# Obtener la info vía la API

# sudo pip3 install spotipy
# oficial api doc: https://developer.spotify.com/web-api/
# python spotify library: https://github.com/plamere/spotipy
# Doc Reference: http://spotipy.readthedocs.io/en/latest/
# Sacar token en Spotify.com: https://developer.spotify.com/my-applications/
# Examples: https://github.com/plamere/spotipy/tree/master/examples

import spotipy
import spotipy.util as util
import json
import urllib3
#import os
#os.environ["SPOTIPY_CLIENT_ID"] = ""
#os.environ["SPOTIPY_CLIENT_SECRET"] = ""
#os.environ["SPOTIPY_REDIRECT_URI"] = "http://localhost:8888/callback"
#username = os.environ["SPOTIFY_USERNAME"]

scope = 'user-read-recently-played'
cantidad = 50
token = util.prompt_for_user_token(username, scope)

# Obtener los últimos 50 tracks escuchados

http = urllib3.PoolManager()
url = 'https://api.spotify.com/v1/me/player/recently-played?limit='+str(cantidad)
headers = {'Content-Type': 'application/json',
           'Authorization': 'Bearer '+token}
r = http.request('GET', url, headers=headers)

# Grabar toda la información en archivos json por separado

#tracks = json.loads(r.data)
#for item in tracks['items']:
#    track = item['track']
#    track['played_at'] = item['played_at']
#    with open(dirname+track['id']+'.json', 'w') as out:
#        out.write(json.dumps(track))

# Grabar el .csv
#x = pd.DataFrame(data=[])
#id_list = list()
#duration_list = list()
#popularity_list = list()
#popularity_list = list()
#track_number_list = list()
#played_at_list = list()
#for subdir, dirs, files in os.walk(files_dir):
#    for f in files:
#        if os.path.splitext(f)[1]==".json":
#            data = json.load( open(files_dir + "/" + f,'r') )
#            id_list.append( os.path.splitext(f)[0] )
#            duration_list.append( data['duration_ms'] )
#            popularity_list.append( data['popularity'] )
#            track_number_list.append( data['track_number'] )
#            played_at_list.append( data['played_at'] )
#x['spotify_id'] = id_list
#x['duration_ms'] = duration_list
#x['popularity'] = popularity_list
#x['track_number'] = track_number_list
#x['played_at'] = played_at_list

#x.to_csv('canciones_escuchadas.csv') # grabar .csv

import pandas as pd
x = pd.DataFrame.from_csv('dataset-spotify/canciones_escuchadas.csv')
x

# Genero una tabla sin el id de spotify ni la fecha+hora de escuchado
xnew = x

del_columns = ['spotify_id', 'played_at']
xnew.drop(del_columns, inplace=True, axis=1)
xnew

from sklearn import datasets
from sklearn.cluster import KMeans

import sklearn.metrics as sm
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.cm as cm

# Se averigua el rango de (número de) tracks posibles 
max_track = max(x.track_number)
print("Rango de tracks: [%i, %i]"%(min(x.track_number), max_track) )

# Plot de duración vs popularidad, anotando el número de track en cada punto/marker...
fig, ax = plt.subplots(figsize=(14,7))
ax.plot(x.duration_ms/1000./60.,x.popularity, ls="", marker="o")
for xi, yi, pidi in zip(x.duration_ms/1000./60.,x.popularity,x.track_number):
    ax.annotate(str(pidi), xy=(xi,yi))

plt.title('Duración (minutos) vs Popularidad (%). Número de track anotado 1 a 12')
plt.show()

colormap_original = cm.viridis_r(np.linspace(0,1,max_track+1))

plt.figure(figsize=(14,7))
plt.scatter(x.duration_ms/1000./60., x.popularity, c=colormap_original[xnew.track_number], s=40)
plt.title('Duración (minutos) vs Popularidad (%). El color denota nro de track') # más oscuro es más alto
plt.show()

model = KMeans(n_clusters=3) 
model.fit(xnew)

plt.figure(figsize=(14,7))
  
colormap_original = cm.viridis_r(np.linspace(0,1,max_track+1))

# Plot the Original Classifications
plt.subplot(2, 2, 1)
plt.scatter(x.duration_ms/1000./60., x.popularity, c=colormap_original[xnew.track_number.as_matrix()], s=40)
plt.title('Original: Color track. Dur vs Pop.')
 
# Plot the Models Classifications
plt.subplot(2, 2, 2)
plt.scatter(x.duration_ms/1000./60., x.popularity, c=colormap_original[model.labels_], s=40)
plt.title('K=3 Classification. Dur vs Pop')

# Plot the Models Classifications
plt.subplot(2, 2, 3)
plt.scatter(x.track_number, x.popularity, c=colormap_original[model.labels_], s=40)
plt.title('K=3 Classification. Track vs Pop')

# Plot the Models Classifications
plt.subplot(2, 2, 4)
plt.scatter(x.track_number, x.duration_ms/1000./60., c=colormap_original[model.labels_], s=40)
plt.title('K=3 Classification. Track vs Dur')

plt.show()

# 5 clusters
model = KMeans(n_clusters=5) 
model.fit(x)

plt.figure(figsize=(14,7))
  
colormap_original = cm.viridis_r(np.linspace(0,1,max_track+1))

# Plot the Original Classifications
plt.subplot(2, 2, 1)
plt.scatter(x.duration_ms/1000./60., x.popularity, c=colormap_original[xnew.track_number.as_matrix()], s=40)
plt.title('Original: Color track. Dur vs Pop.')
 
# Plot the Models Classifications
plt.subplot(2, 2, 2)
plt.scatter(x.duration_ms/1000./60., x.popularity, c=colormap_original[model.labels_], s=40)
plt.title('K=5 Classification. Dur vs Pop')

# Plot the Models Classifications
plt.subplot(2, 2, 3)
plt.scatter(x.track_number, x.popularity, c=colormap_original[model.labels_], s=40)
plt.title('K=5 Classification. Track vs Pop')

# Plot the Models Classifications
plt.subplot(2, 2, 4)
plt.scatter(x.track_number, x.duration_ms/1000./60., c=colormap_original[model.labels_], s=40)
plt.title('K=5 Classification. Track vs Dur')

plt.show()