No obvious way to install without Conda ; this kernel should have a regular `setup.py` and/or be pip-installable (and on Pypi).

[EmmanuelCharpentier]

I don't use Conda, and want to avoid it, since I use and develop Sagemath, which sets up and installs its own virtual environmrnt. In fact, Sage is essentially a (gigantic) venv interfacing Python to a zillion and half mathematical packages.

I tried to mimic manually what install.sh does, but my initial attempt failled, jupyter reporting that there was no such kernel.

I finally succeeded by symlinking the two directories dyalog-kernel dyalog_kernel to the "system" directories of the Sage venv. And killed almost a day in the process.

Since Sage supports onstalling "external" Python packages by a simple use of pip, the shortest path for me would have to just pip dyalog-jupyter-kernel.

I noticed that the Wiki instructions instst on getting Conda, but do not explain how it intervenes in the process. A bit of explanatiion would be useful.

HTH,

[xpqz]

I came across the same problem. This mod to install.sh fixed it for me:

    if [ -n "$VIRTUAL_ENV" ]; then
            SITEDIR="$VIRTUAL_ENV/lib/python$PYVER/site-packages"
    else
            SITEDIR="$(python3 -m site --user-site)"
    fi

I'll put in a PR.