# -*- coding: utf-8 -*-
"""Load Model.ipynb

Automatically generated by Colaboratory.

Original file is located at
    https://colab.research.google.com/drive/1CnM4IpjNR-VDFbqUuaVm96Q0cMkzkeAu
"""

import pandas as pd
import matplotlib.pyplot as plt
import keras
import tensorflow as tf
import requests
import numpy as np
import csv
import gradio as gr

from datetime import date
from datetime import timedelta
from bs4 import BeautifulSoup


"""To get the training data mean and std for the features"""

path = ""

def preprocessdata():
  with open(path + 'data/' + 'preprocessing_data.csv', newline='') as f:
      reader = csv.reader(f)
      data = list(reader)

  version = int(*data[0])
  mean = data[2]
  std = data[3]

  for i in range(len(mean)):
    mean[i] = float(mean[i])
    std[i] = float(std[i])
  return version, mean, std


def normalize(data, mean, std):
  data['Temp (F)'] = (data['Temp (F)'] - mean[0])/std[0]
  data['Humidity'] = (data['Humidity'] - mean[1])/std[1]
  data['Wind Speed (in HG)'] = (data['Wind Speed (in HG)'] - mean[2])/std[2]
  data['Wind Gust (MPH)'] = (data['Wind Gust (MPH)'] - mean[3])/std[3]
  return data


def getdata(mean, std):
  pd_data = pd.read_csv(path + 'data/' + 'weather_data.csv')
  pd_data = pd_data.iloc[-60:]
  pd_data.reset_index(drop=True, inplace=True)

  time_data = pd.DataFrame()
  time_data['Time'] = pd.to_datetime(pd_data['Time PST'])
  pd_data = pd_data.drop(['Time PST'], axis = 1)
  pd_data['Temp (F)'] = pd_data['Temp (F)'].astype(int)
  pd_data['Humidity'] =  pd_data['Humidity'].astype(int)
  pd_data['Wind Speed (in HG)'] = pd_data['Wind Speed (in HG)'].astype(float)
  pd_data['Wind Gust (MPH)'] = pd_data['Wind Gust (MPH)'].astype(float)
  df = normalize(pd_data, mean, std)
  return df


def model_predict(df, model):
  tf_data = tf.convert_to_tensor(df)
  tf_data = tf.reshape(tf_data, [1, *tf_data.shape])
  prediction = model.predict(tf_data, verbose = 0)
  df.loc[len(df.index)] = [*prediction[0]]
  df = df.iloc[-60:]
  return prediction[0][0]


def model_run():
  version, mean, std = preprocessdata()
  df = getdata(mean, std)

  today = date.today() - timedelta(days=1)
  result = pd.DataFrame(columns=['Time PST', 'Temp (F)'])
  model = tf.keras.models.load_model(path + 'model/' + 'LTSM{}.h5'.format(version))

  return_string = "The temperature for the next 24 hrs in Los Angeles: \n"
  for i in range(24):
    temperature = int(model_predict(df, model) * std[0] + mean[0])
    time = ""
    if i < 10:
      time = "{} 0{}:00:00 AM".format(today, i)
      # print(time + ', Temp {} (F)'.format(temperature))
    elif i < 12:
      time = "{} {}:00:00 AM".format(today, i)
      # print(time + ', Temp {} (F)'.format(temperature))
    else:
      time = "{} {}:00:00 PM".format(today, i)
      # print(time + ', Temp {} (F)'.format(temperature))
    return_string += time + ', Temp {} (F)'.format(temperature) + '\n'

  return return_string

def main():
  with gr.Interface(model_run, None, "textbox") as interface:
    interface.launch()

if __name__ == "__main__":
    main()