Shiny for R

The R and Python Shiny packages are quite similar, and as a result if you know how to build a Shiny app in R you are well on your way to building one in Python. All of the main components of Shiny like reactivity, rendering functions, and modules are the same, and both packages use the same JavaScript code. There are, however, a few differences that you need to keep in mind in order to build effective Shiny applications in Python. If you’re reading this, we expect that you are an existing R Shiny user with some Python knowledge.

Shiny Express

Shiny express is a new, more expressive, way to build PyShiny apps. It is not available in R, so the comparisons drawn below are only relevant to core (i.e., non-express) apps.

Getting started

R users tend to use the R console to install and run Shiny while Python requires you to use the terminal. To get started you can do the following (or see the installation instructions for a more in-depth explanation):

In your terminal, create a new directory with mkdir <my_directory> and navigate into it with cd <my_directory>
Install Shiny. We strongly recommend using a virtual environment for this as it will eliminate dependency resolution headaches and make deployment easier.

Install with pip
Install with conda

# Create a virtual environment in the .venv subdirectory
python3 -m venv venv
source venv/bin/activate
# Install shiny
pip install shiny

# Create a conda environment named 'myenv'
conda create --name myenv
conda activate myenv
conda install -c conda-forge shiny

Call shiny create . to create an example shiny app in your directory
Call shiny run --reload to run the app and reload when the source code changes

Takeaways

Always use a virtual environment with Python projects
Install and run shiny from the command line, not a Python process

Syntax differences

There are five main syntax difference between the R and Python versions of Shiny:

Decorators instead of render functions
Function names are used to connect outputs to the UI
More precise namespaces
All inputs are invoked with input.<input_name>()
Some functions have different names

Decorators

Shiny for Python uses decorators instead of top level rendering functions. Decorators are just python functions which take other functions and are invoked by putting @<decorator> above the function definition. While R doesn’t have an exact analog to decorators they are similar to function operators you like purrr::safely.

Takeaways

Use rendering decorators like @render.plot, @render.text, or @render.ui instead of renderPlot(), renderText, or renderUI
Reactive calculations (equivalent to reactive expressions in R) are decorated @reactive.calc, and reactive effects (equivalent to observers in R) are decorated with @reactive.effect.

R

library(shiny)

ui <- fluidPage(
  sliderInput("n", "N", 0, 100, 40),
  verbatimTextOutput("txt")
)

server <- function(input, output, session) {
  output$txt <- renderText({
    paste0("n*2 is ", input$n, " * 2")
  })
}

shinyApp(ui, server)

Python

from shiny import ui, render, App

app_ui = ui.page_fluid(
    ui.input_slider("n", "N", 0, 100, 40),
    ui.output_text_verbatim("txt"),
)

def server(input, output, session):
    @render.text
    def txt():
        return f"n*2 is {input.n() * 2}"

app = App(app_ui, server)

Connecting outputs

Both R an Python use a special object type to connect server computations to UI components, but the interface is quite different. In R, we connect outputs to UI elements by assigning into the output object but since Python renderings are produced with decorators, we instead use the function name to connect the rendered object to its UI component.

Takeaways

In Python, we don’t define outputs by assigning to output$x
Use the function name to connect a server output to a UI element

R

library(shiny)

ui <- fluidPage(
  sliderInput("n", "N", 0, 100, 40),
  verbatimTextOutput("txt")
)

server <- function(input, output) {
  output$txt <- renderText({
    paste0("n*2 is ", input$n," * 2")
  })
}

shinyApp(ui, server)

Python

from shiny import ui, render, App

app_ui = ui.page_fluid(
    ui.input_slider("n", "N", 0, 100, 40),
    ui.output_text_verbatim("txt"),
)

def server(input, output, session):
    @render.text
    def txt():
        return f"n*2 is {input.n() * 2}"

app = App(app_ui, server)

Submodules

All of the Shiny R functions are in a single package namespace. On the Python side we make use of submodules to keep related functions together. Note that “submodules” in this case refers to the generic module which is not the same as shiny modules. For example, instead of sliderInput(), you would call ui.input_slider(), where the ui. refers to a submodule of the main shiny module.

Takeways

Python submodules make autocomplete easier
Some important namespaces include:
- ui for the UI elements
- render includes the rendering decorators
- reactive has reactive expressions and observers

R

library(shiny)

ui <- fluidPage(
  sliderInput("n", "N", 0, 100, 40),
  verbatimTextOutput("txt")
)

server <- function(input, output) {
  output$txt <- renderText({
    paste0("n*2 is ", input$n," * 2")
  })
}

shinyApp(ui, server)

Python

from shiny import ui, render, App

app_ui = ui.page_fluid(
    ui.input_slider("n", "N", 0, 100, 40),
    ui.output_text_verbatim("txt"),
)

def server(input, output, session):
    @render.text
    def txt():
        return f"n*2 is {input.n() * 2}"

app = App(app_ui, server)

Call inputs with `()`

In R reactive values and reactive expressions are retrieved with different syntax. Reactive values like input$value are retrieved like variables while reactive expressions are called like functions my_reactive().

This interface makes it seem like inputs and reactive expressions are different types of thing when in fact they’re the same type of thing. In Python we chose to require that all reactive objects be retrieved with a function call. So instead of calling input.value you use input.value().

R

library(shiny)

ui <- fluidPage(
  sliderInput("n", "N", 0, 100, 40),
  verbatimTextOutput("txt")
)

server <- function(input, output) {

  n <- reactive({input$n})

  output$txt <- renderText({
    paste0("n*2 is ", n()," * 2")
  })
}


shinyApp(ui, server)

Python

from shiny import ui, render, reactive, App

app_ui = ui.page_fluid(
    ui.input_slider("n", "N", 0, 100, 40),
    ui.output_text_verbatim("txt"),
)

def server(input, output, session):
    @reactive.calc
    def n():
      return input.n()

    @render.text
    def txt():
        return f"n*2 is {n() * 2}"

app = App(app_ui, server)

Takeaways

Access input values by calling the object like a function
input.x(), not input$x

Function name changes

The Python function names have been modified to make them easier to discover with tab completion. For example all python output functions start with output_ while the input functions start with input_. This means that you can type ui.ou and hit tab to see all of the available output functions. The Shiny R functions on the other hand all start with the element type (plotOutput, textInput), which makes it hard to see all of the input or output options. For the most part you can follow this naming pattern to find the function you’re looking for, but there are a number of functions that have different names in R and Python, the most important of which are listed below:

R Function	Python Equivalent
`observe`	`@reactive.effect`
`reactive`	`@reactive.calc`
`bindEvent`	`@reactive.event`
`reactiveEvent`	`@reactive.calc` with `@reactive.event`
`observeEvent`	`@reactive.effect` with `@reactive.event`
`htmlTemplate`	`page_template`
`tabPanelBody`	`navs_content`
`*Tab` (`insertTab`, `appendTab`, etc)	`nav_*` (`nav_insert`, `nav_append` etc)
`fluidRow`	`row`

Reactive programming

Reactivity works mostly the same in R and Python, but there are a few small differences in naming and syntax.

New names for `reactive()` and `observe()`

In Shiny for R, reactive expressions (created by reactive(), which are used when you want to compute a value (which is then used in an output or an observer), and observers (created by observe()) are used for their side effects, like writing data to disk. This is a common source of confusion because the names reactive() and observe() do not clearly express when they should be used. To help clarify this confusion we’ve renamed reactive() to @reactive.calc, and observe() to @reactive.effect in Python.

R

library(shiny)

ui <- fluidPage(
  sliderInput("n", "N", 0, 100, 40),
  verbatimTextOutput("txt"),
  actionButton("reset", "Reset")
)

server <- function(input, output, session) {

  val <- reactive({input$n})


  observe({
    input$reset
    updateSliderInput(session, "n", value = 40)
  })

  output$txt <- renderText({
    paste0("n*2 is ", val()," * 2")
  })
}

shinyApp(ui, server)

Python

from shiny import App, reactive, render, ui

app_ui = ui.page_fluid(
    ui.input_slider("n", "N", 0, 100, 40),
    ui.output_text_verbatim("txt"),
    ui.input_action_button("reset", "Reset"),
)

def server(input, output, session):
    @reactive.calc
    def val():
        return input.n()

    @reactive.effect
    def _():
        input.reset()
        ui.update_slider("n", value=40)

    @render.text
    def txt():
        return f"n*2 is {val() * 2}"

app = App(app_ui, server)

Reactive Values

In R, there are two types of reactive objects which store values:

Items in a reactiveValues object. This is a list-like object that contains multiple reactive values. (Note that the input object is a reactiveValues.)
Standalone reactiveVal objects.

In R, the way that you get values from a reactiveValues object differs from how you get it from a reactiveVal. To get the value of an item’s in a reactiveValues object, you would simply access it with input$x. However, for a standalone reactiveVal, you would invoke it like a function, with x().

vals <- reactiveValues(x = 1, y = 2)
z <- reactiveVal(3)

# Retrieve values
print(vals$x)
print(z())

In Shiny for Python, we’ve simplified things in the following ways:

There is no direct analog to R’s reactiveValues.
The analog of R’s standalone reactiveVal is reactive.value. (The input object in Python is a dictionary-like object containing individual reactive.value objects.)
Reactive values have can be retrieved with my_val() or my_val.get() and can be set with my_val.set().

There is no analog of reactiveValues in Python, but you can create something similar by using a dictionary of reactive.value objects.

vals = {
  "x": reactive.value(1),
  "y": reactive.value(2),
}

z = reactive.value(3)

# Retrieve values
print(vals.x())
print(z())

R

library(shiny)

ui <- fluidPage(
  actionButton("minus", "-1"),
  actionButton("plus", "+1"),
  br(),
  textOutput("value")
)

server <- function(input, output, session) {
  value <- reactiveVal(0)



  observeEvent(input$minus, {
    newValue <- value() - 1
    value(newValue)
  })


  observeEvent(input$plus, {
    newValue <- value() + 1
    value(newValue)
  })


  output$value <- renderText({
    value()
  })
}
shinyApp(ui, server)

Python

from shiny import *

app_ui = ui.page_fluid(
    ui.input_action_button("minus", "-1"),
    ui.input_action_button("plus", "+1"),
    ui.br(),
    ui.output_text("value"),
)

def server(input: Inputs, output: Outputs, session: Session):
    val = reactive.value(0)

    @reactive.effect
    @reactive.event(input.minus)
    def _():
        newVal = val() - 1
        val.set(newVal)

    @reactive.effect
    @reactive.event(input.plus)
    def _():
        newVal = val() + 1
        val.set(newVal)

    @render.text
    def value():
        return str(val())


app = App(app_ui, server)

Mutability

One of the biggest differences between R and Python is the mutability of objects.
In R, most objects are immutable. This means, for example, that if you pass a data frame to a function, that function cannot alter your copy of the data frame.
In Python, many objects are mutable. This means that they can be modified in place—modifying an object in one part of a program can cause it to be (unexpectedly) modified in another part of the program.

This difference in mutability has consequences for Shiny applications and for programming in general. In many cases, using .copy() in Python will help avoid bugs due to inadvertently altering mutable objects. See our page on mutability to understand the possible bugs and how to avoid them.

Getting started

Syntax differences

Decorators

R

Python

Connecting outputs

R

Python

Submodules

R

Python

Call inputs with ()

R

Python

Function name changes

Reactive programming

New names for reactive() and observe()

R

Python

Reactive Values

R

Python

Mutability

Call inputs with `()`

New names for `reactive()` and `observe()`