Hot reload/install

[pwaller]

At the moment while getting started on a project I'm finding I'm frequently missing a package I wish I had installed.

I have built up lots of state in jupyter and don't want to kill my kernel (or I have some very long running jobs I want to leave running).

How do I add something to my kernel.python.*.extraPackages and rebuild the environment, but keep my kernel? Is such a thing possible, or am I forced to quit the kernel and come back in? In the "pre nix" world I would pip instal something and then immediately import it but I'm guessing I can't quite do that here.

[djacu]

Hey @pwaller
I don't think such a thing is possible. When the kernel is created, it is given a unique path in the Nix store. When a package is add/removed/modified, a new path is created. This is necessary for reproducibility.

[GTrunSec]

if you want I can add an option of impure in the module to enable the mutable environment. I used to enable that option to debug a kernel or source of the kernel.

The following is the solution:

1. wrap the ~/.local/jupyter/kernels to jupyter path
2. run syncKernels to ~/local/jupyer/<name-impure> https://github.com/GTrunSec/DeSci/blob/main/nix/kernels/entrypoints/default.nix#L13
3. then reload that name of kernel

[GTrunSec]

This request will be solved in the refactor-stage2.