Consider building a GUI about data-describe HOT 4 CLOSED

I think we should start with a JupyterLab (related to #49) extension that adds as a lightweight UI on top of the SDK. The role of the UI should only be to provide UI elements that call existing functions in the Python SDK.

Benefits:

• We provide interactivity for Jupyter users (expecting Jupyter to be main user base)
• We rely on Jupyter's pattern of interfacing with a Python SDK
• Users can use notebook/code side-by-side with the interactive widget. This also helps new users with discovery of the code base, as I imagine we can provide "equivalent Python" tips with the interactive widget to help users learn the SDK
• Jupyter extensions are npm packages, which is a fairly common skill set as opposed to a more obscure GUI toolkit

Disadvantages:

• Users have to open Jupyter to use the package
• The JupyterLab extension is still quite young. I've personally experienced frustration with unsuccessfully getting extensions to install
• There doesn't seem to be a straightforward way to indicate Lab extension dependencies like in requirements.txt

Unknown:

• Is the Lab extension framework is generic enough that we can extend or adapt it to a standalone web app/GUI in the future?

from data-describe.

Plotly has just release a library two days ago called Jupyter Dash that makes it easy to build dashboard apps from Jupyter environments like Jupyter Notebook and Jupyter Lab. This may be something we can leverage? https://github.com/plotly/jupyter-dash

from data-describe.

If we wanted to go with the interactive widgets for jupyter lab, we can leverage voila to render the notebooks into a standalone web application. It also seems to have more community support than Jupyter Dash.

from data-describe.

Another option is Panel:

• Claims to work with any type of visualization
• Works in Jupyter Notebooks and Lab
• Works outside of Jupyter with the same code

from data-describe.