I have not worked on interactive visualizations in general and with bokeh in particular until now and try to derive from the docs what would be the appropriate approach in order to achieve what I am looking for.
I am working on video analysis. What I want is to create a dashboard with three areas. The first is howing a scatter plot which represents the distribution of different saturation values (y-axis) across a movie frame by frame (x-axis). The second and the third area should contain visualizations which depend on the selection of a peculiar circle within the scatter plot. The second area should represent certain color values from the frame of the selected scatterplot circle/saturation value/frame. The third should represent a processed version of the image of the frame that is associated with the selected circle.
The interaction/process model should look like this: A circle on the scatter plot is selected, area 2 and 3 are updated asynchronsly in relation with the selection. However each update first needs computation on the server side since the color values as well as the version of the frame that should be shown needs computation first. The data is not just there, it needs to be produced dynamically from the data that exist.
From the docs (Linking Plots) I can see that it is possible to link plots together so that they dynamically change at the same time. However, from my understanding, such updates operate on the same data source which is sent to the client side. It does not refer back to server. So I am unsure how to achieve my goals. Of cause I could use the tap_tool in combination with OpenURL in order to generate a new website with the corresponding visualization, but that is not nice. The click event of the tap tool should lead to a server side computation of which its results should asynchronously update two of the three plots on the dashboard.
It would be nice if someone could point me into the right direction for this, thx