Using plotly-resampler with streamlit

[lawrencefchan]

I have a streamlit dashboard in which I'm currently using plotly with to plot a fairly large dataset. It can be fairly slow at times, and I suspect that plotly-resampler would be perfect for improving the plotting performance, but I haven't found a way to connect the two together. The issue as I understand it is that streamlit effectively calls fig.show() when displaying plotly charts, which breaks the ability of plotly-resampler to do its job.

Has anyone run into a similar issue, or is able to suggest a workaround?

[jonasvdd]

Hi @lawrencefchan,

You might find this example from our repository helpful: Streamlit App Example.

It demonstrates how to integrate a plotly-resampler figure into a Streamlit application using an iframe.
This approach allows for quick and easy integration, though I’m not certain how well it holds up in a production environment.

Kind regards,
Jonas

[lawrencefchan]

@jonasvdd, thanks for your response. I tested it and it works to an extent, but I believe there is some broken interaction between the dash app and the streamlit app, that stops the dash figure from being updated.

My workaround to this issue is to restart the process responsible for the dash app:

fig = FigureResampler(some_fig)

try:  # close any existing dash app to free resources and allow port to be re-used
    st.session_state.dash_process.terminate()
    st.session_state.dash_process.close()
except AttributeError:
    pass

# Run the visualization (as in the example)
p = Process(target=fig.show_dash, kwargs=dict(mode="external", port=port))

# start a new instance, assign it to the session state so it can be killed next rerun
p.start()
st.session_state.dash_process = p

This is a bit flakey though, so not recommend for production!