridate::year().

_targets.r

library(targets)
library(tarchetypes)
library(future)
library(future.callr)

source("R/test.r")

plan(callr)

tar_option_set(packages = c("lubridate"))

year = 2022
categories = data.frame(category = LETTERS[1:5])


tar_map(
  values = categories,
  names = "category",
  tar_target("test", test_fun(category, year),
    iteration = "list")
)

R/test.r

test_fun = function(category, input_year) {
  data.frame(
    chosen_category = rep(category, 4),
    chosen_year = rep(input_year, 4),
    value = rnorm(4)
  )
}

console

targets::tar_make_future(workers = 2)
## start target test_A
## start target test_B
## error target test_A
## end pipeline
## Error in tar_throw_run(target$metrics$error) :
##   attempt to replicate an object of type 'closure'
## Error in `tar_throw_run()`:
## ! callr subprocess failed: attempt to replicate an object of type 'closure'
## Visit https://books.ropensci.org/targets/debugging.html for debugging advice.
## Run `rlang::last_error()` to see where the error occurred.

[franzbischoff]

Just crossed with this by chance...
I think this may have some relation with tidyevaluation. As the variable year is global, it may collide with the function name. Try to use dplyr::all_of(year) instead and see if it works.

BR.