← Blog

My journey to finding R modules

#R #trick

R is still my go-to language when it comes to data wrangling and analysis tasks. But I have long been frustrated that R does not offer a simple, built-in solution to use modules for organizing code into multiple files, each encapsulates functions/objects to serve a specific purpose. Python has it, JavaScript has it, and many other major languages have it. Why not R?

Yes, there is R package, but creating a package often takes significantly more effort than I’m willing to put even with all the support from devtools and dedicated R Studio features. I don’t want all that boilerplate just to organize some utility functions in separate files and use them without polluting my work environment—especially if they are meant specifically for the project at hand, not some general use cases.

Not finding any satisfying existing solution, I did what many others do in similar situations. I created one—not another package to install but a simple trick (maybe a hack) that relies solely on base R features.

In this post, I’m not going to explain in detail how my solution works. Instead, I’d like to keep a record of my journey to my current solution here.

Chapter one: What works works

My first few attempts at getting modules work involved immediately-invoked function expression, or IIFE, which evaluates what’s inside the function body and outputs the function return value only—all without polluting the global environment. IIFE in R looks like this:

(function() {

  # your code here

})()

Since R function returns the very last object for the evaluated function body, I had to come up with a way to pass multiple functions and objects as a single return value. I first used R environment as a container for functions and objects to export and later switched to using list, which seemed more beginner-friendly to me.

At this point, using a module looked something like this:

# module.R
(function() {
  f <- function() print("Huzzah!")
  list(
    f = f
  )
})()

# main.R
m <- source("module.R")$value

m$f()
#> [1] "Huzzah!"

The R module trick (or hack) was for my own use as much as (potentially) for others’, and mixing IIFE and list served my needs just fine for a while. Module scripts were not looking pretty with clunky IIFE syntax, but I was still able to break long R scripts into smaller pieces and use them without worrying about name collisions.

It worked. And that was enough at first.

Chapter two: For better look and feel

A few months later, however, the aesthetics started to bother me. source("module.R")$value just didn’t look right, and I found myself defining a short utility function like the following:

import_module <- function(path) source(path)$value
m <- import_module("module.R")

m$f()
#> [1] "Huzzah!"

At this point, I realized that I could just put this small function on my “R module trick” repo on GitHub. And while I’m at it, why not add some goodies to make it a bit fancier easier?

So I added attach and name parameters to my import_module(). Setting attach = TRUE (default) would put the module in the search path, making its contents accessible from the global environment. Alternatively, Setting attach = FALSE would assign the module to an object in the global environment. And name value, if provided, would be used instead of the default name, i.e. the module script filename. In addition, I included some error checking in import_module() to give warning messages for invalid parameter inputs as well as an informative message when the module was successfully loaded.

source("https://raw.githubusercontent.com/bobaekang/r-module-trick/master/import_module.R")
import_module("module.R", name = "m")
#> Note: 'module:m' now attached to the search path

f()
#> [1] "Huzzah!"

# alternatively ...
import_module("module.R", name = "m", attach = FALSE)
#> Note: 'm' now available in the global environment

m$f()
#> [1] "Huzzah!"

Hmm, that remote path to the “import_module.R” script seemed too long to my eye. So I created TinyURL for it: ”https://tinyurl.com/r-module.trick/<path>” now would redirect to ”https://raw.githubusercontent.com/bobaekang/r-module-trick/master/<path>“.

Finally, just to help others, like my future self, to get a quick reminder of how the function works, I added another function, import_module_help() to display documentation for import_module().

I realized that I created an entire user experience centered around using my trick. It both looked and felt good.

Chapter three: Less (trick) is more

Only a few days later, I found out about local(), which allows evaluating R expressions in a local environment. While experimenting with using local(), I also learned that source() has local argument, which could keep the evaluated results of source() in a local environment. Wait, I realized, I might not need that ugly IIFE hack after all!

Turns out, the combination of these new findings allowed me to keep the module script simple:

# module.R
f <- function() print("Huzzah!")

# export
list(
  f = f
)

Now that looks like a simple module script! The only requirement here is for the last line to export the list containing module elements.

There was still a trade-off. Now the ugliness was pushed to main.R that imports the module:

# main.R
m <- local(source("module.R", local = TRUE)$value)

m$f()
#> [1] "Huzzah!"

I still liked that the new approach was using one less trick. No IIFE, just normal function calls with what the base R offers. Plus, one can always make the ugly local(source(path, local = TRUE)$value) into a function:

# using DIY import_module
import_module <- function(path) local(source(path, local = TRUE)$value)
m <- import_module("module.R")

m$f()
#> [1] "Huzzah!"

And, of course, everyone is welcome to use my version of import_module() with a few extra features including name, attach and more, which is now living inside a new GitHub repo also with a friendly TinyURL path.

source("https://tinyurl.com/r-module/import_module.R")
import_module("module.R", name = "m")
#> Note: 'module:m' now attached to the search path

f()
#> [1] "Huzzah!"

* * *

So, there it is. My little journey to finding R modules.

Are you an R user? I cordially invite you to give R modules a try, leave a ⭐ if you liked, and open a Github Issue if you find a 🐞.