Regular non-Solara plots not showing in Jupyter Notebooks after Solara has been used

[EwoutH]

For some reason, regular matplotlib and seaborn plots don't show anymore after a Solara page has been initiated. In the code below you can see the first seaborn plot showing well, but the second not showing.

In PyCharm, the first plot is also shown fine, but instead of just not showing the second plot anymore, but gives an error:

ipykernel_29332\2165788151.py:2: UserWarning: FigureCanvasAgg is non-interactive, and thus cannot be shown
  plt.show()

Here's a notebook to reproduce: solara_bug_plotting.zip
And here's the same notebook in Google Colab.

First plot shows fine:

After a Solara component, the second errors or doesn't show anymore:

[EwoutH]

A workaround is adding %matplotlib inline before the second plot.

I still think this is weird behavior.

[maartenbreddels]

Does this also happen after a simple solara import?
After this happens, can you check if solara.server.patch is in sys.modules?

[EwoutH]

Thanks for getting back.

Taking some minimal examples, it looks like everything works fine: solara_minimal_test.zip

Which means probably something is wrong with our implementation. Can you spot anything obvious?

• Implementation: https://github.com/projectmesa/mesa/blob/74e2df6fdfc2d5172f84eb96469b277b35bfd795/mesa/visualization/jupyter_viz.py#L92
• Usage tutorial: https://mesa.readthedocs.io/en/latest/tutorials/visualization_tutorial.html

[maartenbreddels]

https://github.com/projectmesa/mesa/blob/74e2df6fdfc2d5172f84eb96469b277b35bfd795/mesa/visualization/jupyter_viz.py#L39 seems suspicious.
Maybe you have added that because that was needed before #540 to run under solara server.
Now things should just run.

[EwoutH]

Thanks! Let me check it out.

[EwoutH]

I think that help, I at least don't need the %matplotlib inline anymore (which is quite logical in hindsight).

The JetBrains (PyCharm) implementation is also still not rock solid:

But it now works most of the time, with reloading a cell or two!

Thanks for the quick responses, really appreciated!