[Bug]: Including matplotlib in 'extraPackages' leads to infinite recursion

[collinarnett]

Current Behavior

{
  name,
  availableKernels,
  extraArgs,
}:
availableKernels.python {
  name = "jupyterWith-testing"; # must be unique
  displayName = "jupyterWith-testing";
  extraPackages = ps: [ ps.numpy ps.matplotlib ];
}

adding matplotlib to the example in the tutorial

leads to

$ nix run
warning: Using saved setting for 'extra-substituters = https://tweag-jupyter.cachix.org' from ~/.local/share/nix/trusted-settings.json.
warning: Using saved setting for 'extra-trusted-public-keys = tweag-jupyter.cachix.org-1:UtNH4Zs6hVUFpFBTLaA4ejYavPo5EFFqgd7G7FxGW9g=' from ~/.local/share/nix/trusted-settings.json.
warning: creating lock file '/home/collin/projects/jupyterWith-test/flake.lock'
error: infinite recursion encountered

       at /nix/store/fr68dr66gmggn62jir7vp1bcnignwxp9-source/pkgs/stdenv/generic/make-derivation.nix:311:7:

          310|       depsBuildBuild              = lib.elemAt (lib.elemAt dependencies 0) 0;
          311|       nativeBuildInputs           = lib.elemAt (lib.elemAt dependencies 0) 1;
             |       ^
          312|       depsBuildTarget             = lib.elemAt (lib.elemAt dependencies 0) 2;
(use '--show-trace' to show detailed location information)

Expected Behavior

Jupyter notebook with matplotlib and numpy available.

Steps To Reproduce

Add ps.matplotlib to extraPackages

Version

$ nix-shell -p nix-info --run "nix-info -m
"
 - system: `"x86_64-linux"`
 - host os: `Linux 5.15.59, NixOS, 22.11 (Raccoon), 22.11.20221024.afd4752`
 - multi-user?: `yes`
 - sandbox: `yes`
 - version: `nix-env (Nix) 2.11.0`
 - channels(collin): `""`
 - channels(root): `"nixos-21.11.335130.386234e2a61"`
 - nixpkgs: `/nix/store/msv8qjxnga8h81biz33gpa4pyw9kg0j1-source

Additional Context

This is probably related to nix-community/poetry2nix#750. I'm not sure how I feel about having poetry2nix under jupyterWith. In the past I've had many issues using poetry2nix. Trying to diagnose nix issues + poetry issues is a real headache with nix's non-descriptive error messages. The previous jupyterWith was convenient because I could just specify what packages I wanted without caring about what versions I wanted. I understand the documentation is still under construction so maybe there's a simple method to use jupyterWith like I did previously.

Relevant log output

No response

[djacu]

Hey 👋
Thanks for the bug report!

We noticed there were some issues recently and figure it is because we had not updated the top level flake.lock file in a while. We are actively working on updating all the kernels so they work again. You can see the progress in this branch:
https://github.com/tweag/jupyterWith/tree/update_flake_lock_action

I also had trouble getting matplotlib working on the new branch and using extraPackages. As you can see from this output.

$ nix run
error: attribute 'format' missing

       at /nix/store/cc1lfn4pzkl0b4g13qr943ddv8w8lsrs-source/pkgs/development/tools/poetry2nix/poetry2nix/overrides/default.nix:1463:40:

         1462|             ++ lib.optionals (self.isPyPy) [ tk xorg.libX11 ];
         1463|           preConfigure = lib.optional (old.format != "wheel") preConfigure;
             |                                        ^
         1464|         }
(use '--show-trace' to show detailed location information)

However, I was able to get it working using the other method mentioned in the TUTORIALS file.

kernels/mypython/default.nix

{
  name,
  availableKernels,
  extraArgs,
}:
availableKernels.python {
  projectDir = ./.;
  displayName = "my-python";
  name = "my-python";
}

kernels/mypython/pyproject.toml

[tool.poetry]
name = "my-python-kernel"
version = "0.1.0"
description = ""
authors = []

[tool.poetry.dependencies]
python = "^3.9"
ipykernel = "*"
numpy = "*"
scipy = "*"
matplotlib = "*"

[tool.poetry.dev-dependencies]

[build-system]
requires = ["poetry-core>=1.0.0"]
build-backend = "poetry.core.masonry.api"

I was able to use all 3 packages (numpy, scipy, and matplotlib) in a notebook successfully.

[djacu]

This should be fixed with #415.
A new example kernel was added to the CI that pulls in numpy, scipy, and matplotlib via extraPackages. All seems to be working now.