Updating Source in Shiny

A common scenario when using shinyCohortBuilder in custom Shiny application is to allow a user to switch between multiple data sources.

The goal of this document is to explain how shinyCohortBuilder can be used in the two scenarios.

Whenever you use different data in your Shiny application you should use cohortBuilder::update_source method to update underlying data.

The method not only replaces Source in the Cohort object, but also triggers set of specific operations in filtering panel.

The operations depends on the way you store filters configuration and extra options passed to the method.

Below we describe two most common solutions when dealing with the above scenarios.

Consistent data

When working with consistent datasets you usually want to keep filtering panel unchanged for each data version. In this situation:

Configure filtering steps only once when initializing cohort.
When updating source, use update_source(cohort_object, keep_steps = TRUE).

As a result:

configured filtering steps will remain unchanged,
no additional rendering in filtering panel will take place in Shiny what results faster update,
all the filters in panel will be updated having new data filtering options (choices, ranges etc.).

Below application presents the described approach:

library(shiny)
library(cohortBuilder)
library(shinyCohortBuilder)

mtcars_list = list(
  "0" = dplyr::filter(mtcars, am == 0),
  "1" = dplyr::filter(mtcars, am == 1)
)

ui <- fluidPage(
  sidebarLayout(
    sidebarPanel(
      radioButtons("version", "Version", choices = c("0", "1")),
      cb_ui("mtcars")
    ),
    mainPanel(
      verbatimTextOutput("cohort_data")
    )
  )  
)

server <- function(input, output, session) {
  init_source <- set_source(tblist(mtcars = mtcars_list[["0"]]))
  mt_cohort <- cohort(
    init_source,
    filter("range", id = "mpg", dataset = "mtcars", variable = "mpg", active = FALSE),
    filter("range", id = "qsec", dataset = "mtcars", variable = "qsec", active = FALSE)
  )
  
  cb_server("mtcars", mt_cohort)
  
  observeEvent(input$version, {
    new_source <- set_source(tblist(mtcars = mtcars_list[[input$version]]))
    update_source(mt_cohort, new_source, keep_steps = TRUE)
  })
}

shinyApp(ui, server)

Inconsistent data

When filters configuration is different across multiple data sources you want to render the filtering panel from scratch every time the source is updated. In this scenario:

Keep filtering steps configuration within a source.
When updating source, use update_source(cohort_object, keep_steps = FALSE).

As a result:

configured filtering steps in Source will be used to update Cohort configuration,
the previous filtering panel will be removed,
the new filtering panel will be rendered based on the new configuration.

Below application presents the described approach:

library(shiny)
library(cohortBuilder)
library(shinyCohortBuilder)

source_mtcars <- set_source(
  tblist(mtcars = mtcars),
  filter("range", id = "mpg", dataset = "mtcars", variable = "mpg", active = FALSE),
  filter("range", id = "qsec", dataset = "mtcars", variable = "qsec", active = FALSE)
)

source_iris <- set_source(
  tblist(iris = iris),
  filter("discrete", id = "species", dataset = "iris", variable = "Species", active = FALSE),
  filter("range", id = "petal_length", dataset = "iris", variable = "Petal.length", active = FALSE)
)

ui <- fluidPage(
  sidebarLayout(
    sidebarPanel(
      radioButtons("dataset", "Dataset", choices = c("mtcars", "iris")),
      cb_ui("data_panel")
    ),
    mainPanel(
      verbatimTextOutput("cohort_data")
    )
  )  
)

server <- function(input, output, session) {
  cohort_object <- cohort(source_mtcars)
  cb_server("data_panel", cohort_object)
  
  observeEvent(input$dataset, {
    if (input$dataset = "mtcars") {
      update_source(cohort_object, source_mtcars, keep_steps = FALSE)  
    }
    if (input$dataset = "iris") {
      update_source(cohort_object, source_iris, keep_steps = FALSE)  
    }
  })
}

shinyApp(ui, server)