Enviroment variables --> configuration file

[chillenzer]

I usually refrain from using environment variables. If we want users to configure their runs without changing the notebook, wouldn't a configuration file be nicer?

[edbennett]

I don't particularly like environment variables, but I don't particularly like hardcoding the name of a configuration file either. Both are workarounds to the fact that jupyter nbconvert doesn't give a way to pass parameters or input to the notebook.

Could go for the best (and/or worst) of both worlds, with a configuration filename defined via environment variable?

[chillenzer]

Maybe that reveals the deeper question I have: Why is it a notebook? I personally prefer scripts.

[edbennett]

A few reasons:

• I prefer notebooks for interactive exploration of data, which this was created as.
• Having this as a notebook allows someone to load it into a service like MyBinder and continue exploring it easily.
• There is the possibility (which would be good to do before the repo is released, and I should make an issue about it) to make the document literate, with some narrative mixed in with not only the code (as could be done with comments or docstrings) but also the plots. Mixing narrative with plots would otherwise necessitate something like Rmarkdown, Org-mode in Emacs, or just having a separate TeX document that pulls in PDFs generated by the script.