Which matplotlib.axes methods are safe to call from a background thread?

Question

I’m writing an interactive application using wxPython and matplotlib. I have a single graph embedded in the window as a FigureCanvasWxAgg instance. I’m using this graph to display some large data sets, between 64,000 and 512,000 data points, so matplotlib’s rendering takes a while. New data can arrive every 1–2 seconds so rendering speed is important to me.

Right now I have an update_graph_display method that does all of the work of updating the graph. It handles updating the actual data as well as things like changing the y axis scale from linear to logarithmic in response to a user action. All in all, this method calls quite a few methods on my axes instance: set_xlim, set_ylabel, plot, annotate, and a handful of others.

The update_graph_display is wrapped in a decorator that forces it to run on the main thread in order to prevent the UI from being modified from multiple threads simultaneously. The problem is that all of this graph computation and drawing takes a while, and since all of this work happens on the main thread the application is unresponsive for noticeable periods of time.

To what extent can the computation of the graph contents be done on some other thread? Can I call set_xlim, plot, and friends on a background thread, deferring just the final canvas.draw() call to the main thread? Or are there some axes methods which will themselves force the graph to redraw itself?

Answer 1

I will reproduce @tcaswell’s comment:

No Axes methods should force a re-draw (and if you find any that do please report it as a bug), but I don't know enough about threading to tell you they will be safe. You might get some traction using blitting and/or re-using artists as much as possible (via set_data calls), but you will have to write the logic to manage that your self. Take a look at how the animation code works.