Arch Linux User Repository

The kernel here is very outdated. To update, you also need to update jupyter-metakernel. For the octave_kernel, I have an udpated PKGBUILD here: https://gitlab.com/three-comrades/arch-aur/blob/master/jupyter-octave_kernel/PKGBUILD

@latjv
I'm not quite sure, but according to my (previous) experience, you can manually add a kernel spec file:

Create file `$HOME/.local/share/jupyter/kernels/octave/kernel.json`, and add the following:

{"argv": ["/usr/bin/python", "-m", "octave_kernel", "-f", "{connection_file}"], "display_name": "Octave", "language": "octave", "mimetype": "text/x-octave", "name": "octave"}

[vincent@vincent-static ~]$ yaourt octave_kernel
1 aur/jupyter-octave_kernel 0.19.9-2 [installed] (2) (1.25)

[vincent@vincent-static ~]$ jupyter-kernelspec list
Available kernels:
python3 /usr/lib/python3.6/site-packages/ipykernel/resources
julia-0.5 /home/vincent/.local/share/jupyter/kernels/julia-0.5

How can I help fix this?

@xantares
Hmm, interesting point. I never realise that.
Maybe you are right as I uninstalled python-oct2py and the notebook still works.

However, since I never use it anymore, I'm not sure about everything... Until today have I realised the package (on 0.19.*) seems not to function well...
Anyway, the dependency will be removed.

About package name, I'm not sure which is better.
I myself think naming it as 'jupyter-*' would be better because it is an extension to Jupyter, not to Python (i.e. a library); and there are many packages with naming schema 'jupyter-*' in AUR.

Are there any perticular reasons for a Python2 version (since jupyter itself depends on python3 stuff)? This kernel should have nothing to do with Python itself, except for pip to install (and maybe there is a way not to use pip, which I don't know)...

also I think it should be named python-octave_kernel
could you provide a python2 split package too ?

hi,
are you sure it depends python-oct2py ?
imho it's the other way around
xan.

Hi @bachp ,

Thanks for mentioning, I wasn't aware metakernel is also a dependency.

As for 'gnuplot', yes it should be added. (I though it was already added but I was wrong)
However, I was always hesitating whether to count it as a (hard) dependency or an optional dependency. Octave and this kernel seem to work well without gnuplot (without the ability to make plots, which is quite useful [and maybe important], I see).

I had to add `jupyter-metakernel` and `gnuplot` to make it work.