rclnodeJs installation broken: "Could not locate the bindings file"

Description We’ve been transitioning an electron app that uses roslibjs over to a ros2 interface. But, currently pulling our hair out trying to get things playing nicely.

Here’s an MVP application (https://github.com/nckswt/ros2-electron-app-mvp) to recreate the issue. The main issue we’re facing is this error:

Uncaught Error: Could not locate the bindings file. Tried:
 → /home/nick/ros2-electron-app-mvp/build/rclnodejs.node
 → /home/nick/ros2-electron-app-mvp/build/Debug/rclnodejs.node
 → /home/nick/ros2-electron-app-mvp/build/Release/rclnodejs.node
 → /home/nick/ros2-electron-app-mvp/out/Debug/rclnodejs.node
 → /home/nick/ros2-electron-app-mvp/Debug/rclnodejs.node
 → /home/nick/ros2-electron-app-mvp/out/Release/rclnodejs.node
 → /home/nick/ros2-electron-app-mvp/Release/rclnodejs.node
 → /home/nick/ros2-electron-app-mvp/build/default/rclnodejs.node
 → /home/nick/ros2-electron-app-mvp/compiled/14.16.0/linux/x64/rclnodejs.node
 → /home/nick/ros2-electron-app-mvp/addon-build/release/install-root/rclnodejs.node
 → /home/nick/ros2-electron-app-mvp/addon-build/debug/install-root/rclnodejs.node
 → /home/nick/ros2-electron-app-mvp/addon-build/default/install-root/rclnodejs.node
 → /home/nick/ros2-electron-app-mvp/lib/binding/node-v87-linux-x64/rclnodejs.node
    at bindings (bindings.js:126)
    at Object../node_modules/rclnodejs/lib/clock.js (clock.js:17)
    at __webpack_require__ (bootstrap:24)
    at fn (hot module replacement:61)
    at Object../node_modules/rclnodejs/index.js (index.js:17)
    at __webpack_require__ (bootstrap:24)
    at fn (hot module replacement:61)
    at Object../src/index.js (index.js:11)
    at __webpack_require__ (bootstrap:24)
    at startup:6

This PR makes me think we should be able to load rclnodejs into the electron renderer.

• Library Version: 0.20.1
• ROS Version: galactic
• Platform / OS: Ubuntu 20.04.3 LTS
• Node version: v12.22.7

Steps To Reproduce

git clone [e-mail]:nckswt/ros2-electron-app-mvp.git
npm i
npm run start:dev

Expected Behavior

I’d expect the node to run without crashing.

Actual Behavior See error above.

Wondering if there’s something I need to do to generate the bindings file? Or if there’s an incompatibility with my NodeJS setup somehow? Or if it’s even an upstream issue?

Please let me know if there’s more info I can provide for debugging.

[anmilleriii]

I apologize but I was not able to spend the time to debug an ENOENT for ../app.asar/build/preload.js error. This error is related to the webpack.prd.config.js, package.json, and the location of preload.js.

While I wanted to quickly solve the npm run start:dist, instead I’ve provided some high-level comments which provide context on whether a preload.js script should be used at all, among other items which I hope you find helpful.

Comments

• extraResources can be helpful for keeping some production files writeable (e.g., a database), but as you have identified, it is not appropriate to include unbundled node_modules/. I personally have not encountered an issue including rclnodejs in the AppImage as well as in development - did you trace the error directly to rclnodejs being missing or was it the preload.js error that I alluded to above (which would make rclnodejs undefined in production) that stopped you? Also see this issue.
• By convention build/ is not generated but rather for assets to be included in the build (such as logos, etc.)
• By convention public/ is solely for the static files through which the rest of the application is mounted (i.e., usually just an index.html file and favicon.icon)
• Electron main AND renderer source code should be in src/. While the structure is up to you, often people use src/main for the main process file (i.e., main.js) and src/renderer for the renderer process - i.e., your React app source code. The big idea is simply that source code doesn’t live in public/.
• Restructuring your project to the above will require a few changes to your existing Webpack configurations which I didn’t have the time to look through, but I think you will find these changes to adhere to conventions worthwhile.
• As a design-level data point, running rclnodejs in the view (e.g., React app) renderer process is not necessary to distribute it with Electron. You can have multiple “renderer” (or more aptly, “worker”) Electron proccesses which run in Node and communicate with the view over WebSocket connection, ipcRenderer, or other communication bridge. Our team found this encapsulation of rclnodejs to be simpler to work with, since it decouples rclnodejs and ROS from your view layer. See electron-ros below for more details.
• My advice to use a preload.js isn’t a best practice with nodeIntegration enabled, since the purpose of a preload script is to use Node modules in the browser with nodeIntegration disabled. Was just a quick fix to run the project as-is.
• You may already be aware of this, but in production the end-user device must have ROS installed and you will need a launch script which sources ROS to run the AppImage in production (since you won’t have an interactive terminal to do it). See this issue. My personal solution to this is documented here

Resources

As I probably won’t be able to provide too much of a further look, hopefully these are helpful:

• electron-react-boilerplate - Boilerplate for Electron/React. Not suggesting you use this, but you might find a few useful structural ideas from it.
• electron-ros repository - Sample implementation of rclnodejs in Electron which covers many rclnodejs specific issues.
• Foxglove Studio repository - Example of a very well-structured Electron app with self-managed Webpack configuration. Well beyond the scope of your requirements probably, but studying this may be helpful, as it has been for me.

Lastly @wayneparrott if @nckswt is okay with it, I might suggest closing this issue or at least removing the bug label, as the original bindings file issue is more of a limitation of rclnodejs integration with Webpack than a bug with rclnodejs itself. Once again sorry I couldn’t immediately fix the npm run start:dist issue.

[wayneparrott]

I was hoping someone with more electron experience than me would pick up this issue. I’ll give it a go tomorrow and see what I learn. @nckswt Thx for the mvp. Please share any new info you may have on the matter.