#!/usr/bin/env python
# -*- coding: utf-8 -*-
import pandas as pd

def wczytaj_dane():
    return pd.read_csv('mieszkania.csv')

def most_common_room_number(dane):
    return dane['Rooms'].mode()[0]

def cheapest_flats(dane, n):
    return dane.sort_values(by=[u'Expected'], na_position='last').head(n)

def find_borough(desc):
    dzielnice = ['Stare Miasto',
                 'Wilda',
                 'Jeżyce',
                 'Rataje',
                 'Piątkowo',
                 'Winogrady',
                 'Miłostowo',
                 'Dębiec']
    histogram = {districtName: desc.find(districtName) for districtName in dzielnice if desc.find(districtName) != -1}
    return 'Inne' if not histogram else min(histogram, key=histogram.get)


def add_borough(dane):
    dane['Borough'] = dane.apply( lambda row: find_borough( row['Location']), axis = 1)

def write_plot(dane, filename):
    dane['Borough'].value_counts().plot.bar().get_figure().savefig(filename)

def mean_price(dane, room_number):
    return dane.loc[ dane['Rooms'] == room_number ].mean()[1]

def find_13(dane):
    return dane.loc[dane['Floor'] == 13]['Borough'].unique()

def find_best_flats(dane):
    return dane.loc[(dane['Borough'] == 'Winogrady') & (dane['Rooms'] == 3) & (dane['Floor'] == 1)]

def predict(dane, col_name):
    from sklearn import linear_model
    from sklearn.metrics import mean_squared_error, r2_score
    d_X = pd.DataFrame(dane[col_name])
    print('Dane z kolumny ', col_name)
    print(d_X.head())
    d_X_train = d_X[4000:]
    d_X_test = d_X[:4000]
    d_y = pd.DataFrame(dane['Expected'])
    d_y_train = d_y[4000:]
    d_y_test = d_y[:4000]
    regr = linear_model.LinearRegression()
    regr.fit(d_X_train, d_y_train)
    y_pred = regr.predict(d_X_test)
    print('MODEL(%s): pred_y = %f * x + %f' % (col_name, regr.coef_[0], regr.intercept_) )
    print('Mean squared error: %.2f' % mean_squared_error(d_y_test, y_pred))

    import matplotlib.pyplot as plt

    plt.clf()
    dataLine, =  plt.plot(d_X_test, d_y_test, 'ro', label='collected data')
    predLine, = plt.plot(d_X_test, y_pred, color='blue', linestyle='--', linewidth = 2, label='predictions')
    ax = plt.gca().add_artist(plt.legend(handles=[dataLine], loc=1))
    plt.legend(handles=[predLine], loc=4)
    plt.xticks(())
    plt.yticks(())
    plt.xlabel(col_name)
    plt.ylabel('Price')
    plt.show()

def main():
    dane = wczytaj_dane()
    print(dane[:5])

    print("Najpopularniejsza liczba pokoi w mieszkaniu to: {}"
          .format(most_common_room_number(dane)))

    print("7 najtańszych mieszkań to: ")
    print(cheapest_flats(dane, 7))

    print("{} to najłądniejsza dzielnica w Poznaniu.".format(find_borough("Grunwald i Jeżyce")))

    add_borough(dane)
    write_plot(dane, 'tmp_borough_hist.png')
    for i in dane['Rooms'].unique():
        print("Średnia cena mieszkania {}-pokojowego, to: {}".format(i, mean_price(dane, i)))
    print('Dzielnice z mieszkaniami na 13 piętrze: {}'.format(find_13(dane)))
    print('"Najlepsze" mieszkania: ')
    print(find_best_flats(dane))

    predict(dane, 'Rooms')
    predict(dane, 'SqrMeters')


if __name__ == "__main__":
    main()