kernel.json has hardcoded python path

[isuruf]

This should just be python or the path to sys.executable replaced by setup.py

[redtrumpet]

I ran into trouble with this, because jupyter is running with python3.6 on my Ubuntu system, but python is actually Python 2.7. I got the error message

/usr/bin/python: No module named jupyter_kernel_singular

in my jupyter log, and fixed it (for now) by adjusting the "hardcoded" path to /usr/bin/python3.6. Fortunately it's not really hardcoded, but configurable. Still not optimal though.

[rjacuna]

The fixed code needs to be pushed to the pip3 repository, I ran into the issue and it took me a couple of hours to figure out what went wrong.

[sebasguts]

Sorry for wastig your time. I uloaded the fixed code to PyPI. Hope it works now.

[EmmanuelCharpentier]

Sorry for wastig your time. I uloaded the fixed code to PyPI. Hope it works now.

Unfortunately not. I'm trying to install it in Sage, where it is packaged with a very thin wrapper around pip, and the resultant kernel.json still is

{
    "argv": ["/usr/bin/python", "-m", "jupyter_kernel_singular", "-f", "{connection_file}"],
    "codemirror_mode": "singular",
    "display_name": "Singular",
    "language": "singular"
}

Since /usr/bin/python is still Python 2, the kernel isn't found.

Contrast this with the installatuion procedure of the Gap kernel, which builds a custom kernel.json (including the current executable (sys.executable): in the Sage case, it finds ths correct Sage script) and dumps it in the relevant directory before installing.

Could you check what is installable via pip ?