python – sqlalchemy many-to-many, but inverse? – Education Career Blog

I’m sorry if inverse is not the preferred nomenclature, which may have hindered my searching. In any case, I’m dealing with two sqlalchemy declarative classes, which is a many-to-many relationship. The first is Account, and the second is Collection. Users “purchase” collections, but I want to show the first 10 collections the user hasn’t purchased yet.

from sqlalchemy import *
from sqlalchemy.orm import scoped_session, sessionmaker, relation
from sqlalchemy.ext.declarative import declarative_base

Base = declarative_base()

engine = create_engine('sqlite:///:memory:', echo=True)
Session = sessionmaker(bind=engine)

account_to_collection_map = Table('account_to_collection_map', Base.metadata,
                                Column('account_id', Integer, ForeignKey('')),
                                Column('collection_id', Integer, ForeignKey('')))

class Account(Base):
    __tablename__ = 'account'

    id = Column(Integer, primary_key=True)
    email = Column(String)

    collections = relation("Collection", secondary=account_to_collection_map)

    # use only for querying?
    dyn_coll = relation("Collection", secondary=account_to_collection_map, lazy='dynamic')

    def __init__(self, email): = email

    def __repr__(self):
        return "<Acc(id=%s email=%s)>" % (,

class Collection(Base):
    __tablename__ = 'collection'

    id = Column(Integer, primary_key=True)
    slug = Column(String)

    def __init__(self, slug):
        self.slug = slug

    def __repr__(self):
        return "<Coll(id=%s slug=%s)>" % (, self.slug)

So, with account.collections, I can get all collections, and with dyn_coll.limit(1).all() I can apply queries to the list of collections…but how do I do the inverse? I’d like to get the first 10 collections that the account does not have mapped.

Any help is incredibly appreciated. Thanks!


I would not use the relationship for the purpose, as technically it it not a relationship you are building (so all the tricks of keeping it synchronized on both sides etc would not work).
IMO, the cleanest way would be to define a simple query which will return you the objects you are looking for:

class Account(Base):
    # please note added *backref*, which is needed to build the 
    #query in Account.get_other_collections(...)
    collections = relation("Collection", secondary=account_to_collection_map, backref="accounts")

    def get_other_collections(self, maxrows=None):
        """ Returns the collections this Account does not have yet.  """
        q = Session.object_session(self).query(Collection)
        q = q.filter(~Collection.accounts.any(
        # note: you might also want to order the results
        return q:maxrows if maxrows else q.all()

Leave a Comment