I have been building Shiny apps in R for what seems like a long time. Recently, I’ve been dabbling with Shiny for Python and for one particular reason. With Shinylive these apps can be hosted without the need for a dedicated server - they just run in the web browser. For me this opens up two interesting possibilities. Firstly, it should mean I can embed an app in this blog (more on this below) and secondly, I can create apps for others and they won’t need to use a Shiny server (or similar) setup to run them.

To experiment with embedding an app in a blog post I created a small Shiny for Python app using the plotnine plotting package. I use ggplot2 for the vast majority of plotting in R and so I wanted to experiment with a similar result in python.

This app just has a quick filter in the number of diamonds to plot and renders them as quick barplot.

from shiny import App, render, ui
from plotnine import ggplot, geom_bar, aes
from plotnine.data import diamonds

app_ui = ui.page_fluid(
    ui.h2("Quick Diamonds Example"),
    ui.input_slider("n", "Number of diamonds", 10000, 20000, 1000),
    ui.output_plot("diamonds_plot")
)


def server(input, output, session):
    @output
    @render.plot
    def diamonds_plot():
        p = (ggplot(diamonds.sample(n = input.n()),
                   aes(x = "cut", fill = "cut")) 
            + geom_bar())
        return p


app = App(app_ui, server)

Having written and tested the app (in my case inventively caused diamonds), we can build it (presuming everything required has been installed) via

shinylive export diamonds site

This creates a folder called site containing all the bits and pieces we need. Finally, we can embed this in our blog via

knitr::include_app("site/index.html", height = "600px")