In reply to @raitobezarius:matrix.org
or not using nixpkgs

Any easy way of setting up an ad-hoc Python environment with a few packages on the command line with nix run

I'm able to get nix run nixpkgs#jupyter working but can't get NumPy into a kernel without writing a shell.nix or similar first. Was thinking nix run 'nixpkgs#python3.withPackages(ps: with ps; [torch])' could work but it doesn't seem to.

Guess this is one area where nix-shell is still better?

nix-shell -p 'python310.withPackages (p: [ p.numpy ])' does what I expect.

and you can use the flake registry entry like this, but it's more verbose:

nix shell --impure --expr '(builtins.getFlake "nixpkgs").legacyPackages.x86_64-linux.python3.withPackages (p: [p.numpy p.seaborn])'

maybe this is slightly better?

nix shell --impure --expr '(import (builtins.getFlake "nixpkgs") {}).python3.withPackages (p: [p.numpy])'

and you can use the flake registry entry like this, but it's more verbose:

nix shell --impure --expr '(builtins.getFlake "nixpkgs").legacyPackages.x86_64-linux.python3.withPackages (p: [p.numpy])'

maybe this is slightly better because you don't have to remember what system you're using?

nix shell --impure --expr '(import (builtins.getFlake "nixpkgs") {}).python3.withPackages (p: [p.numpy])'

In reply to @brodriguesco:matrix.org
fellow Nixers, I'm a beginner and started a series of blog posts that might interest some of you. It focuses on R but I believe a lot of what I describe could be used for any language