The fewer the dependencies, and the less they depend on the OS, the easier the packaging.

In SOFA Statistics, dependencies, especially image-processing dependencies, made it very hard to package cross-platform.

Minimum:

• Python
• wxPython - for GUI
• matplotlib - for images that overload SVG / JS based toolkits such as Dojo
• pandas (inc numpy) - for misc data processing including in core stats engine

All dependencies are distro / system dependencies, not pip dependencies (even if we develop in a venv).

This is to make deb packaging easy where we list system dependencies.

wxPython - https://shanemcd.org/2020/05/03/how-to-install-wxpython-in-a-python-virtual-environment-on-debian-buster/

A couple of the key steps (there are others but you may have already done them):

sudo apt-get install build-essential python3-dev \
  libwebkit2gtk-4.0-dev libtiff-dev libnotify-dev \
  freeglut3-dev libsdl1.2-dev libgstreamer-plugins-base1.0-dev

And in venv (when activated):

pip install wxPython