Shiny, R - Error: cannot derive coordinates from non-numeric matrix

Question

I am attempting to use the Shiny package for R to plot points on a map as they appear at different points throughout the day.

An unexpected problem arises when the Shiny server is told to take the coordinates (lon and lat) of a subsetted dataframe:

coordinates(selected.tweet.points.spdf) <- ~ lon + lat

The interesting thing is that this works outside of Shiny, but is not working for me within Shiny.

The error message displayed in the Shiny window is:

Error: cannot derive coordinates from non-numeric matrix

This is the same error message that would appear if you were to call coordinates() on an object that is not a spatial object without telling it what to consider coordinates, like so:

coordinates(selected.tweet.points.spdf)

Here is a minimally reproducible example (sans the actual plotting):

library(shiny)
library(tigris)
library(dplyr)
library(leaflet)
library(sp)
library(ggmap)
library(maptools)
library(broom)
library(httr)
library(rgdal)
library(tidyr)

nyc.neighborhoods.url <- GET('http://data.beta.nyc//dataset/0ff93d2d-90ba-457c-9f7e-39e47bf2ac5f/resource/35dd04fb-81b3-479b-a074-a27a37888ce7/download/d085e2f8d0b54d4590b1e7d1f35594c1pediacitiesnycneighborhoods.geojson')
nyc.neighborhoods.spdf <- readOGR(content(nyc.neighborhoods.url, 'text'), 'OGRGeoJSON', verbose = F)
nyc.neighborhoods.df <- tidy(nyc.neighborhoods.spdf)

tweet.points.df <- data.frame(class = c("Not Private", "Not Private", "Private", "Private", "Private", "Private"),
                        lat = c(40.65514, 40.65514, 42.74662, 42.74662, 40.65514, 40.57238),
                        lon = c(-73.94878, -73.94878, -75.77004, -75.77004, -73.94878, -74.15395),
                        time = c("14:00", "14:00", "14:30", "14:30", "14:30", "14:30"))

shinyApp(
  # User interface
  ui = fluidPage(
    titlePanel("Social Media Post Privacy Classifier"),
    sidebarLayout(
      sidebarPanel(
        sliderInput("time", "Time of day (by half hour)",   
                    min = as.POSIXlt("2017-01-01 00:00:00", tz = "GMT"),   
                    max = as.POSIXlt("2017-01-01 23:30:00", tz = "GMT"),   
                    value = as.POSIXct("2017-01-01 00:00:00", tz = "GMT"),   
                    timeFormat="%H:%M", timezone = "+0000", step = 60 * 30, animate = T)
      ),
      mainPanel(
        textOutput("out")
      )
    )
  ),
  # Server
  server = function(input, output) {
    choropleth.selection <- reactive({
      selected.time <- format(input$time, '%H:%M')
      selected.tweet.points.df <- tweet.points.df[tweet.points.df$time == selected.time, ]
      selected.tweet.points.spdf <- selected.tweet.points.df
      coordinates(selected.tweet.points.spdf) <- ~ lon + lat
      proj4string(selected.tweet.points.spdf) <- proj4string(nyc.neighborhoods.spdf)
      selected.tweet.points.df.matches <- over(selected.tweet.points.spdf, nyc.neighborhoods.spdf)
      selected.tweet.points.df <- cbind(selected.tweet.points.df, selected.tweet.points.df.matches)
      selected.tweet.points.df <- selected.tweet.points.df %>% drop_na()
      print(selected.tweet.points.df$class)
    })
    output$out <- renderPrint({ choropleth.selection() })
  }
)

This is an example that proves that the same technique works outside of Shiny:

selected.time <- "14:30"
points.df <- data.frame(class = c("Not Private", "Not Private", "Private", "Private", "Private", "Private"),
                        lat = c(40.65514, 40.65514, 42.74662, 42.74662, 40.65514, 40.57238),
                        lon = c(-73.94878, -73.94878, -75.77004, -75.77004, -73.94878, -74.15395),
                        time = c("14:00", "14:00", "14:30", "14:30", "14:30", "14:30"))
points.df <- data.frame(points.df[points.df$time == selected.time, ])
coordinates(points.df) <- ~ lon + lat

Answer 1

The error arises when selected.tweet.points.spdf is empty, ie when there's no matching time. This is the case in the initial run, when selected.time is 00:00 - there's no matching tweet.points.df$time for that.

One possible solution would be to change your server function like this:

server = function(input, output) {
    choropleth.selection <- reactive({
      selected.time <- format(input$time, '%H:%M')
      selected.tweet.points.df <- tweet.points.df[tweet.points.df$time == selected.time, ]
      selected.tweet.points.spdf <- selected.tweet.points.df
      validate(need(nrow(selected.tweet.points.df)>0, "No match")) # <=======
      coordinates(selected.tweet.points.spdf) <- ~ lon + lat
      proj4string(selected.tweet.points.spdf) <- proj4string(nyc.neighborhoods.spdf)
      selected.tweet.points.df.matches <- over(selected.tweet.points.spdf, nyc.neighborhoods.spdf)
      selected.tweet.points.df <- cbind(selected.tweet.points.df, selected.tweet.points.df.matches)
      selected.tweet.points.df <- selected.tweet.points.df %>% drop_na()
      print(selected.tweet.points.df$class)
    })
    output$out <- renderPrint({ choropleth.selection() })
  }