remove same-named relationship warning

The warning below:

                util.warn("Warning: relationship '%s' on mapper "
                          "'%s' supersedes the same relationship "
                          "on inherited mapper '%s'; this can "
                          "cause dependency issues during flush"
                          % (self.key, self.parent, inheriting))

dates back to 4f5aa12547264fb6a2e2f3af58d02ff09949d302 , in December of 2007 in version 0.4. the commit has no test nor refers to any bug report and appears to be theoretical; "added a warning when a relation() is added to an inheriting mapper that is present on a super-mapper; multiple DependencyProcessors are not expected during the flush process". This was years before the unit of work rewrite.

in modern day, a test illustrating this use case:

from sqlalchemy import *
from sqlalchemy.orm import *
from sqlalchemy.ext.declarative import declarative_base

Base = declarative_base()


class A(Base):
    __tablename__ = 'a'
    id = Column(Integer, primary_key=True)
    bs = relationship("B")


class ASub(A):
    __tablename__ = 'a_sub'
    id = Column(Integer, ForeignKey('a.id'), primary_key=True)
    bs = relationship("B")


class B(Base):
    __tablename__ = 'b'
    id = Column(Integer, primary_key=True)
    a_id = Column(ForeignKey('a.id'))


e = create_engine("sqlite://", echo=True)
Base.metadata.create_all(e)

s = Session(e)


s.add_all([
    A(bs=[B(), B(), B()]),
    ASub(bs=[B(), B(), B()])
])
s.commit()

for a in s.query(A):
    assert len(a.bs) == 3

putting a pdb.set_trace() into OneToManyDP we can see:

(Pdb) self
OneToManyDP(ASub.bs)
(Pdb) cont
> /home/classic/dev/sqlalchemy/lib/sqlalchemy/orm/dependency.py(450)presort_saves()
-> for state in states:
(Pdb) self
OneToManyDP(A.bs)
(Pdb)

which illustrates that we do in fact have multiple DependencyProcessor objects for the same-named relationship.

It's not clear if there are in fact other effects from having same-named relationships on inheritance, however propose we drop the warning. If this encourages this pattern, then we'd find out if there are in fact cases where this is a problem and we can test for them.

Comments (2)