Currently, the kernel code is manually installed by copying the contents of src into $JUPYTER_PATH/kernels/guile and replacing the absolute path of guile-jupyter-kernel.scm inside kernel.json. Ideally, a Jupyter kernel should be able to run without making any changes to the source code, thus the project should be refactored accordingly, providing a Makefile or some similar installation script.
The best choice imho would be providing the kernel script as a Guile module. Additionally, I have looked into including the Jupyter path relative variable into kernel.json, but jupyter/jupyter_client provides no way to add in variables other than connection_file. Perhaps looking into how guix-kernel does it could be helpful?
Currently, the kernel code is manually installed by copying the contents of
src
into$JUPYTER_PATH/kernels/guile
and replacing the absolute path ofguile-jupyter-kernel.scm
insidekernel.json
. Ideally, a Jupyter kernel should be able to run without making any changes to the source code, thus the project should be refactored accordingly, providing a Makefile or some similar installation script.The best choice imho would be providing the kernel script as a Guile module. Additionally, I have looked into including the Jupyter path relative variable into
kernel.json
, but jupyter/jupyter_client provides no way to add in variables other thanconnection_file
. Perhaps looking into howguix-kernel
does it could be helpful?