Revolving errors in qmd with robservable and {ojs} cells

[allisonhorst]

Hi everybody! [UPDATE: solved - using both robservable and {ojs} cells in a .qmd created conflicts!]

In a Quarto doc I'm trying to show different ways that folks can pull in content from Observable notebooks to a Quarto doc, or work in {ojs} cells. Repo: https://github.com/allisonhorst/robservable-testing

Sometimes it shows up perfectly, but more often I get errors throughout - and the error messages will sometimes change when I refresh the page. For example, below are four different outcomes for the same cell, only refreshing the page between:

My coworker pointed me to this discussion (which is a bit over my head, to be honest): https://talk.observablehq.com/t/quarto-module-require/6573 (UPDATE): They also shared an idea: "robservable conflicts with the quarto-ojs-runtime because they both package d3-require" -- so this could be an issue with robservable?

Any ideas / recommendations for making this thing work consistently? Am I missing something important that I need in my Quarto doc? Thank you so much!

I'm using:

• RStudio Version 2023.05.0-daily+313 (2023.05.0-daily+313)
• Quarto Version 1.3.340

[allisonhorst]

OK some follow up: once I broke the notebook up with the robservable example separate from the {ojs} examples, then this works great! So it does seem to be a conflict between the two. I'm gonna close this and follow up with the robservable developers.

[cscheid]

I don't know how robservable works, but I would be very surprised if it interacted well (or at all) with our implementation!

They also shared an idea: "robservable conflicts with the quarto-ojs-runtime because they both package d3-require" -- so this could be an issue with robservable?

This is true, and in fact worse than you might expect: in order for our implementation to work well with jupyter widgets, we run a fork of d3-require. So it's almost certainly the case that if you run both robservable and quarto-ojs-runtime on the same page, they're going to step on each other's toes :(

Are there robservable features that don't exist in our ojs engine support that you'd need to use?

I should note that if you use robservable by itself on a page without an ojs code cell, then things should work, because we only include quarto-ojs-runtime on pages we're positive require quarto-ojs-runtime.

[allisonhorst]

Thanks for following up, I really appreciate it! This is really helpful.

I don't think I have any real-world need for robservable + ojs in a single .qmd except for this one case where I was trying to show the different ways that people can work between Observable and Quarto -- but splitting it up works fine!

Thanks again for all of your work on Quarto and the ojs engine, it's really awesome and I appreciate your thoughtful response.

[allisonhorst]

Also just in case it's of interest, my awesome coworker Fabian shared this notebook and said it might be a useful alternate workaround because it "prevents another instance from grabbing window.define": https://observablehq.com/@mootari/require-conflict

[cscheid]

I'm going to have to study that notebook more carefully! Our issue is not (only, maybe?) with multiple instances of d3-require, but with how d3-require doesn't play nicely with Jupyter widget's expectation of the js environment. Our specific fork is here but because it's deep inside a bunch of git subtree nonsense, it's hard to diff against the upstream, I apologize.