Datashader interactivity

[JohnTigue]

WIP deploy on Colab: datashader_on_colab.ipynb

A goal of this project is to handle large datasets. This mean 2D, 3D, 4D embedding of millions of cells. Datashader is a good idea for that.

Another goal is to be highly interactive.

• It seems Bokeh works well with Datashader
• Datashader Interactivity
• Bokeh comes pre-installed on Colab
• Ergo, Datashader and Bokeh.

(Perhaps even without a running kernel? If served via HTTP, Colab will cache images (~20 MB?). But unless all "tiles" pre-rendered, coverage will be spotty.)

[JohnTigue]

Here's blois, a Google support person saying:

When navigating to the running server it is currently failing because it's attempting to connect with a websocket.

Comment links to code

[JohnTigue]

Bokeh docs

• Running a Bokeh Server
• Bokeh Server Architecture
• Running a Bokeh server from Python, not CLI

[JohnTigue]

Plotly

Perhaps Plotly is better than Bokeh in that a server is not required?

Plotly: DataShader Case Study in Python/v3

[JohnTigue]

Critical is to call once:

output_notebook()