# Import gradio
import gradio as gr

# Data import from open-meteo.com
import openmeteo_requests               
import requests_cache
from retry_requests import retry
    
# Data management and visualization
import pandas as pd
import numpy as np
import datetime
import pickle
import matplotlib
import matplotlib.pyplot as plt
import matplotlib.dates as mdates

# Machine Learning
import sklearn

# Print version info
print("Version info:")
print('pandas: %s' % pd.__version__)
print('numpy: %s' % np.__version__)
print('sklearn: %s' % sklearn.__version__)
print(" ")


def makeprediction():
#%% 1. User Inputs
    # These inputs may be changed by the user
    model_file_name = "Solar RF Model_2024-01-03T08-53-53_.pkl"    # This is the name of the file generated by the script "Build_solar_RF_model.py" (must be located in same folder)
    horizon = 2                                                    # Forecast horizon in days, limit is 16 days.
                                                                   # Note that the further we look into the future, the less accurate the weather forecast is likely to be
    PV_capacity = 15                                               # The maximum capacity of your PV installation in kW, only used for plotting.
    
    #%% 2. Read Model
    # Read in the model
    f_model = open(model_file_name, 'rb')               # Opens the file
    model_dict = pickle.load(f_model)                   # Reads the dictionary from the file
    f_model.close()                                     # closes the file again
    
    #%% 3. Fetch weather forecast
    
    # Setup the Open-Meteo API client with cache and retry on error
    cache_session = requests_cache.CachedSession('.cache', expire_after = -1)
    retry_session = retry(cache_session, retries = 5, backoff_factor = 0.2)
    openmeteo = openmeteo_requests.Client(session = retry_session)
    
    # Make sure all required weather variables are listed here
    url = "https://api.open-meteo.com/v1/forecast"
    
    # Calcualte start and end date
    start_date = datetime.datetime.now()
    end_date = start_date + datetime.timedelta(days=horizon)
    sdate_str = str(start_date)[0:10]                       # First date in file as string (yyyy-mm-dd)
    edate_str = str(end_date)[0:10]                         # Last date in file as string (yyyy-mm-dd)
    
    # Use the same parameters as was used for training the model.
    params = {
    	"latitude": model_dict['API Request Params']['latitude'],
    	"longitude": model_dict['API Request Params']['longitude'],
    	"hourly": model_dict['API Request Params']['hourly'],
        "start_date": sdate_str,
        "end_date": edate_str,
    }
    responses = openmeteo.weather_api(url, params=params)
    
    # Process & print request info.
    response = responses[0]
    print("Request info:")
    print(f"Coordinates {response.Latitude()}°E {response.Longitude()}°N")
    print(f"Elevation {response.Elevation()} m asl")
    print(f"Timezone {response.Timezone()} {response.TimezoneAbbreviation()}")
    print(f"Timezone difference to GMT+0 {response.UtcOffsetSeconds()} s")
    #print(f"From: " + sdate_str + " To: " + edate_str)
    print(" ")
    
    # Process hourly data. The order of variables needs to be the same as requested.
    hourly = response.Hourly()                                  # API Response
    hourly_data = {"date": pd.date_range(                       # Dictionary that we add the API respnse to
    	start = pd.to_datetime(hourly.Time(), unit = "s"),
    	end = pd.to_datetime(hourly.TimeEnd(), unit = "s"),
    	freq = pd.Timedelta(seconds = hourly.Interval()),
    	inclusive = "left"
    )}
    
    # We iterate through the variables and add the API response data to our dictionary
    print("Adding variables to dataframe...")
    index_variable = 0
    for variable in params['hourly']:
        hourly_data[variable] = hourly.Variables(index_variable).ValuesAsNumpy()    # Add the variable to dataframe
        print("Added " + variable)
        index_variable += 1                                                         # Increment counter
        
    print(" ")
    # Create dataframe
    weather_data = pd.DataFrame(data = hourly_data)    
    weather_data = weather_data.set_index('date')    # Set index to be dates
    
    #%% 4. Data manipulation & encoding
    # Create a new dataframe to hold all data, also encode new features.
    
    # Create a main dataframe that holds all data
    main_df = weather_data.copy()
    
    for index, row in main_df.iterrows():
        main_df.loc[index,'month'] = index.month
        main_df.loc[index,'day'] = index.day
        main_df.loc[index,'hour'] = index.hour
        main_df.loc[index,'sine month'] = np.sin((index.month - 1)*np.pi/11)
        main_df.loc[index,'cos month'] = np.cos((index.month - 1)*np.pi/11)
        main_df.loc[index,'sine hour'] = np.sin((index.month - 1)*np.pi/23)
        main_df.loc[index,'cos hour'] = np.cos((index.month - 1)*np.pi/23)
    
    # Index to keep track of which hours the sun is up for.        
    Sun_index = main_df[main_df['is_day'] == 1].index        
    
    # Create a new dataframe with only relevant data, this will be used for statistical 
    sun_is_up_data = main_df.copy()                                 # Create copy
    sun_is_up_data = sun_is_up_data.loc[Sun_index]                  # Only include rows when the sun is up
    sun_is_up_data = sun_is_up_data[model_dict['Input features']]   # Only include columns needed for model
    
    
    #%% 5. Make Predictions
    print("Making predictions...")
    # Create dataframe to hold predictions and initialize with 0.
    predictions = pd.DataFrame(0, index = weather_data.index, columns = ['Main Forecast', 'Lower Bound', 'Upper Bound'])
    predictions['Date'] = predictions.index.tolist()             # Make dates also to a column for easier plotting
    
    # Make predictions
    predictions.loc[Sun_index,'Main Forecast'] = model_dict['Random Forest Model'].predict(sun_is_up_data)
    predictions.loc[Sun_index,['Lower Bound', 'Upper Bound']] = model_dict['Random Forest Quantile Model'].predict(sun_is_up_data, quantiles=[0.1, 0.9], interpolation='linear')
    print(" ")
    #%% 6. Make Plot
    
    # Plotting
    plt.style.use('_mpl-gallery')                   # Set plot style
    matplotlib.rc('font', **{'size'   : 20})        # Change font size
    fig, ax = plt.subplots()                        # Create plot
    fig.set_size_inches(18.5, 10.5, forward=True)   # Set figure size
    fig.set_dpi(100)                                # Set resolution
    ax.fill_between(predictions['Date'], predictions['Lower Bound'], predictions['Upper Bound'], alpha=.5, linewidth=0) # Plot prediction interval
    ax.plot(predictions['Date'], predictions['Main Forecast'], linewidth=4, color='black')                              # Plot main forecast
    
    # Calculate x-ticks
    step_length = int(predictions.shape[0]/32)
    ax.set(ylim = [0,PV_capacity*1.05], xticks = predictions['Date'].iloc[::step_length])
    ax.xaxis.set_major_formatter(mdates.DateFormatter('%m-%d %H:%M'))
    plt.ylabel('Solar power [kW]')
    plt.xticks(rotation=90)
    plt.legend(['Prediction Interval (10 - 90%)', 'Main Forecast'])
    plt.title('Solar Power Forecast from ' + sdate_str + ' to ' + edate_str)
    # plt.show()
    
    return fig


iface = gr.Interface(fn=makeprediction, inputs=None, outputs=gr.Plot())
iface.launch()