use a subset of sort tables when we only need to sort on inheritance chains

[Former user]

Hello,

I'm getting a strange CircularDependencyError with SQLAlchemy 0.7.8, PostgreSQL 9.0.8. I'm using the joined inheritance and the autoload feature.

I've added a test case in attachment

Julien

[Former user]

db script

[Former user]

test script (please replace user/password in the DSN)

[Former user]

I digged a little bit more, and it seems that the problem is that ''use_alter=True'' must be set on the foreign keys which have circular dependencies.

I don't know if this should be considered as a bug in the autoload feature? In other words: Should the autoload feature detects foreign keys which have circular dependencies and automatically set ''use_alter=True'' on those?

So it works with the following:

    table_content = schema.Table('content', db_meta,
        schema.Column('container_id', types.Integer(),
                  schema.ForeignKey('folder.content_id', use_alter=True,
                                    name='fk_folder')),

        schema.Column('icon_content_id', types.Integer(),
                  schema.ForeignKey('data.content_id', use_alter=True,
                                    name='fk_data')),
        autoload=True, 
        extend_existing=True, 
        autoload_replace=True,
        autoload_with = db_engine)

(check test_use_alter.py)

[Former user]

this one works

[Mike Bayer]

um ok, here's a tip. I cannot run a test that relies upon reflection, I do not have your database here. If the error is just the use_alter for the FK, no that's not a bug, it's documented.

I'm going to make the test run and assuming the cirtuclardep is just your FK, this is a "worksforme" (this is why its essential to send along stack traces when issues are first reported).

[Mike Bayer]

sorry, this test names eight tables that aren't defined here so this is not a useful test case. use_alter is required for cycles in foreign keys, here's the docs which introduce that from an ORM perspective: http://docs.sqlalchemy.org/en/rel_0_7/orm/relationships.html#rows-that-point-to-themselves-mutually-dependent-rows

[Mike Bayer]

I'm just seeing your SQL script, sorry. But in the future I really need succinct test cases that I can run in one step, thanks.

[Mike Bayer]

I see where the use_alter is being required in an unusual place, this can be improved.

[Mike Bayer]

here is the test 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)
    b_id = Column(Integer, ForeignKey('b.id'))

class B(A):
    __tablename__ = "b"

    id = Column(Integer, ForeignKey('a.id'), primary_key=True)

    __mapper_args__ = {'inherit_condition':id==A.id}

e = create_engine("sqlite://", echo=True)

Base.metadata.create_all(e)

s = Session(e)

s.add(B())
s.flush()

[Former user]

ok.. I'll try to write a more useful test case next time

[Mike Bayer]

the reflection is not strictly necessary, however you'd need to know what's going on here to know that, so in this case i can't fault you for that (though you can try removing all those extra tables next time).

I didn't see the sql scripts since I went straight to the test case in the email, sorry.

[Mike Bayer]

this is 0.8 only since it's destabilizing and is in b297b40fca923a03e3c34094e5298d6524944c39 . sorry for the noise.

[Mike Bayer]

Removing milestone: 0.8.0b1 (automated comment)