|
| 1 | +from django.conf import settings |
| 2 | +from django.contrib.auth.models import User |
1 | 3 | from django.db import models |
| 4 | +from django.db.models import Avg, F |
2 | 5 |
|
3 | | -# Create your models here. |
| 6 | + |
| 7 | +MAX_VOTES = getattr(settings, 'MAX_VOTES', 5) |
| 8 | +TECH_WEIGHT = 0.2 |
| 9 | +PERSONAL_WEIGHT = 0.8 |
| 10 | + |
| 11 | +VOTES = [(i, str(i)) for i in range(1, MAX_VOTES + 1)] |
| 12 | + |
| 13 | + |
| 14 | +class Vote(models.Model): |
| 15 | + application = models.ForeignKey('application.Application', on_delete=models.CASCADE) |
| 16 | + user = models.ForeignKey(settings.AUTH_USER_MODEL, null=True, on_delete=models.SET_NULL) |
| 17 | + tech = models.IntegerField(choices=VOTES, null=True) |
| 18 | + personal = models.IntegerField(choices=VOTES, null=True) |
| 19 | + calculated_vote = models.FloatField(null=True) |
| 20 | + |
| 21 | + def save(self, force_insert=False, force_update=False, using=None, update_fields=None): |
| 22 | + """ |
| 23 | + We are overriding this in order to standarize each review vote with the |
| 24 | + new vote. |
| 25 | + Also we store a calculated vote for each vote so that we don't need to |
| 26 | + do it later. |
| 27 | +
|
| 28 | + Thanks to Django awesomeness we do all the calculations with only 3 |
| 29 | + queries to the database. 2 selects and 1 update. The performance is way |
| 30 | + faster than I thought. |
| 31 | + If improvements need to be done using a better DB than SQLite should |
| 32 | + increase performance. As long as the database can handle aggregations |
| 33 | + efficiently this will be good. |
| 34 | +
|
| 35 | + By casassg |
| 36 | + """ |
| 37 | + super().save(force_insert, force_update, using, update_fields) |
| 38 | + |
| 39 | + # only recalculate when values are different than None |
| 40 | + if not self.personal or not self.tech: |
| 41 | + return |
| 42 | + |
| 43 | + # Retrieve averages |
| 44 | + avgs = User.objects.filter(id=self.user_id).aggregate( |
| 45 | + tech=Avg('vote__tech'), |
| 46 | + pers=Avg('vote__personal')) |
| 47 | + p_avg = round(avgs['pers'], 2) |
| 48 | + t_avg = round(avgs['tech'], 2) |
| 49 | + |
| 50 | + # Calculate standard deviation for each scores |
| 51 | + sds = User.objects.filter(id=self.user_id).aggregate( |
| 52 | + tech=Avg((F('vote__tech') - t_avg) * (F('vote__tech') - t_avg)), |
| 53 | + pers=Avg((F('vote__personal') - p_avg) * |
| 54 | + (F('vote__personal') - p_avg))) |
| 55 | + |
| 56 | + # Alternatively, if standard deviation is 0.0, set it as 1.0 to avoid |
| 57 | + # division by 0.0 in the update statement |
| 58 | + p_sd = round(sds['pers'], 2) or 1.0 |
| 59 | + t_sd = round(sds['tech'], 2) or 1.0 |
| 60 | + |
| 61 | + # Apply standarization. Standarization formula: |
| 62 | + # x(new) = (x - u)/o |
| 63 | + # where u is the mean and o is the standard deviation |
| 64 | + # |
| 65 | + # See this: http://www.dataminingblog.com/standardization-vs- |
| 66 | + # normalization/ |
| 67 | + personal = PERSONAL_WEIGHT * (F('personal') - p_avg) / p_sd |
| 68 | + tech = TECH_WEIGHT * (F('tech') - t_avg) / t_sd |
| 69 | + Vote.objects.filter(user=self.user).update(calculated_vote=(personal + tech) * MAX_VOTES / 10) |
| 70 | + |
| 71 | + class Meta: |
| 72 | + unique_together = ('application', 'user') |
0 commit comments