library(shiny)

ui <- fluidPage(
  fluidRow(
    column(12,
           fileInput("data",
                     "Upload Smoker Data",
                     accept = c("text/xlsx", ".xlsx")
           ), align="center")
  ),

  navbarPage("App Title",

    tabPanel("User stats",
      fluidRow(
        column(12, numericInput("userIndex",
                                label = h3("Select User"),
                                value = 1, max = 32, min = 1),
               align="center")
      ),

      fluidRow(
        column(4, h3("Cigarettes mode"), align="center"),
        column(4, h3("Number of cigarettes per day"), align="center"),
        column(4, h3("Mean smoking pattern per weekday"), align="center")
      ),

      fluidRow(
        column(4, plotOutput(outputId ="plotUserCigarettesModes")),
        column(4, plotOutput(outputId ="plotUserDaySmoked")),
        column(4, plotOutput(outputId = "plotUserMeanWeekday"))
      ),

      fluidRow(
        column(4, h3("Cigarettes mode (Pie)"), align="center"),
        column(4, h3("Number of cigarettes per interval"), align="center"),
        column(4, h3("Improvements"), align="center")
      ),

      fluidRow(
        column(4, plotOutput(outputId ="plotPieModes")),
        column(4, plotOutput(outputId="plotIntervalStats")),
        column(2, numericInput("improvementWeek", label = h3("Weekly Improvements"), value = 1, max = 12, min = 1), align="center"),
        column(2, htmlOutput(outputId = "userImprovements"))

      )



    ),

    tabPanel("General stats",
      fluidRow(column(12, h1("General Hourly Mean per Weekday"), align="center")),
      fluidRow(column(12, h4("Loading can take a few seconds..."))),
      fluidRow(column(3), column(6, plotOutput(outputId = "plotMeanWeekday")))
    )

  )
)

# Define server logic required to draw a histogram ----
server <- function(input, output) {
  # library imports
  library(readxl)
  library(ggplot2)

  Sys.setlocale("LC_TIME", "C")

  dataInput <- reactive({
    validate(
      need(input$data != "", "Please select a data set")
    )

    inFile <- input$data

    userdata <- read_excel(inFile$datapath)
    userdata$Weekday <- weekdays(as.Date(userdata$Time))
    userdata$Weekday <- as.factor(userdata$Weekday)
    userdata$User <- as.factor(userdata$User)
    userdata$Day <- as.Date(userdata$Time,format="%d-%m-%y")
    userdata$Type <- as.factor(userdata$Type)

    userdata
  })

  userData <- reactive({
    validate(
      need(!is.null(input$userIndex), "Please select a user")
    )

    userdata <- dataInput()
    userdata[userdata == input$userIndex,]
  })


  smokedData <- reactive({
    userdata <- userData()
    smoked <- userdata[! userdata$Type %in% c("Auto skipped", "Friend", "Skipped"),]
    smoked
  })

  smokedData <- reactive({
    userdata <- userData()
    smoked <- userdata[! userdata$Type %in% c("Auto skipped", "Friend", "Skipped"),]
    smoked$Hour <- format(smoked$Time,"%H")
    smoked$Interval <- "0-5:59"
    smoked[as.numeric(smoked$Hour) %in% c(6:11),]$Interval <- "6-11:59"
    smoked[as.numeric(smoked$Hour) %in% c(12:17),]$Interval <- "12-17:59"
    smoked[as.numeric(smoked$Hour) %in% c(18:23),]$Interval <- "18-23:59"

    smoked$Interval = factor(smoked$Interval, levels= c("0-5:59", "6-11:59", "12-17:59", "18-23:59"))
    smoked <- smoked[order(smoked$Interval), ]

    smoked
  })

  daySmokedData <- reactive({
    smoked <- smokedData()
    day.smoked <- aggregate(smoked$User, by = list(smoked$User, smoked$Day, smoked$Weekday), FUN = length)
    names(day.smoked) <- c("User", "Day", "Weekday", "count")
    day.smoked$Weekday = factor(day.smoked$Weekday, levels= c("Monday", "Tuesday", "Wednesday", "Thursday", "Friday", "Saturday", "Sunday"))
    day.smoked[order(day.smoked$Weekday), ]
    day.smoked
  })

  meanUserSmokedData <- reactive({
    day.smoked <- daySmokedData()
    day.smoked.mean <- as.data.frame(aggregate(day.smoked$count, by = list(day.smoked$User, day.smoked$Weekday), FUN = mean))
    day.smoked.sd <- as.data.frame(aggregate(day.smoked$count, by = list(day.smoked$User, day.smoked$Weekday), FUN = sd))
    names(day.smoked.mean) <- c("User", "Weekday", "mean")
    names(day.smoked.sd) <- c("User ", "Weekday", "sd")
    day.smoked.sd$sd[is.na(day.smoked.sd$sd)] <- 0
    day.smoked.mean$sd <- day.smoked.sd$sd
    day.smoked.mean
  })

  improvements <- reactive({
    validate(
      need(!is.null(input$improvementWeek), "Select a week")
    )

    df <- smokedData()
    d <- aggregate(df$Day, by=list(df$Type, df$User), FUN=max)
    names(d) <- c("Type", "User", "End")

    week <- data.frame(row.names = c("User", "Type", "Time", "Weekday", "Day"))
    week <- na.omit(df[
                    (df$Day > d[d$Type == "Observation week",]$End+8*(input$improvementWeek-1)) &
                    (df$Day <= d[d$Type == "Observation week",]$End + 8*(input$improvementWeek))
                  ,])

    if(nrow(week) == 0){
      return("<h1>0%</h1>")
    }


    week.sum <- aggregate(week$User, by=list(week$User), FUN=length)$x

    observation.sum <- aggregate(df[df$Type=="Observation week",]$User, by=list(df[df$Type=="Observation week",]$User), FUN=length)$x
    week.improvements <- (observation.sum - week.sum) / observation.sum

    if(week.improvements > 0)
      paste(c("<h1 style='color:green'>",as.character(round(week.improvements * 100, digit=2)),"%</h1>"), sep = " ")
    else
      paste(c("<h1 style='color:red'>",as.character(round(week.improvements * 100, digit=2)),"%</h1>"), sep = " ")
  })

  output$userImprovements <- renderText({
    improvements()
  })


  output$plotUserCigarettesModes <- renderPlot({
    ggplot(userData()) + aes(x=Type, fill = Type) + geom_bar() + xlab("Type of cigarette") + ylab("Number of Cigarettes") + labs(fill = "Used mode")
  })

  output$plotUserDaySmoked <- renderPlot({
    ggplot(meanUserSmokedData(),aes(x=Weekday, y=mean, group=User, colour=mean)) + geom_line(size=.5) + geom_point(size=5) +
      geom_errorbar(aes(ymin=mean-sd, ymax=mean+sd), width=.2, position=position_dodge(.9))

  })

  output$plotUserMeanWeekday <- renderPlot({
    smoked = smokedData()

    smoked$Hour <- format(smoked$Time,"%H")
    hour.smoked <- as.data.frame(aggregate(smoked$User, by = list(smoked$User, smoked$Day, smoked$Weekday, smoked$Hour), FUN = length))
    names(hour.smoked) <- c("User", "Day", "Weekday", "Hour", "count")
    hour.smoked.mean <- as.data.frame(aggregate(hour.smoked$count, by = list(hour.smoked$Weekday, hour.smoked$Hour), FUN = mean))
    names(hour.smoked.mean) <- c("Day", "Hour", "mean")
    ggplot(hour.smoked.mean) + aes(x=Hour, y=mean, group=Day, color=Day) + stat_smooth(size=1, method = 'lm', formula = y ~ poly(x,5), se=FALSE) + ggtitle("Mean smoking pattern per weekday")
  })

  output$plotMeanWeekday <- renderPlot({
    smoked = dataInput()

    smoked$Hour <- format(smoked$Time,"%H")
    hour.smoked <- as.data.frame(aggregate(smoked$User, by = list(smoked$User, smoked$Day, smoked$Weekday, smoked$Hour), FUN = length))
    names(hour.smoked) <- c("User", "Day", "Weekday", "Hour", "count")
    hour.smoked.mean <- as.data.frame(aggregate(hour.smoked$count, by = list(hour.smoked$Weekday, hour.smoked$Hour), FUN = mean))
    names(hour.smoked.mean) <- c("Day", "Hour", "mean")
    ggplot(hour.smoked.mean) + aes(x=Hour, y=mean, group=Day, color=Day) + stat_smooth(size=1, method = 'lm', formula = y ~ poly(x,6), se=FALSE) + ggtitle("Mean smoking pattern per weekday")
  })

  output$plotPieModes <- renderPlot({
    userdata <- userData()
    userdata$Type <- as.factor(userdata$Type)
    lbls <- levels(userdata$Type)
    lbls <- paste(lbls, round(prop.table(table(userdata$Type))*100))
    lbls <- paste(lbls, "%", spe="")
    modes <- with(userdata, aggregate(User, by=list(Type), FUN=length))
    names(modes) <- c("Mode", "count")
    pie(modes$count, labels = lbls, main="Mode distribution")
  })

  output$plotIntervalStats <- renderPlot({
    ggplot(smokedData()) + aes(x=Interval, fill = Type) + geom_bar() + xlab("Intervals") + ylab("Number of Cigarets") + labs(fill = "Used mode")
  })

}

shinyApp(ui = ui, server = server)