Hello, given the following base:

class Bar(Base):
    __tablename__ = 'bars'
    id = Column(Integer, primary_key=True)

class Foo(Base):
    __tablename__ = 'foos'

    id = Column(Integer, primary_key=True)

    bar_id = Column(Integer, ForeignKey(Bar.id))
    bar = relationship(Foo, backref='bar')

I wonder why sqlalchemy allows this kind of ORM queries to be executed:

session.query(Foo).filter(Bar.id == 42)

that results in the execution of this SQL:

SELECT foos.id AS foos_id, foos.bar_id AS foos_bar_id
FROM foos, bars
WHERE bars.id = ?

This is way more often a mistake than what the developer originally intended. I am asking this question because it is quite easy to add a filter condition to a model that was not originally joined, effectively voiding the filter condition:

session.query(Foo).join(Foo.bar).filter(Bar.id==42)  # Good query
session.query(Foo).filter(Bar.id==42)  # Oops, I forgot to join Foo.bar.

In the above example, the second query will fetch all the Foos from the database. This is both a very slow operation and it also return unwanted records. Maybe it should raise instead?

EDIT: Fixed the last example