Globals bug using future_map

[abdellah19jan]

Hi,

I didn't receive a response to my last post in HenrikBengtsson/globals#77 since 28 days, if you could give me a solution to work around the problem (see code below), it will break my deadlock!!!

future_lapply didn't solve the problem.

Thanks in advance!

library(future.apply)
library(furrr)
library(tidyverse)

plan(multisession, workers = 2)

fun <- function(arg) arg + 1
dic <- c("mod" = "fun")

#=================================================#
# REFERRING TO THE FUNCTION AS A CHARACTER STRING #
#=================================================#

f <- function(x) invoke_map(dic[[x]], 2)

map("mod", f)
#> [[1]]
#> [[1]][[1]]
#> [1] 3

future_map("mod", f)
#> Error in fun(2) : could not find function "fun"

future_lapply("mod", f)
#> Error in fun(2) : could not find function "fun"

#===========================================#
# PASSING THE ACTUAL FUNCTION TO invoke_map #
#===========================================#

f <- function(x) invoke_map(eval(parse(text = dic[[x]])), 2)

map("mod", f)
#> [[1]]
#> [[1]][[1]]
#> [1] 3

future_map("mod", f)
#> Error in eval(parse(text = dic[[x]])) : object 'fun' not found

future_lapply("mod", f)
#> Error in eval(parse(text = dic[[x]])) : object 'fun' not found

[DavisVaughan]

Make the values of your dic dictionary the actual function object, rather than just a character string of their name. That ensures that globals can find them.

library(furrr)
#> Loading required package: future
library(purrr)

fun <- function(arg) arg + 1
dic <- c("mod" = fun) # <- CHANGE IS HERE

f <- function(x) invoke_map(dic[[x]], 2)

map("mod", f)
#> [[1]]
#> [[1]][[1]]
#> [1] 3

plan(multisession, workers = 2)

future_map("mod", f)
#> [[1]]
#> [[1]][[1]]
#> [1] 3

plan(sequential)

^{Created on 2021-11-24 by the reprex package (v2.0.1)}