Python Preview Extensions

[MarcSkovMadsen]

Request

Make it easy to create python based VS Code Extensions for previewing .py, .csv, .parquet etc. files.

Motivation

Discussion on Twitter

Motivated by https://twitter.com/luumelo14/status/1556683030571261952?s=20&t=OgR9ab9qU8RYkVsAHCweOQ.

Data App Previewers

Python has many easy to use data app and web frameworks that can be used to create tool and interactive visualization. Many of these have "previews" for Jupyter.

I think it would be useful to be able to have the same in VS Code enabling end to end development of data apps.

Especially if you are running some version of Code on a server, it can sometimes be quite tricky to serve the apps and open them in your browser.

At my work we have a jupyter hub with code-server deployed and being able to offer this combination is really powerful as these two solutions appeals to different people and use cases. I can panel serve name_of_file.py in a terminal and with some help of the jupyter-panel-proxy based on jupyter-server-proxy I can open a live app with autoreload. But this requires forming a specific url manually, opening a browser tab and pasting in the url which takes time. So I really would prefer a "one click preview" of my Panel apps in Code.

Examples

• Panel: https://blog.holoviz.org/panel_0.12.0.html#JupyterLab-previews
• Voila: https://www.kylabendt.com/blog/setting-up-jupyterlab-and-voila/
• Dash: https://stackoverflow.com/questions/45490002/how-to-use-dash-within-jupyter-notebook-or-jupyterlab
• Streamlit: not that I know of

panel-jupyter-preview.mp4

Data Previewers

Using the above mentioned frameworks its also very easy to create custom, domain specific tools for visualizing data. These tools could be turned into "preview" tools in VS Code if it was just easy to do.

It would also make it easy to turn a lot of pythons auto-viz and auto-ml tools into VS code extensions.

When these apps are started they should have the path to the active file as input. The image above shows a custom previewer for a netcdf file.

Launch pad

With Jupyter Server Proxy you actually create entry points in your Python packages, that can add any web app to the Jupyter Launcher and serve it. Something similar could be a possibility for VS Code to make it really, really easy to enable the packages installed in the active virtual environment open web apps inside VS Code by the click of a button.