Chapter 16 The shiny Framework

Adding interactivity to a data report is a highly effective way of communicating that information and enabling users to explore a data set. In this chapter, you will learn about the Shiny framework for building interactive applications in R. Shiny provides a structure for communicating between a user-interface (i.e., a web-browser) and an R session, allowing users to interactively change the “code” that is run and the data that are output. This not only enables developers to create interactive graphics, but provides a way for users to interact directly with an R session (without writing any code!).

16.1 Creating Shiny Apps

Shiny is a web application framework for R. As opposed to a simple (static) web page like you’ve created with R Markdown, a web application is an interactive, dynamic web page—the user can click on buttons, check boxes, or input text in order to change the presentation of the data. Shiny is a framework in it provides the “code” for producing and enabling this interaction, while you as the developer simply “fill in the blanks” by providing variables or functions that the provided code will utilize to create the interactive page.

shiny is another external package (like dplyr and ggplot2), so you will need to install and load it in order to use it:

install.packages("shiny")  # once per machine

This will make all of the framework functions and variables you will need to worth with available.

16.1.1 Application Structure

Shiny applications are divided into two parts:

  1. The User Interface (UI) defines how the application will be displayed in the browser. The UI can render R content such as text or graphics just like R Markdown, but it can also include widgets, which are interactive controls for your application (think buttons or sliders). The UI can specify a layout for these components (e.g., so you can put widgets above, below, or beside one another).

    The UI for a Shiny application is defined as a value, usually one returned from calling a layout function. For example:

    # The ui is the result of calling the `fluidPage()` layout function
    my.ui <- fluidPage(
      # A widget: a text input box (save input in the `username` key)
      textInput('username', label="What is your name?"),
      # An output element: a text output (for the `message` key)

    This UI defines a fluidPage (where the content flows “fluidly” down the page), that contains two content elements: a text input box where the user can type their name, and some outputted text based on the message variable.

  2. The Server defines the data that will be displayed through the UI. You can think of this as an interactive R script that the user will be able to “run”: the script will take in inputs from the user (based on their interactions) and provide outputs that the UI will then display. The server user reactive expressions, which are like functions that will automatically be re-run whenever the input changes. This allows the output to be dynamic and interactive.

    The Server for a Shiny application is defined as a function (as opposed to the UI which is a value). This function takes in two lists as argments: an input and output. It then uses render functions and reactive expressions that assign values to the output list based on the input list. For example:

    # The server is a function that takes `input` and `output` args
    my.server <- function(input, output) {
      # assign a value to the `message` key in `output`
      # argument is a reactive expression for showing text
      output$message <- renderText({
        # use the `username` key from input and and return new value
        # for the `message` key in output
        return(paste("Hello", input$username))

Combined, this UI and server will allow the user to type their name into an input box, and will then say “hello” to whatever name is typed in.

More details about the UI and server components can be found in the sections below. Combining UI and Server

There are two ways of combining the UI and server:

The first (newer) way is to define a file called app.R. This file should call the shinyApp() function, which takes a UI value and Server function as arguments. For example:

# pass in the variables defined above
shinyApp(ui = my.ui, server = my.server)

Executing the shinyApp() function will start the App (you can also click the “Run App” button at the top of RStudio).

  • Note: if you change the UI or the Server, you do not need to stop and start the app; you can simply refresh the browser or viewer window and it will reload with the new UI and server.

  • If you need to stop the App, you can hit the “Stop Sign” icon on the RStudio console.

Using this function allows you to define your entire application (UI and Server) in a single file (which must be named app.R). This approach is good for simple applications that you wish to be able to share with others, since the entire application code can be listed in a single file.

However, it is also possible to define the UI and server as separate files. This allows you to keep the presentation (UI) separated from the logic (server) of your application, making it easier to maintain and change in the future. To do this, you define two separate files: ui.R for the UI and server.R for the Server (the files must be named ui.R and server.R). In these files, you must call the functions shinyUI() and shinyServer() respectively to create the UI and server, and then RStudio will automatically combine these files together into an application:

# In ui.R file
my.ui <- fluidPage(
  # define widgets

# In server.R file
my.server <- function(input, output) {
  # define output reactive expressions


You can then run the app by using the “Run App” button at the top of RStudio:

Use RStudio to run a shiny app defined in separate UI and Server files.

Use RStudio to run a shiny app defined in separate UI and Server files.

This chapter will primarily use the “single file” approach for compactness and readability, but you are encouraged to break up the UI and server into separate files for your own, larger applications.

  • Note that it is also possible to simply define the (e.g.) my.ui and my.server variables in separate files, and then use source() to load them into the app.R file and pass them into shinyApp().

16.1.2 The UI

The UI defines how the app will be displayed in the browser. You create a UI by calling a layout function such as fluidPage(), which will return a UI definition that can be used by the shinyUI() or shinyApp() functions.

You specify the “content” that you want the layout to contain (and hence the app to show) by passing each content element (piece of content) as an argument to that function:

# a "pseudocode" example, calling a function with arguments
ui <- fluidPage(element1, element2, element3)

Content elements are defined by calling specific functions that create them: for example h1() will create an element that has a first-level heading, textInput() will create an element where the user can enter text, and textOutput will create an element that can have dynamic (changing) content. Usually these content elements are defined as nested (anonymous) variables, each on its own line:

# still just calling a function with arguments!
ui <- fluidPage(
  h1("My App"),  # first argument
  textInput('username', label="What is your name?"),  # second argument
  textOutput('message')  # third argument

Note that layout functions themselves return content elements, meaning it is possible to include a layout inside another layout. This allows you to create complex layouts by combining multiple layout elements together. For example:

ui <- fluidPage(   # UI is a fluid page
  titlePanel("My Title"),  # include panel with the title (also sets browser title)

  sidebarLayout(   # layout the page in two columns
    sidebarPanel(  # specify content for the "sidebar" column
      p("sidebar panel content goes here")
    mainPanel(     # specify content for the "main" column
      p("main panel content goes here")

See the Shiny documentation and gallery for details and examples of doing complex application layouts.

Fun Fact: much of Shiny’s styling and layout structure is based on the Bootstrap web framework.

You can include static (unchanging) content in a Shiny UI layout—this is similar to the kinds of content you would write in Markdown (rather than inline R) when using R Markdown. However, you usually don’t specify this content using Markdown syntax (though it is possible to include a markdown file’s content). Instead, you include content functions that produce HTML, the language that Markdown is converted to when you look at it in the browser. These functions include:

  • p() for creating paragraphs, the same as plain text in Markdown
  • h1(), h2(), h3() etc for creating headings, the same as # Heading 1, ## Heading 2, ### Heading 3 in Markdown
  • em() for creating emphasized (italic) text, the same as _text_ in Markdown
  • strong() for creating strong (bolded) text, the same as **text** in Markdown
  • a(text, href='url') for creating hyperlinks (anchors), the same as [text](url) in Markdown
  • img(text, src='url') for including images, the same as ![text](url) in Markdown

There are may other methods as well, see this tutorial lesson for a list. If you are familiar with HTML, then these methods will seem familiar; you can also write content in HTML directly using the tag() content function. Control Widgets and Reactive Outputs

It is more common to include control widgets as content elements in your UI layout. Widgets are dynamic (changing) control elements that the user can interact with. Each stores a value that the user has entered, whether by typing into a box, moving a slider, or checking a button. When the user changes their input, the stored value automatically changes as well.

Examples of control widgets (image from

Examples of control widgets (image from

Like other content elements, widgets are created by calling an appropriate function. For example:

  • textInput() creates a box in which the user can enter text
  • sliderInput() creates a slider
  • selectInput() creates a dropdown menu the user can choose from
  • checkboxInput() creates a box the user can check (using checkboxGroupInput() to group them)
  • radioButtons() creates “radio” buttons (which the user can select only one of at a time)

See the documentation, and this tutorial lesson for a complete list.

All widget functions take at least two arguments:

  • A name (as a string) for the widget’s value. This will be the “key” that will allow the server to be able to access the value the user has inputted (think: the key in the input list).
  • A label (a string or content element described above) that will be shown alongside the widget and tell the user what the value represents. Note that this can be an empty string ("") if you don’t want to show anything.

Other arguments may be required by a particular widget—for example, a slider’s min and max values:

# this function would be nested in a layout function (e.g., `fluidPage()`)
sliderInput('age',              # key this value will be assigned to
            "Age of subjects",  # label
            min = 18,           # minimum slider value
            max = 80,           # maximum slider value
            value = 42          # starting value

Widgets are used to provide inputs to the Server; see the below section for how to use these inputs, as well as examples from the gallery.

In order to display outputs from the Server, you include a reactive output element in your UI layout. These are elements similar to the basic content elements, but instead of just displaying static (unchanging) content they can display dynamic (changing) content produced by the Server.

As with other content elements, reactive outputs are creating by calling an appropriate function. For example:

  • textOutput() displays output as plain text (note this output can be nested in a content element for formatting)
  • tableOutput() displays output as a data table (similar to kable() in R Markdown). See also dataTableOutput() for an interactive version!
  • plotOutput() displays a graphical plot, such as one created with ggplot2

Each of these functions takes as an argument the name (as a string) of the value that will be displayed. This is the “key” that allows it to access the value the Server is outputting. Note that the functions may take additional arguments as well (e.g., to specify the size of a plot); see the documentation for details.

16.1.3 The Server

The Server defines how the data inputted by the user will be used to create the output displayed by the app—that is, how the control widgets and reactive outputs will be connected. You create a Server by defining a new function (not calling a provided one):

server <- function(input, output){
  # assign values to `output` here

Note that this is just a normal function that happens two take lists as arguments. That means you can include the same kinds of code as you normally would—though that code will only be run once (when the application is first started) unless defined as part of a reactive expression.

The first argument is a list of any values defined by the control widgets: each name in a control widget will be a key in this list. For example, using the above sliderInput() example would cause the list to have an age key (referenced as input$age). This allows the Server to access any data that the user has inputted, using the key names defined in the UI. Note that the values in this list will change as the user interacts with the UI’s control widgets.

The purpose of the Server function is to assign new values to the output argument list (each with an appropriate key). These values will then be displayed by the reactive outputs defined in the UI. To make it so that the values can actually be displayed by by the UI, the values assigned to this list need to be the results of Render Functions. Similar to creating widgets or reactive outputs, different functions are associated with different types of output the server should produce. For example:

  • renderText() will produce text (character strings) that can be displayed (i.e., by textOutput() in the UI)
  • renderTable() will produce a table that can be displayed (i.e., by tableOutput() in the UI)
  • renderPlot() will produce a graphical plot that can be displayed (i.e., by plotOutput() in the UI)

Render functions take as an argument a Reactive Expression. This is a lot like a function: it is a block of code (in braces {}) that returns the value which should be rendered. For example:

output$msg <- renderText({
  # code goes here, just like any other function
  my.greeting <- "Hello"

  # code should always draw upon a key from the `input` variable
  message <- paste(my.greeting, input$username)

  # return the variable that will be rendered

The only difference between writing a reactive expression and a function is that you only include the block (the braces and the code inside of them): you don’t use the keyword function and don’t specify a set of arguments.

This technically defines a closure, which is a programming concept used to encapsulate functions and the context for those functions.

These reactive expressions will be “re-run” every time one of the input values that it references changes. So if the user interacts with the username control widget (and thereby changes the value of the input list), the expression in the above renderText() will be executed again, returning a new value that will be assigned to output$msg. And since output$msg has now changed, any reactive output in the UI (e.g., a textOutput()) will update to show the latest value. This makes the app interactive! Multiple Views

It is quite common in a Shiny app to produce lots of output variables, and thus to have multiple reactive expressions. For example:

server <- function(input, output) {
  # render a histogram plot
  output$hist <- renderPlot({
    uniform.nums <- runif(input$num, 1, 10)  # random nums between 1 and 10
    return( hist(uniform.nums) )  # built-in plotting for simplicity

  # render the counts
  output$counts <- renderPrint({
    uniform.nums <- runif(input$num, 1, 10)  # random nums between 1 and 10
    counts <- factor(cut(uniform.nums, breaks=1:10))  # factor
    return( summary(counts) )  # simple vector of counts

If you look at the above example though, you’ll notice that each render function produces a set of random numbers… which means each will produce a different set of numbers! The histogram and the table won’t match!

This is an example of where you want to share a single piece of data (a single model) between multiple different renditions (multiple views). Effectively, you want to define a shared variable (the uniform.nums) that can be referenced by both render functions. But since you need that shared variable to be able to update whenever the input changes, you need to make it be a reactive expression itself. You can do this by using the reactive() function:

server <- function(input, output) {
  # define a reactive variable
  uniform.nums <- reactive({
    return( runif(input$num, 1, 10) )  # just like for a render function

  # render a histogram plot
  output$hist <- renderPlot({
    return( hist(uniform.nums()) )  # call the reactive variable AS A FUNCTION

  # render the counts
  output$counts <- renderPrint({
    counts <- factor(cut(uniform.nums(), breaks=1:10))  # call the reactive variable AS A FUNCTION
    return( summary(counts) )

The reactive() function lets you define a single “variable” that is a reactive function which can be called from within the render functions. Importantly, the value returned by this function (the uniform.nums()) only changes when a referenced input changes. Thus as long as input$num stays the same, uniform.nums() will return the same value.

This is very powerful for allowing multiple views of a single piece of data: you can have a single source of data displayed both graphically and textually, with both views linked off of the same processed data table. Additionally, it can help keep your code more organized and readable, and avoid needing to duplicate any processing.

16.2 Publishing Shiny Apps

Sharing a Shiny App with the world is a bit more involved than simply pushing the code to GitHub. You can’t just use GitHub pages to host the code because, in addition to the HTML UI, you need an R interpreter session to run the Server that the UI can connect to (and GitHub does not provide R interpreters)!

While there are a few different ways of “hosting” Shiny Apps, in this course you’ll use the simplest one: hosting through is a platform for hosting and running Shiny Apps; while large applications cost money, anyone can deploy a simple app (like the ones you’ll create in this course) for free.

In order to host your app on, you’ll need to create a free account. Note that you can sign up with GitHub or your Google/UW account. Follow the site’s instructions to

  1. Select an account name (use something professional, like you used when siging up with GitHub)
  2. Install the required rsconnect package (may be included with RStudio)
  3. Set your authorization token (“password”). Just click the green “Copy to Clipboard” button, and then paste that into the Console in RStudio. You should only need to do this once.

    Don’t worry about “Step 3 - Deploy”; you’ll do that through RStudio directly!

After you’ve set up an account, you can Run your application (as above) and hit the Publish button in the upper-right corner:

How to publish a running Shiny App to

How to publish a running Shiny App to

This will put your app online, available at

Important Publishing to is one of the major “pain points” in working with Shiny. For the best experience, be sure to:

  1. Always test and debug your app locally (e.g., on your own computer, by running the App through RStudio). Make sure it works on your machine before you try to put it online.

  2. Use correct folder structures and relative paths. All of your app should be in a single folder (usually named after the project). Make sure any .csv or .R files referenced are inside the app folder, and that you use relative paths to refer to them. Do not include any setwd() statements in your code; you should only set the working directory through RStudio (because will have its own working directory).

  3. It is possible to see the logs for your deployed app, which may include errors explaining any problems that arise when you deploy your app.

For more options and details, see the documentation.