Handling memory issues in R: terminal use, arguments, storing model objects by regular expressions

models.R

# HOW TO MANAGE MEMORY ISSUES IN R?

# A common problem of data- and computation-intensive projects
# in R is memory management.

# Suppose you would like to estimate a series of models,
# but estimating all of them would exceed your available
# memory.
#
# One solution could be to have individual R scripts
# for each model, and run them separately from the 
# terminal (e.g., using RScript script1.R).
#
# A disadvantage of this approach is that you'll 
# quickly end up with a zillion of script1.R-script1zillion.R,
# and you lose track about which model is the 
# "right model".
#
# So, how can we attain the benefits in memory handling
# from running "only 1 model", AND the benefits from
# having all of the models in one script?
#
# The solution is quite simple. We're gonna tell 
# our script WHICH models to run, directly from the
# terminal.
#
# For example, `Rscript models.R m1` will estimate
# model 1, and `Rscript models.R m1 m2` will estimate models
# 1-2.
#
# Here's the script!

# Read arguments from the command line (e.g., `m1`, `m2`, `m3`)`
args = commandArgs(trailingOnly=TRUE)

# Print arguments to the console, so you know they've been read correctly
print(args)

# MODEL 1

# run only if required
if ('m1' %in% args) {
  print('running m1')
  
  # not a model, I know... but this is where your model code is supposed to go
  m1 <- rnorm(100)
}


if ('m2' %in% args) {
  print('running m2')
  m2 <- rnorm(100)
}


if ('m3' %in% args) {
  print('running m3')
  m3 <- rnorm(100)
}

# Well, how to we save the objects that we've created, without
# a priori knowing their names?
# We can use the following function to scan the global environment for occurence of regular expression (`regex`), 
# and save all objects in `filename`.
save_by_regex <- function(regex, filename) {
  lscall = ls(envir=.GlobalEnv)
  stuff_to_save = grep(regex, lscall, value=T)
  if (length(stuff_to_save)>0) {
    cat('saving...\n')
    cat(paste0('(', paste0(stuff_to_save, collapse=', '), ')\n'))
    save(list=stuff_to_save , file = filename)
    cat('...done.\n') } else {
      cat('No objects to save. Verify regular expression.\n')
    }
}

# So... this one is going to save all objects starting with `m` (note it's appending those names to the result object)
save_by_regex('^m', paste0('results_', paste(args, collapse='_'),'.RData'))

# And... how to run it from the command line?
# Just go to the command line/terminal, and type

# `RScript models.R m1`
# `RScript models.R m1 m3`
# `RScript models.R m2`
# etc.