Kernel specs with custom env variables not listed in list of kernel specs

[DonJayamanne]

• Assume you have a conda env abc
• Create a kernel spec in the above conda env named xyz
  • This should be a kernel spec with custom env variables
  • This should be located in the env folder & not globally
• We do not list this kernel spec

[DonJayamanne]

@IanMatthewHuff weird it works for you, this doesn't work for me in stable or insiders.
This is the issue I was referring to today.

[IanMatthewHuff]

@DonJayamanne Just took a look. In my setup I used a conda env to install kernelspecs globally not locally in the env folder. That was fixed for me a release or two ago, and has been ok since then.

[rchiodo]

This is working for me now.

[roblourens]

Anything to verify here?

[roblourens]

This should be a kernel spec with custom env variables

If this should be verified with the original steps, how do I do that?

[rchiodo]

Kernelspecs normally look like so (you can find yours with jupyter kernelspec list --json):

{
      "spec": {
        "argv": [
          "python",
          "-m",
          "ipykernel_launcher",
          "-f",
          "{connection_file}"
        ],
        "env": {},
        "display_name": "Python 3 (ipykernel)",
        "language": "python",
        "interrupt_mode": "signal",
        "metadata": {
          "debugger": true
        }
      }
    }

For one that matches a specific interpreter, the argv[0] would be the path to the interpreter:

{
      "spec": {
        "argv": [
          "c:\\users\\aku91\\miniconda3\\envs\\foobar\\python.exe",
          "-m",
          "ipykernel_launcher",
          "-f",
          "{connection_file}"
        ],
        "env": {},
        "display_name": "MY CUSTOM FOOBAR",
        "language": "python",
        "interrupt_mode": "signal",
        "metadata": {
          "debugger": true
        }
      }
    }

That shouldn't show up though (at least not yet) as it doesn't have an 'env' entry. To add an 'env' entry, just do this:

{
      "spec": {
        "argv": [
          "c:\\users\\aku91\\miniconda3\\envs\\foobar\\python.exe",
          "-m",
          "ipykernel_launcher",
          "-f",
          "{connection_file}"
        ],
        "env": {},
        "display_name": "MY CUSTOM FOOBAR",
        "language": "python",
        "interrupt_mode": "signal",
        "metadata": {
          "debugger": true
        }
       "env": {
            "FOOBAR": 1
       }
      }
    }

[roblourens]

Sorry, where do I put this JSON?

[rchiodo]

If you type jupyter kernelspec list it will show where your other kernels are. Mine are here:

Available kernels:
  .net-csharp                                                                               C:\Users\aku91\AppData\Roaming\jupyter\kernels\.net-csharp
  .net-fsharp                                                                               C:\Users\aku91\AppData\Roaming\jupyter\kernels\.net-fsharp
  .net-powershell                                                                           C:\Users\aku91\AppData\Roaming\jupyter\kernels\.net-powershell
  .venvkernel                                                                               C:\Users\aku91\AppData\Roaming\jupyter\kernels\.venvkernel
  .venvnokernel                                                                             C:\Users\aku91\AppData\Roaming\jupyter\kernels\.venvnokernel
  dupeenvdos                                                                                C:\Users\aku91\AppData\Roaming\jupyter\kernels\dupeenvdos
  dupeenvuno                                                                                C:\Users\aku91\AppData\Roaming\jupyter\kernels\dupeenvuno
  julia-1.7                                                                                 C:\Users\aku91\AppData\Roaming\jupyter\kernels\julia-1.7
  python3101jvsc74a57bd02303460301e1fef0aa1e694dd88ed9c989d4570ff871b238892fc427e0c167de    C:\Users\aku91\AppData\Roaming\jupyter\kernels\python3101jvsc74a57bd02303460301e1fef0aa1e694dd88ed9c989d4570ff871b238892fc427e0c167de
  special_venv_kernel                                                                       C:\Users\aku91\AppData\Roaming\jupyter\kernels\special_venv_kernel
  python3                                                                                   C:\Users\aku91\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.10_qbz5n2kfra8p0\LocalCache\local-packages\share\jupyter\kernels\python3

[roblourens]

I added it to one of my existing kernelspecs, then I try to access the var like os.environ.get('FOO'), but it's empty. Should it be there or does it get exposed somewhere else?

[rchiodo]

One of your existing kernelspecs may not be using the kernelspec to launch the kernel. I guess it depends.

But yes that's how you'd test it with a custom kernelspec.

[roblourens]

Why would that be? How do I find one that uses the kernelspec?

[rchiodo]

Here's my list:

You can see that I have a global python kernel and two other kernelspecs that map to the same interpreter. The two other ones are listed as jupyter kernels. That's where you'd start one that's a custom kernel.

It might be you edited a kernelspec that had the same name as an interpreter and so you started the interpreter instead of your custom kernelspec. You need to give it a unique name for it to show up as a different entry.

[greazer]

Verified