On Fri, Aug 2, 2019, 1:56 PM Chris Angelico <rosuav@gmail.com> wrote:

Also a bit old-school (it took me many years to learn the value of syntax highlighting), and an educator, and I've seen students start out with Jupyter. As an alternative to the vanilla REPL, I think it's awesome if a little expensive on low-end machines; but as an alternative to a text editor, it's an attractive nuisance. Yes, you can try things out and see the results instantly, AND you can save it, edit, rerun, etc; the cost is that debugging becomes a nightmare. But for discoverability, incl tab completion? It's great, and I probably should start using it more.

There are definitely things for which Jupyter is perfect, and things for which it is a terrible choice. A lot of data scientists (of whom I am one, but via a winding path) don't know anything outside Jupyter. The result is that they throw rambling code that should be a script or module into a giant notebook with many unrelated cells stuck in the middle (and no way to add unit tests, code checkers and linters have poor support, execution order is not transparent, etc). That said, the interactive notebook is a huge win for many things. I recently produced a bunch of somewhat related graphs. Each sliced and diced data a little bit differently, then rendered then with with different visual options. I didn't come in with known requirements or specification, just a general sense of what I wanted to illustrate. So modify a cell, rerun the graph, repeat a hundred times. Doing that in a plain code editor with a separate image file viewer would have been WAY harder. Jupyter is right for this. On the other hand, I wound up with the same blocks of 40-ish lines of code repeated 5 times in different places, which is terrible. But at least I then took the time to move all of that to a utility module, factoring differences in behavior into some function arguments. So the notebook is small, but allows visualizing many variations without switching tool context.