Fix actions for Julia

[Robinlovelace]

Upstream issue that may be relevant: quarto-dev/quarto-cli#10837

Any ideas @evetion? Hard to debug at this stage..

[cderv]

Upstream issue that may be relevant: quarto-dev/quarto-cli#10837

Just a heads up that this upstream issue is not relevant as your document is not using engine: julia from Quarto.

You are running Julia cell in a document with Python and R too. This means that by default engine: knitr will be used by Quarto (because of R cell presence).

It implies

• Python cells are run through reticulate
• Julia celles are run through JuliaCall by default.

You need to be sure both are correctly setup regarding how to find Julia. Especiall JuliaCall needs to work specifically with the R version through RCall. It may be interested to add a configure step in your workflow where R would call julia <- julia_setup() to be sure it finds all the necessary tool.

Hope it helps

[Robinlovelace]

Thanks @cderv that is v. helpful, will give it a go..

[Robinlovelace]

Trying using the Julia kernel in #7 https://quarto.org/docs/computations/julia.html

[cderv]

I don't know exactly what you are trying to do here but whatever the solution, you need to do the correct setup.

• If using knitr, I would look at julia_setup() R function (https://non-contradiction.github.io/JuliaCall/reference/julia_setup.html) and call it once before doing quarto render (or call it in a setup chunk) so that configuration is done ok. Especially julia_setup(rebuild = TRUE) may be needed, or JULIA_HOME needs to be set.

  • Doing this in a specific CI step with R will give your confirmation that R is correctly setup to call Julia. This is prerequisite for knitr to call it, and so for Quarto to work.

• If Using jupyter to use Julia, you need to setup the kernels. We document this at https://quarto.org/docs/computations/julia.html#installation. This is required for Jupyter to find the right kernal with JUpyter which will call Julia. Then Julia package needs to be available in the right environment.

• If Using quarto engine: julia, there is no jupyter involved. Julia needs to be installed, and JUlia environment project needs to be configured at the Quarto project level. The julia engine will call julia --project=@. and so use the closest Project.toml, and assumes good practice to have an environment at project level.

I hope this helps clarify. Note that you can't mix 3 options in one document. Either engine: knitr or engine: <julia-kernel-name> or engine: julia.

[Robinlovelace]

Thanks for this @cderv. FYI @evetion has it working on mac os actions: https://github.com/evetion/cross_language_projects

Seems to us that maybe there's something about R/Julia interop that fails on Linux, probably not Quarto's fault, will be good to properly diagnose and with Julia people like Marten onboard we have hopes of doing that.

[Robinlovelace]

Results looking good, rendering R, Julia, Python versions, go Quarto! https://www.evetion.nl/cross_language_projects/demo.html

Will be really good to get this working on Linux one day, if anyone has any ideas how let us know.

[cderv]

Seems to us that maybe there's something about R/Julia interop that fails on Linux, probably not Quarto's fault, will be good to properly diagnose and with Julia people like Marten onboard we have hopes of doing that.

Oh this reminds me of

• Issue with julia_setup on Ubuntu 22.10 JuliaInterop/JuliaCall#192
  and Quarto report
• julia chunk errors when using knitr quarto-dev/quarto-cli#7752

So possibly this is the problem.

This new package https://github.com/stefan-m-lenz/JuliaConnectoR could be a better alternative, but we still need to add a knitr engine to it (stefan-m-lenz/JuliaConnectoR#28)

[Robinlovelace]

This sounds really great, thanks for sharing, following with interest!