Yes it recommends using ``no_autoflush`` and:
with DBSession.no_autoflush:
dep1.users.append(user2)
fixes it.
So is this the correct way appending to the collection, and should I wrap
all code that is manipulating ``Department.users`` with a ``no_autoflush``?
Erkan Ozgur Yilmaz
On Thu, Nov 27, 2014 at 2:13 AM, Michael Bayer <mike...@zzzcomputing.com>
wrote:
> take a look at the stack trace (or send it along). Do you see autoflush
> in it? If so, you should use “with session.no_autoflush” around the area
> where you are likely triggering a lazy load of an attribute or collection.
>
>
> http://docs.sqlalchemy.org/en/rel_0_9/orm/session.html#sqlalchemy.orm.session.Session.no_autoflush
>
>
> On Nov 26, 2014, at 6:00 PM, Erkan Özgür Yılmaz <eoyil...@gmail.com>
> wrote:
>
> As I understand it is not related with the __init__() methods or how I
> initialize the attributes
>
> Erkan Ozgur Yilmaz
>
> On Thu, Nov 27, 2014 at 12:58 AM, Erkan Özgür Yılmaz <eoyil...@gmail.com>
> wrote:
>
>> Ok my test case was not reproducing the error as in my previous post, but
>> consider this one it is generating the IntegrityError
>>
>> if __name__ == "__main__":
>> database_engine_settings = {
>> "sqlalchemy.url": "sqlite:///:memory:",
>> "sqlalchemy.echo": False,
>> }
>>
>> engine = engine_from_config(database_engine_settings, 'sqlalchemy.')
>> DBSession.configure(
>> bind=engine,
>> extension=None
>> )
>> Base.metadata.create_all(engine)
>>
>> user1 = User()
>> dep1 = Department(users=[user1])
>> DBSession.add(user1)
>> DBSession.add(dep1)
>> DBSession.commit()
>>
>> assert user1 in dep1.users
>> assert dep1 in user1.departments
>>
>> dep_users = DBSession.query(DepartmentUser).all()
>> assert len(dep_users) == 1
>>
>> # user2 = User(departments=[dep1])
>> user2 = User()
>> DBSession.add(user2)
>> DBSession.commit()
>>
>> dep1.users.append(user2)
>>
>>
>>
>> Erkan Ozgur Yilmaz
>>
>> On Thu, Nov 27, 2014 at 12:39 AM, Erkan Özgür Yılmaz <eoyil...@gmail.com>
>> wrote:
>>
>>> Hi,
>>>
>>> I've encountered a situation where I don't know if this is a bug or it
>>> is the correct behavior. It is about association_proxy. Please see the
>>> User.__init__ and Department.__init__ methods.
>>>
>>> from sqlalchemy import *
>>> from sqlalchemy.orm import *
>>> from sqlalchemy.ext.declarative import declarative_base
>>> from sqlalchemy.ext.associationproxy import association_proxy
>>>
>>>
>>> Base = declarative_base()
>>>
>>> DBSession = scoped_session(
>>> sessionmaker(
>>> extension=None
>>> )
>>> )
>>>
>>>
>>> class Entity(Base):
>>> __tablename__ = 'Entities'
>>>
>>> id = Column(Integer, primary_key=True)
>>> entity_type = Column(String(128), nullable=False)
>>>
>>> __mapper_args__ = {
>>> "polymorphic_on": entity_type,
>>> "polymorphic_identity": "SimpleEntity"
>>> }
>>>
>>>
>>> class User(Entity):
>>> __tablename__ = 'Users'
>>> __mapper_args__ = {"polymorphic_identity": "User"}
>>> user_id = Column('id', Integer, ForeignKey('Entities.id'),
>>> primary_key=True)
>>>
>>> departments = association_proxy(
>>> 'department_role',
>>> 'department',
>>> creator=lambda d: DepartmentUsers(department=d)
>>> )
>>>
>>> department_role = relationship(
>>> 'DepartmentUser',
>>> back_populates='user',
>>> cascade='all, delete-orphan',
>>> primaryjoin='Users.c.id
>>> <http://users.c.id/>==Department_Users.c.uid'
>>> )
>>>
>>> def __init__(self, departments=None):
>>> super(User, self).__init__()
>>> if departments is None:
>>> departments = []
>>>
>>> # using this creates an auto flush and raises an IntegrityError
>>> # self.departments = departments
>>>
>>> # where as using this is OK
>>> for department in departments:
>>> self.department_role.append(
>>> DepartmentUser(user=self, department=department)
>>> )
>>>
>>>
>>> class Department(Entity):
>>> __tablename__ = 'Departments'
>>> __mapper_args__ = {"polymorphic_identity": "Department"}
>>> department_id = Column('id', Integer, ForeignKey('Entities.id'),
>>> primary_key=True)
>>>
>>> users = association_proxy(
>>> 'user_role',
>>> 'user',
>>> creator=lambda u: DepartmentUser(user=u)
>>> )
>>>
>>> user_role = relationship(
>>> "DepartmentUser",
>>> back_populates="department",
>>> cascade='all, delete-orphan',
>>> primaryjoin='Departments.c.id
>>> <http://departments.c.id/>==Department_Users.c.did'
>>> )
>>>
>>> @validates('user_role')
>>> def _validate_user_role(self, key, user_role):
>>> """validates the given user_role variable
>>> """
>>> return user_role
>>>
>>> def __init__(self, users=None):
>>> if users is None:
>>> users = []
>>> # again using this creates an auto flush and raises an
>>> IntegrityError
>>> # self.users = users
>>>
>>> # where as using this is OK
>>> for user in users:
>>> self.user_role.append(
>>> DepartmentUser(department=self, user=user)
>>> )
>>>
>>>
>>> class Role(Entity):
>>> __tablename__ = 'Roles'
>>> __mapper_args__ = {"polymorphic_identity": "Role"}
>>> role_id = Column('id', Integer, ForeignKey('Entities.id'),
>>> primary_key=True)
>>>
>>>
>>> class DepartmentUser(Base):
>>> __tablename__ = 'Department_Users'
>>>
>>> user_id = Column(
>>> 'uid',
>>> Integer,
>>> ForeignKey('Users.id'),
>>> primary_key=True
>>> )
>>>
>>> user = relationship(
>>> 'User',
>>> back_populates='department_role',
>>> primaryjoin='DepartmentUser.user_id==User.user_id'
>>> )
>>>
>>> department_id = Column(
>>> 'did',
>>> Integer,
>>> ForeignKey('Departments.id'),
>>> primary_key=True
>>> )
>>>
>>> department = relationship(
>>> 'Department',
>>> back_populates='user_role',
>>> primaryjoin='DepartmentUser.department_id==Department.department_id'
>>> )
>>>
>>> role_id = Column(
>>> 'rid',
>>> Integer,
>>> ForeignKey('Roles.id'),
>>> nullable=True
>>> )
>>>
>>> role = relationship(
>>> 'Role',
>>> primaryjoin='DepartmentUser.role_id==Role.role_id'
>>> )
>>>
>>> def __init__(self, department=None, user=None, role=None):
>>> self.department = department
>>> self.user = user
>>> self.role = role
>>>
>>>
>>> if __name__ == "__main__":
>>> database_engine_settings = {
>>> "sqlalchemy.url": "sqlite:///:memory:",
>>> "sqlalchemy.echo": False,
>>> }
>>>
>>> engine = engine_from_config(database_engine_settings, 'sqlalchemy.')
>>> DBSession.configure(
>>> bind=engine,
>>> extension=None
>>> )
>>> Base.metadata.create_all(engine)
>>>
>>> user1 = User()
>>> dep1 = Department()
>>> DBSession.add(user1)
>>> DBSession.add(dep1)
>>> DBSession.commit()
>>>
>>> dep1.users.append(user1)
>>>
>>> assert user1 in dep1.users
>>> assert dep1 in user1.departments
>>>
>>> dep_users = DBSession.query(DepartmentUser).all()
>>> assert len(dep_users) == 1
>>>
>>> user2 = User()
>>> dep1.users.append(user2)
>>> DBSession.add(user2)
>>> DBSession.commit()
>>>
>>> So, I get an IntegrityError if I set the **self.departments** or
>>> **self.users** directly, so apparently creating the association objects
>>> triggers an auto-flush but if I create the association objects instead of
>>> directly setting the attributes there is no error. I'm ready to use the
>>> work around of creating the association object but I'm also curious if this
>>> is a bug or normal way of using association_proxy.
>>>
>>> Thanks,
>>>
>>> Erkan Ozgur Yilmaz
>>>
>>
>>
>
> --
> You received this message because you are subscribed to the Google Groups
> "sqlalchemy" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to sqlalchemy+unsubscr...@googlegroups.com.
> To post to this group, send email to sqlalchemy@googlegroups.com.
> Visit this group at http://groups.google.com/group/sqlalchemy.
> For more options, visit https://groups.google.com/d/optout.
>
>
> --
> You received this message because you are subscribed to the Google Groups
> "sqlalchemy" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to sqlalchemy+unsubscr...@googlegroups.com.
> To post to this group, send email to sqlalchemy@googlegroups.com.
> Visit this group at http://groups.google.com/group/sqlalchemy.
> For more options, visit https://groups.google.com/d/optout.
>
--
You received this message because you are subscribed to the Google Groups
"sqlalchemy" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to sqlalchemy+unsubscr...@googlegroups.com.
To post to this group, send email to sqlalchemy@googlegroups.com.
Visit this group at http://groups.google.com/group/sqlalchemy.
For more options, visit https://groups.google.com/d/optout.
