Is there any more elegant way to add a value sensitive unique together constraint in Django Model?

Question

Here is the problem:

I have a model like this:

class UserBook(models.Model):
    user = models.ForeignKey(User)
    book = models.ForeignKey(Book)
    is_active = models.BooleanField(default=False)

    class Meta:
        unique_together = ("user", "book")

Obviously, this model already has a unique together constraint for field user and book. And probably there will be some entries like this in the database:

    ------------------------------
    |user_id  book_id  is_active |
    |      1        1          0 |
    |      1        2          0 |
    |      1        3          1 |
    ------------------------------

And I have one more constraint to add, which is each user can have at most one entry that the value of is_active field is 1(True).

Currently I solve this problem by changing the model into this:

class UserBook(models.Model):
    user = models.ForeignKey(User)
    book = models.ForeignKey(Book)
    is_active = models.BooleanField(default=False)
    key = models.charFeild(max_length=255, unique=True)

    class Meta:
        unique_together = ("user", "book")

    def save(self, *args, **kwargs):
        if self.is_active:
            self.key = "%s_%s" %(self.user_id, self.is_active)
        else:
            self.key = "%s_%s_%s" %(self.user_id, self.is_active, self.book_id)

Add a field key, and customize the save method of this model.

But the max_length cannot be greater than 255 in this approach(which is no need to worry in my case, but sometimes the key field may be very long).

So, I would like to know if there is any more elegant approach to solve this kind of problem.

Thanks!

Answer 1

Redefine the is_active to be as follows:

# Equals user ID if active; otherwise null.
is_active = models.IntegerField(null = True, unique = True)

The user IDs will be unique in the column (satisfying your desired constraint) and the many null values in the column won't violate the constraint, as discussed here.