Here's a test which generates essentially the same form and runs fine, I'll try
to simulate more of exactly what you're doing. Or if you had a real test case
ready to go, would save me a ton of time.
from sqlalchemy import *
from sqlalchemy.orm import *
from sqlalchemy.ext.declarative import declarative_base
Base= declarative_base()
class A(Base):
__tablename__ = "a"
id = Column(Integer, primary_key=True)
bs = relationship("B")
class B(Base):
__tablename__ = "b"
id = Column(Integer, primary_key=True)
a_id = Column(Integer, ForeignKey('a.id'))
e = create_engine("sqlite://", echo=True)
Base.metadata.create_all(e)
s = Session(e)
s.add_all([
A(bs=[B(), B()])
])
s.commit()
print s.query(A.bs.any()).select_from(A).from_self().all()
SQL:
SELECT anon_1.anon_2 AS anon_1_anon_2
FROM (SELECT EXISTS (SELECT 1
FROM b
WHERE a.id = b.a_id) AS anon_2
FROM a) AS anon_1
2012-02-28 11:41:19,912 INFO sqlalchemy.engine.base.Engine ()
[(True,)]
On Feb 28, 2012, at 11:37 AM, Michael Bayer wrote:
>
> it appears here the "anon_2" is a label being given to your otherwise unnamed
> FirstThing.moved_by.any() call, which is a subquery.
>
> you're not showing me the full query being rendered but I would imagine the
> important bits are:
>
> SELECT anon_1.anon_2 AS anon_1_anon_2 FROM
> (SELECT EXISTS (...) AS anon_2) AS anon_1
>
> which is valid. The query would fail to execute if it weren't.
>
> NoSuchColumnError here would likely be alleviated if you just said
> FirstThing.moved_by.any().label("some_label").
>
> I'll look into seeing why an anonymous any() subquery doesn't get targeted by
> Query correctly here.
>
>
>
> On Feb 28, 2012, at 11:02 AM, naktinis wrote:
>
>> I should have pointed out that I got a NoSuchColumnError because of
>> "anon_1.anon_2". There is no column "anon_2" in any of the tables. It's just
>> an alias name of a derived table.
>>
>> Is "table_name_1.table_name_2" supposed to mean anything?
>>
>> On Tuesday, February 28, 2012 5:53:42 PM UTC+2, Michael Bayer wrote:
>>
>> On Feb 28, 2012, at 9:40 AM, naktinis wrote:
>>
>>> Column "anon_1.anon_2" is generated in the following scenario:
>>>
>>> dbsession.query(FirstThing, FirstThing.moved_by.any(User.id ==
>>> user_id)).options(joinedload_all('some_property'))
>>> query = query.join(SecondThing, SecondThing.first_thing_id == FirstThing.id)
>>> query = query.order_by(OneThing.ordering_field).limit(count)
>>>
>>> Also, it is important that both FirstThing and SecondThing polymorphically
>>> inherit from Thing.
>>>
>>> Effectively, query.all() generates a query like
>>>
>>> SELECT ... anon_1.anon_2 AS anon_1_anon_2 ...
>>> FROM
>>> (SELECT first_thing.id AS first_thing.id, EXISTS (SELECT 1 FROM
>>> first_thing_moves, users ...) AS anon_2
>>> FROM thing JOIN first_thing ON ... JOIN (SELECT ... FROM thing JOIN
>>> second_thing) AS anon_3 ON ... ORDER BY ... LIMIT ...) AS anon_1 ORDER BY
>>> ...
>>>
>>> Why would "anon_1.anon_2" column be generated there - it is, I think, not
>>> even a valid syntax?
>>
>> it's valid, "anon_1" is the label applied to a subquery, you can see where
>> it has "(SELECT .... ) AS anon_1". "anon_1" becomes what we sometimes call
>> a "derived table" in the query and is then valid like any other alias name.
>>
>> The join is because when we have a joined inheritance class B inherits from
>> A, then we join to it from C, we are effectively joining:
>>
>> SELECT * FROM C JOIN (A JOIN B ON A.id=B.id) ON C.x=A.y
>>
>> That is valid SQL, however, it doesn't work on SQLite, and also doesn't work
>> on MySQL versions before 5. It also may or may not have issues on some
>> other backends. So SQLAlchemy turns "A JOIN B" into a subquery:
>>
>> SELECT * FROM C JOIN (SELECT * FROM A JOIN B ON A.id=B.id) AS anon_1 ON
>> C.x=anon_1.y
>>
>> as it turns out, this approach generalizes much more nicely than just
>> putting "A JOIN B" in there. Suppose classes B1 and B2 inherit from A in a
>> concrete fashion, using tables "B1" and "B2" to represent the full row.
>> Then you wanted to join from C to A. SQLAlchemy would have you doing a
>> "polymorphic union" which means you select from the UNION of B1 and B2:
>>
>> SELECT * FROM C JOIN (SELECT * FROM B1 UNION SELECT * FROM B2) AS anon_1 ON
>> C.x=anon_1.y
>>
>> where "anon_1.y" here would be "y" from B1 unioned to "y" from B2.
>>
>> Anyway, SQLAlchemy is very quick to wrap up a series of rows in a subquery,
>> applying an alias to it, since that syntax works the most consistently
>> across not only all backends but across a really wide range of scenarios.
>>
>>
>>
>>
>>
>> --
>> You received this message because you are subscribed to the Google Groups
>> "sqlalchemy" group.
>> To view this discussion on the web visit
>> https://groups.google.com/d/msg/sqlalchemy/-/g5juNMWd4moJ.
>> 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.
>
>
> --
> 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.
--
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.
