Otago Study Group SYSKA

Recently I have been going through some old code that I needed to rerun. As part of doing so I was taking the opportunity to reimplement it with some new tricks that I have discovered since I wrote it ~2 years ago.

This particular script has many, many for loops which I now wanted to replace with purrr::map to make the code cleaner and easier to maintain.

As part of using purrr::map I wondered if I could also make the code go faster by parallelising the loop. The short answer is yes, with the furrr package, using furrr::future_map.

The long answer comes from https://byuistats.github.io/M335/parallel_furrr.html

but in simplicity

library(purrr)

purrr::map(list_object, function)

becomes:

library(furrr)
no_cores <- availableCores() - 1
plan(multicore, workers = no_cores)

furrr::future_map(list_object, function)

where each item of the list if processed in parallel