import matplotlib.cm as colormap
from matplotlib.pyplot     import figure, show, legend, gca, setp
from matplotlib.gridspec   import GridSpec
from numpy                 import sort, log10, histogram, pi, arange, where, ma, zeros, append
from numpy                 import rot90, flipud, histogram2d, logspace, linspace, size, loadtxt
from numpy                 import array, argsort, nditer, mean
from mpl_toolkits.mplot3d  import Axes3D
from src.read              import select_events, ReadProfile
from src.analytic          import degre, yr, Analytic_delay_vs_theta, Analytic_observables_vs_energy

import matplotlib as mpl
label_size = 20
mpl.rcParams['xtick.labelsize'] = label_size 
mpl.rcParams['ytick.labelsize'] = label_size 

def weighted_median(values, weights):
    ''' compute the weighted median of values list. The weighted median is computed as follows:
    1- sort both lists (values and weights) based on values.
    2- select the 0.5 point from the weights and return the corresponding values as results
    e.g. values = [1, 3, 0] and weights=[0.1, 0.3, 0.6] assuming weights are probabilities.
    sorted values = [0, 1, 3] and corresponding sorted weights = [0.6,     0.1, 0.3] the 0.5 point on
    weight corresponds to the first item which is 0. so the weighted     median is 0.'''

    #convert the weights into probabilities
    sum_weights = sum(weights)
    weights = array([(w*1.0)/sum_weights for w in weights])
    #sort values and weights based on values
    values = array(values)
    sorted_indices = argsort(values)
    values_sorted  = values[sorted_indices]
    weights_sorted = weights[sorted_indices]
    #select the median point
    it = nditer(weights_sorted, flags=['f_index'])
    accumulative_probability = 0
    median_index = -1
    while not it.finished:
        accumulative_probability += it[0]
        if accumulative_probability > 0.5:
            median_index = it.index
            return values_sorted[median_index]
        elif accumulative_probability == 0.5:
            median_index = it.index
            it.iternext()
            next_median_index = it.index
            return mean(values_sorted[[median_index, next_median_index]])
        it.iternext()

    return values_sorted[median_index]



#=========================== CORELATION BETWEEN OBSERVABLES ===========================#
def observables_vs_energy(energy,theta,delay,weight,Erange=[1e-3,1e5],median=False,nbBins=-1):#28):
   # nbBins=16, Erange=[1e-1,1e3] <=> Arlen 2014 (energy binning)
   if (nbBins==-1):
      bins=append(logspace(log10(Erange[0]),log10(0.9),100),logspace(0,log10(Erange[1]),16))
      nbBins=size(bins)-1
   if (nbBins==-2): # <=> Arlen 2014 (energy binning)
      nbBins=16 
      Erange=[1e-1,1e3]
      bins=logspace(log10(Erange[0]),log10(Erange[1]),nbBins+1)
   else:
      bins=logspace(log10(Erange[0]),log10(Erange[1]),nbBins+1)
   ener = (bins[1:nbBins+1]*bins[0:nbBins])**0.5
   #dtheta, dE = histogram(energy,bins,weights=weight*theta)
   #dt,     dE = histogram(energy,bins,weights=weight*delay)
   #dN,     dE = histogram(energy,bins,weights=weight)
   dt = zeros(nbBins,dtype="float64")
   dtheta = zeros(nbBins,dtype="float64")
   if median==False:
      for i in arange(nbBins):
         mask = (energy >= bins[i]) & (energy < bins[i+1])
         w = sum(weight[mask])
         t = sum(weight[mask]*delay[mask])
         angle = sum(weight[mask]*theta[mask])
         if w !=0:
            dt[i]=t/w
            dtheta[i]=angle/w
   else:
      for i in arange(nbBins):
         mask = (energy >= bins[i]) & (energy < bins[i+1])
         if True in mask:
            dt[i] = weighted_median(delay[mask],weight[mask])
            dtheta[i] = weighted_median(theta[mask],weight[mask])
         else:
            dt[i] = 0
            dtheta[i] = 0
   return ener, dtheta, dt

def observables_vs_delay(energy,theta,delay,weight,time_range=[10**-0.5,10**8.5],median=False,nbBins=100):#9): 
   # <=> Taylor 2011 (time binning)
   bins=logspace(log10(time_range[0]),log10(time_range[1]),nbBins+1)
   time = (bins[1:nbBins+1]*bins[0:nbBins])**0.5
   dE = zeros(nbBins,dtype="float64")
   dtheta = zeros(nbBins,dtype="float64")
   if median==False:
      for i in arange(nbBins):
         mask = (delay >= bins[i]) & (delay < bins[i+1])
         w = sum(weight[mask])
         E = sum(weight[mask]*energy[mask])
         angle = sum(weight[mask]*theta[mask])
         if w !=0:
            dE[i]=E/w
            dtheta[i]=angle/w
   else:
      for i in arange(nbBins):
         mask = (delay >= bins[i]) & (delay < bins[i+1])
         if True in mask:
            dE[i] = weighted_median(energy[mask],weight[mask])
            dtheta[i] = weighted_median(theta[mask],weight[mask])
         else:
            dE[i] = 0
            dtheta[i] = 0
   return time, dtheta, dE

def drawObservables(fileId,psf=180,Nb=28,plot_generation_density=False,plot_others_codes=False,one_figure=True):
   #ax = figure(figsize=(12,9)).add_subplot(111,projection='3d')
   if one_figure:
      fig = figure(figsize=(20,18))
      gs = GridSpec(2, 2, height_ratios=[1,1], width_ratios=[1,1]) 
      fig.subplots_adjust(hspace=0,wspace=0)
      ax0 = fig.add_subplot(gs[0])
      ax2 = fig.add_subplot(gs[2],sharex=ax0)
      ax3 = fig.add_subplot(gs[3],sharey=ax2)
   else:
      fig0 = figure(figsize=(12,9))
      ax0 = fig0.add_subplot(111)
      fig2 = figure(figsize=(12,9))
      ax2 = fig2.add_subplot(111)
      fig3 = figure(figsize=(12,9))
      ax3 = fig3.add_subplot(111)

   fileId = "Simulations/"+fileId

   i=0
   nbBins = 250
   theta_range = [1e-5,psf]
   delay_range = [1e-4,5e9]
   energy_range = [1e-3,1e4]
   weight,energy,delay,theta,phi,Esource,generation = select_events(fileId,Erange=energy_range)
   delay /= yr
   theta /= degre

   print "theta range:  [",min(theta),", ",max(theta),"] (degre)"
   print "delay range:  [",min(delay),", ",max(delay),"] (yr)"
   print "energy range: [",min(energy),", ",max(energy),"] (GeV)"
   print "weight range: [",min(weight),", ",max(weight),"]"

   colors=['b','g']

   cond = (theta>0) & (delay>0)
   if plot_generation_density: 
      cmaps=[colormap.Blues,colormap.Greens,colormap.Reds]
      for gen in [2,4]:#sort(list(set(generation))):
         cond = cond & (generation==gen)
         H, xedges, yedges = histogram2d(log10(energy[cond]),log10(delay[cond]),bins=nbBins,weights=weight[cond])
         H = flipud(rot90(ma.masked_where(H==0,H)))
         im1 = ax0.pcolormesh(10**xedges,10**yedges,log10(H),cmap=cmaps[i])   

         H, xedges, yedges = histogram2d(log10(energy[cond]),log10(theta[cond]),bins=nbBins,weights=weight[cond])
         H = flipud(rot90(ma.masked_where(H==0,H)))
         im2 = ax2.pcolormesh(10**xedges,10**yedges,log10(H),cmap=cmaps[i])      

         H, xedges, yedges = histogram2d(log10(delay[cond]),log10(theta[cond]),bins=nbBins,weights=weight[cond])
         H = flipud(rot90(ma.masked_where(H==0,H)))
         im3 = ax3.pcolormesh(10**xedges,10**yedges,log10(H),cmap=cmaps[i])

         ener,angle,dt = observables_vs_energy(energy[cond],theta[cond],delay[cond],weight[cond],nbBins=Nb,median=False)
         ax0.plot(ener,dt,color=colors[i],linewidth=2,label="gen = %.0f"%(i+1))
         ax2.plot(ener,angle,color=colors[i],linewidth=2,label="gen = %.0f"%(i+1))
         dt,angle,ener = observables_vs_delay(energy[cond],theta[cond],delay[cond],weight[cond],nbBins=Nb,median=False)
         ax3.plot(dt,angle,color=colors[i],linewidth=2,label="gen = %.0f"%(i+1))
         i+=1
            
   else:
      H, xedges, yedges = histogram2d(log10(energy[cond]),log10(delay[cond]),bins=nbBins,weights=weight[cond])
      H = flipud(rot90(ma.masked_where(H==0,H)))
      im1 = ax0.pcolormesh(10**xedges,10**yedges,log10(H),cmap=colormap.YlOrBr)   

      H, xedges, yedges = histogram2d(log10(energy[cond]),log10(theta[cond]),bins=nbBins,weights=weight[cond])
      H = flipud(rot90(ma.masked_where(H==0,H)))
      im2 = ax2.pcolormesh(10**xedges,10**yedges,log10(H),cmap=colormap.YlOrBr)      
    
      H, xedges, yedges = histogram2d(log10(delay[cond]),log10(theta[cond]),bins=nbBins,weights=weight[cond])
      H = flipud(rot90(ma.masked_where(H==0,H)))
      im3 = ax3.pcolormesh(10**xedges,10**yedges,log10(H),cmap=colormap.YlOrBr)

      for gen in [2,4]:#sort(list(set(generation))):
         cond = (generation==gen)
         ener,angle,dt = observables_vs_energy(energy[cond],theta[cond],delay[cond],weight[cond],nbBins=Nb,median=False)
         ax0.plot(ener,dt,color=colors[i],linewidth=2,label="gen = %.0f"%(i+1))
         ax2.plot(ener,angle,color=colors[i],linewidth=2,label="gen = %.0f"%(i+1))
         dt,angle,ener = observables_vs_delay(energy[cond],theta[cond],delay[cond],weight[cond],nbBins=Nb,median=False)
         ax3.plot(dt,angle,color=colors[i],linewidth=2,label="gen = %.0f"%(i+1))
         i+=1       
            
   ener,angle,dt = observables_vs_energy(energy,theta,delay,weight,nbBins=Nb,median=False)
   ax0.plot(ener,dt,color='k',linewidth=2,label="all gen")
   ax2.plot(ener,angle,color='k',linewidth=2,label="all gen")
   dt,angle,ener = observables_vs_delay(energy,theta,delay,weight,nbBins=Nb,median=False)
   ax3.plot(dt,angle,color='k',linewidth=2,label="all gen")

   nE = 5000 
   Emin = 1.e-3
   Emax = 1e4
   E = Emin*(Emax/Emin)**(arange(nE)/(nE-1.))
   theta_fit, delay_fit, theta_fit2, delay_fit2 = Analytic_observables_vs_energy(E,"../"+fileId)
   ax0.plot(E,delay_fit/yr,'--b',linewidth=2)
   ax0.plot(E,delay_fit2/yr,'--g',linewidth=2)
   ax2.plot(E,theta_fit,'--b',linewidth=2)
   ax2.plot(E,theta_fit2,'--g',linewidth=2)
   
   ax3.plot(delay_fit[delay_fit.argsort()]/yr,theta_fit[delay_fit.argsort()],'--b',linewidth=2)
   ax3.plot(delay_fit2[delay_fit2.argsort()]/yr,theta_fit2[delay_fit2.argsort()],'--g',linewidth=2)
   #ax3.scatter(delay_fit/yr,theta_fit,color='b',marker="+")
    
   if plot_others_codes:
      # Results from Arlen 2014 - fig. 2a, 2b
      # =====================================
      #   - binning in energy : log, 16 bins between 1e-1 et 1e3 GeV
      #   - PSF = 10 deg
      data = loadtxt(fileId+'/Arlen2014-fig2a.csv', delimiter=',')
      ax0.plot(data[:,0],data[:,1],color="m",linestyle='--',linewidth=2,label="Arlen 2014 - no EBL")
      data = loadtxt(fileId+'/Arlen2014-fig2b.csv', delimiter=',')
      ax0.plot(data[:,0],data[:,1],color="m",linestyle='-.',linewidth=2,label="Arlen 2014 - EBL")

      # Results from Taylor 2011 - fig. 2
      # =================================
      data = loadtxt(fileId+'/Taylor2011-fig2.csv', delimiter=',')
      ax0.plot(data[:,0]*1e-9,data[:,1],color="m",linestyle=':',linewidth=2,label="Taylor 2011")  

   #ax.legend(loc="best")
   #ax.set_xlabel("$E$ [GeV]")
   #ax.set_ylabel("$\\theta$ [deg]")
   #ax.set_zlabel("$\\Delta t$ [s]")

   #ax3.set_title(fileId+" - selection %0.1f"%psf)
   if one_figure:
      ax0.legend(bbox_to_anchor=(1.05, 1), loc=2, borderaxespad=0.)

   ax0.grid(b=True,which='major')
   ax0.set_xscale('log')
   ax0.set_yscale('log')
   ax0.set_xlim(energy_range)
   ax0.set_ylim(delay_range)
   ax0.set_ylabel("$\\Delta t$ [yrs]",fontsize=label_size)
   if one_figure:
      setp(ax0.get_xticklabels(), visible=False)
   else:
      ax0.set_xlabel("$E$ [GeV]",fontsize=label_size)
      ax0.legend(loc="best",fontsize=label_size)
      cbar1=fig0.colorbar(im1, ax=ax0)
      cbar1.ax.set_ylabel("counts [log]",fontsize=label_size)

   ax2.grid(b=True,which='major')
   ax2.set_xscale('log')
   ax2.set_yscale('log')
   ax2.set_xlim(energy_range)
   ax2.set_ylim(theta_range)
   ax2.set_xlabel("$E$ [GeV]",fontsize=label_size)
   ax2.set_ylabel("$\\theta$ [deg]",fontsize=label_size)
   if not one_figure:
      ax2.legend(loc="best",fontsize=label_size)
      cbar2=fig2.colorbar(im2, ax=ax2)
      cbar2.ax.set_ylabel("counts [log]",fontsize=label_size)

   ax3.grid(b=True,which='major')
   ax3.set_xscale('log')
   ax3.set_yscale('log')
   ax3.set_xlim(delay_range)
   ax3.set_ylim(theta_range)
   ax3.set_xlabel("$\\Delta t$ [yrs]",fontsize=label_size)
   if one_figure:
      setp(ax3.get_yticklabels(), visible=False)
   else:
      ax3.set_ylabel("$\\theta$ [deg]",fontsize=label_size)
      ax3.legend(loc="best",fontsize=label_size)
      cbar3=fig3.colorbar(im3, ax=ax3)
      cbar3.ax.set_ylabel("counts [log]",fontsize=label_size)

   show()

def drawDelays_vs_energy(files,psf=180,plot_others_codes=False):
   ax1 = figure(figsize=(12,9)).add_subplot(111)
   nbBins = 100

   for fileId0 in files:
      fileId = "Simulations/simple case/"+fileId0
      weight, energy, delay, arrival_angle , theta, phi, gen = select_events(fileId)
      #dt,dtheta,ener = observables_vs_delay(energy,arrival_angle/degre,delay/yr,weight)
      #ax1.plot(ener,dt,marker='*',linestyle=":",linewidth=2) 
      cond = (arrival_angle/degre < psf) & (gen%2==0) #PSF_Taylor2011(energy) 
      energy = energy[cond]
      arrival_angle = arrival_angle[cond]
      delay = delay[cond]
      weight = weight[cond]
      ener,dtheta,dt = observables_vs_energy(energy,arrival_angle/degre,delay/yr,weight,nbBins=-2)
      p0, =ax1.plot(ener,dt,marker='*',linewidth=2,label=fileId0) 
      
      nth = 5000
      thmin = 0.1
      thmax = 1e4
      E = thmin*(thmax/thmin)**(arange(nth)/(nth-1.))
      #ax1.plot(E,Analytic_observables_vs_energy(E,fileId)[1]/yr,color=p0.get_color(),linestyle="-")
                                                     
   if plot_others_codes:
      # Results from Arlen 2014 - fig. 2a, 2b
      # =====================================
      #   - binning in energy : log, 16 bins between 1e-1 et 1e3 GeV
      #   - PSF = 10 deg
      data = loadtxt(fileId+'/Arlen2014-fig2a.csv', delimiter=',')
      p1,= ax1.plot(data[:,0],data[:,1],color="r",linestyle='--',linewidth=2,label="Arlen 2014 - fig 2a")
      data = loadtxt(fileId+'/Arlen2014-fig2b.csv', delimiter=',')
      p2,= ax1.plot(data[:,0],data[:,1],color="r",linestyle='-.',linewidth=2,label="Arlen 2014 - fig 2b")
   
      # Results from Taylor 2011 - fig. 2
      # =================================
      #data = loadtxt(fileId+'/Taylor2011-fig2.csv', delimiter=',')
      #p3,= ax1.plot(data[:,0]*1e-9,data[:,1],color=p0.get_color(),linestyle=':',linewidth=2)
   
      #leg1 = legend([p0,p1,p2], ["our code","Arlen 2014 - fig 2a",
      #    "Arlen 2014 - fig 2b"], loc=1)
                                                              
   ax1.grid(b=True,which='major')  
   ax1.set_xscale('log')
   ax1.set_yscale('log')
   #ax1.set_ylim([1,1e8])
   #ax1.set_xlim([1e-1,1e3])
   ax1.set_xlabel("$E$ [GeV]",fontsize=label_size)
   ax1.set_ylabel("$\\Delta t$ [yrs]",fontsize=label_size) 
   ax1.legend(loc="best",fontsize=label_size)#2)
   #ax1 = gca().add_artist(leg1)

   show()