How do I join three tables with SQLalchemy and keeping all of the columns in one of the tables?

Question

So, I have three tables:

The class defenitions:

engine = create_engine(\'sqlite://test.db\', echo=False)
SQLSession = sessionmaker(bind=engine)
Base = de         


        

Answer 1

Don't query from the user. Query from the Channel.

user = query(User).filter_by(id=1).one()
for channel in query(Channel).all():
    print channel.id, channel.title, user in channel.subscriptions.user

That way you get all the channels, not just the ones that are associated with the user in question.

Answer 2

Option-1:

Subscription is just a many-to-many relation object, and I would suggest that you model it as such rather then as a separate class. See Configuring Many-to-Many Relationships documentation of SQLAlchemy/declarative.

You model with the test code becomes:

from sqlalchemy import create_engine, Column, Integer, DateTime, String, ForeignKey, Table
from sqlalchemy.orm import relation, scoped_session, sessionmaker, eagerload
from sqlalchemy.ext.declarative import declarative_base

engine = create_engine('sqlite:///:memory:', echo=True)
session = scoped_session(sessionmaker(bind=engine, autoflush=True))
Base = declarative_base()

t_subscription = Table('subscription', Base.metadata,
    Column('userId', Integer, ForeignKey('user.id')),
    Column('channelId', Integer, ForeignKey('channel.id')),
)

class Channel(Base):
    __tablename__ = 'channel'

    id = Column(Integer, primary_key = True)
    title = Column(String)
    description = Column(String)
    link = Column(String)
    pubDate = Column(DateTime)

class User(Base):
    __tablename__ = 'user'

    id = Column(Integer, primary_key = True)
    username = Column(String)
    password = Column(String)
    sessionId = Column(String)

    channels = relation("Channel", secondary=t_subscription)

# NOTE: no need for this class
# class Subscription(Base):
    # ...

Base.metadata.create_all(engine)


# ######################
# Add test data
c1 = Channel()
c1.title = 'channel-1'
c2 = Channel()
c2.title = 'channel-2'
c3 = Channel()
c3.title = 'channel-3'
c4 = Channel()
c4.title = 'channel-4'
session.add(c1)
session.add(c2)
session.add(c3)
session.add(c4)
u1 = User()
u1.username ='user1'
session.add(u1)
u1.channels.append(c1)
u1.channels.append(c3)
u2 = User()
u2.username ='user2'
session.add(u2)
u2.channels.append(c2)
session.commit()


# ######################
# clean the session and test the code
session.expunge_all()

# retrieve all (I assume those are not that many)
channels = session.query(Channel).all()

# get subscription info for the user
#q = session.query(User)
# use eagerload(...) so that all 'subscription' table data is loaded with the user itself, and not as a separate query
q = session.query(User).options(eagerload(User.channels))
for u in q.all():
    for c in channels:
        print (c.id, c.title, (c in u.channels))

which produces following output:

(1, u'channel-1', True)
(2, u'channel-2', False)
(3, u'channel-3', True)
(4, u'channel-4', False)
(1, u'channel-1', False)
(2, u'channel-2', True)
(3, u'channel-3', False)
(4, u'channel-4', False)

Please note the use of eagerload, which will issue only 1 SELECT statement instead of 1 for each User when channels are asked for.

Option-2:

But if you want to keep you model and just create an SA query that would give you the columns as you ask, following query should do the job:

from sqlalchemy import and_
from sqlalchemy.sql.expression import case
#...
q = (session.query(#User.username, 
                   Channel.id, Channel.title, 
                   case([(Subscription.channelId == None, False)], else_=True)
                  ).outerjoin((Subscription, 
                                and_(Subscription.userId==User.id, 
                                     Subscription.channelId==Channel.id))
                             )
    )
# optionally filter by user
q = q.filter(User.id == uid()) # assuming uid() is the function that provides user.id
q = q.filter(User.sessionId == id()) # assuming uid() is the function that provides user.sessionId
res = q.all()
for r in res:
    print r

The output is absolutely the same as in the option-1 above.

Answer 3

To make this a little easyer I've added relationships to your model, that way you can just do user.subscriptions to get all the subscriptions.

engine = create_engine('sqlite://test.db', echo=False)
SQLSession = sessionmaker(bind=engine)
Base = declarative_base()

class Channel(Base):
    __tablename__ = 'channel'

    id = Column(Integer, primary_key = True)
    title = Column(String)
    description = Column(String)
    link = Column(String)
    pubDate = Column(DateTime)

class User(Base):
    __tablename__ = 'user'

    id = Column(Integer, primary_key = True)
    username = Column(String)
    password = Column(String)
    sessionId = Column(String)

class Subscription(Base):
    __tablename__ = 'subscription'

    userId = Column(Integer, ForeignKey('user.id'), primary_key=True)
    user = relationship(User, primaryjoin=userId == User.id, backref='subscriptions')
    channelId = Column(Integer, ForeignKey('channel.id'), primary_key=True)
    channel = relationship(channel, primaryjoin=channelId == channel.id, backref='subscriptions')

results = session.query(
    Channel.id,
    Channel.title,
    Channel.subscriptions.any().label('subscribed'),
)

for channel in results:
    print channel.id, channel.title, channel.subscribed