Dynamic Plot on Session List View

[tchamberlin]

Instead of the static datashader bitmap, the Session List View (/disk/sessions) should display an interactive plot, via Holoviews/bokeh/datashader.

This will be a good starting point: https://datashader.org/getting_started/Interactivity.html#holoviews-datashader-bokeh

Each time you zoom or pan, the callback will be given the new viewport that’s now visible, and datashader will render a new image to update the display. The result makes it look as if all of the data is available in the web browser interactively, while only ever storing a single image at any one time. In this way, full interactivity can be provided even for data that is far too large to display in a web browser directly

Here's a working example of the basic behavior that we are looking for: https://nyc-taxi.pyviz.demo.anaconda.com/dashboard

[tchamberlin]

Note: I can't get this notebook to work https://datashader.org/getting_started/Interactivity.html#holoviews-datashader-bokeh

Not sure what's going on -- perhaps mismatched package versions on my end. But also possible that the example is out of date

[KaseyW31]

Hm the notebook works for me, but I'm still trying to figure out how to have a live Python process running for the interactive updating. I've been looking into Bokeh servers and possibly embedding one into the Django template, for example by using bokeh.client.pull_session and saving the session script as a context variable (https://holoviews.org/user_guide/Deploying_Bokeh_Apps.html). Though I keep getting errors on IOLoops that I haven't been able to fix yet

[tchamberlin]

I think that's the correct approach. Django and Flask are pretty much the same for the purposes of that example

[KaseyW31]

The IOLoop issue last week was solved by starting the Bokeh server in a separate script, outside of Django.

I ran into another issue earlier this week with Django appending the current session page URL to the CSS resource URLs generated from pull_session, which caused 404 errors. I tested embedding one of the Bokeh server examples (https://github.com/bokeh/bokeh/blob/3.2.0/examples/server/api/standalone_embed.py), which uses Bokeh's Server() constructor and Applications instead of panel.serve(), and it worked on the session page

I'll be working on trying to put dynamic plots into a pure Bokeh application/server

[KaseyW31]

I now have a Bokeh server embedded in the session detail view, but the session is currently hardcoded. I spent time yesterday reading through Bokeh server documentation, and I'm trying to implement a full server that can process queries with handlers. There don't seem to be concrete examples to work off of, so I'll try to build it up gradually

[KaseyW31]

After several attempts, I was able to start a server using Bokeh's Server constructor instead of renderer.server_doc, and I roughly followed an example (https://holoviews.org/reference/apps/bokeh/player.html) to add the Holoviews plot to the Bokeh document. I also used a FunctionHandler (https://docs.bokeh.org/en/latest/docs/reference/application/handlers/function.html) to create the application and modify the document.

It seems to work when I access the server from a new browser window, but getting the server document through Django caused weird errors that I haven't been able to figure out yet. I'll work on trying to understand/fix them, and if nothing else works, I might try upgrading bokeh to see what changes

[KaseyW31]

I found that I can pass arguments to Bokeh while still using renderer.server_doc to start the Bokeh server instead of the Server constructor, so I no longer need to deal with the weird errors from before. In alda's views.py, I get the Bokeh plot using script = server_document(url=<url>, arguments={"session":session})).