Overview

Do you guys use Spotify? Spotify song data contains information about what the song has. In this article, I will try to create a playlist of only the songs that are judged to be "Major" from the songs registered as favorites. To use the API, use the python library spotipy.

Access authentication

First, perform access authentication. You can get the client ID etc. from here.

import spotipy
import spotipy.util as util

username = 'User name'
client_id = 'Client ID'
client_secret = 'Client secret'
redirect_uri = 'http://example.com'

#Used to grant app permissions
#https://developer.spotify.com/documentation/general/guides/scopes/
scope = 'user-library-read user-read-playback-state playlist-read-private user-read-recently-played playlist-read-collaborative playlist-modify-public playlist-modify-private'
token = util.prompt_for_user_token(username, scope, client_id, client_secret, redirect_uri)
sp = spotipy.Spotify(auth = token)

Get song ID and song title

You can get the songs you have registered as favorites with current_user_saved_tracks.


import pandas as pd

def track_info(limit = 20):
    saved_tracks = sp.current_user_saved_tracks(limit = limit)

    track_id = lambda i: saved_tracks['items'][i]['track']['id']
    track_name = lambda i: sp.track(track_id(i))['name']

    info = [[track_id(i), track_name(i)] for i in range(limit)]
    return pd.DataFrame(info, columns = ['track_id','track_name'])

print(track_info())

                 track_id                        track_name
0   3xkGcE52Hpr8zd34ws2D0A                     hot chocolate
1 5Hqh7J6Oy0Uz85DSx5JUsw Strobo Nights
2   6yK4KPKZa0xX0h58bY96VY                             Miree
3   753nzL4g1fBQt3RuHVHVDQ                 You're Everything
4   0albLGvYx6ftOUJ7KF0i5I                         STAY TUNE
5   0oCP1Vu9LGVP88atKzOswC                         ambiguous
6 0cSkn2l67csUljEy0EEBPn Flame
7   3BSPKTravgBDjmuG1rIXvw                     BLOODY STREAM
8 2iOvMtA1u9s95NF6mJDOuT Perfume
9   0E5pNzt2pwd3bOqQsOnfQv  Tick Tock Skip Drop - Mili remix
...

It looks like it's taken well! (Otaku song selection feeling

Get the characteristics of the song

Features can be obtained from audio_features. Bright songs are output as 1 and dark songs as 0 in the column mode. -Get Audio Features for a Track


def track_features(track_ids, limit = 20):
    return pd.DataFrame(sp.audio_features(track_ids))

print(track_features(track_info()['track_id']).drop(['track_href', 'analysis_url', 'id', 'uri'], axis=1))

    danceability  energy  key  loudness  mode  speechiness  acousticness  instrumentalness  liveness  valence    tempo            type  duration_ms  time_signature
0          0.563   0.958    1    -4.514     1       0.0476      0.173000          0.003240    0.2250    0.620  160.421  audio_features       193508               4
1          0.550   0.879    0    -4.496     1       0.0490      0.000826          0.000059    0.1540    0.143  124.989  audio_features       345171               4
2          0.717   0.688   10    -5.886     0       0.0270      0.022500          0.000000    0.1070    0.715  124.015  audio_features       242600               4
3          0.614   0.443    8   -12.060     0       0.0335      0.354000          0.004890    0.1070    0.291   98.974  audio_features       227762               4
4          0.681   0.938    5    -4.246     1       0.0557      0.028800          0.000264    0.0634    0.600  122.991  audio_features       297333               4
5          0.441   0.962    0    -2.299     1       0.1080      0.142000          0.000103    0.3170    0.592  178.201  audio_features       272480               4
6          0.477   0.685    2    -4.554     1       0.0325      0.105000          0.000000    0.2770    0.308  152.040  audio_features       275000               4
7          0.274   0.903    8    -3.125     1       0.0640      0.008600          0.000000    0.3150    0.662  129.999  audio_features       261920               4
8          0.848   0.352    3    -6.838     1       0.0410      0.824000          0.000000    0.0906    0.776  121.833  audio_features       252594               4
9          0.453   0.466    7    -9.435     1       0.0310      0.538000          0.000010    0.0663    0.312   79.987  audio_features       308962               4
...

Get only bright songs

Combine the music name and ID table and the feature table to narrow down the conditions.


def get_track_from_mode(mode = 1):
    info_with_features = pd.concat([track_info(), track_features(track_info()['track_id'])], axis = 1)
    return info_with_features[info_with_features['mode'] == mode]

print(get_track_from_mode()[['track_id', 'track_name', 'mode']])

                  track_id                        track_name  mode
0   3xkGcE52Hpr8zd34ws2D0A                     hot chocolate     1
1 5Hqh7J6Oy0Uz85DSx5JUsw Strobo Nights 1
4   0albLGvYx6ftOUJ7KF0i5I                         STAY TUNE     1
5   0oCP1Vu9LGVP88atKzOswC                         ambiguous     1
6 0cSkn2l67csUljEy0EEBPn Flame 1
7   3BSPKTravgBDjmuG1rIXvw                     BLOODY STREAM     1
8 2iOvMtA1u9s95NF6mJDOuT Perfume 1
9   0E5pNzt2pwd3bOqQsOnfQv  Tick Tock Skip Drop - Mili remix     1
10 3xf3Ji3kDHH62sYK662mwS Alianza de Sangre 1
12  5QWtG4MLn5yRVRv0OVPjCl                      Main Heroine     1
...

Only songs with mode of 1 could be acquired.

Create playlist

Finally, create a playlist.

from nested_lookup import nested_lookup

def make_playlist(playlist_name):
    user_id = sp.current_user()['id']

    if playlist_name not in nested_lookup('name', sp.current_user_playlists()):
        sp.user_playlist_create(user_id, playlist_name)
        sp.user_playlist_add_tracks(user = user_id, playlist_id = sp.current_user_playlists()['items'][0]['id'], tracks = get_track_from_mode()['track_id'])

make_playlist('test')

名称未設定.png It's done!