guest post proposal: using pyodide to port pandastutor educational tool client-side

[pgbovine]

Good morning!

I am a professor at UC San Diego working with my grad student Sam Lau on Pandas Tutor, an educational visualization tool that shows how your pandas data transformation code works step-by-step: https://pandastutor.com/

We have recently been porting Pandas Tutor over to using Pyodide so that it can run directly in students' browsers without needing us to host or manage a webserver to run server-side Python. This will be greatly beneficial to teaching large classes in particular (e.g., 500+ students!) since it would destroy our server if that many students tried using the tool all at once during class; with Pyodide our tool can scale effortlessly since each student will be running it in their own laptop's web browser (and can cache the JS/WASM bundle downloads from the server).

We would love to write a guest blog post talking about the benefits of porting over to Pyodide, and some of the optimizations we have been implementing. Our changes haven't been pushed to the main https://pandastutor.com/ site yet, but if this blog post is approved, then by the time it gets published, all the changes should be on the live site. Here's a brief proposal for the post:

• The current Pandas Tutor server-side sandboxed Python/Pandas execution setup, based on Docker hosted on a Linux cloud server. And all the limitations and slowdowns that come with maintaining a server setup.
• Our decision to port over to Pyodide and reasons for doing so
• Initial challenges with porting over to Pyodide and how we overcame them
• Performance comparison of new Pyodide in-browser execution vs. the old server-side Docker-based execution
• The new kinds of interactions that Pyodide enables us to do, such as live programming mode (edit the code and instantly re-run and visualize it!) and downloading/caching data files from the web client-side
• Conclusion: why Pyodide is essential for teaching data science in educational settings due to its client-side nature and being able to distribute workloads across everyone's computer

Thanks for your consideration, Philip Guo

[rth]

Hello Philip,

this sounds like a very interesting blog post and pandastutor is also great. We would be very happy to include it.

Don't hesitate to open a PR with your post once it's ready. We can probably even make pandastutor run interactively as an example in the post.

Thanks!

[pgbovine]

Awesome, thanks! Also a great idea to try to get pandastutor embedded and running interactively in the post itself :)

[pgbovine]

Just opened a pull request here: https://github.com/pyodide/pyodide-blog/pull/27

We'd love to get this posted in a week or so if possible (e.g,. Sat, May 14 or a bit later). We didn't have time to direct-embed pandastutor into the post itself since it was already getting long, but we've linked to the live site. Thanks again for your help!

[pgbovine]

great, thanks, @rth -- our tweet with blog post is here, feel free to close this issue https://twitter.com/samlau95/status/1525097276192047104

[rth]

Thanks a lot for writing this blog post!