# install.packages("profvis")
# install.packages("scales")
# install.packages("reticulate")
# install.packages("shiny")
# install.packages("tidyverse")
# devtools::install_github("rstudio/rsconnect", ref='737cd48')

library(foreign)
library(shiny)
library(ggplot2)
library(reticulate)
library(scales)
library(tidyverse)

virtualenv_create(envname = "python_environment", python= "python3")
#virtualenv_install("python_environment", packages = c('keras', 'pandas','numpy','scipy','scikit-learn', 'tensorflow'))
reticulate::use_virtualenv("python_environment", required = TRUE)

# Define UI for the health predictor tool

ui <- fluidPage(
  titlePanel("Health Outcome Predictor Tool"),

  # Sidebar layout with a input and output definitions
  sidebarLayout(

    # Inputs
    sidebarPanel(

      # Select race

      h3("Health Outcome Selection", align = "center"),

      #have to figure out how to structure the =x to make it so that it will be able to call the correct neural net
      selectInput(inputId = "health_outcome",
                  label = "Select the health outcome of interest:",
                  choices = c("Anxiety",
                              "Overweight",
                              "Asthma",
                              "Binge Drinking",
                              "Diabetes",
                              "Depression",
                              "Smoking"),
                  selected = "Overweight"),

      h3("Demographics and ACE Data", align = "center"),

      # Select race
      selectInput(inputId = "race",
                  label = "Select your race:",
                  choices = c("White"=1,
                              "Black"=2,
                              "Asian"=3,
                              "Native Hawaiian or other Pacific Islander"=4,
                              "American Indian or Alaskan Native"=5,
                              "Other"=6),
                  selected = "Other"),

      # Select gender
      selectInput(inputId = "gender",
                  label = "Select your gender:",
                  choices = c("Male" = 0,
                              "Female" = 1),
                  selected = 1),

      #Select if vetran or not
      checkboxInput(inputId = "vetran",
                    label = "Check if you ever served on active duty in the United States Armed Forces"),

      # Mark all of the ACEs you have experiences
      checkboxGroupInput(inputId = "aces",
                         label = "Check all of the following that you experienced between the ages of 0 and 18:",
                         choices = c( "Lived with someone who was depressed, mentally ill, or suicidal" = 1,
                                      "Lived with someone who was a problem drinker or alcoholic"= 2,
                                      "Lived with someone who used illegal street drugs or who abused prescription medications" = 3,
                                      "Live with someone who served time or was sentenced to serve time in a prison, jail, or other correctional facility"= 4,
                                      "Parents separated or divorced" = 4,
                                      "Parents or adults in your home slapped, hit, kicked, punched or beat each other up" = 6,
                                      "Parent or adult in your home hit, beat, kicked, or physically hurt you in any way" = 7,
                                      "Parent or adult in your home swore at you, insulted you, or put you down" = 8,
                                      "Someone at least 5 years older than you touched you sexually" = 9,
                                      "Someone at least 5 years older than you tried to make you touch them sexually" =10,
                                      "Someone at least 5 years older than you forced you to have sex" =11)),


      #Input age
      numericInput(inputId = "age",
                   label = "How old are you?",
                   value = 18,
                   min = 18,
                   max = 100,
                   step = 1),

      # Select level of education
      selectInput(inputId = "education",
                  label = "Select the highest grade or year of school you have completed",
                  choices =c("Never attended school or only kindergarten" = 1,
                             "Grades 1 through 8 (Elementary)" = 2,
                             "Grades 9 through 11 (Some high school)" = 3,
                             "Grade 12 or GED (High school graduate)" = 4,
                             "College 1 year to 3 years (Some college or technical school)" = 5,
                             "College 4 years or more (College graduate)" = 6)),

      # Select income level
      selectInput(inputId = "income",
                  label = "What is your annual household income (from all sources)?",
                  choices = c("Less than $10,000" = 1,
                              "Between $10,000 and $15,000" = 2,
                              "Between $15,000 and $20,000" = 3,
                              "Between $20,000 and $25,000" = 4,
                              "Between $25,000 and $35,000" = 5,
                              "Between $35,000 and $50,000" = 6,
                              "Between $50,000 and $75,000" = 7,
                              "$75,000 or more" = 8))

    ),

    # Outputs
    mainPanel(
      uiOutput(outputId = "nnet")
      #,br(),
      #consider how we are going to add text and what it will be
      #compared to people who are like you
      #plotOutput(outputId = "bar")

    )
  )
)

# Define server function
server <- function(input, output) {
  # Define some reactive components
  # These use the inputs to create vectors that are used in the neural net prediction


  numbers <- reactive({
    validate(
      need(is.numeric(input$age), "Please input your age (in numbers)")
    )
  })

  # aces reactive
  aces_reactive <- reactive({
    aces_list <- replicate(11, 0)
    aces_list[as.integer(input$aces)] = 1
    return(aces_list)
  })

  #vetran reactive
  vetran_reactive <- reactive({
    ifelse(input$vetran, vetran_value <- 1, vetran_value <- 0)
    return(vetran_value)
  })

  # race reactive
  race_reactive <- reactive({
    vec <- replicate(6, 0)
    vec[(as.integer(input$race))]=1
    return(vec)})

  # define the neural network model and python
  source_python("testing the models new.py")
  py_run_file("testing the models new.py")

  #A REACTIVE VALUE TO TELL ME WHICH predict function to use then just call that, deifned by input
  #will NEED TO ADD MORE here
  #function reactive
  func <- reactive({

    if (input$health_outcome == "Overweight") {
      return(predict_overweight_nn)
    }
    if (input$health_outcome == "Anxiety") {
      return(predict_anxiety_nn)
    }
    if (input$health_outcome == "Asthma") {
      return(predict_asthma_nn)
    }
    if (input$health_outcome == "Binge Drinking") {
      return(predict_binge_drinking_nn)
    }
    if (input$health_outcome == "Diabetes") {
      return(predict_diabetes_nn)
    }
    if (input$health_outcome == "Depression") {
      return(predict_diabetes_nn)
    }
    if (input$health_outcome == "Smoking") {
      return(predict_anxiety_nn)
    }
  })

  age_reactive <- reactive({
    age <- (input$age-18)/(99-18)
    return(age)
  })

  #score reactive
  #still need this, this is for with ACEs
  score_reactive <- reactive({
    # uses neural net to predict p(outcome) given the information
    predictions = c(aces_reactive(),vetran_reactive(),age_reactive(),input$education,input$income,input$gender,race_reactive())
    score <- (func()(predictions))
    return((score))
  })

  #still need this, this is for no ACEs
  none_reactive <- reactive({
    # uses neural net to predict p(overwieght) given the same demographics but no ACEs
    pred_none = c(replicate(11, 0),vetran_reactive(),age_reactive(),input$education,input$income,input$gender,race_reactive())
    score_none <- (func()(pred_none))
    score_none <- (score_none)
    return(score_none)
  })

  #make a reactive for yes with ACEs
  with_aces_reactive <- reactive({
    x <- score_reactive()
    if(0 <=x && x < 0.05){
      return("very unlikely")
    }
    if(0.05 <=x && x < 0.25){
      return("unlikely")
    }
    if(0.25 <=x && x < 0.75){
      return("moderately likely")
    }
    if(0.75 <=x && x < 0.95){
      return("likely")
    }
    if(0.95 <=x && x <= 1){
      return("very likely")
    }

  })

  #make a reactive for without ACEs


  #won't need this...sad
  relative_reactive <- reactive({
    relative <- (score_reactive()/none_reactive())
    return(relative)
  })

  #may end up getting rid of this but right now the graphic relies on it
  diff_reactive <- reactive({
    # # calculates difference in real value and if with no ACEs
    diff <- ((score_reactive())-none_reactive())
    return(diff)
  })

  #gotta change these too
  #your risk for
  names_reactive <- reactive({
    if (input$health_outcome == "Overweight") {
      return("being overweight")
    }
    if (input$health_outcome == "Anxiety") {
      return("having clinical anxiety")
    }
    if (input$health_outcome == "Asthma") {
      return("having asthma")
    }
    if (input$health_outcome == "Binge Drinking") {
      return("binge drinking")
    }
    if (input$health_outcome == "Depression") {
      return("having clinical depression")
    }
    if (input$health_outcome == "Diabetes") {
      return("having diabetes")
    }
    if (input$health_outcome == "Smoking") {
      return("smoking cigarettes")
    }
  })

  #this will go
  incdec_reactive <- reactive({
    if (relative_reactive()-1 > 0) {
      return("higher than")
    }
    if (relative_reactive()-1 < 0) {
      return("lower than")
    }
    if (relative_reactive()-1 == 0) {
      return("which is the same as")
    }
  })

  #this will go
  incdecnew_reactive <- reactive({
    if (relative_reactive()-1 > 0) {
      return(paste(percent(relative_reactive()-1)," higher than",sep=""))
      return("higher than")
    }
    if (relative_reactive()-1 < 0) {
      return(paste(percent(abs(relative_reactive()-1))," lower than",sep=""))
    }
    if (relative_reactive()-1 == 0) {
      return("the same as")
    }
  })



  # create the output of the app
  output$nnet <- renderUI({

    # gives a custom error message if a number is not entered in the age box
    numbers()

    if (sum(aces_reactive()) > 0){
      paste("According to the model and given your information your risk for",
            names_reactive(),
            "is: ",
            paste(with_aces_reactive(),".",sep=""),
            "This risk is",
            incdecnew_reactive(),
            "it would be relative to if you had 0 ACEs.")
    } else {
      paste("According to the model and given your information your risk for",
            names_reactive(),
            "is: ",
            paste(percent(score_reactive()),".",sep=""))
    }
  })



  #this makes the visualization
  #it would be better to have the names
  #can bring this back but right now it feels irrelevant
  # output$bar <- renderPlot({
  #   viz <- data.frame(Type = c(paste("Probability of not",names_reactive()), paste("Increased probability of",names_reactive(), "given selected ACEs"), paste("Probability of", names_reactive(), "with your demographics and no ACEs")),
  #                     Proportion = c((1-score_reactive()), diff_reactive(), (score_reactive()-diff_reactive())))
  #   viz <- viz %>%
  #     mutate(ace = "Probability of Being Overweight")
  #
  #   viz$Type <- factor(viz$Type, levels = viz$Type[c(0:3)])
  #
  #   ggplot(viz, aes(x = ace, y = Proportion, fill = Type)) +
  #     geom_col() +
  #     coord_flip() +
  #     #could prob just put in percent(number) instead of proportion + % and hopefully would work
  #     geom_text(aes(label = paste0(percent(Proportion))),
  #               position = position_stack(vjust = 0.5)) +
  #     scale_fill_brewer(palette = "Set2") +
  #     theme_minimal(base_size = 14) +
  #     ylab("Percentage") +
  #     xlab(NULL)
  #
  # })

}

# Create a Shiny app object
shinyApp(ui = ui, server = server)