Plotly performance in shiny app

[annasophie]

Hi everyone,

I am creating a shiny app with large input data - around 6 Mio. observations. The app is used for interactive analysis. In each tab I have maybe 5 plots - a total of max. 15 plots. I have used plotly, b/c at some point it was faster than ggplot. However, I also have quite a lot of filters for my data and depending on the analyses, so a lot of rendering is going on.

At this point I'm only working with 10% of the data and am already having performance issues. I will have to use 100% of the data at some point.

I have two stacked bar plots and than that mostly histograms. I have looked into possibilities to increase performance and have stumbled onto toWebGL() and partial_bundle() but I'm not quite sure how they work and how I would apply them to my plots.

Any feedback and help is welcome regarding performance. If needed I can also provide a reprex.

Thanks! Anna

Edit: I was thinking if conditional filters would work, so not all filters are available at a time. Has anyone had any experience with that?

[andresrcs]

Does your data come from a sql database? If so you could consider plotting from the database (which should be faster) with dbplot package, I would also suggest renderCachedPlot() but if you have a lot of filters then that wouldn't be practical.

[annasophie]

Thanks for your reply!

The data does not come from a database but from a .csv file. I will first try different formats before trying to use a database

[annasophie]

I have a follow up question. Does renderCachedPlot() work with plotly? How would I replace plotlyOutput() in the UI and renderPlotly() in the server respectively?

[cpsievert]

renderCachedPlot() doesn't work with plotly...perhaps someday plotly will have it's own caching version, but it's not in our immediate plans