A few years ago, I wrote about making GUI applications in Python. However, in my experience, most of the time what we really need are just dashboards, simple Python apps that visualize data, with interactive filtering and faceting. One path of low (if not least) resistance is of course iPython notebooks, but there are increasingly more feature-rich options, depending on your particular needs and constraints.

HoloViews

Typically, output in iPython notebooks like dataframes or figures are rendered as static inline “chunks”. HoloViews allows these chunks to contain interactive plots. I actually wrote about HoloViews briefly in the past.

Individual plots can be composed together in the same chunk using the + operator used in the “grammar of graphics” style. Facetting is also possible. Individual Holoview elements can be exported as HTML files, but the overall experience still feels like a notebook since the underlying Python code REPL is still visible.

Plotly Dash

Dash is built on top of Flask, Plotly and React. If you have used R and Shiny previously, Dash will feel very familiar. The entire dashboard is invoked using a single Dash() call, visual widgets are added and interactivity is handled via callbacks. The experience feels more like a dashboard: by default no REPL is exposed, and code changes require the entire app to be rebuilt and restarted.

Streamlit

The latest visualization framework is Streamlit (started 2019). Its widgets look a bit more modern with sensible defaults that contribute to an easier learning curve and faster prototyping. Compared to Dash, Steamlit has a smaller (but rapidly growing) user community.