Name conflict on instantiation of inherited class

[Florian Schulze]

In certain cases when a class is instanciated, there can be a name conflict which causes wrong values on attributes. Here is a diff with a test case against current head (should apply to at least 0.7.x):

diff -r cd6e5d6dea83d5d5317765f14b15641fdf54ecc0 test/orm/inheritance/test_polymorph2.py
--- a/test/orm/inheritance/test_polymorph2.py   Tue Jan 17 19:35:55 2012 -0500
+++ b/test/orm/inheritance/test_polymorph2.py   Wed Jan 18 14:05:42 2012 +0100
@@ -1263,4 +1263,43 @@
             }
         )
         assert Dude.supervisor.property.direction is MANYTOONE
-        self._dude_roundtrip()
\ No newline at end of file
+        self._dude_roundtrip()
+
+class NameConflictTest(fixtures.TestBase):
+    def test_name_conflict(self):
+        from sqlalchemy.ext.declarative import declarative_base
+        metadata = MetaData(testing.db)
+        Base = declarative_base(metadata=metadata)
+
+        class Content(Base):
+            __tablename__ = u'content'
+
+            id = Column(Integer, primary_key=True,
+                                test_needs_autoincrement=True)
+
+            type = Column(String(30))
+
+            __mapper_args__ = dict(polymorphic_on=type)
+
+        class Foo(Content):
+            __tablename__ = u'foo'
+            __mapper_args__ = dict(polymorphic_identity='foo')
+
+            id = Column(Integer, ForeignKey('content.id'),
+                                primary_key=True)
+
+            content_type = Column(String(30))
+
+        configure_mappers()
+        metadata.create_all()
+
+        sess = create_session()
+        f = Foo()
+        f.content_type = u'bar'
+        sess.add(f)
+        sess.flush()
+        f_id = f.id
+        print f.content_type
+        del f
+        sess.expunge_all()
+        assert sess.query(Content).get(f_id).content_type == u'bar'

[Florian Schulze]

A standalone version of the test:

from sqlalchemy import create_engine
from sqlalchemy import Column, ForeignKey, Integer, String
from sqlalchemy.ext.declarative import declarative_base
from sqlalchemy.orm import sessionmaker


engine = create_engine('sqlite:///:memory:', echo=True)

Base = declarative_base(engine)
metadata = Base.metadata
Session = sessionmaker(bind=engine)


class Content(Base):
    __tablename__ = u'content'

    id = Column(Integer, primary_key=True)

    type = Column(String(30))

    __mapper_args__ = dict(polymorphic_on=type)


class Foo(Content):
    __tablename__ = u'foo'
    __mapper_args__ = dict(polymorphic_identity='foo')

    id = Column(Integer, ForeignKey('content.id'),
                        primary_key=True)

    content_type = Column(String(30))

#configure_mappers()
metadata.create_all()

sess = Session()
f = Foo()
f.content_type = u'bar'
sess.add(f)
sess.flush()
f_id = f.id
print f.content_type
del f
sess.expunge_all()
assert sess.query(Content).get(f_id).content_type == u'bar'

[Mike Bayer]

ugh....OK...this is not the fix but this confirms it's the key_fallback that's doing this:

diff -r 8c05a3bf6599f92bbf8d2246123597e8966f3a52 lib/sqlalchemy/engine/base.py
--- a/lib/sqlalchemy/engine/base.py Wed Jan 18 12:42:54 2012 -0500
+++ b/lib/sqlalchemy/engine/base.py Wed Jan 18 14:42:27 2012 -0500
@@ -2642,7 +2642,7 @@
         # fallback for targeting a ColumnElement to a textual expression
         # this is a rare use case which only occurs when matching text()
         # constructs to ColumnElements, and after a pickle/unpickle roundtrip
-        elif isinstance(key, expression.ColumnElement):
+        elif False: #isinstance(key, expression.ColumnElement):
             if key._label and key._label.lower() in map:
                 result = map[key._label.lower()](key._label.lower())
             elif hasattr(key, 'name') and key.name.lower() in map:

this is a high priority issue.

[Mike Bayer]

the fix is very involved. not sure when I can have all the edge cases worked out, general idea is attached.

[Mike Bayer]

that patch gets everything plus adds tests to orm and sql. a little more review and it'll be good.

[Mike Bayer]

09a503e49724b61a8119f0b7855a990a29fc1521

[Mike Bayer]

also see 37773f34488a5f5d65bf31d5d1b899ea56e6af8a which refines this a bit more.

[Mike Bayer]

Removing milestone: 0.7.5 (automated comment)