I'd like to understand the behavior of association_proxy when creating new
records across a many-to-many table where, in some cases, the joining table
will have additional attribute values beyond the two primary keys. In the
below example, there is a many-many between Person and Events represented
by the PersonEvents table. If I create and commit the person record, create
and commit the event record, and then add the event to the person and try
to commit, I get a key error. If I create the Person record, create the
Event record, and then create the PersonEvent record before committing, I
get no error and my tests pass.
There are two differences in the test approaches that I think I understand;
in the failing case the PersonEvent object is being created implicitly
through the association proxy definition, and the commit order is
different. However, as I read the documentation, I don't understand why the
results are different. I'd like to avoid building more complex logic until
I understand whether the association_proxy has use case limitations I don't
understand.
Here's the data model code:
class Person(Base):
__tablename__ = 'Person'
__table_args__ = {'mysql_charset':'utf8'}
id = Column(Integer, primary_key=True)
full_name = Column(String(240))
email = Column(String(120),unique=True)
other_data = Column(JSON)
events = association_proxy('PersonEvent','Event')
class PersonEvent(Base):
__tablename__ = 'PersonEvent';
__tableargs__ = {'mysql_charset':'utf8'}
person_id = Column(Integer, ForeignKey('Person.id'), primary_key=True)
event_id = Column(Integer, ForeignKey('Event.id'), primary_key = True)
role = Column(String(40))
# bi-directional attribute / collection of "Person" / "Event"
person = relationship('Person',
backref=backref("PersonEvent",
cascade="all, delete-orphan"))
# reference to the Event object
event = relationship('Event')
class Event(Base):
__tablename__ = 'Event'
__table_args__ = {'mysql_charset':'utf8'}
id = Column(Integer, primary_key=True)
start = Column(DateTime)
end = Column(DateTime)
short_name = Column(String(40))
name = Column(String(240))
other_data = Column(JSON)
The following code in the test setup method throws a KeyError on the last
commit from emit_backref_from_scalar_set_event child_impl =
child_state.manager[key].impl.
session = DBSession()
p1 = Person(id=1, first_name='Eric', last_name='Wittle',
full_name='Eric L. Wittle', email='e...@wittle.net')
p1.set('favorite_color','red')
session.add(p1)
session.commit()
logger.debug('Added person 1')
e1 = Event(id=1, name='Birth',
start=datetime.strptime('01/25/1976',"%m/%d/%Y"),
end=datetime.strptime('01/25/1976',"%m/%d/%Y"))
session.add(e1)
session.commit()
logger.debug('Added event 1')
p1.events.append(e1)
session.add(p1)
session.commit()
However, if I replace it with the code below, the data setup completes
without error, and the test passes fine (looking up person with id 1 and
testing if the property events[0].name == 'Birth'):
session = DBSession()
p1 = Person(id=1, first_name='Eric', last_name='Wittle',
full_name='Eric L. Wittle', email='e...@wittle.net')
p1.set('favorite_color','red')
e1 = Event(id=1, name='Birth',
start=datetime.strptime('01/25/1976',"%m/%d/%Y"),
end=datetime.strptime('01/25/1976',"%m/%d/%Y"))
pe1 = PersonEvent(person=p1, event=e1, role = 'Owner')
session.add(p1)
session.commit()
The first set of code that fails seems more similar to the example in the
association proxy documentation, section "Simplifying Association Objects"
than the code that passes.
Thanks in advance for any advice & guidance. I've been really impressed
with the sqlalchemy orm so far, and am trying to use more of it.
-Eric
--
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 https://groups.google.com/group/sqlalchemy.
For more options, visit https://groups.google.com/d/optout.
