Parallelize 'foreach' functions

The CRAN 'foreach' package + The 'futurize' hexlogo = The 'future' logo

The futurize package allows you to easily turn sequential code into parallel code by piping the sequential code to the futurize() function. Easy!

TL;DR

library(futurize)
plan(multisession)
library(foreach)

slow_fcn <- function(x) {
  message("x = ", x)
  Sys.sleep(0.1)  # emulate work
  x^2
}

xs <- 1:10
ys <- foreach(x = xs) %do% slow_fcn(x) |> futurize()

Introduction

This vignette demonstrates how to use this approach to parallelize functions such as foreach() and times() of the foreach package. For example, consider:

library(foreach)
xs <- 1:1000
ys <- foreach(x = xs) %do% slow_fcn(x)

This foreach() construct is resolved sequentially. We can use the futurize package to tell foreach to hand over the orchestration of parallel tasks to futureverse. All we need to do is to pass the expression to futurize() as in:

library(foreach)

library(futurize)
plan(multisession) ## parallelize on local machine

xs <- 1:1000
ys <- foreach(x = xs) %do% slow_fcn(x) |> futurize()
#> x = 1
#> x = 2
#> x = 3
#> ...
#> x = 10

Note how messages produced on parallel workers are relayed as-is back to the main R session as they complete. Not only messages, but also warnings and other types of conditions are relayed back as-is. Likewise, standard output produced by cat(), print(), str(), and so on is relayed in the same way. This is a unique feature of Futureverse - other parallel frameworks in R, such as parallel, foreach with doParallel, and BiocParallel, silently drop standard output, messages, and warnings produced on workers. With futurize, your code behaves the same whether it runs sequentially or in parallel: nothing is lost in translation.

The built-in multisession backend parallelizes on your local computer and it works on all operating systems. There are other parallel backends to choose from, including alternatives to parallelize locally as well as distributed across remote machines, e.g.

plan(future.mirai::mirai_multisession)

and

plan(future.batchtools::batchtools_slurm)

Here is another example that parallelizes times() of the foreach package via the futureverse ecosystem:

library(foreach)
library(futurize)
ys <- times(10) %do% rnorm(3) |> futurize()

Supported Functions

The futurize() function supports parallelization of the following foreach functions:

Progress Reporting via progressr

For progress reporting, please see the [progressr] package. It is specially designed to work with the Futureverse ecosystem and provide progress updates from parallelized computations in a near-live fashion. See the vignette("futurize-11-apply", package = "futurize") for more details and an example.