library(shiny)
library(mlbench)
library(plotly)
library(shinythemes)
library(dplyr)

# Load data
data(BreastCancer)

# Remove NAs
BreastCancer <- na.omit(BreastCancer)

# Remvove ID
BreastCancer <- BreastCancer[,-1]

# Store features and actual class in seprate variables
featureList <- colnames(BreastCancer)[-10]
class <- BreastCancer$Class

# Convert to numeric
BreastCancer[,1:9] <- apply(BreastCancer[,-10], 2, as.numeric)

# ui.R definition
ui <- fluidPage(
  # Set theme
  theme = shinytheme("spacelab"),

  # Some help text
  h2("Coupled events in plotly charts using Shiny"),
  h4("This Shiny app showcases coupled events using Plotly's ", tags$code("event_data()"), " function."),
  tags$ol(
    tags$li("The first chart showcases", tags$code("plotly_selected")),
    tags$li("The third chart showcases", tags$code("plotly_click"))
  ),

  # Vertical space
  tags$hr(),

  # Feature selection
  fixedRow(
    column(3, selectInput(inputId = "featureInput1", label = "Select first feature", choices = featureList, selected = "Cell.Size")),
    column(4, selectInput(inputId = "featureInput2", label = "Select second feature (observed event)", choices = featureList, selected = "Epith.c.size"))),

  # First row
  fixedRow(
    column(6, plotlyOutput("Plot1", height = "600px")),
    column(6, plotlyOutput("Plot2", height = "600px"))),

  tags$hr(),
  tags$blockquote("First drag a selection box in the scatter plot to populate the barchart. Then select one of the bars in the barchat
    to populate the boxplot"),


  # Second row
  fixedRow(
    column(3, plotlyOutput("Plot3", height = "600px")),
    column(9, plotlyOutput("Plot4", height = "600px"))))

# server.R definition
server <- function(input, output){

  # Observes the second feature input for a change
  observeEvent(input$featureInput2,{

    # Create a convenience data.frame which can be used for charting
    plot.df <- data.frame(BreastCancer[,input$featureInput1],
                          BreastCancer[,input$featureInput2],
                          Class = BreastCancer$Class)

    # Add column names
    colnames(plot.df) <- c("x", "y", "Class")

    # Do a plotly contour plot to visualize the two featres with
    # the number of malignant cases as size
    # Note the use of 'source' argument
    output$Plot1 <- renderPlotly({
      plot_ly(plot.df, x = ~x, y = ~y, mode = "markers", type = "scatter", color = ~Class, source = "subset",
              marker = list(size = 30)) %>%
        layout(title = paste(input$featureInput1, "vs ", input$featureInput2),
               xaxis = list(title = input$featureInput1),
               yaxis = list(title = input$featureInput2),
               dragmode =  "select",
               plot_bgcolor = "6A446F")
    })

    # Create a contour plot of the number of malignant cases
    output$Plot2 <- renderPlotly({

      plot.df %>%
        group_by(x, y, Class) %>%
        summarize(Count = n()) %>%
        filter(Class == "malignant") %>%
        plot_ly(x = ~x, y = ~y, z = ~Count, type = "contour") %>%
        layout(title = "Contour map of number of malignant cases",
               xaxis = list(title = input$featureInput1),
               yaxis = list(title = input$featureInput2))

    })

    # Assign to parent environment
    plot.df <<- plot.df
  })

  # Coupled event 1
  output$Plot3 <- renderPlotly({

    # Get subset based on selection
    event.data <- event_data("plotly_selected", source = "subset")

    # If NULL dont do anything
    if(is.null(event.data) == T) return(NULL)

    # Get number of malignant and benign cases from selection
    malig.class <- subset(plot.df, Class == "malignant")[subset(event.data, curveNumber == 0)$pointNumber + 1,]
    benign.class <- subset(plot.df, Class == "benign")[subset(event.data, curveNumber == 1)$pointNumber + 1,]

    # Combine
    plot.subset <- rbind(malig.class, benign.class)

    # Summarize
    plot.summ <- plot.subset %>%
      group_by(x, y, Class) %>%
      summarize(Count = n())

    # Assign to parent frame
    plot.summ <<- plot.summ

    # Plot
    plot_ly(plot.summ, x = ~Class, y = ~Count, type = "bar", source = "select", color = ~Class) %>%
      layout(title = "No. of Malignant and Benign cases <br> in Selection",
             plot_bgcolor = "6A446F",
             yaxis = list(domain = c(0, 0.9)))
  })

  # Coupled event 2
  output$Plot4 <- renderPlotly({

    # Get subset based on selection
    event.data <- event_data("plotly_click", source = "select")

    # If NULL dont do anything
    if(is.null(event.data) == T) return(NULL)

    # If Malignant
    if(event.data[3] == "malignant"){

      tab <- subset(plot.summ, Class == "malignant")

      p1 <- plot_ly(tab, x = x, y = Count, type = "box", showlegend = F) %>%
        layout(yaxis = list(title = "Count"),
               xaxis = list(title = input$featureInput1))

      p2 <- plot_ly(tab, x = y, y = Count, type = "box", showlegend = F) %>%
        layout(title = "Box plot for Malignant cases",
               yaxis = list(title = "Count"),
               xaxis = list(title = input$featureInput2))

      subplot(p1, p2)
    }else{
      tab <- subset(plot.summ, Class == "benign")

      p1 <- plot_ly(tab, x = ~x, y = ~Count, type = "box", showlegend = F) %>%
        layout(yaxis = list(title = "Count"),
               xaxis = list(title = input$featureInput1))

      p2 <- plot_ly(tab, x = ~y, y = ~Count, type = "box", showlegend = F) %>%
        layout(title = "Box plot for Benign cases",
               yaxis = list(title = "Count"),
               xaxis = list(title = input$featureInput2))

      subplot(p1, p2)
    }

  })

}