Reproduce Serial Loop Output with Default RNGkind

[keatonwilson]

Hi there - thanks for all the work on this package! I'm working on adapting some legacy code using base R loops (and the default RNG type) to have parallel implementation (solid speed increases). Unfortunately, I need to replicate the outputs from previous runs of the code, and can't just swap over to L'Ecuyer-CMRG. From the documentation, it looks like it may be possible to replicate the old code by providing seeds explicitly:

.options.RNG with complete sequence of seeds: the complete description of the sequence of seeds to be used may be passed via options.RNG, as a list or a matrix
with the seeds in columns. This is useful to seed a loop exactly as desired, e.g. using
an RNG other than "L’Ecuyer-CMRG", or using different RNG kinds in each iteration, which probably have different seed length, in order to compare their stochastic
properties. It also allows to reproduce %dorng% loops without knowing their seeding
details:

But I can't seem to figure out the appropriate way to do this. Some example code:

library(doParallel)
library(foreach)
library(doSNOW)
library(doRNG)

# Traditional Seed and For Loop Approach ----------------------------------

set.seed(42)
out_1 <- list()

for(i in 1:3) {
  out_1[[i]] <- runif(3)
}

# foreach in serial -------------------------------------------------------

set.seed(42)

out_2 <- foreach(i = 1:3) %dopar% {
  runif(3)
}

identical(out_1, out_2)

# foreach in parallel -----------------------------------------------------

cluster <- snow::makeCluster(2)
doSNOW::registerDoSNOW(cluster)

# this is where I tried manually generating a seed matrix/list, but couldn't come 
# up with something that generated matching output

out_3 <- foreach(i = 1:3, options.RNG = manual_list) %dorng% {
  runif(3)
}

snow::stopCluster(cluster)
identical(out_1, out_3[1:3])

Any help you could provide would be amazing - and again, thanks so much for this package (and for the great vignette!).