RShiny UI checkbox for suboptions?

Question

I have a basic RShiny app that has a reactive checkbox which plots timeseries data based on the data (column of df) selected in the checkbox. My current code produces a UI with checkbox inputs like this:

    # Load R packages
library(shiny)
library(shinyBS)

##example df in similar format to the data I'm working with
Both <- data.frame(
  Year  = c("1990", "1991", "1992", "1993"),
  SST_anomaly_GOM = c("-1.1", "0.23", "0.87", "-0.09"),
  SST_anomaly_GB = c("-1.1", "0.23", "0.87", "-0.09"),
  SST_anomaly_MAB = c("-1.1", "0.23", "0.87", "-0.09"),
  BT_anomaly_GOM = c("-2.5", "0.55", "1.20", "-0.19"),
  BT_anomaly_GB = c("-1.1", "0.05", "1.24", "-0.29"),
  BT_anomaly_MAB = c("-1.1", "-1.08", "0.67", "-2.40")
)

# Define UI
ui <- fluidPage(
  # useShinyBS
    "Visualizing Indicators", #app title
    tabPanel("",      # tab title
             sidebarPanel(width=6,
                          
                          checkboxGroupInput("variable", label = "Checkbox", choiceNames  = gsub("_", " ", colnames(Both[2:7])), 
                                                choiceValues = colnames(Both[2:7]), 
                                                ),
             ), # sidebarPanel
    ), #tabPanel
) # fluidPage

#Define Server:
server<- function (input,output){
   output$rendered <-   renderUI({
    })
}
# Create Shiny object
shinyApp(ui = ui, server = server)

This Produces an interface like this:

This is fine, but a little repetitive, and with more timeseries variables I eventually want to include to this list, this can get cumbersome for the user to sift through and will take up a lot of the space on the UI to list everything in this way.

My question is how can I adjust my code such that it produces an interface with unique variables listed, then checkboxes for each sub-region of interest? (GOM, BG, MAB, etc.) An example of what I have in mind is an interface that looks more like this:

Is this possible? Is this possible with a df in the format as I currently have (such as my example df called "Both").

Thanks!

Pterous

How about using a dropdown menu using selectInput with multiple = TRUE for the regions? – Jehol

@Jehol I thought about that, but I like the idea of this format better because this way the user can select multiple options at once and can easily see their sections without having to click on something to see the options they chose – Pterous

Answer 1

To create an answer for your solution, I've implemented a checkbox group input using the DT package. The solution comes in two parts: 1. The Helper functions. 2. The App.

Example Image

Helper Functions

The first helper function creates a data table with checkbox inputs, each with a unique id that is a combination of the rowname and column name.

The second helper function evaluates the 'checked' status of each of the checkboxes in the constructed table, returning a matrix with TRUE/FALSE values for each of the cells in the checkbox table.

App

The app code is pretty straight forward.

First, we create an example table using the first helper function.

Then, we render the table with DT, making sure to disable escape (so the checkboxes can be rendered), sorting, paging, and selection on the table. Most importantly, we send preDrawCallback and drawCallback JS functions to make sure the checkboxes are registered with shiny.

Lastly, any time the user interacts with the table, we call our second helper function to evaluate the checkbox statuses. You can do whatever you please with that information.

Code

# Checkbox Table Demo

library(shiny)
library(DT)


#### Helper Functions ####
#' Construct a checkbox table for an app.
construct_checkbox_table <- function(rows,
                                     cols,
                                     rownames,
                                     colnames) {
  checkbox_table <- matrix(
    character(),
    nrow = rows,
    ncol = cols,
    dimnames = list(rownames, colnames)
  )
  
  for (i in seq_len(rows)) {
    for (j in seq_len(cols)) {
      checkbox_table[i, j] <-
        sprintf(
          '<input id="%s,%s" type="checkbox" class="shiny-bound-input" />',
          rownames[[i]],
          colnames[[j]]
        )
    }
  }
  
  checkbox_table
}

#' Get the status of checkboxes in a checkbox table.
evaluate_checkbox_table_status <- function(input, input_table) {
  table_status <-
    matrix(
      logical(),
      nrow = nrow(input_table),
      ncol = ncol(input_table),
      dimnames = list(rownames(input_table), colnames(input_table))
    )
  
  table_rownames <- rownames(input_table)
  table_colnames <- colnames(input_table)
  
  for (i in seq_len(nrow(input_table))) {
    for (j in seq_len(ncol(input_table))) {
      table_status[i, j] <-
        input[[sprintf("%s,%s", table_rownames[[i]], table_colnames[[j]])]]
    }
  }
  
  table_status
}
#### End Helper Functions ####


#### App ####
# Create an example checkbox input table to use for the app
example_checkbox_table <-
  construct_checkbox_table(
    2,
    4,
    rownames = c("Annual Bottom Temp Absolute", "Bottom Temp Anomoly"),
    colnames = c("GOM", "GB", "MAB", "SS")
  )

ui <- fluidPage(DT::DTOutput("selection_table"),
                verbatimTextOutput("table_selections"),)

server <- function(input, output, session) {
  output$selection_table <- DT::renderDT({
    DT::datatable(
      example_checkbox_table,
      escape = FALSE,
      selection = "none",
      options = list(
        dom = "t",
        ordering = FALSE,
        paging = FALSE,
        preDrawCallback = JS(
          'function() { Shiny.unbindAll(this.api().table().node()); }'
        ),
        drawCallback = JS(
          'function() { Shiny.bindAll(this.api().table().node()); } '
        )
      )
    )
  }, server = FALSE)
  
  observeEvent(input$selection_table_cell_clicked, {
    output$table_selections <- renderPrint({
      evaluate_checkbox_table_status(input, example_checkbox_table)
    })
  })
}
#### End App ####

shinyApp(ui, server)

Unsound

thanks @Jacob Bumgarner, this looks good for the UI part. However because it creates a new table instead of directly calling from the dataframe "Both", how would I modify the code such that for example, checking the checkbox in the table corresponding to "Bottom Temperature Anomaly" for the GOM region, would call the "BT_anomaly_GOM" column in the dataset? I am looking to eventually plot the timeseries in the server based on the checkbox selections rather than having a table that identifies which boxes were checked as you wrote in your code. Thanks! – Pterous

There are a variety of ways that you could do this, but it's mostly up to you. One option would be to modify the code so that the id of each of the checkboxes matches one of your Both dataframe columns. Another option would be to encode the Both column names into a matrix of the same dimension as the example_checkbox_table (let's call it encoded_matrix). You could then mask the encoded_matrix with the output of the evaluate_checkbox_table_status function to identify the Both column names you'd like to select. Do either of those options make sense? – Unsound

@Jamie_B: can you post the solution if you end up getting it to work? Thanks! – Jehol

@Jehol - if you can indicate which solution you prefer, I'd be happy to edit my answer to improve it. Just let me know which solution you'd rather see implemented – Unsound

@JacobBumgarner: actually both options you mentioned make sense to me. Maybe the 1st option is easier for you to implement? Thanks – Jehol

@Tung, yes I can! I haven't had much time lately to take an additional stab at it, but if get something working, I will definitely post it. – Pterous