Yep, works excellent (the "Why??" I asking about why is it wrong to
invoke "prop.do_init()" instead of StrategizedProperty.do_init(prop))

On Mar 5, 11:31 am, Michael Bayer <[email protected]> wrote:
> On Mar 5, 2011, at 11:18 AM, Kent wrote:
>
> > Thank you!
>
> > I don't disagree: I've been brainstorming how to work it out upfront,
> > but I think I'd need your topological sort to put the mappers in the
> > correct dependency order and since it is legacy support, I'm ok with
> > the non public API and potential consequences.
>
> > I had tried prop.do_init() in place of
> > StrategizedProperty.do_init(prop), but it failed.
>
> > Why???  (would invoke RelationshipProperty.do_init(), but I would have
> > guess that was the correct method instead of StrategizedProperty's)
>
> here it is:
>
> from sqlalchemy import *
> from sqlalchemy.orm import *
> from sqlalchemy.ext.declarative import declarative_base
> Base = declarative_base()
>
> class Parent(Base):
>     __tablename__ = 'parent'
>     id = Column(Integer, primary_key=True)
>
>     children = relationship("Child", lazy='joined', backref=backref('parent', 
> lazy='joined'))
>
> class Child(Base):
>     __tablename__ = 'child'
>     id = Column(Integer, primary_key=True)
>     parent_id = Column(Integer, ForeignKey('parent.id'))
>
> compile_mappers()
>
> from sqlalchemy.orm import strategies, interfaces
>
> for prop in (Parent.children.property, Child.parent.property):
>     prop.strategy_class = strategies.factory('subquery')
>     interfaces.StrategizedProperty.do_init(prop)
>
> e = create_engine('sqlite://', echo=True)
> Base.metadata.create_all(e)
> s = Session(e)
>
> s.add(Parent(children=[Child(), Child()]))
> s.commit()
>
> print "----------------------"
> s.query(Parent).all()
> s.close()
>
> print "----------------------"
> s.query(Child).all()
>
>
>
> > On Mar 5, 11:04 am, Michael Bayer <[email protected]> wrote:
> >> On Mar 5, 2011, at 10:37 AM, Kent wrote:
>
> >>> Oracle 8 strikes again.  But our client's current legacy application
> >>> requires it (until we can get them off the app).
>
> >>> Anyway, when Oracle 8 is detected, I wish to convert certain mapper
> >>> properties' lazy attribute from False => 'subquery' because oracle 8
> >>> isn't smart enough to run the query anywhere near efficiently (but 9i
> >>> is).
>
> >>> So, after all the mappers are compiled (I need backrefs also), I'm
> >>> looping through the _mapper_registry and detecting which properties
> >>> need to be converted if the oracle is 8i.
>
> >>> Unfortunately for me:
> >>>      prop.lazy = 'subquery'
> >>>      prop.strategy_class = strategies.factory('subquery')
>
> >>> isn't enough because the prop.strategy was already initialized I
> >>> surmise.
>
> >>> Because of potential circular references and complications with
> >>> backrefs and future mapped classes not being mapped to a table until
> >>> after mapper() is invoked for the class, I do not think I can figure
> >>> out whether lazy should be False vs. 'subquery' during the mapper()
> >>> invocation...(at least trivially), so I am waiting until after all
> >>> mappers are compiled.
>
> >>> Can you think of any solutions for me?  Any way to change a properties
> >>> lazy attribute after it's been instantiated?  Any way to clone the
> >>> property and replace it with a new one with a difference lazy
> >>> attribute?
>
> >> If it were me I'd still try to solve the problem of deciding which 
> >> relationships/backrefs need the setting up front.   You can limit it to 
> >> those who are setting up lazy="joined" I assume, and i'd consider getting 
> >> ugly too with some hardcoding, since this is for legacy support anyway.
>
> >> Otherwise the internal API magic you need would be:
>
> >>         from sqlalchemy.orm.interfaces import StrategizedProperty
>
> >>         prop.strategy_class = strategies.factory('subquery')
> >>         StrategizedProperty.do_init(prop)
>
> >> which would reset the "self.strategy" attribute and the collection of 
> >> alternate strategies.
>
> > --
> > You received this message because you are subscribed to the Google Groups 
> > "sqlalchemy" group.
> > To post to this group, send email to [email protected].
> > To unsubscribe from this group, send email to 
> > [email protected].
> > For more options, visit this group 
> > athttp://groups.google.com/group/sqlalchemy?hl=en.
>
>

-- 
You received this message because you are subscribed to the Google Groups 
"sqlalchemy" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to 
[email protected].
For more options, visit this group at 
http://groups.google.com/group/sqlalchemy?hl=en.

Reply via email to