search

nbgallery / dashboards

Notes and experiments in Jupyter dashboarding
MIT License
16 stars 3 forks source link

Best dashboarding solution with support for `ipython_blocking` #3

Open dataist opened 5 years ago

dataist commented 5 years ago

Great collection of stuff here!

Very aligned with how we are making use of notebooks internally here at frame.ai. I'm curious if you've found any dashboarding solutions that play nicely with ipython_blocking as we also use that heavily for soliciting input from non-tech users before continuing with execution?

somedave commented 5 years ago

@kafonek

kafonek commented 5 years ago

@dataist thanks for the kind words on ipython_blocking! I'm glad that library has been useful for you. I'm very happy that I was able to write it with help from the core Jupyter devs (@minrk in particular) during the sprints at Jupytercon last year.

Unfortunately, ipython_blocking does not work with Voila, our preferred Dashboarding solution, because Voila disables execute_request messages to the Kernel and overriding execute_request handlers is how ipython_blocking works under the hood.

During the recent Jupyter Dashboarding conference in Paris, I got the chance to discuss "MVC in Jupyter" with the Voila developers (Maarten, Sylvain, etc). Through those conversations, I started work on a still-in-early-development library notebook_restified that hopes to hit some of the same goals as ipython_blocking.

Instead of having one notebook that has the user-friendly GUI pieces(ipywidgets) and the application logic (ipython_blocking-enabled), you would have two separate Notebooks. The GUI piece is the View Notebook, the application logic is the Model Notebook. The View Notebook can get user input parameters and execute the parameterized Model Notebook then display the result.

I'd definitely appreciate your feedback on the ideas behind the library and how Jupyter/MVC fits in to frame.ai's work.

Thanks.