Hello,

This is not exactly a bug report, nor a feature request, but rather more of a development dilemma that I'd like to discuss with you. I hope it's OK to discuss it here and not on discourse, as I feel here much more comfortable with code snippets and attachments, and I think that after a discussion this issue will be actionable as a feature request or a bug report.

I managed to create the following "poor man's GUI" to view results of many simulations in 1 figure:

Where the black vertical line in the lower right plot shows the progress of the simulation. This seems to behave great right? Apparently, I had to call an animation's _start() function to achieve this. Here are details of the progress:

I used the Data browser code as a starting point. Basically, when self.update is called (every time a simulation is picked), an animation is created using self.anim = animation.FuncAnimation(...). The weird behavior I experienced before using self.anim._start(), is demonstrated in the following video:

As you can see, the animation never restarts it self, although the self.anim object is assigned a new value every time a simulation is changed - the black time marker in the lower-right plot demonstrates this. Sometimes I experienced that pressing n or p (events handled by the class), in a certain timing, does make the simulation restart, but this is not well defined.

Here is the commit in my code that reproduces the issue I'm describing:

https://gitlab.com/doronbehar/lab-ion-trap-simulations/-/commit/2edc2abbe874ebd2af458f1d4f80710754d8625a

If you really want to reproduce the issue with my data, you will have to trust me with the pickle data I uploaded to this repository, to the branch matplotlib-_start-issue-reproducing.

Sorry for not creating a minimal working example 😬, but I hope you'd be able to discuss the issue although the code that reproduces it is involved in many irrelevant details.

Thanks for your attention!