How to get ggplot in shiny to recognize input$ as a vector?

[kate1]

Hi all - This is my first post ever, so apologies for any missteps in this question. I am a statistical programmer in SAS, which is quite different from R and Shiny, and I am getting stuck on language and syntax. For this question, I have already googled around on this and other boards, and have tried many things. My gut tells me I'm missing something simple.

In my app, I want the user to select a variable (aka column, feature) from the dataset, and then have my scatter plot (aka geom_point) color the points according to that variable. These variables are all factor types and have different numbers of levels. Also, I want them each to use a different palette, which I set manually.

In the ggplot call, I try to refer to the name of the palette with a reactive (aka dynamic, macro) variable. But it is interpreted as a string of length 1, not as a reference to the palette vector I have already set up. (The equivalent in SAS would be a double ampersand (&&) reference to a macro variable within a macro.)

I have created a toy example and used reprex() to paste below.

Thank you so much,
Kate

library("shiny")
library("ggplot2")
library("tidyverse")

# Make a toy dataset
    
    x <- c(1,2,3,4,5)
    y <- c(52,49,19,15,31)
    grouping1 <- as.factor(c(1,1,1,2,2))
    grouping2 <- as.factor(c(1,2,3,4,5))
    toydataset <- data.frame(x,y,grouping1,grouping2)
    
# Make palettes to apply to each grouping
    
    palette1 <- c("blue","red")
    palette2 <- c("orange","yellow","green","blue","purple")
    
# the UI bit:
ui <- fluidPage(
    titlePanel("My question"),
    sidebarLayout(
        sidebarPanel(
            selectInput(inputId = "selectedvariable",
                        label = "Select a variable", 
                        choices = c("grouping1","grouping2")),
                    ),
        mainPanel(
           plotOutput("myplot")
        )
    )
)

# the server bit:
server <- function(input, output) {
    
    currentvariable <- reactive({input$selectedvariable})
    currentpalette <- reactive({
        if (input$selectedvariable == "grouping1"){palette1}
        else if (input$selectedvariable == "grouping2"){palette2}
        })
    output$myplot <- renderPlot({
        ggplot() +
            geom_point(data = toydataset,
                       aes(x=x, y=y, color = currentvariable() )) +
            scale_color_manual(values= (currentpalette() ))
            })
}

# Run it 
shinyApp(ui = ui, server = server)

Shiny applications not supported in static R Markdown documents

^{Created on 2019-12-08 by the reprex package (v0.3.0)}

[andresrcs]

You can use aes_string()

library("shiny")
library("ggplot2")
library("tidyverse")

# Make a toy dataset

x <- c(1,2,3,4,5)
y <- c(52,49,19,15,31)
grouping1 <- as.factor(c(1,1,1,2,2))
grouping2 <- as.factor(c(1,2,3,4,5))
toydataset <- data.frame(x,y,grouping1,grouping2)

# Make palettes to apply to each grouping

palette1 <- c("blue","red")  
palette2 <- c("orange","yellow","green","blue","purple")

# the UI bit:
ui <- fluidPage(
    titlePanel("My question"),
    sidebarLayout(
        sidebarPanel(
            selectInput(inputId = "selectedvariable",
                        label = "Select a variable", 
                        choices = c("grouping1","grouping2")),
        ),
        mainPanel(
            plotOutput("myplot")
        )
    )
)

# the server bit:
server <- function(input, output) {

    currentpalette <- reactive({
        if (input$selectedvariable == "grouping1"){palette1}
        else if (input$selectedvariable == "grouping2"){palette2}
    })
    output$myplot <- renderPlot({
        ggplot() +
            geom_point(data = toydataset,
                       aes_string(x = "x", y = "y", color = input$selectedvariable)) +
            scale_color_manual(values = currentpalette())
    })
}

# Run it 
shinyApp(ui = ui, server = server)

[kate1]

Hooray! That works wonderfully, thank you so much.
Can I ask a quick followup? I did try to play with aes_string because I'd found it elsewhere as a possible solution. But I was putting "input$selectedvariable" into quotes as well, and it didn't work.
Is there a conceptual reason why, under aes_string, the x and y aesthetics are in quotes but the color is not? How is R intepreting quoted content?
I am imagining that aes_string basically says to R: anything in quotes you should interpret as an object or vector, and anything not in quotes you should replace with its reactive value, and then interpret THAT thing as an object or vector.

[izahn]

The key is to pay attention to what the argument is rather than whether it is assigned or created directly. Here is a simpler example:

xvar <- "x"
yvar <- "y"
input <- list(selectedvariable = "grouping1")

ggplot() +
  geom_point(data = toydataset,
             aes_string(x = xvar, y = yvar, 
                        color = input$selectedvariable))

Note that aes_string("color = input$selectedvariable") doesn't work for the same reason that aes_string(x = "xvar", y = "yvar") doesn't work: the string should contain the name of the variable, not the name of the object containing the variable.