start

#!/bin/bash

set -ex

# Write all server logs to ~/.jupyter-server-log.txt for debugging
echo "Notebook server start command: $@" | tee -a ~/.jupyter-server-log.txt

# mybinder sends a complex command to detect jupyter notebook or lab, using standard python3 executable:
#   arg1: python3
#   arg2: -c
#   arg3: import os ... (long Python code, which adds "--NotebookApp.default_url=/lab/" to the argument list)
#   arg4: --ip=0.0.0.0
#   arg5: --port=8888
#   arg6: --NotebookApp.base_url=/user/lassoan-slicernotebooks-1g6ow0gs/
#   arg7: --NotebookApp.token=sw77G_ZwRdCCE8qy9-R_Tg
#   ... (there may be more arguments)

# The docker image does not have system python, just Slicer's therefore we collect the arguments
# that configure the notebook and pass those to Slicer jupyter.

jupyter_args=()
for arg in "$@"
do
    if [[ "$arg" =~ ^--ip=.+|^--port.+|^--NotebookApp\..+  ]]; then
        jupyter_args+=("$arg")
    fi
done

# Recent-enough jupyterlab is available, make it the default UI
jupyter_args+=("--NotebookApp.default_url=/lab/")

echo "Found jupyter notebook arguments: ${jupyter_args[@]}" 2>&1 | tee -a ~/.jupyter-server-log.txt

/home/sliceruser/run.sh

# Make output redirection of server logs more robust
export PYTHONUNBUFFERED=1

# Use bash as shell in Jupyter terminal (much more convenient than the default sh)
export SHELL=/bin/bash

exec /home/sliceruser/Slicer/bin/PythonSlicer -m jupyter notebook "${jupyter_args[@]}" 2>&1 | tee -a ~/.jupyter-server-log.txt

# If launching fails for *any reason* then all the information mybinder provides is something like this:
#
#   2021-10-16T05:57:55Z [Normal] Created container notebook
#   2021-10-16T05:57:55Z [Normal] Started container notebook
#
# ...and connection to the jupyter server times out.
#
# To debug the startup issues, the easiest method is to install system Python
# (similarly to this repository: https://github.com/jupyter/docker-stacks/tree/master/minimal-notebook)
# because mybinder surely works with that, and use the command terminal to access logs (~/.jupyter-server-log.txt)
# and test things interactively.