def subsetDataFrame(df, tracks):
## subset dataframe by spotify ids
track_ids = []
for track in tracks:
track_ids.append(sptfy.getSpotifyTrackIDs(track)[0])
db_subset = df[df.spotify_id.isin(track_ids)]
return db_subset
def subsetDataFrame(df, tracks):
## subset dataframe by spotify ids
track_ids = []
for track in tracks:
track_ids.append(sptfy.getSpotifyTrackIDs(track)[0])
db_subset = df[df.spotify_id.isin(track_ids)]
return db_subset
def main():
# ## load tracks in playlist
fives = loadFile("input", "fives.txt")
fours = loadFile("input", "fours.txt")
threes = loadFile("input", "threes.txt")
twos = loadFile("input", "twos.txt")
ones = loadFile("input", "ones.txt")
## load database of song ratings
db = loadFile("../Databases", "song_ratings_db.csv")
album_ratings = loadFile("../Databases", "album_ratings_db.csv")
config = loadFile("../config", "config.csv", True)
token = sptfy.authSpotipy()
rating = 0
for ls in [ones, twos, threes, fours, fives]:
## playlists are looped in this order such that if a song is in multiple lists it's rating will end up being the highest one
rating = rating + 1
for song in ls:
track_id, spotify_uri = sptfy.getSpotifyTrackIDs(song)
if not lookupSongBySpotifyID(track_id, db):
track = sptfy.pullSpotifyTrack(track_id, token=token)
db = db.append([{
'spotify_id': track_id,
'artist': track['artist'],
'album': track['album'],
'spotify_album_id': track['spotify_album_id'],
'song': track['title'],
'rating': rating
}])
db = db.sort(['artist', 'album', 'rating'])
saveDataFrame(db, "../Databases", "song_ratings_db.csv")
for album in pd.unique(zip(db.artist, db.album)):
artist = album[0]
album = album[1]
if not lookupAlbumBySpotifyID(album, album_ratings):
album_id = pd.unique(db[db.artist == artist][db.album == album]
['spotify_album_id'])[0]
album_data = sptfy.pullSpotifyAlbum(album_id, token=token)
ratings = db[db.artist == artist][db.album ==
album]['rating'].tolist()
if len(ratings) < 3:
## don't make album ratings for singles
continue
score = 0
countNot3 = len(np.where(r != 3))
countMoreThan2 = len(np.where(r > 2))
countMoreThan3 = len(np.where(r > 3))
for r in ratings:
if r == 5:
score = score + 100 * 1.0
scoreNot3 = score + 100 * 1.0
elif r == 4:
score = score + 80 * 1.2
scoreNot3 = score + 80 * 1.2
elif r == 3:
score = score + 60 * 1.0
elif r == 2:
score = score + 40 * 1.2
scoreNot3 = score + 40 * 1.2
elif r == 1:
score = score + 20 * 1.0
scoreNot3 = score + 20 * 1.0
std = np.std(ratings)
if std == 0.0:
std = 0.25
if countNot3 == 0:
adjMean = 3
else:
adjMean = scoreNot3 / countNot3
prop4or5 = countMoreThan3 / len(r)
adj1 = (adjMean - 3) * prop4or5
adj2 = adj1 + countMoreThan2 * 0.03
score = np.mean(ratings) + adj2
if prop4or5 == 0:
adjSD = std * 0.05
else:
adjSD = std * prop4or5 / len(r)
score = score - adjSD
## min possible score: (mean of 1-star)
min1 = 1.0
## max possible score: Radiohead "OK Computer"
max1 = 5.662521
min2 = -1.0
max2 = -0.125
scaledScore = (score - min1) / (max1 - min1)
## transform (curves the linear scores to inflate higher scores and reduce lower)
transformedScore = -1 * (8 ^ (-1 * scaledScore))
scaledScore = (transformedScore - min2) / (max2 - min2)
album_score = (round(scaledScore * 1000)) / 1.0
if album_score > 1000:
album_score = 1000
if album_score >= 965:
album_rating = 5.0
elif album_score >= 890:
album_rating = 4.5
elif album_score >= 750:
album_rating = 4.0
elif album_score >= 690:
album_rating = 3.5
elif album_score >= 625:
album_rating = 3.0
elif album_score >= 420:
album_rating = 2.5
elif album_score > 325:
album_rating = 2.0
elif album_score > 235:
album_rating = 1.5
elif album_score >= 100:
album_rating = 1.0
elif album_score < 100:
album_rating = 0.5
album_ratings = album_ratings.append([{
'spotify_album_id': album_id,
'artist': artist,
'album': album,
'year': album_data['year'],
'album_rating': album_rating,
'album_score': album_score
}])
saveDataFrame(album_ratings, "../Databases", "album_ratings_db.csv")
def lookupSongBySpotifyID(song, df):
track_id = sptfy.getSpotifyTrackIDs(song)[0]
return any(df.spotify_id == track_id)
def lookupSongBySpotifyID(song, df):
track_id = sptfy.getSpotifyTrackIDs(song)[0]
return any(df.spotify_id == track_id)
def pullEchoNestSong(auth, track, album=None, local_link=None):
url_base = "https://api.spotify.com/v1/audio-features/"
if 'spotify' in track:
track_id, spotify_uri = sptfy.getSpotifyTrackIDs(track)
## due to echonest using 2 different bucket params and url encoding the ampersand, payload cannot be used
# payload = {'api_key' : api_key, 'track_id' : spotify_uri, 'bucket' : "audio_summary&bucket=id:spotify", 'format' : "json"}
url_suffix = "%s" % sptfy.stripSpotifyURI(spotify_uri)
else:
pdb.set_trace()
## it's an echonest id and can be accessed directly
url_suffix = "?api_key=%s&id=%s&bucket=audio_summary&bucket=id:spotify&format=json" % (
auth, track)
url = url_base + url_suffix
data = mhlpr.callAPI(url, headers=auth)
## if response is a success
if int(data['response']['status']['code']) == 0 and len(
data['response']['songs']) > 0:
song = mhlpr.flattenDictCustom(data['response']['songs'][0])
if 'spotify' in track:
track = sptfy.pullSpotifyTrack(track_id)
song['album'] = track['album']
song['spotify_artist_id'] = track['spotify_artist_id']
## add spotify uri to song data
song['spotify_id'] = track_id
else:
song['album'] = album
song['spotify_id'] = local_link.strip()
## pop off unneeded data and flatten dict
song.pop('audio_md5', None)
song.pop('analysis_url', None)
song['echonest_artist_id'] = song.pop('artist_id')
if 'artist_foreign_ids' in song:
song.pop('artist_foreign_ids')
## rename keys as necessary
song['echonest_id'] = song.pop('id')
song['artist'] = song.pop('artist_name')
elif int(data['response']['status']['code']) == 5:
## the song cannot be found by the spotify id
url = "http://developer.echonest.com/api/v4/song/search"
if 'spotify' in track:
track = sptfy.pullSpotifyTrack(track_id)
artist = track['artist']
title = track['title']
payload = {
'api_key': api_key,
'artist': artist,
'title': title,
'bucket': "audio_summary",
'format': "json"
}
data = mhlpr.callAPI(url, payload)
if len(data['response']['songs']) > 0:
## pop off unneeded data and flatten dict
song = mhlpr.flattenDictCustom(data['response']['songs'][0])
if 'spotify' in track:
song['album'] = track['album']
song['spotify_artist_id'] = track['spotify_artist_id']
## check to be sure it's the correct song -- fuzzy string match of at least .75 levenshtein ratio
if fuzzyMatch(song['artist'], track['artist'], song['title'],
track['title']):
song.pop('audio_md5', None)
song.pop('analysis_url', None)
song['echonest_artist_id'] = song.pop('artist_id')
if 'artist_foreign_ids' in song:
song.pop('artist_foreign_ids')
## add spotify uri to song data
song['spotify_id'] = track_id
## rename keys as necessary
song['echonest_id'] = song.pop('id')
song['artist'] = song.pop('artist_name')
else:
## pop off unneeded data and flatten dict
song.pop('audio_md5', None)
song.pop('analysis_url', None)
song['echonest_artist_id'] = song.pop('artist_id')
if 'artist_foreign_ids' in song:
song.pop('artist_foreign_ids')
## rename keys as necessary
song['echonest_id'] = song.pop('id')
song['album'] = album
song['spotify_id'] = track_id
song['artist'] = song.pop('artist_name')
else:
print "Song not found via EchoNest search: {}".format(spotify_uri)
return None
else:
pdb.set_trace()
"Unrecognized error code."
return song
def pullEchoNestSong(auth, track, album = None, local_link = None):
url_base = "https://api.spotify.com/v1/audio-features/"
if 'spotify' in track:
track_id, spotify_uri = sptfy.getSpotifyTrackIDs(track)
## due to echonest using 2 different bucket params and url encoding the ampersand, payload cannot be used
# payload = {'api_key' : api_key, 'track_id' : spotify_uri, 'bucket' : "audio_summary&bucket=id:spotify", 'format' : "json"}
url_suffix = "%s" % sptfy.stripSpotifyURI(spotify_uri)
else:
pdb.set_trace()
## it's an echonest id and can be accessed directly
url_suffix = "?api_key=%s&id=%s&bucket=audio_summary&bucket=id:spotify&format=json" % (auth, track)
url = url_base + url_suffix
data = mhlpr.callAPI(url, headers = auth)
## if response is a success
if int(data['response']['status']['code']) == 0 and len(data['response']['songs']) > 0:
song = mhlpr.flattenDictCustom(data['response']['songs'][0])
if 'spotify' in track:
track = sptfy.pullSpotifyTrack(track_id)
song['album'] = track['album']
song['spotify_artist_id'] = track['spotify_artist_id']
## add spotify uri to song data
song['spotify_id'] = track_id
else:
song['album'] = album
song['spotify_id'] = local_link.strip()
## pop off unneeded data and flatten dict
song.pop('audio_md5', None)
song.pop('analysis_url', None)
song['echonest_artist_id'] = song.pop('artist_id')
if 'artist_foreign_ids' in song:
song.pop('artist_foreign_ids')
## rename keys as necessary
song['echonest_id'] = song.pop('id')
song['artist'] = song.pop('artist_name')
elif int(data['response']['status']['code']) == 5:
## the song cannot be found by the spotify id
url = "http://developer.echonest.com/api/v4/song/search"
if 'spotify' in track:
track = sptfy.pullSpotifyTrack(track_id)
artist = track['artist']
title = track['title']
payload = {'api_key' : api_key, 'artist' : artist, 'title' : title, 'bucket' : "audio_summary", 'format' : "json"}
data = mhlpr.callAPI(url, payload)
if len(data['response']['songs']) > 0:
## pop off unneeded data and flatten dict
song = mhlpr.flattenDictCustom(data['response']['songs'][0])
if 'spotify' in track:
song['album'] = track['album']
song['spotify_artist_id'] = track['spotify_artist_id']
## check to be sure it's the correct song -- fuzzy string match of at least .75 levenshtein ratio
if fuzzyMatch(song['artist'], track['artist'], song['title'], track['title']):
song.pop('audio_md5', None)
song.pop('analysis_url', None)
song['echonest_artist_id'] = song.pop('artist_id')
if 'artist_foreign_ids' in song:
song.pop('artist_foreign_ids')
## add spotify uri to song data
song['spotify_id'] = track_id
## rename keys as necessary
song['echonest_id'] = song.pop('id')
song['artist'] = song.pop('artist_name')
else:
## pop off unneeded data and flatten dict
song.pop('audio_md5', None)
song.pop('analysis_url', None)
song['echonest_artist_id'] = song.pop('artist_id')
if 'artist_foreign_ids' in song:
song.pop('artist_foreign_ids')
## rename keys as necessary
song['echonest_id'] = song.pop('id')
song['album'] = album
song['spotify_id'] = track_id
song['artist'] = song.pop('artist_name')
else:
print "Song not found via EchoNest search: {}".format(spotify_uri)
return None
else:
pdb.set_trace()
"Unrecognized error code."
return song
def main():
# ## load tracks in playlist
fives = loadFile("input", "fives.txt")
fours = loadFile("input", "fours.txt")
threes = loadFile("input", "threes.txt")
twos = loadFile("input", "twos.txt")
ones = loadFile("input", "ones.txt")
## load database of song ratings
db = loadFile("../Databases", "song_ratings_db.csv")
album_ratings = loadFile("../Databases", "album_ratings_db.csv")
config = loadFile("../config", "config.csv", True)
token = sptfy.authSpotipy()
rating = 0
for ls in [ones, twos, threes, fours, fives]:
## playlists are looped in this order such that if a song is in multiple lists it's rating will end up being the highest one
rating = rating + 1
for song in ls:
track_id, spotify_uri = sptfy.getSpotifyTrackIDs(song)
if not lookupSongBySpotifyID(track_id, db):
track = sptfy.pullSpotifyTrack(track_id, token = token)
db = db.append([{'spotify_id' : track_id, 'artist' : track['artist'], 'album' : track['album'], 'spotify_album_id' : track['spotify_album_id'], 'song' : track['title'], 'rating' : rating}])
db = db.sort(['artist', 'album', 'rating'])
saveDataFrame(db, "../Databases", "song_ratings_db.csv")
for album in pd.unique(zip(db.artist, db.album)):
artist = album[0]
album = album[1]
if not lookupAlbumBySpotifyID(album, album_ratings):
album_id = pd.unique(db[db.artist == artist][db.album == album]['spotify_album_id'])[0]
album_data = sptfy.pullSpotifyAlbum(album_id, token = token)
ratings = db[db.artist == artist][db.album == album]['rating'].tolist()
if len(ratings) < 3:
## don't make album ratings for singles
continue
score = 0
countNot3 = len(np.where(r != 3))
countMoreThan2 = len(np.where(r > 2))
countMoreThan3 = len(np.where(r > 3))
for r in ratings:
if r == 5:
score = score + 100 * 1.0
scoreNot3 = score + 100 * 1.0
elif r == 4:
score = score + 80 * 1.2
scoreNot3 = score + 80 * 1.2
elif r == 3:
score = score + 60 * 1.0
elif r == 2:
score = score + 40 * 1.2
scoreNot3 = score + 40 * 1.2
elif r == 1:
score = score + 20 * 1.0
scoreNot3 = score + 20 * 1.0
std = np.std(ratings)
if std == 0.0:
std = 0.25
if countNot3 == 0:
adjMean = 3
else:
adjMean = scoreNot3 / countNot3
prop4or5 = countMoreThan3 / len(r)
adj1 = (adjMean - 3) * prop4or5
adj2 = adj1 + countMoreThan2 * 0.03
score = np.mean(ratings) + adj2
if prop4or5 == 0:
adjSD = std * 0.05
else:
adjSD = std * prop4or5 / len(r)
score = score - adjSD
## min possible score: (mean of 1-star)
min1 = 1.0
## max possible score: Radiohead "OK Computer"
max1 = 5.662521
min2 = -1.0
max2 = -0.125
scaledScore = (score - min1) / (max1 - min1)
## transform (curves the linear scores to inflate higher scores and reduce lower)
transformedScore = -1 * (8 ^ (-1 * scaledScore))
scaledScore = (transformedScore - min2) / (max2 - min2)
album_score = (round (scaledScore * 1000)) / 1.0
if album_score > 1000:
album_score = 1000
if album_score >= 965:
album_rating = 5.0
elif album_score >= 890:
album_rating = 4.5
elif album_score >= 750:
album_rating = 4.0
elif album_score >= 690:
album_rating = 3.5
elif album_score >= 625:
album_rating = 3.0
elif album_score >= 420:
album_rating = 2.5
elif album_score > 325:
album_rating = 2.0
elif album_score > 235:
album_rating = 1.5
elif album_score >= 100:
album_rating = 1.0
elif album_score < 100:
album_rating = 0.5
album_ratings = album_ratings.append([{'spotify_album_id' : album_id, 'artist' : artist, 'album' : album, 'year' : album_data['year'], 'album_rating' : album_rating, 'album_score' : album_score}])
saveDataFrame(album_ratings, "../Databases", "album_ratings_db.csv")