Open discussions on specific topics selected by the Software Working Group and selected from the list of SWG Topics For Discussion.
Wednesday, May 5, 2021 - Visualization in Notebooks - Moderated by Matt Berry
Recording: https://uofi.box.com/s/r03icyhwbr08e60xjra2r9noxrqo9509
Attendees
- Luigi Marini
- Elizabeth Yanello
- Mark Vanmoer
- Mikolaj Kowalik
- Jong Lee
- Peter Groves
- Wenjie Zhu
- Charles Blatti
- Jeff Terstriep
- Chen Wang
- Steven Peckins
- Stephen Pietrowicz
- Michael Bobak
- Sandeep Puthanveetil Satheesan
- James Phillips
- Todd Nicholson
- Kaveh Karimi Asli
- Yong Wook Kim
- Dukyun Nam
- Michal Ondrejcek
- Stuart Levy
- Christopher Navarro
Computational notebooks present new opportunities and challenges for visualizing data. A discussion about tips, tools, pitfalls, and future directions. Matt shared a visualization in notebooks with sticky notes with several considerations of what people may have used visualization in notebooks. It appears that most in the SD are using Python 2 & 3 with some using R and JavaScript.
A survey presented of how to expect our use of visualization in notebooks to change Most chose More for frequency, complexity, and novelty.
What are the major challenges and risks? Most chose UX and DX as well as interactivity and providing data to browser process. Peter posted a sticky about underlying hardware. Jeff mentioned secure shell tunnels are not as secure.
Code portability is important. Code developed in one application should be able to work in another applications.
Outgrowing notebooks is an issue. IN-CORE started using libraries and packages. We have found that people often don't have the experience in notebooks to properly use it. When you move code out of a notebook, you lose some of the visualization.
What visualization libraries do most use? ggplot; matplotlib; pandas; seaborn. Some use geospatial visualization.
What are we using to build novel visualizations? Pure Python; custom Jupyter; JavaScript; Jupyter ServerProxy; Jupyter Comms. Discussion ensued about which worked best.
Why Notebooks? Best platform for visualization; best platform for the rest of the content or workflow; Best place to reach the audience; Only place to reach the audience; Prototyping; A good option for publishing code along with a paper; Replacement for Python REPL. Most chose Prototyping
Why are we making notebooks? Ourselves? Other researchers? Students?
What's the future?
What are we excited about?
What do we want to contribute?
Julia is another option
Links:
https://jamboard.google.com/d/1ONvYULENzLC2vNp1LoAUnZ_EnSw6_Wr5Ie4Fn4Q5tUg/edit?usp=sharing
https://jupyter.org/widgets https://towardsdatascience.com/introduction-to-papermill-2c61f66bea30
https://www.nature.com/articles/d41586-021-01174-w