In python/moose/init.py is dirtied with many utility functions, especially model loaders.
- Unless a thin wrapper over a C++ core function, these increase the loading time quite a bit (e.g., neuroml2)
- Better to keep these as submodules so that users are not burdened with stuff they do not need: SBML user does not need to load NeuroML2 and vice versa
- Examples of this approach: numerous submodules of pytorch, scipy.stats, scipy.fft, etc.
In python/moose/init.py is dirtied with many utility functions, especially model loaders.