All Jupyter Notebooks are supposed to live in this directory
Since poetry installs this package into environment in development mode (aka
editable mode, which means package contents loads dinamically from disk) you are
able to import it directly:
from ds_project.utils import data_dir
Notebooks are used to experiment with data and derive robust approaches to a
given task. Then theese approaches are being added to .py files to become a
part of main pipeline. Some crucial boilerplate notebooks may be added to
repository under dvc management (because they change frequently and a lot).
One have to be careful with experimenting in notebooks and move code to .py
files as soon as modules developed have proven their usfulness.
Important rule: Execute notebook from top to bottom only. If you change previous cells (upper than you run last) then restart kernel and execute all from the start. All in all it's a good practice to restart notebook time after time.
Consider installing Jupyter Notebook extensions and configurator to facilitate your productivity. Particularly usefuls extensions are: Collapsible Headings, Ruler on 80, 100 and 120 columns, Table of Contents (2), ExecuteTime, Autopep8. Although theese extensions are handy, be carful not to overload you renderer especially on weak client machines.
Anoter useful tool is automatical conda kernels discovery. It lets you launch only one notebook server and don't bother yourself with manual kernels addition.