Jupyter Dashboards Layout Extension | Hacker News Reader

Jupyter Dashboards Layout Extension

jupyter-dashboards-layout.readthedocs.io

32 pointsstared6 comments

In my team we use jupyter as a means of socializing somewhat-digested data intel among educated team members. The idea of using jupyter as dashboard material is ideal for us. It makes a quick and powerful information radiator.

That being said there are many directions jupyter can expand from here. Ive invested time experimenting with numerous jupyter extensions only to see several of them die on the vine.

Given the high amount of flux in jupyter/ipython I'd be reticent to adopt this tech in earnest unless I saw it within some greater roadmap of long-term jupyter support. I want to avoid hitching my wagon to soon-to-be-abandoned great ideas.

Good: Jupyter will hopefully stick around for a long time. I just spent a ton of working writing a new separate implementation of the whole compatible Jupyter stack for CoCalc, and Kyle Kelly at Netflix has similarly written another implementation for their use (see nteract).

Ugly: My impression is that the current “jupyter extension” API is actually being deprecated in favor a completely different presumably better API that JupyterLab will use. I think the Jupyter devs have a vision that involves everybody rewriting all their extensions against a new API... Regarding extensions, CoCalc also can’t possibly work with Classical Jupyter notebook extensions, so I personally plan to rewrite several popular extensions for CoCalc (I’ve done some already).

Bad: The Jupyter project seems to be very heavily funded and reliant on Moore/Sloane/NSF. I know that Fernando Perez put a huge amount of effort into helping to get some of these grants, which are I think well over 10million total by now, and I think these grants fund numerous fulltime positions. I’m personally also worried that the grants will run out, not get extended, and the project will be in major trouble as a result. Fortunately, at least soem developers like Jason Grout and Chris Colbert are fulltime at companies (e.g., Bloomberg, maybe Continuum Analytics), so they have more stable funding. Also, CoCalc (my company) isn’t dependent on grants, and we will be around for a long time.

I think it will stick around for a while, if only because academics have adopted in great numbers. And academics are verrryy slow to change their tools ;).

Big fan of having an easy to find "understanding the use-case" section in the docs!

Looking at the examples provided, you still need to implement your own dashboard widgets by hand using HTML/Polymer (outside of simple libraries with native support like Bokeh), which is the actual time-consuming part of making dashboards that working with Jupyter doesn't mitigate.

That said, the use case of sharing dashboards is interesting, albeit not a new feature relative to existing solutions like Tableau/Looker.

Lovely stuff! Any working examples / demos of this anywhere on the web?

Reply on news.ycombinator.com