SQLAlchemy ORM: Polymorphic Single Table Inheritance, with fallback to parent class if "polymorphic_identity" is not found

Question

Using Python 3.5 and SQLAlchemy 1.0.14 (ORM).

I have a table of items declared as such:

from sqlalchemy.ext.declarative.api import declarative_base

Base = declarative_base()

class Item(Base):
    __tablename__ = 'items'

    id = Column(Integer, primary_key=True)
    type = Column(String)
    # other non relevant attributes

My Items can be of many different types, the type identifier being stored in type. For a few of those objects types, I need to have specific methods or attributes available.

To achieve that I tried to use single table inheritance with several SpecialisedItem as subclass of Item:

class Item(Base):
    __tablename__ = 'items'

    id = Column(Integer, primary_key=True)
    type = Column(String, index=True)
    # other non relevant attributes

    __mapper_args__ = {
        'polymorphic_on': type,
    }

class SpecialisedItem(Base):
    __mapper_args__ = {
        'polymorphic_identity': 'specialitem',
    }

    def specialised_method(self):
        return "I am special"

Now when I load my items, I'd want all specialised items (having type=='specialitem') to be loaded as such, while any other type value would result in the parent class Item being loaded. That doesn't work, I get AssertionError: No such polymorphic_identity 'normal' is defined when loading the items.

I would like to avoid creating inherited classes that do nothing just to cover all possible type values, instead having "unmapped" type falling back to the parent class Item.

Is there any way to achieve that effect ?

Minimal test case for reference:

from sqlalchemy.engine import create_engine
from sqlalchemy.ext.declarative.api import declarative_base
from sqlalchemy.orm.session import sessionmaker
from sqlalchemy.sql.schema import Column
from sqlalchemy.sql.sqltypes import Integer, String


Base = declarative_base()

class Item(Base):
    __tablename__ = 'items'

    id = Column(Integer, primary_key=True)
    type = Column(String, index=True)
    # other non relevant attributes

    __mapper_args__ = {
        'polymorphic_on': type,
    }

class SpecialisedItem(Item):
    __mapper_args__ = {
        'polymorphic_identity': 'special',
    }

    specialAttribute = Column(String)

    def specialised_method(self):
        return "I am special"


engine = create_engine("sqlite:///:memory:")
Base.metadata.create_all(engine)
Session = sessionmaker(bind=engine)
session = Session()

session.add(Item(type='normal'))
session.add(Item(type='special'))
session.commit()
# loading only specialized items works
for item in session.query(Item).filter_by(type="special"):
    print(item.specialised_method())

# loading other items fails
for item in session.query(Item):
    print(item.type)

Thanks,

Guillaume

Answer 1

A reusable decorator solution, based on @r-m-n answer. Custom class is also replaced with collections.defaultdict that actually does same thing.

def receive_mapper_configured(mapper, class_):
    mapper.polymorphic_map = defaultdict(lambda: mapper, mapper.polymorphic_map)
    # to prevent 'incompatible polymorphic identity' warning, not necessary
    mapper._validate_polymorphic_identity = None


def polymorphic_fallback(mapper_klass):
    event.listens_for(mapper_klass, 'mapper_configured')(receive_mapper_configured)
    return mapper_klass

Then in your code you can just add this decorator to base classes:

@polymorphic_fallback
class Item:
    ...

class SpecificItem(Item):
    ...