Plotly graph misbehaving in Shiny - r

Question

When I select options in my Shiny application, I expect it to produce a graph in plotly and it does. It look like this:
When I unselect the options, it SHOULD look like this:
However, the output is misbehaving in a way I don't understand why. This is output I get:
As you can see, the x and y axis still remains. If I hover over the plot with no options selected, the plot still shows the coordinates.
Here is my code:
# UI
box(width = NULL, uiOutput("plotTitle"),
plotlyOutput('plot1', height="730px"), collapsed = F, title = "Figure",
status = "warning", solidHeader = T, height = "830px")
# SERVER
output$plot1 <- renderPlotly({
validate(
need(input$crops, "INSTRUCTIONS: \n\n1) Select Crops Of Interest.\n2)
Use 'Farm' tab to refine results.\n3) Hover over graph to see values.")
)
crop.data <- reactive({
crop.list <- input$crops
z.df <- dat
c.df <- subset(z.df, Crop %in% crop.list)
c.df <- unique(c.df)
})
output$plotTitle <- renderUI({
if(length(crop.data()) > 0) {
div(style='background-color:green; color:white; text-align:center;
font-size:22px; font-family:"Open Sans",verdana,arial,sans-serif',
input$metric)
}
})
# s.df is the Dataframe
s.df <- subset(s.df, Crop %in% input$crops)
if(length(crop.data()) > 0) {
if(input$metric == "Percentage") {
p <- plot_ly(s.df, x = ~Farm, y = ~Percentage, color = ~Crop) %>%
layout(barmode = 'stack', xaxis = list(title = ''), margin = list(b = 140))
} else if(input$metric == "Acreage") {
p <- plot_ly(s.df, x = ~Farm, y = ~Acreage, color = ~Crop) %>%
layout(barmode = 'stack', xaxis = list(title = ''), margin = list(b = 140))
} else {
p <- plot_ly(s.df, x = ~Farm, y = ~Exposure_Costs, color = ~Crop) %>%
layout(barmode = 'stack', xaxis = list(title = ''),
yaxis = list(title = 'Exposure Costs'), margin = list(b = 140))
}
p
}
})
Is there a specific reason as to why this happens? Also, this error doesn't exist on my local instance but occurs on the shiny server and another's local instance. How do I fix it?

Answer 1

You're using length(crop.data()) to check whether or not to plot but this will return a value greater than 0 for an "empty" data.frame because it is counting the number of columns.
e.g.
> (x <- iris[0, ])
[1] Sepal.Length Sepal.Width Petal.Length Petal.Width Species
<0 rows> (or 0-length row.names)
> length(x)
[1] 5
You really should untangle your mangling of the reactives, renderUI and renderPlotly but replacing length(crop.data()) with nrow(crop.data()) should solve your problem.
This reproducible example shows that your desired behavior does occur for a simple example:
library("shiny")
library("plotly")
set.seed(100)
d <- diamonds[sample(nrow(diamonds), 1000), ]
ui <- fluidPage(
titlePanel("Null Plotly Object"),
sidebarLayout(
sidebarPanel(
checkboxGroupInput("cuts", "Cut", choices = sort(unique(d$cut)))
),
mainPanel(
plotlyOutput("distPlot")
)
)
)
server <- function(input, output) {
plot_dat <- reactive({
shiny::validate(
need(input$cuts, "Select a cut")
)
d[d$cut %in% input$cuts, ]
})
output$distPlot <- renderPlotly({
plot_ly(plot_dat(), x = ~carat, y = ~price, color = ~carat, type = "scatter",
mode = "markers", size = ~carat, text = ~paste("Clarity: ", clarity))
})
}
shinyApp(ui = ui, server = server)