Contributing `jlab-scaffold` - a more advanced cookiecutter

[daviseford]

Hi folks,

I'm Davis Ford, I'm a principal software engineer at Capital One's Center for Machine Learning.

I submitted this talk to JupyterCon 2020: Creating Extensions - Faster and have gotten a lot of great feedback from core developers. I've met with various core maintainers over the past few days, and met the team at their October 21 meeting, where I loosely proposed my idea.

We would like to contribute the lessons we've learned (and the code that comes along with it) to the Jupyter organization. We'd like to maintain and continue to contribute to the codebase in whatever capacity works best for the Jupyter org. I can tell you I am personally very excited by this project and I can anticipate a very high level of commitment from myself, and potentially other team members.

What is jlab-scaffold?

Much like create-react-app, jlab-scaffold is an CLI tool that asks the user a few questions about their project and then automatically spawns a new JupyterLab extension codebase.

Watch the talk to get a feel for it.

It's similar to cookiecutter in terms of functionality, but jlab-scaffold gets the end user way closer to production.

The automatically generated codebase is somewhat opinionated, but has a lightweight code footprint, so it's easy to tweak or make changes right out of the gate.

Features:

• React/ Typescript/Redux (via redux-toolkit)
• Automatic conda environment creation (uses mamba for faster env solving if installed)
• Helpful dev scripts (launches yarn watch and jupyter lab --watch in separate tabs with correct conda environments enabled)
• Well-typed React code with lots of documentation
• Sample React unit tests
• Server extension sample POST and GET requests
• Extremely thorough documentation
  • Best practices
  • Tips/warnings
  • Troubleshooting
  • Publishing tips (helps your devs get their packages published to pip and npm)
  • We even automatically generate a USER_GUIDE.md that you can give to your users, complete with all of the instructions they'll need to install your new extension.

jlab-scaffold is what we use at Capital One to roll out production-ready JupyterLab extensions as quickly as possible.

It takes about 5 minutes for a dev to go from 0 -> fully created dev environment. It's a godsend.

I look forward to discussing this idea with you all :) Thanks!

[trajamsmith]

+1 to Davis's excitement from our team at Capital One — I'd personally love to contribute more to this tool.

[saulshanabrook]

This sounds great! Is there a link to the library?

[echarles]

Thx @daviseford This looks promising and amazing.

I need to go AFK, but had a quick scan through your youtube video. I have seen stuff like jupyterContext, useJupyterlab, one way signals (is it lumino signal? are there 2 way signals`. Can you tell us a bit more here before we listen to your talk?

Is there a link to a jlab-scaffold repo already?

[tgeorgeux]

@daviseford I came here to ask if there is a repo already out somewhere we can look at, but it seems somebody else beat me to it 🤣

Thanks for submitting this I'd love to get a gander at the library.

[jasongrout]

This looks really great - I loved your JupyterCon talk!

Can we get the repo in the public first so we can see it, before deciding about adopting it as an official Project Jupyter repo?

[daviseford]

This sounds great! Is there a link to the library?

Unfortunately, no. It is currently an internal project and I'm restricted in what I can share.

Can we get the repo in the public first so we can see it

Working on it :)

[jasongrout]

Working on it :)

That's great! You can see there is a lot of interest in the community to see and help, and a possibility of adoption in the Jupyter org, if that helps make the case for open-sourcing it.

[manurampandit]

@daviseford - Curious to know if you were able to share this for public ?