ggvis plot disappears at random Shiny - r

Question

I have a strange problem in Shiny. My shiny app has one ggvis plot with layer_points() and several options to manipulate the plot . When I run my app sometimes everything works good even if I change all options, but sometimes ( I suppose there is no specific rule) plot disappers. Plot comes back when I change one of options but it is not cool.
I study this issue but I do not really know whether it is a solution for my problem.
When the plot disappears my Shiny app looks like:
This my code:
ui.R
library(ggvis)
library(markdown)
library(shiny)
library(dplyr)
library(magrittr)
shinyUI(
fluidPage(
h3("Title"),
fluidRow(
column(3,
wellPanel(
radioButtons("radio",h5("Select"),choices=list("All values","Selected values"),
selected="All values"),
conditionalPanel(
condition = "input.radio != 'All values'",
checkboxGroupInput("checkGroup",label = "",
choices,
selected = c("AT1","AT2"))
),
hr(),
radioButtons("dataset", label = h5("Drilldown"),
choices = list("2 Level" = "df1", "3 Level" = "df2")
),
hr(),
h5("Choice"),
selectInput("xvar", h6(""),
axis_vars_x,
selected = "value"),
selectInput("yvar", h6(""),
axis_vars_y,
selected = "number2"),
hr(),
uiOutput("slider")
)
),
column(9,
ggvisOutput("plot")
)
)
)
)
server.R
library(shiny)
shinyServer(function(input, output,session) {
datasetInput <- reactive({
switch(input$dataset,
df2 = df2,
df1 = df1)
})
axis_vara_y <- reactive({
switch(input$yvar,
number = 2,
number2 = 3)
})
output$slider <- renderUI({
sliderInput("inslider",h5(""), min = round(min(datasetInput()[,axis_vara_y()]),0)-1,
max = round(max(datasetInput()[,axis_vara_y()]),0)+1,
value = c(round(min(datasetInput()[,axis_vara_y()]),0)-1,
round(max(datasetInput()[,axis_vara_y()]),0)+1),
step = 0.5)
})
data <- reactive({
filteredData <- datasetInput()
axisData <- axis_vara_y()
if(!is.null(input$inslider)){
if(input$radio == "All values"){
filteredData <- filteredData %>%
filter(filteredData[,axisData] >= input$inslider[1],
filteredData[,axisData] <= input$inslider[2])
}
else {
filteredData <- filteredData %>%
filter(value %in% input$checkGroup,
filteredData[,axisData] >= input$inslider[1],
filteredData[,axisData] <= input$inslider[2])
}
}
return(filteredData)
})
data_point <- reactive({
data() %>%
mutate(id = row_number())
})
xvar <- reactive(as.symbol(input$xvar))
yvar <- reactive(as.symbol(input$yvar))
dotpoint_vis <- reactive({
xvar_name <- names(axis_vars_x)[axis_vars_x == input$xvar]
yvar_name <- names(axis_vars_y)[axis_vars_y == input$yvar]
data_point_detail <- data_point()
plot <- data_point_detail %>%
ggvis(x = xvar(),y = yvar()) %>%
layer_points(size := 120,fill = ~value) %>%
add_axis("x", title = xvar_name) %>%
add_axis("y", title = yvar_name) %>%
set_options(width = 750, height = 500, renderer = "canvas")
})
dotpoint_vis %>% bind_shiny("plot")
})
global.R
choices <- list("Value1" = "AT1", "Value2" = "AT2",
"Value3" = "AT3", "Value4" = "AT4",
"Value5" = "AT5", "Value6" = "RT1",
"Value7" = "AT6", "Value8" = "AT7",
"Value9" = "AT8", "Value10" = "AT9",
"Value11" = "AT10", "Value12" = "RT2")
levele <- c("AT1","AT2","AT3","AT4","AT5","RT1","AT6","AT7","AT8","AT9","AT10","RT2")
df1 <- data.frame(value = levele,number = seq(2,46,4), number2 = seq(2,24,2),order = 1:12)
df2 <- data.frame(value = levele,number = rep(4:15), number2 = rep(4:9,each = 2),order = 1:12)
df1$value <- factor(df1$value, levels = levele)
df2$value <- factor(df2$value, levels = levele)
axis_vars_y <- c("number","number2")
axis_vars_x <- c("value", "order","number","number2")
update
I also do not know what happened with animation in ggvis.

Answer 1

The problem was difficult to reproduce at first, but I found I can reproduce it by clicking back and forth between All Values and Selected Values. The graph disappears or reappears after some number of switches between the two radio buttons, but it varies seemingly randomly -- sometimes it takes 4 clicks to make the graph disappear or reappear and other times it takes 2 clicks or some other number of clicks.
There must be a bug in bind_shiny() or ggvisOutput(), because the following changes do create a graphic that does not seem to disappear:
In ui.R, make this change:
# ggvisOutput("plot")
plotOutput('plot')
In server.R, make this change:
plot(data_point_detail[ , c(input$xvar, input$yvar)], xlab=xvar_name, ylab=yvar_name)
# plot <- data_point_detail %>%
# ggvis(x = xvar(),y = yvar()) %>%
# layer_points(size := 120,fill = ~value) %>%
# add_axis("x", title = xvar_name) %>%
# add_axis("y", title = yvar_name) %>%
# set_options(width = 750, height = 500, renderer = "canvas")
# plot
and
output$plot <- renderPlot(dotpoint_vis())
# dotpoint_vis %>% bind_shiny("plot")