#!/usr/bin/env python
# Copyright (c) 2019, Anthony Latorre <tlatorre at uchicago>
#
# This program is free software: you can redistribute it and/or modify it
# under the terms of the GNU General Public License as published by the Free
# Software Foundation, either version 3 of the License, or (at your option)
# any later version.
#
# This program is distributed in the hope that it will be useful, but WITHOUT
# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
# FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
# more details.
#
# You should have received a copy of the GNU General Public License along with
# this program. If not, see <https://www.gnu.org/licenses/>.

from __future__ import print_function, division
import yaml
try:
    from yaml import CLoader as Loader
except ImportError:
    from yaml.loader import SafeLoader as Loader

import numpy as np
from scipy.stats import iqr
from matplotlib.lines import Line2D

# on retina screens, the default plots are way too small
# by using Qt5 and setting QT_AUTO_SCREEN_SCALE_FACTOR=1
# Qt5 will scale everything using the dpi in ~/.Xresources
import matplotlib
matplotlib.use("Qt5Agg")

IDP_E_MINUS  =    20
IDP_MU_MINUS =    22

SNOMAN_MASS = {
    20: 0.511,
    21: 0.511,
    22: 105.658,
    23: 105.658
}

def plot_hist(x, label=None):
    # determine the bin width using the Freedman Diaconis rule
    # see https://en.wikipedia.org/wiki/Freedman%E2%80%93Diaconis_rule
    h = 2*iqr(x)/len(x)**(1/3)
    n = max(int((np.max(x)-np.min(x))/h),10)
    bins = np.linspace(np.min(x),np.max(x),n)
    plt.hist(x, bins=bins, histtype='step', label=label)

def plot_legend(n):
    plt.figure(n)
    ax = plt.gca()
    handles, labels = ax.get_legend_handles_labels()
    new_handles = [Line2D([],[],c=h.get_edgecolor()) for h in handles]
    plt.legend(handles=new_handles,labels=labels)

def get_stats(x):
    """
    Returns a tuple (mean, error mean, std, error std) for the values in x.

    The formula for the standard error on the standard deviation comes from
    https://stats.stackexchange.com/questions/156518.
    """
    mean = np.mean(x)
    std = np.std(x)
    n = len(x)
    u4 = np.mean((x-mean)**4)
    error = np.sqrt((u4-(n-3)*std**4/(n-1))/n)/(2*std)
    return mean, std/np.sqrt(n), std, error

if __name__ == '__main__':
    import argparse
    import matplotlib.pyplot as plt
    import numpy as np

    parser = argparse.ArgumentParser("plot fit results")
    parser.add_argument("filenames", nargs='+', help="input files")
    args = parser.parse_args()

    for filename in args.filenames:
        print(filename)
        with open(filename) as f:
            data = yaml.load(f.read(),Loader=Loader)

        dx = []
        dy = []
        dz = []
        dT = []
        thetas = []
        likelihood_ratio = []
        t_electron = []
        t_muon = []
        psi = []
        for event in data['data']:
            # get the particle ID
            id = event['mcgn'][0]['id']

            if 'ev' not in event:
                continue

            if 'fit' not in event['ev'][0]:
                # if nhit < 100 we don't fit the event
                continue

            if id not in event['ev'][0]['fit']:
                continue

            fit = event['ev'][0]['fit']

            mass = SNOMAN_MASS[id]
            # for some reason it's the *second* track which seems to contain the
            # initial energy
            true_energy = event['mcgn'][0]['energy']
            # The MCTK bank has the particle's total energy (except for neutrons)
            # so we need to convert it into kinetic energy
            ke = true_energy - mass
            energy = fit[id]['energy']
            dT.append(energy-ke)
            true_posx = event['mcgn'][0]['posx']
            posx = fit[id]['posx']
            dx.append(posx-true_posx)
            true_posy = event['mcgn'][0]['posy']
            posy = fit[id]['posy']
            dy.append(posy-true_posy)
            true_posz = event['mcgn'][0]['posz']
            posz = fit[id]['posz']
            dz.append(posz-true_posz)
            dirx = event['mcgn'][0]['dirx']
            diry = event['mcgn'][0]['diry']
            dirz = event['mcgn'][0]['dirz']
            true_dir = [dirx,diry,dirz]
            true_dir = np.array(true_dir)/np.linalg.norm(true_dir)
            theta = fit[id]['theta']
            phi = fit[id]['phi']
            dir = [np.sin(theta)*np.cos(phi),np.sin(theta)*np.sin(phi),np.cos(theta)]
            dir = np.array(dir)/np.linalg.norm(dir)
            thetas.append(np.degrees(np.arccos(np.dot(true_dir,dir))))
            if IDP_E_MINUS in fit and IDP_MU_MINUS in fit:
                fmin_electron = fit[IDP_E_MINUS]['fmin']
                fmin_muon = fit[IDP_MU_MINUS]['fmin']
                likelihood_ratio.append(fmin_muon-fmin_electron)
            if IDP_E_MINUS in fit:
                t_electron.append(fit[IDP_E_MINUS]['time'])
            if IDP_MU_MINUS in fit:
                t_muon.append(fit[IDP_MU_MINUS]['time'])

            if 'nhit' in event['ev'][0]:
                nhit = event['ev'][0]['nhit']
                psi.append(fit[id]['psi']/nhit)

        if len(t_electron) < 2 or len(t_muon) < 2:
            continue

        mean, mean_error, std, std_error = get_stats(dT)
        print("dT      = %.2g +/- %.2g" % (mean, mean_error))
        print("std(dT) = %.2g +/- %.2g" % (std, std_error))
        mean, mean_error, std, std_error = get_stats(dx)
        print("dx      = %4.2g +/- %.2g" % (mean, mean_error))
        print("std(dx) = %4.2g +/- %.2g" % (std, std_error))
        mean, mean_error, std, std_error = get_stats(dy)
        print("dy      = %4.2g +/- %.2g" % (mean, mean_error))
        print("std(dy) = %4.2g +/- %.2g" % (std, std_error))
        mean, mean_error, std, std_error = get_stats(dz)
        print("dz      = %4.2g +/- %.2g" % (mean, mean_error))
        print("std(dz) = %4.2g +/- %.2g" % (std, std_error))
        mean, mean_error, std, std_error = get_stats(thetas)
        print("std(theta) = %4.2g +/- %.2g" % (std, std_error))

        plt.figure(1)
        plot_hist(dT, label=filename)
        plt.xlabel("Kinetic Energy difference (MeV)")
        plt.figure(2)
        plot_hist(dx, label=filename)
        plt.xlabel("X Position difference (cm)")
        plt.figure(3)
        plot_hist(dy, label=filename)
        plt.xlabel("Y Position difference (cm)")
        plt.figure(4)
        plot_hist(dz, label=filename)
        plt.xlabel("Z Position difference (cm)")
        plt.figure(5)
        plot_hist(thetas, label=filename)
        plt.xlabel(r"$\theta$ (deg)")
        plt.figure(6)
        plot_hist(likelihood_ratio, label=filename)
        plt.xlabel(r"Log Likelihood Ratio ($e/\mu$)")
        plt.figure(7)
        plot_hist(np.array(t_electron)/1e3/60.0, label=filename)
        plt.xlabel(r"Electron Fit time (minutes)")
        plt.figure(8)
        plot_hist(np.array(t_muon)/1e3/60.0, label=filename)
        plt.xlabel(r"Muon Fit time (minutes)")
        if len(psi):
            plt.figure(9)
            plot_hist(psi, label=filename)
            plt.xlabel(r"$\Psi$/Nhit")

    plot_legend(1)
    plot_legend(2)
    plot_legend(3)
    plot_legend(4)
    plot_legend(5)
    plot_legend(6)
    plot_legend(7)
    plot_legend(8)
    if len(psi):
        plot_legend(9)
    plt.show()