KeyError if there is no ggsize

[schakalakka]

If I have a plot without ggsize like lp.ggplot(df, lp.aes(x='foo, y='bar')) + lp.geom_bar() then streamlit_letsplot throws an KeyError: 'ggsize' with a throwsback to File "C:\ProgramData\Anaconda3\envs\efleets39\lib\site-packages\streamlit_letsplot\__init__.py", line 24, in st_letsplot height=plot_dict["ggsize"]["height"]

Solution:
Adding lp.ggsize(500, 250) to the plot works.

[schakalakka]

I am not sure why but now I am not able to get it working again but get instead another error:
generateStaticHtmlPage() - Exception: Trying to access top level value not marked as @ThreadLocal or @SharedImmutable from non-main thread even when ggsize is added.

If I remove the ggsize again it throws the KeyError first.

[randyzwitch]

Thanks for letting me know @schakalakka! I'm only using the default HTML generating method from LetsPlot, but maybe I have a version compatibility issue of some sort

[randyzwitch]

I think I see the issue here, I'm assuming too much from the library in terms of base data. I'll add a default, so that if ggsize isn't provided then it will default to something.

[alshan]

Hi @randyzwitch ! I'm not familiar with the specifics of streamlit but nevertheless would like to share an info on one hidden feature of lets-plot related to the plot size:

If you define a data attribute "letsPlotPreferredWidth" (i.e. data-letsPlotPreferredWidth=500) in the plot's top container (i.e. iframe or, otherwise body), then the plot's width will be 500 px and plot height will be computed according to the plot's aspect ratio, which is either default or specified explicitly via ggsize().

I hope that might make easier for you to manage lets-pot size in streamlit apps.