Skip to contents

Prepare Packages for Parallel Processing with Future

Source: R/general_load_packages_future.R
load_packages_future.Rd

Prepares a list of packages for use in parallel processing with the future package, determining whether to load packages in the main process or pass them to parallel workers based on the specified future strategy. This function is designed to minimize package-loading messages in SLURM environments, especially for multicore.

Usage

load_packages_future(packages = character(), strategy = "multisession")

Arguments

packages

Character vector of package names to load, or NULL to indicate no packages are needed (returns NULL).

strategy

Character. The parallel processing strategy to use. Valid options are "sequential", "multisession" (default), "multicore", and "cluster". See future::plan().

Value

A value depending on strategy:

  • sequential: NULL (no workers; packages not loaded).

  • multicore (non-Windows): NULL (packages loaded in the main process, inherited by forks).

  • multicore (Windows), multisession, or cluster: packages (character vector of package names to load in workers, e.g., via future.packages).

  • If packages is NULL: NULL (no packages to load).

Details

This function helps manage package loading for parallel processing with the future package. It ensures efficient package handling and minimizes package-loading messages in SLURM environments, particularly for multicore on non-Windows systems, where packages are loaded in the main process to avoid redundant messages in worker forks. For multisession, cluster, or multicore on Windows (where multicore falls back to multisession), it returns the package names for loading in workers, typically via the future.packages argument in functions like future.apply::future_lapply().

Author

Ahmed El-Gabbas

Examples

(pkg_init <- loaded_packages())
#>  [1] "tidyterra" "tidyr"     "qs2"       "arrow"     "car"       "carData"  
#>  [7] "rworldmap" "raster"    "sp"        "purrr"     "terra"     "fs"       
#> [13] "tibble"    "png"       "sf"        "ggplot2"   "dplyr"     "ecokit"   
#> [19] "magrittr"  "stats"     "graphics"  "grDevices" "utils"     "datasets" 
#> [25] "methods"   "base"     
pkg_to_load <- c("tidyterra", "lubridate", "tidyr", "sf", "scales")

# sequential
load_packages_future(pkg_to_load, "sequential")
#> NULL
setdiff(loaded_packages(), pkg_init)
#> character(0)

# multisession
load_packages_future(pkg_to_load, "multisession")
#> [1] "tidyterra" "lubridate" "tidyr"     "sf"        "scales"   
setdiff(loaded_packages(), pkg_init)
#> character(0)

# multicore
load_packages_future(pkg_to_load, "multicore")
#> NULL
setdiff(loaded_packages(), pkg_init)
#> [1] "scales"    "lubridate"