SolaraViz: Fast option for dynamic plots

[EwoutH]

It would be useful to have a fast, simple option for dynamic plot updating (streaming) capabilities in our Solara-based visualization module, allowing for the addition of data points each timestep without redrawing the entire plot, ideally with a fluent animation between.

Options

Some options I found:

1. Matplotlib:

   • We currently use it, maybe it can be made faster.
   • Use FuncAnimation or similar methods to update plots.

2. Altair:

   • Also currently already there.
   • Update the chart data periodically and re-render using Solara.
   • Limitation: Might involve partial re-drawing due to Altair's design.
   • https://altair-viz.github.io/user_guide/interactions.html

3. Plotly:

   • Utilize FigureWidget for efficient updates without full re-rendering.
   • Seems recommended for smoother updates and better performance with complex plots.
   • Maybe https://dash.plotly.com/live-updates

4. Bokeh:

   • Leverage ColumnDataSource to stream data directly into plots.
   • Offers more control over the streaming process and seamless integration with Python backends.

Steps

• Evaluate each library's performance and ease of integration with Solara.
• Implement the most suitable solution and ensure compatibility with existing modules.
• Update documentation with examples and usage guidelines.

[EwoutH]

@quaquel could you write up what you told me in the meeting today? The stuff about Altair and sending dataframes to it.