How to Write an R Package Wrapping a NodeJS Module

Mr A Bichat was looking for a way tobundle a NodeJS module inside n R package. Here is an attempt at a reproducible example, that might also help others!

About NodeJS Packages

There are two ways to install NodeJS packages: globally and locally.

The idea with local dependencies is that when writing your application or your script, you bundle inside one big folder everything needed to make that piece of JavaScript code run. That way, you can have various versions of a Node module on the same computer without one interfering with another. On a production server, that also means that when publishing your app, you don’t have to care about some global libraries,or about putting an app to prod with a module version that might break another application.

I love the way NodeJS allows to handle dependencies, but that’s the subject for another day.

Node JS inside an R package

To create an app or cli in NodeJS, you will be following these steps:

• Creating a new folder
• Inside this folder, run npm init -y (the -y pre-fills all thefields), which creates package.json
• Create a script (app.js, index.js, whatever.js) which will contain your JavaScript logic ; this file takes command lines arguments that will be processed inside the script
• Install external modules with npm install module, which add elements to package.json, and creates package-lock.json ; here, the whole module and its deps are downloaded and put inside anode_modules/ folder

Once your software is built, be it an app or a cli, you will be sharing to the world the package.json, package-lock.json, and all the files that are required to run the tool, but not the node_modules/ folder.

It can then be shared on npm, the Node package manager, or simply put on git, so that users git clone it, and install with npm installinside the folder.

Let’s create a small example:

cd /tmp
mkdir nodeexample
cd nodeexample
npm init -y


Wrote to /private/tmp/nodeexample/package.json:

{
  "name": "nodeexample",
  "version": "1.0.0",
  "main": "index.js",
  "scripts": {
    "test": "echo \"Error: no test specified\" && exit 1"
  },
  "keywords": [],
  "author": "",
  "license": "ISC",
  "dependencies": {
    "chalk": "^4.0.0"
  },
  "devDependencies": {},
  "description": ""
}


touch whatever.js
npm install chalk



npm WARN nodeexample@1.0.0 No description
npm WARN nodeexample@1.0.0 No repository field.

+ chalk@4.0.0
updated 1 package and audited 7 packages in 6.686s

1 package is looking for funding
  run `npm fund` for details

found 0 vulnerabilities


echo "const chalk = require('chalk');" >> whatever.js
echo "console.log(chalk.blue('Hello world'));" >> whatever.js
cat whatever.js


const chalk = require('chalk');
console.log(chalk.blue('Hello world'));

Now this can be run with Node:

node /tmp/nodeexample/whatever.js


Hello world

Here is our current file structure:

fs::dir_tree("/tmp/nodeexample", recurse= 1)


/tmp/nodeexample
└── node_modules
    ├── @types
    ├── ansi-styles
    ├── chalk
    ├── color-convert
    ├── color-name
    ├── has-flag
    └── supports-color

As you can see, you have a node_modules folder that contains all the modules, installed with the requirements of your machine.

Let’s now move this file to another folder (imagine it’s a git clone), where we won’t be sharing the node_modules folder: the users will have to install it to there machine.

mkdir /tmp/nodeexample2
mv /tmp/nodeexample/package-lock.json /tmp/nodeexample2/package-lock.json
mv /tmp/nodeexample/package.json /tmp/nodeexample2/package.json
mv /tmp/nodeexample/whatever.js /tmp/nodeexample2/whatever.js

But if we try to run this script:

node /tmp/nodeexample2/whatever.js


node /tmp/nodeexample2/whatever.js
internal/modules/cjs/loader.js:979
  throw err;
  ^

Error: Cannot find module 'chalk'
Require stack:
- /private/tmp/nodeexample2/whatever.js
    at Function.Module._resolveFilename (internal/modules/cjs/loader.js:976:15)
    at Function.Module._load (internal/modules/cjs/loader.js:859:27)
    at Module.require (internal/modules/cjs/loader.js:1036:19)
    at require (internal/modules/cjs/helpers.js:72:18)
    at Object.<anonymous> (/private/tmp/nodeexample2/whatever.js:1:15)
    at Module._compile (internal/modules/cjs/loader.js:1147:30)
    at Object.Module._extensions..js (internal/modules/cjs/loader.js:1167:10)
    at Module.load (internal/modules/cjs/loader.js:996:32)
    at Function.Module._load (internal/modules/cjs/loader.js:896:14)
    at Function.executeUserEntryPoint [as runMain] (internal/modules/run_main.js:71:12) {
  code: 'MODULE_NOT_FOUND',
  requireStack: ['/private/tmp/nodeexample2/whatever.js']
}

We have a “Module not found” error: that’s because we haven’t installed the dependencies yet.

cd /tmp/nodeexample2 && npm install


npm WARN nodeexample@1.0.0 No description
npm WARN nodeexample@1.0.0 No repository field.

added 7 packages from 4 contributors and audited 7 packages in 2.132s

2 packages are looking for funding
  run `npm fund` for details

found 0 vulnerabilities


fs::dir_tree("/tmp/nodeexample2", recurse= 1)


/tmp/nodeexample2
├── node_modules
│ ├── @types
│ ├── ansi-styles
│ ├── chalk
│ ├── color-convert
│ ├── color-name
│ ├── has-flag
│ └── supports-color
├── package-lock.json
├── package.json
└── whatever.js


cd /tmp/nodeexample2 && node whatever.js


Hello world

Tada 🎉!

Ok, but how can we bundle this into an R package? Here is what we will do:

• on our machine, we will create the full, working script into theinst/ folder of the package, and share everything but our node_modules folder
• Once the users install our package on their machine, they will have something that will look like the first version of our/tmp/nodeexample2 inside their package installation folder
• Then, from R, they will run an npm install inside the package installation folder, i.e inside system.file(package = "mypak")
• Once the installation is completed, we will call the script with the working directory being our installed package ; this script will take command line arguments passed from R

node-minify

While I’m at it, let’s try to use something that I might use in the future: node-minify, a node library which can minify CSS, notably through the clean-css extension:https://www.npmjs.com/package/@node-minify/clean-css.

If you don’t know what the minification is and what it’s used for, it’s the process of removing every unnecessary characters from a file so that it’s lighter. Because you know, on the web every byte counts.

See https://en.wikipedia.org/wiki/Minification_(programming) for more info.

Step 1, create the package

I won’t expand on that, please refer to online documentation.

Step 2, initiate npm infrastructure

Once in the package, here is the script to initiate everything:

mkdir -p inst/node
cd inst/node 
npm init -y
npm install @node-minify/core @node-minify/clean-css

touch app.js

This app.js will do one thing: take the path to a file and an output file, and then run the node-minify on this file.

Step 3, creating the NodeJS script

Here is app.js:

const compressor = require('node-minify');

compressor.minify({
  compressor: 'gcc',
  input: process.argv[2], // processing the script argument
  output: process.argv[3],
  callback: (err, min) => {} // not adding any callback but you should
});

Let’s now create a dummy css file:

echo "body {" >> test.css
echo " color:white;" >> test.css
echo "}" >> test.css

And it can be processed it:

node app.js test.css test2.css

Nice, we now have a script in inst/ that can be run with Node! How to make it available in R?

Step 4, building functions

Let’s start by ignoring the node_modules folder.

usethis::use_build_ignore("inst/node/node_modules/")

Then, create a function that will install the Node app on the users’machines, i.e where the package is installed.

minifyr_npm_install <- function(
  force = FALSE
){
  # Prompt the users unless they bypass (we're installing stuff on their machine)
  if (!force) {
    ok <- yesno::yesno("This will install our app on your local library.
                       Are you ok with that? ")
  } else {
    ok <- TRUE
  }

  # If user is ok, run npm install in the node folder in the package folder
  # We should also check that the infra is not already there
  if (ok){
    processx::run(
      command = "npm", 
      args = c("install"), 
      wd = system.file("node", package = "minifyr")
    )
  }
}

Let’s now build a function to run the minifyer:

minifyr_run <- function(
  input,
  output
){
  input <- path_abs(input)
  output <- path_abs(output)
  run(
    command = "node",
    args = c(
      "app.js",
      input,
      output
    ),
    wd = system.file("node", package = "minifyr")
  )
  return(output)
}

And here it is! With some extra package infrastructure, we’ve got everything we need :)

Step 5, try on our machine

Let’s run the build package on our machine:

# To do once
minifyr::minifyr_npm_install()

Then, if we have a look at our local package lib:

fs::dir_tree(
  system.file(
    "node",
    package = "minifyr"
  ), 
  recurse = FALSE
)


/Library/Frameworks/R.framework/Versions/3.6/Resources/library/minifyr/node
├── app.js
├── node_modules
├── package-lock.json
└── package.json

Let’s try our function:

# Dummy CSS creation
echo "body {" > test.css
echo " color:white;" >> test.css
echo "}" >> test.css
cat test.css


body {
  color:white;
}


minifyr::minifyr_run(
  "test.css", 
  "test2.css"
)


/Users/colin/Seafile/documents_colin/R/site/colinfaypointme/_posts/test2.css


cat test2.css


body{color:#fff}

Tada 🎉!

Result package at: https://github.com/ColinFay/minifyr

Step 6, one last thing

Of course, one cool thing would be to test that npm and Node are installed on the user machine. We can do that by running a dummy node command, and check if the result of system() is either 0 or 127, 127 meaning that the command failed to run.

node_available <- function(){
  test <- suppressWarnings(
    system(
      "npm -v",
      ignore.stdout = TRUE,
      ignore.stderr = TRUE
    )
  )
  attempt::warn_if(
    test, 
    ~ .x != 0, 
    "Error launching npm"
  )
    test <- suppressWarnings(
    system(
      "node -v",
      ignore.stdout = TRUE,
      ignore.stderr = TRUE
    )
  )
  attempt::message_if(
    test, 
    ~ .x != 0,
    "Error launching Node"
  )
}