# -*- coding: utf-8 -*-
"""
Created on Sat Mar  7 10:41:51 2020

@author: Johan Rene van Dorp
"""

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import Dist_Library as dl
from sklearn.neighbors import KernelDensity

df = pd.read_csv('SoftwareFailure.csv')
FailTimes = df["FailureTimes"]
n_data = len(FailTimes)

# Plotting a two panel histogram plot of the durations
# the waiting times

plt.rc('font', family='serif', size='10')
plt.figure(figsize=(10, 5))
Figure = plt.figure()
Figure.subplots_adjust(hspace=0.4, wspace=0.4)
off_set_x = 250

# Calculating the histogram data for the left panel
Number_of_bins = 15
LB = min(FailTimes)
UB = max(FailTimes)
bins = np.linspace(LB, UB, Number_of_bins+1) 
hist_data = dl.Estimate_empirical_histogram_table(bins,FailTimes)

# Adding the left top panel with the theoretical pmf
the_bounds = np.append(hist_data["LB"][0],hist_data["UB"])
the_density = np.append(0,hist_data["Bin PDF"])
x_lims = (LB-off_set_x,UB+off_set_x)
y_lims = (0,0.002)
Panel = Figure.add_subplot(1,2,1)
dl.add_pmf_density_hist_to_figure_panel(Panel,the_bounds,the_density,'indianred')
plt.xlim(x_lims)
plt.ylim(y_lims)
plt.xlabel('Software Failure Times')
plt.ylabel('Density')
text_str = '# Bins '+str(Number_of_bins)
plt.text(2250,0.0017,text_str,color = 'red',size = 10)
# Adding a Kernel Density to the plot
FT = FailTimes.to_numpy()
kde = KernelDensity(bandwidth=100, kernel='gaussian')
kde.fit(FT[:, None])
x_d = np.linspace(LB-off_set_x, UB+off_set_x, 1000)
logprob = kde.score_samples(x_d[:, None])
plt.plot(x_d, np.exp(logprob), alpha=1, color = 'navy',label ='Kernel Density')
plt.legend(loc = 'best')

# Plotting the Empirical CDF of the Waiting Times

FailTimes.ordered = np.sort(FailTimes)
FailTimes.ordered = np.append(LB-off_set_x,FailTimes.ordered)
FailTimes.ordered = np.append(FailTimes.ordered,UB+off_set_x)
F = np.arange(1,(n_data+1))/n_data
F = np.append(0,F)
F = np.append(F,1)

x_lims = (LB-off_set_x,UB+off_set_x)
y_lims = (-0.05,1.1)
Panel = Figure.add_subplot(1,2,2)
plt.step(FailTimes.ordered,F,lw = 1,color ='skyblue')
plt.xlim(x_lims)
plt.ylim(y_lims)
plt.xlabel('Software Failure Times')
plt.ylabel('Empirical CDF $\hat{F}(t) = Pr(T \leq t)$')
plt.axvline(0, lw=2, ls = '-',color = 'lightgray',alpha=0.5)
plt.axhline(0, lw=2, ls = '-',color = 'lightgray',alpha=0.5)
plt.axhline(1, lw=2, ls = ':',color = 'lightgray',alpha=0.5)
text_str = 'n = '+str(n_data)
plt.text(100,1.025,text_str,color = 'red',size = 10)

q_level = 0.8
emp_quantile, q_index, x_k, x_k_plus_1 = dl.empirical_quantile(FailTimes,q_level)

Panel.vlines(emp_quantile, 0, q_level, color='black', linestyles=':', lw=1)
Panel.hlines(q_level, 0, emp_quantile, color='black', linestyles=':', lw=1)

text_str = 'p = ' + str(q_level)
plt.text(emp_quantile+115,q_level,text_str, color = 'red')

text_str = '$\hat{x}_p = $' + f'{emp_quantile:5.1f}'
plt.text(emp_quantile+115,0.025,text_str, color = 'red')

Figure.suptitle('Histogram and ECDF Software Failure Data',size='14')
plt.savefig('SoftwareFailure.png', dpi=1200)