How to deal with “false many-to-many” relationships in SQLAlchemy?

问题

I'm mapping a database using SQLAlchemy that have multiple cases of "false" many-to-many relationships. What I mean by this is, suppose I have the following objects:

class User(Base):
    __tablename__ = 'user'
    id = Column(Integer, primary_key=True)
    addresses = relationship('Address', secondary='user_address')

class Address(Base):
    __tablename__ = 'address'
    id = Column(Integer, primary_key=True)
    users = relationship('User', secondary='user_address')

class UserAddressLink(Base):
    __tablename__ = 'user_address'
    id = Column(Integer, primary_key=True)
    user_id = Column(Integer, ForeignKey('user.id'))
    address_id = Column(Integer, ForeignKey('address.id'))

So, a simple many-to-many relationship, right? But there's one catch: it was never intended to be many-to-many. This is actually a one-to-one relationship that someone decided to design like this in the database for whatever reason. There's only one Address per User and vice-versa. I have no control over database design (in fact, I'm only reading from this database and never writing on it) so I can't change this.

Is there a standard way of dealing with this on SQLAlchemy? It automatically assumes that this is a many-to-many relationships and treat the User.adresses and Address.users as lists.

The way I'm dealing with it is creating properties:

class User(Base):
    __tablename__ = 'user'
    id = Column(Integer, primary_key=True)
    _addresses = relatioship('Address', secondary='user_address')

    @property
    def address(self):
        return self.addresses[0] if len(self.addresses) > 0 else None

    @address.setter
    def address(self, value):
        self.addresses = [value]

And so on.

Is this the best way to deal with this or is there any other workaround?

回答1:

There is a very straightforward way to define such a relationship by using uselist = False as done in One-to-One relationship definition on both side of the relationship:

class User(Base):
    __tablename__ = 'user'
    id = Column(Integer, primary_key=True)
    # other columns
    name = Column(String)


class Address(Base):
    __tablename__ = 'address'
    id = Column(Integer, primary_key=True)
    # other columns
    name = Column(String)

    # relationship(
    user = relationship(
        User,
        secondary='user_address',
        uselist=False,
        backref=backref('address', uselist=False),
    )

user_address = Table(
    'user_address', Base.metadata,
    Column('id', Integer, primary_key=True),
    Column('use_id', Integer, ForeignKey('user.id')),
    Column('address_id', Integer, ForeignKey('address.id')),
)

Then you can use the code as you desire to:

# add some data
u1 = User(name='JJ', address=Address(name='superstreet'))
a2 = Address(name='LA')
a2.user = User(name='John')
session.add(u1)
session.add(a2)
session.commit()
session.expunge_all()

# get users and preload addresses as well in one query
q = session.query(User).options(joinedload(User.address))
for u in q.all():
    print(u)
    print("  {}".format(u.address))

Few more notes on your code:

you should not define the relationship on both sides, just use backref for this
you should not define the whole mapped class for the user_address table, table definition as above is

来源：https://stackoverflow.com/questions/27449462/how-to-deal-with-false-many-to-many-relationships-in-sqlalchemy

标签

database-design

sqlalchemy

many-to-many