Cannot render preview with Julia Kernel

[EvoArt]

Bug description

Hi, I am hitting this error, trying to render a .qmd file, using Quarto in PowerShell (same with VSCode) on Windows 10. I can use the julia-1.7 kernel in jupyter lab, no probs. And I can render a .qmd using the python3 kernel in quarto

 Starting julia-1.7 kernel...Traceback (most recent call last):
   File "C:\Program Files\Quarto\share\jupyter\jupyter.py", line 21, in <module>
     from notebook import notebook_execute, RestartKernel
   File "C:\Program Files\Quarto\share\jupyter\notebook.py", line 17, in <module>
     from nbclient import NotebookClient
   File "C:\Users\arn203\.julia\conda\3\lib\site-packages\nbclient\__init__.py", line 6, in <module>
     from .client import NotebookClient, execute  # noqa: F401
   File "C:\Users\arn203\.julia\conda\3\lib\site-packages\nbclient\client.py", line 14, in <module>
     from jupyter_client import KernelManager
   File "C:\Users\arn203\.julia\conda\3\lib\site-packages\jupyter_client\__init__.py", line 6, in <module>
     from .asynchronous import AsyncKernelClient  # noqa
   File "C:\Users\arn203\.julia\conda\3\lib\site-packages\jupyter_client\asynchronous\__init__.py", line 1, in <module>
     from .client import AsyncKernelClient  # noqa
   File "C:\Users\arn203\.julia\conda\3\lib\site-packages\jupyter_client\asynchronous\client.py", line 6, in <module>
     from jupyter_client.channels import HBChannel
   File "C:\Users\arn203\.julia\conda\3\lib\site-packages\jupyter_client\channels.py", line 12, in <module>
     import zmq.asyncio
   File "C:\Users\arn203\.julia\conda\3\lib\site-packages\zmq\__init__.py", line 103, in <module>
     from zmq import backend
   File "C:\Users\arn203\.julia\conda\3\lib\site-packages\zmq\backend\__init__.py", line 31, in <module>
     raise original_error from None
   File "C:\Users\arn203\.julia\conda\3\lib\site-packages\zmq\backend\__init__.py", line 26, in <module>
     _ns = select_backend(first)
   File "C:\Users\arn203\.julia\conda\3\lib\site-packages\zmq\backend\select.py", line 31, in select_backend
     mod = import_module(name)
   File "C:\Users\arn203\.julia\conda\3\lib\importlib\__init__.py", line 127, in import_module
     return _bootstrap._gcd_import(name[level:], package, level)
   File "C:\Users\arn203\.julia\conda\3\lib\site-packages\zmq\backend\cython\__init__.py", line 6, in <module>
     from . import (
 ImportError: DLL load failed while importing _device: The specified module could not be found.

I'm pretty lost with this at this point! Any help would be much appreciated.

Checklist

• Please include a minimal, fully reproducible example in a single .qmd file? Please provide the whole file rather than the snippet you believe is causing the issue.
• Please format your issue so it is easier for us to read the bug report.
• Please document the RStudio IDE version you're running (if applicable), by providing the value displayed in the "About RStudio" main menu dialog?
• Please document the operating system you're running. If on Linux, please provide the specific distribution.

[TS-CUBED]

This bug seems to be back with the new versions of Julia.

Anything newer than Julia 1.8.3 will not work.

Julia 1.8.4, 1.8.5, 1.9beta2, all fail with:

Starting julia-1.8 kernel...ERROR: 

Kernel died before replying to kernel_info

If I change the version number in ~/.local/share/jupyter/kernels/julia-1.8 from 1.8.4 to 1.8.3 it works.

All of the tested kernels work in all other instances where I use them:

• Jupter notebook starts them properly
• Emacs Babel (emacs-jupyter) starts the Jupyter kernel as expected.

So, only quarto seems to have this problem.

[cscheid]

@TS-CUBED I'm trying to understand this better. I can't reproduce it on my side:

---
format: html
jupyter: julia-1.8
---

```{julia}
versioninfo()
```

The suggestion you provided changes information in jupyter:

If I change the version number in ~/.local/share/jupyter/kernels/julia-1.8 from 1.8.4 to 1.8.3 it works.

I don't have a ~/.local/share/jupyter folder in my setup, but I suspect this is causing your system trouble.

[TS-CUBED]

Of course you can't reproduce it on a MacOS system, since it's a Linux problem ;-)

Although the OP had it on Windows.

The kernelspec info will be in a different place on a Mac, I guess, not in .local/share/jupyter

[TS-CUBED]

This problem has also been reported here:

https://discourse.julialang.org/t/quarto-julia-cli-kernel-died-before-replying-to-kernel-info/91416

quarto-dev/quarto-julia#3

#1652

It seems to be a flaky issue, as in it does show up on a variety of systems (Mac, Win, Linux) with various installations of Jupyter (built-in, packaged, pip, conda).

The strange thing is that is sounds like a Python issue, but only occurs in quarto, the respective kernels work flawless in other clients that connect to a Jupyter server:

• Jupiter itself (notebook and lab)
• VSCode
• Emacs with Babel/Org-Mode

Is there a way to make quarto output useful debugging info?

[TS-CUBED]

Running your file I get:

Julia Version 1.8.3
Commit 0434deb161e (2022-11-14 20:14 UTC)
Platform Info:
  OS: Linux (x86_64-linux-gnu)
  CPU: 8 × Intel(R) Core(TM) i7-4770 CPU @ 3.40GHz
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-13.0.1 (ORCJIT, haswell)
  Threads: 4 on 8 virtual cores
Environment:
  JULIA_NUM_THREADS = 4

Envious of your machine, btw ;-)

[TS-CUBED]

Kernel setup Jupyter (this is what ] build IJulia creates in .local/share/jupyter/kernels/julia-1.8/kernel.json):

{
  "display_name": "Julia 1.8.5",
  "argv": [
    "/home/thor/.julia/juliaup/julia-1.8.5+0.x64.linux.gnu/bin/julia",
    "-i",
    "--color=yes",
    "--project=@.",
    "/home/thor/.julia/packages/IJulia/6TIq1/src/kernel.jl",
    "{connection_file}"
  ],
  "language": "julia",
  "env": {},
  "interrupt_mode": "signal"
}

Results in:

Starting julia-1.8 kernel...ERROR: 

Kernel died before replying to kernel_info

Process quarto-preview-/home/thor/test.qmd exited abnormally with code 1

When I install the 1.8.3 kernel:

{
  "display_name": "Julia 1.8.3",
  "argv": [
    "/home/thor/.julia/juliaup/julia-1.8.3+0.x64.linux.gnu/bin/julia",
    "-i",
    "--color=yes",
    "--project=@.",
    "/home/thor/.julia/packages/IJulia/6TIq1/src/kernel.jl",
    "{connection_file}"
  ],
  "language": "julia",
  "env": {},
  "interrupt_mode": "signal"
}

I get:

Starting julia-1.8 kernel...Done

Executing 'test.ipynb'
  Cell 1/1...Done

pandoc 
  to: html
  output-file: test.html
  standalone: true
  section-divs: true
  html-math-method: mathjax
  wrap: none
  default-image-extension: png
  
metadata
  document-css: false
  link-citations: true
  date-format: long
  lang: en
  jupyter: julia-1.8
  
Output created: test.html

Watching files for changes
Browse at http://localhost:5920/

and

versioninfo()

Julia Version 1.8.3
Commit 0434deb161e (2022-11-14 20:14 UTC)
Platform Info:
  OS: Linux (x86_64-linux-gnu)
  CPU: 8 × Intel(R) Core(TM) i7-4770 CPU @ 3.40GHz
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-13.0.1 (ORCJIT, haswell)
  Threads: 4 on 8 virtual cores
Environment:
  JULIA_NUM_THREADS = 4

[prosoitos]

Having the same issue (also on Linux), so I can confirm that it is reproducible.

Julia Version 1.8.4
Commit 00177ebc4fc (2022-12-23 21:32 UTC)
Platform Info:
  OS: Linux (x86_64-linux-gnu)
  CPU: 16 × Intel(R) Core(TM) i7-10875H CPU @ 2.30GHz
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-13.0.1 (ORCJIT, skylake)
  Threads: 1 on 16 virtual cores

[prosoitos]

In the Discourse post, the user cannot render. I can render (although it takes a surprisingly long time for the Julia kernel to start). What leads to the Kernel died before replying to kernel_info in my case is when I try to preview.

[TS-CUBED]

To rule out any hangup from my system config, I have just set up a new installation of Ubuntu 22.04.1 LTS

After the installation:

• Install jupyter:

sudo apt install python3-pip curl
python3 -m pip install jupyter

• Install Julia

curl -fsSL https://install.julialang.org | sh

• Install quarto-1.2.313-linux-amd64.deb from Quarto website.

This will install Julia 1.8.5.

In Julia REPL:

] add IJulia
] build IJulia

After this a quarto render of the file @cscheid put up (MWE for a Julia .qmd) gives the Kernel died before replying ... error.

---

julia add 1.8.3
julia +1.8.3

] add IJulia
] build IJulia

will install the 1.8.3 version as the Julia-1.8 kernel in Jupyter, after which Quarto renders the file without a problem.

---

Also tested on the new Ubuntu installation:

Kernel 1.8.5 and 1.9beta2 both work in the Jupyter notebook itself.

[prosoitos]

If I change the version number in ~/.local/share/jupyter/kernels/julia-1.8 from 1.8.4 to 1.8.3 it works.

In my ~/.local/share/jupyter/kernels/julia-1.8, I actually have 1.8.3, which is puzzling since I am using 1.8.4. So I must be using a different Jupyter installation.

Note that I did not install Python and Jupyter through Conda.jl.

[TS-CUBED]

In the Discourse post, the user cannot render. I can render (although it takes a surprisingly long time for the Julia kernel to start). What leads to the Kernel died before replying to kernel_info in my case is when I try to preview.

Interesting. I can't render either.

In my ~/.local/share/jupyter/kernels/julia-1.8, I actually have 1.8.3, which is puzzling since I am using 1.8.4. So I must be using a different Jupyter installation.

It's quite easy to get out of sync between the Julia version you use in the REPL and the one that Jupyter uses. In order to udate the version Jupyter uses you need to do

] add IJulia
] build IJulia

or it will continue to use the old one.