query() should interpret a many-to-one relation() as an entity?

[Former user]

The following query:

list(session.query(Employee, Employee.company))

unexpectedly returns 5-element tuples

[e1, 1, u'e1', u'employee', 1),
 (Employee e2, 2, u'e2', u'employee', 1),
 (Employee e3, 3, u'e3', u'employee', 2)]((Employee)

IMHO no one expects such a result.

Full test case attached.

[Mike Bayer]

this is an enhancement, that you'd like query() to intelligently interpret a relation().

But the documented way to do this is:

query(Employee, Company)

additionally its unclear if query(Employee, Employee.company) should automatically "join" the two tables together. Once again, there is a clear method for this already:

query(Employee, Company).join(Employee.company)

So I think the best behavior here would be for query(some_relation()) to just raise an error.

[Former user]

I agree, that raising an error is a good solution.

[Mike Bayer]

lately I've been swinging the other way on this. if you query(Parent.child), it really is obvious what you're looking for. just wondering about implementation.

[Former user]

Here's a patch that adds some tests and an implementation. It touches a fair amount of code, but a lot of that is just to add a simple way for queries to join themselves.

Also, the only tests this has are the basic query(Parent.child) case and the add_entity(Parent.child) case. Are there others?

Final thought: should this work for one-to-many relations as well? e.g. query(Child.parents)

-Scott Torborg

[Former user]

Can someone add [e-mail] to the cc list for this ticket please?

[Former user]

(original author: ged)

[Former user]

Is this related to my stackoverflow question?

http://stackoverflow.com/questions/5298498/sqlalchemy-adds-unwanted-columns-to-query-when-a-foreign-key-table-is-added

sounds like it, and I sure didn't expect it.

[Mike Bayer]

consolidating under #2846