UI extension - allowing users to see existing table/columns in a connected db

[idomic]

This will probably be a widget to allow users view without querying the existing data model in the db, and existing columns.
We probably need more feedback here before building it.

[edublancas]

A few thoughts:

The most challenging part here is communicating with the JupyterLab front end. We tried doing something similar in Ploomber a while ago: we wanted to show an alert when a specific event in the Python kernel happened, but we needed help finding an easy solution. It's certainly possible since the visual debugger does precisely that, but I bet it'll require some heavy work.

As a first version, we can create an interactive HTML that allows users to explore the tables/columns. Example:

%sqlreport 

# or a Python API?
from SQL import report
report()

This could generate an HTML that users can download or interact with directly in the cell. See a list of tables, columns for each table, and maybe even get some summary statistics (note that this overlaps with ploomber/jupysql#66)

Since each DB implements a different API for listing tables/columns, we can leverage SQLAlchemy to use a single API compatible with many databases.

[idomic]

Reopening as not done

[psychemedia]

Some years ago, I explored various magics for visualising schemas from Postgres database connections; for example: https://github.com/innovationOUtside/ipython_magic_sqlalchemy_schemadisplay which we still use in a course today (the visualisations are a bit ropey, but useful as quick sketches). The dependencies may be quite heavy though...

[edublancas]

moved this to jupysql-plugin