Shiny/ggvis Reactivity to Subset Plot Data

Question

After seeing a talk on GGVIS, I have been trying my hand unsuccessfully at creating my first Shiny/ggvis app. My plot works in R, but when I try to migrate it into a Shiny app for display on the web I get nothing. The radio buttons are displayed and from what I can tell seem to be working (I tested using the table from rStudio/Shiny/Reactivity tutorial, but it doesn't seem to place nice with my ggvis plot). I have been following the tutorials on Rstudio and using pieces from the demo folder in the ggvis source code. I am trying to create a simple kernel density plot that subsets data based on user input and displays the distribution. Attached is a reproducible example of what I have thusfar. I would appreciate any insight as to what I am missing here.

ui.R

# clear memory & load packages
      rm(list=ls())

      library(shiny)
      library(ggvis)

# Define UI for distribution application
    shinyUI(fluidPage(

# Application title
         titlePanel("Diamond Carats by Color/Cut"),

# Sidebar with controls to select subset
          sidebarLayout(
          sidebarPanel(
            radioButtons("cut", "Diamond Cut:",
                         c("Ideal" = "IDEAL",
                           "Premium" = "IDEAL",
                           "Good" = "GOOD",
                           "Very Good" = "VGOOD"))
          ),

#  Display your plot created by GGvis        
            mainPanel(ggvis_output("my_plot"))

          )
        ))

---

server.R

# clear memory & load packages    
        rm(list=ls())

        library(shiny)
        library(ggvis)

# Define server logic for distribution application

        shinyServer(function(input, output, session) {

# load your data      
          dataset <- diamonds

# Reactive expression to generate the subset.      
          datasetInput <- reactive({                
            selection <-switch(input$cut,
                               IDEAL = "Ideal",
                               PREM = "Premium",
                               GOOD = "Good",
                               VGOOD = "Very Good") 

            subset(dataset, cut == selection)                   
          })


# Generate your plot using GGvis and your reactive inputs      
          gv <- reactive({

            ggvis(datasetInput(), by_group(color),
                  props(x = ~carat,
                        stroke = ~color,
                        fill = ~color,
                        fillOpacity := 0.2,
                        fillOpacity.hover := 0.7)) +
              layer_density()                                
          })


# necessary additions for ggvis integration to shiny        
          output$controls <- renderControls(gv)
          observe_ggvis(gv, "my_plot", session)               
        })

EDIT: To follow up on this question, I have noticed that if I remove the following:

            by_group(color)

and

            stroke = ~color, 
            fill = ~color,

from my call to create the graph in server.R, the graph runs (albeit without the multicolored groupings, which is what I was after in the first place) and the radio buttons successfully select my data subset. As mentioned, when I am just running the graph in R, without the use of the reactive subsetting function, the graph is able to run and vary the stroke/fill by diamond color. Is this feature currently not supported in Shiny to ggvis interaction at this time? Or am I just misinterpreting this function somehow? Here is my updated code, so you can see my result:

ui.R

# clear memory & load packages
      rm(list=ls())

      library(shiny)
      library(ggvis)

# Define UI for distribution application
    shinyUI(fluidPage(

# Application title
         titlePanel("Diamond Carats by Color/Cut"),

# Sidebar with controls to select subset
          sidebarLayout(
          sidebarPanel(
            radioButtons("cut", "Diamond Cut:",
                         c("Ideal" = "IDEAL",
                           "Premium" = "IDEAL",
                           "Good" = "GOOD",
                           "Very Good" = "VGOOD"))
          ),

#  Display your plot created by GGvis        
            mainPanel(ggvis_output("my_plot"))

          )
        ))

---

server.R

# clear memory & load packages    
        rm(list=ls())

        library(shiny)
        library(ggvis)

# Define server logic for distribution application

        shinyServer(function(input, output, session) {

# load your data      
          dataset <- diamonds

# Reactive expression to generate the subset.      
          datasetInput <- reactive({                
            selection <-switch(input$cut,
                               IDEAL = "Ideal",
                               PREM = "Premium",
                               GOOD = "Good",
                               VGOOD = "Very Good") 

            subset(dataset, cut == selection)                   
          })


# Generate your plot using GGvis and your reactive inputs      
          gv <- reactive({

            ggvis(datasetInput(),
                  props(x = ~carat,
                        fillOpacity := 0.2,
                        fillOpacity.hover := 0.7)) +
              layer_density() 

          })


# necessary additions for ggvis integration to shiny        
          output$controls <- renderControls(gv)
          observe_ggvis(gv, "my_plot", session)               
        })

Answer 1

Further search would indicate that this is a known bug for ggvis: https://github.com/rstudio/ggvis/issues/71