There is a lot of work and attention invested in JupyterLab to improve the user experience. It's great !
However there is a type of workflow I have not heard much, but I see very often in my company: People develop Python packages partly in notebooks, partly in IDEs (VS Code, PyCharm, etc). When a piece of code is more or less ok, it moves from a notebook cell to the .py file part of a package (IDEs are better suited to deal with large Python packages). This package is imported in the top cell of the notebook. Because developing is a lot of trial and error, we use the autoreload IPython magic a lot: "%load_ext autoreload, %autoreload 2". Unfortunately this magic works 80% of the time. If the package has a new file of a class a new method typically, it won't work. Then a dev has to restart the kernel.
My question is: How difficult would it be to create an "%autoreload 3" that would work 100% of the time ?
Has anybody else got this problem ?
Does anybody else use Jupyter and and IDE in parallel ? If so any tip ?
Thx