Flask implementation of datashaded bokeh app

[benvddriessche]

I am working on a dashboard for visualization of large time series. The combination of Jupyter + Bokeh + Holoviews/Datashader is perfect for prototyping and data exploration. However, I would like to embed some graphs on more user-friendly web pages using Flask.

• Deploying bokeh apps without datashading using script, div = components(bokeh_figure) in Flask works perfectly fine.
• Visualizing bokeh apps with datashading in Jupyter or using bokeh serve also works fine, e.g. with doc = hv.renderer('bokeh').server_doc(datashaded_hvObject).
• Where I'm running into issues is combining these two steps. I can render a datashaded bokeh app in Flask like so:
  renderer = hv.renderer('bokeh').instance(mode='server')
hvplot = renderer.get_plot(datashaded_hvObject)
script_data, div_data = components(hvplot.state)

... but this of course only shows a single rasterized frame and is missing the callback to update the plot when someone is interacting with it.

Based on some tutorials, I'm running a bokeh server from my Flask script as follows:

bokeh_process = subprocess.Popen(['bokeh', 'serve', '--allow-websocket-origin=localhost:5000','test.py'], stdout=subprocess.PIPE)

With the main Flask @app.route set as follows:
session = pull_session(session_id=None)
bokeh_script = server_session(None, session_id=session.id)
return render_template('test.html', script_data=bokeh_script)

This is not working though as it doesn't appear to connect to the correct session ID: Cannot pull session document because we failed to connect to the server (to start the server, try the 'bokeh serve' command). I'm also not sure this would actually get me a working interactive session.

I realize this is not a strict Holoviews question but I figured it may be useful to other Datashader users. Let me know if it's out of scope!

[jlstevens]

I don't know offhand how to resolve the issue you have pulling the session document but I can say that a docs PR explaining how to use holoviews/datashader together with flask would be very welcome!

[philippjfr]

There is already a PR for a flask app here #2907, which could maybe be adapted.

[benvddriessche]

Thank you @philippjfr, the PR put me on the right track!