Graphs and tables for your Spotify account.
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
 
 
 

82 lines
2.4 KiB

from django.db import models
# id's are 22 in length in examples but set to 30 for buffer
MAX_ID = 30
# Artist {{{ #
class Artist(models.Model):
class Meta:
verbose_name = "Artist"
verbose_name_plural = "Artists"
artist_id = models.CharField(primary_key=True, max_length=MAX_ID)
# unique since only storing one genre per artist right now
name = models.CharField(unique=True, max_length=50)
genre = models.CharField(max_length=20)
def __str__(self):
return self.name
# }}} Artist #
# User {{{ #
class User(models.Model):
class Meta:
verbose_name = "User"
verbose_name_plural = "Users"
user_id = models.CharField(primary_key=True, max_length=MAX_ID) # the user's Spotify ID
# username = models.CharField(max_length=30) # User's Spotify user name, if set
def __str__(self):
return self.user_id
# }}} User #
# Track {{{ #
class Track(models.Model):
class Meta:
verbose_name = "Track"
verbose_name_plural = "Tracks"
track_id = models.CharField(primary_key=True, max_length=MAX_ID)
# artist = models.ForeignKey(Artist, on_delete=models.CASCADE)
artists = models.ManyToManyField(Artist, blank=True)
year = models.PositiveSmallIntegerField()
popularity = models.PositiveSmallIntegerField()
runtime = models.PositiveSmallIntegerField()
name = models.CharField(max_length=75)
users = models.ManyToManyField(User, blank=True)
def __str__(self):
return self.name
# }}} Track #
# AudioFeatures {{{ #
class AudioFeatures(models.Model):
class Meta:
verbose_name = "AudioFeatures"
verbose_name_plural = "AudioFeatures"
track = models.OneToOneField(Track, on_delete=models.CASCADE, primary_key=True,)
acousticness = models.DecimalField(decimal_places=3, max_digits=3)
danceability = models.DecimalField(decimal_places=3, max_digits=3)
energy = models.DecimalField(decimal_places=3, max_digits=3)
instrumentalness = models.DecimalField(decimal_places=3, max_digits=3)
loudness = models.DecimalField(decimal_places=3, max_digits=6)
speechiness = models.DecimalField(decimal_places=3, max_digits=3)
tempo = models.DecimalField(decimal_places=3, max_digits=6)
valence = models.DecimalField(decimal_places=3, max_digits=3)
def __str__(self):
return super(AudioFeatures, self).__str__()
# }}} AudioFeatures #