Gordon Shotwell - Understanding reactive execution in Shiny | PyData Global 2023

Discover the power of reactive execution in Shiny, a framework that infers relationships between components, minimizing re-execution and enabling efficient caching and notification of downstream dependencies.

Key takeaways

Reactive execution in Shiny is appealing because it allows for efficient re-executions without needing to manually define callbacks.
Shiny’s strategy is to infer relationships between components, building a computation graph, and then using that graph to minimally re-execute the application.
Shiny apps use reactive calculations (like reactiveCalc) to represent values, and reactive effects (like reactiveEffect) to mark effects, such as fetching data from a database.
Shiny’s reactive graph can help with caching and notify downstream dependencies when upstream parts of the graph change.
Shiny’s reactivity is great for applications with complex dynamics and changing user inputs.
Streamlit, Dash, and other frameworks may not perform as well in these situations, and programmers may need to manually manage state.
Shiny’s reactivity allows for automatic rendering of plots and widgets without needing to manually update state.
The tourism example demonstrates a simple Shiny app that changes its behavior based on user input.
ReactiveIsolate can be used to prevent re-reactions to certain inputs.
Shiny has a “range” of capabilities, from simple to complex, and can handle all the layers of a data science application.

Gordon Shotwell - Understanding reactive execution in Shiny | PyData Global 2023

More talks