Tips and Tricks

It can be challenging to troubleshoot or design the output of reactive data elements in a flexdashboard. Very often, to see a change, the builder has to run the dashboard in order to see the output. This package is designed to alleviate this challenge by turning your reactive objects into functions you can interact with in the console. Here are a few tips that will help you (and others) troubleshoot your code:

Create a chunk to hold dummy input values

If you create a chunk like the code below, you will be able to simulate reactive values like input$displ and input$year without having to run the whole dashboard. The input object created here is a list containing 3 elements (displ, year, and drv). By creating the list named input in a chunk of your code, you can simulate the way the values will work when it actually runs. Do note, you will need to use eval = FALSE in the chunk so this part of your code will be ignored when the dashboard is run.

```{r input_demo, eval = FALSE}
input <-list(
  displ = 1.8,
  year = 2008,
  drv = "f"
)

```

With the dummy input object created earlier, I am able to run the df and subsequent ggplot() code locally in the console. As a note, running all of renderPlot() will only show text in the console but if you run the df <- section and the ggplot() section, you will have access to df in your environment and see the bar chart in the plot pane of RStudio.

library(tidyverse)
library(shiny)

raw_data <- mpg

renderPlot({
  df <- 
    raw_data %>% 
    filter(
      displ >= input$displ,
      year == input$year,
      drv == input$drv
    )

  ggplot(df, aes(class)) +
    geom_bar()
})

Put reactive objects within their reactive outputs

I often see something like the code below. Notice that there are two reactive steps: reactive() to create the reactive data frame and then renderPlot().

reactive_df <- reactive(
  raw_data %>% 
    filter(displ >= input$displ)
)

renderPlot(
  ggplot(reactive_df(), aes(class)) +
    geom_bar()
)

If this reactive data frame is created only for this one plot, you can embed the data manipulation within the renderPlot() function. If this data will be used thin two or more outputs in the dashboard and has a lot of data manipulation, a reactive dataframe is a good idea.

renderPlot({
  df <-
    raw_data %>% 
    filter(displ >= input$displ)
    
  ggplot(df, aes(class)) + geom_bar()
})

When you add curly braces {...} inside the renderPlot() it creates a mini environment where you can create multiple objects. This is similar to how you might do the same in a function() {...} or a loop for(i in 1:10) {...}

In summary