# -*- coding: utf-8 -*-
"""
Created on Wed Dec 28 14:46:30 2016

@author: valerie
"""

#import os
#cwd = os.getcwd()


#ratings_data_contents.movie[0:3]
user = ratings_data_contents.user
movie = ratings_data_contents.movie
#ii = np.ravel(movie == movie[0])
#nb = sum(ii)

# Statistiques descriptives
# Nombre de vote par film
#nb = np.zeros((1,max(movie)))
#for k in range(min(movie),max(movie)):
#    ii = np.ravel(movie == k)
#    nb[k-1] = sum(ii)
    
#movielist = np.unique(movie)  
#nb = list()
#for k in movielist:
#    ii = np.ravel(movie == k)
#    nb.append(sum(ii)) # on ajoute un élément à la liste np
    

# Normalisation by members


# type de films
# on créé une matriee de 0 et de 1 pour lister les types de film pour chaque film
f = open('/Users/valerie/Dropbox/ENSEIGNEMENT/EXPOSES/2016-2017/MOVIES/movies.dat', 'r')
movietype = list()
for line in f:
    movietype.append(repr(line))
    
"""
ex = movietype[0]
np.ravel(ex=="(")

if "(" in ex:
    print "trouvé" 
"""
    
liste_types = ["Action","Adventure","Animation","Children's","Comedy","Crime",
               "Documentary","Drama","Fantasy","Horror","Musical","Mystery",
               "Romance","Sci-Fi","Thriller","War"]
lastmovie = 3952
type_mat = np.zeros((lastmovie,len(liste_types)),int)
annee = list()
typemovie = list()
for k in range(0,len(movietype)-1):
    ex = movietype[k]
    i=0
    for c in range(0,len(ex)):
        i += 1
        if ex[c]==")" and ex[c+1]==":": break
    annee.append(ex[i-5:i-1])
    typemovie.append(ex[i+2:(len(ex)-3)])
    j=0
    for c in ex:
        j += 1
        if c==":": break
    for ktyp in range(0,len(liste_types)-1):
        if liste_types[ktyp] in ex[i+7:(len(ex)-3)]: type_mat[int(ex[1:j-1]),ktyp] = 1

#################################
# Créer  un fichier votant-films, type de film, demo, note
# np.concatenate
f = open('/Users/valerie/Dropbox/ENSEIGNEMENT/EXPOSES/2016-2017/MOVIES/users.dat', 'r')
userdemo = list()
for line in f:
    userdemo.append(repr(line))
    
user_mat = np.zeros((len(userdemo),4),int)
for k in range(0,len(userdemo)-1):
    ex = userdemo[k]
    i = 0
    for c in ex:
        i += 1
        if c==":": break
    user_mat[k,0] = np.array(ex[i+1]=='M')*1
    if ex[i+5]==":":
        user_mat[k,1] = ex[i+4:i+5]
        j = i+6
    else:
        user_mat[k,1] = ex[i+4:i+6]
        j = i+7
    if (ex[j+2]==":"):
        user_mat[k,2] = ex[j+1:j+2]
        i = j+4
    else:
        user_mat[k,2] = ex[j+1:j+3]
        i=j+5
    user_mat[k,3] = ex[i:i+5]
    
mat_all = np.concatenate((np.reshape(user,(len(user),1)),
                          np.reshape(movie,(len(user),1)),
                          type_mat[movie-1,],user_mat[user-1,]),axis=1)

# Pour chaque votant, on compte les types de films qu'il a noté
#           
rating = ratings_data_contents.rating
votant = np.unique(user)          
            
votant_mat = np.zeros((len(votant),len(liste_types)),int)
votant_sd = np.zeros((len(votant),len(liste_types)),int)
best_rate = [0] * len(mat_all)
i=0
for  k  in votant:
    elt = np.where(user==k)
    #elt = np.array(elt[0])
    tmp = rating[elt[0]]
    index = elt[0]
    votant_mat[i,] = np.dot(tmp,type_mat[movie[elt[0]]-1,])
    votant_sd[i,] = np.dot(tmp**2,type_mat[movie[elt[0]]-1,])/sum(type_mat[movie[elt[0]]-1,])
    i = i+1
    nv = len(elt[0])
    nv5 = int(round(nv*0.05,0))
    output = np.argsort(-tmp)
    if nv5>0:
        for ii in output[0:nv5]:
            best_rate[index[int(ii)]] = 1
    else:
        ii = output[0:1]
        best_rate[index[int(ii)]]
    #for i, x in enumerate(sorted(range(nv), key=lambda y: input[y])):
    #   output[x] = i

mat_all = np.concatenate((np.reshape(user,(len(user),1)),
                          np.reshape(movie,(len(user),1)),
                          type_mat[movie-1,],user_mat[user-1,],
                          np.reshape(best_rate,(len(user),1))),axis=1)   
    



    
# Pb : cette matrice ne tient pas compte des votes... 
# Calculer une moyenne pondérée des votes pour chaque type de film? A voir
    
# Clustering
    
# Ensemble d'apprentissage/validation
    
#from sklearn.cross_validation import train_test_split    

#X_train, X_test = train_test_split(votant_mat) 
#from sklearn.cluster import KMeans
 
#kmeans = KMeans(n_clusters=5, random_state=0).fit(X_train)
#kmeans.cluster_centers_
#kmeans.labels_
#cc = kmeans.predict(X_test) # ???