Utility functions for relaying stdout and conditions when using mclapply()/parLapply(), ...

[HenrikBengtsson]

I'm adding this one to this package, but it probably belongs to a 'parallel.extras' package:

Add some type of utility functions for relaying stdout and conditions when using mclapply(), parLapply(), ... of the 'parallel' package. Not sure what such an API would look like, but it might help us support 'progressr' also in those cases, because current the following will relay nothing from the cluster nodes:

library(progressr)
options(progressr.interval = 0.0, progressr.delay = 0.01)

cl <- parallel::makeCluster(3)
with_progress({
  p <- progressor(10)
  parallel::clusterExport(cl, c("p", "slow_sum"))
  y <- parallel::parLapply(cl, X = 1:10, function(x) {
    p()
    slow_sum(x, stdout=TRUE, message=TRUE)
  })
})
parallel::stopCluster(cl)

Maybe something that will allow us to do:

res <- parallel::parLapply(cl, X = 1:10, function(x) record_output({
  p()
  slow_sum(x, stdout=TRUE, message=TRUE)
}))
y <- value(res)  ## cf. how futures work

[HenrikBengtsson]

Value or values?

EDIT 2020-05-06: value() which is in line with how the future package now calls it.