testing parallel dplyr::do

test.R

library("nycflights13")
library("dplyr") 
library("mgcv")

by_dest <- flights %>% group_by(dest) %>% filter(n() > 100)
  
system.time(
by_dest %>% do(smooth = gam(arr_delay ~ s(dep_time) + month, data = .))
)

init_cluster(parallel::detectCores())
system.time(
  by_dest %>% do(smooth = gam(arr_delay ~ s(dep_time) + month, data = .))
)
stop_cluster()

Note: dplyr::init_cluster() and friends now appear in the documentation: http://rpackages.ianhowson.com/cran/dplyr/man/dplyr-cluster.html but don't seem to be implemented in the do method at this time.