David Gardner wrote:
> I re-worked my test a little bit to use logging, and to switch table
> definitions based on a command line argument.
> I was able to trim down my schema to just the two tables in question,
> although the behavior is the same regardless
> of the related table, and attached is that schema along with some data
> to make the test work.
I have a theory what this is. Can you try this patch against 0.5.6:
Index: lib/sqlalchemy/sql/expression.py
===================================================================
--- lib/sqlalchemy/sql/expression.py (revision 6377)
+++ lib/sqlalchemy/sql/expression.py (working copy)
@@ -2011,7 +2011,11 @@
the same type.
"""
- return isinstance(other, _BindParamClause) and
other.type.__class__ == self.type.__class__ and self.value == other.value
+ return isinstance(other, _BindParamClause) and \
+ len(set(other.type.__class__.__mro__).intersection(
+ self.type.__class__.__mro__
+ )) > 3 \
+ and self.value == other.value
def __getstate__(self):
"""execute a deferred value for serialization purposes."""
if that works let's make a trac ticket to add tests for this.
>
> Also I'm using SA 0.5.6, Postgres 8.4.1, psycopg2 2.0.12, python 2.5.4
> on Debian Squeeze.
>
> dgard...@cssun32 ~/assetdb/one-time:$ python eager_loadtest.py
> using PGText
> INFO:sqlalchemy.orm.strategies.LazyLoader:AssetRelation.Source lazy
> loading clause asset.path = %(param_1)s
> INFO:sqlalchemy.orm.strategies.LazyLoader:AssetRelation.Source will use
> query.get() to optimize instance loads
> INFO:sqlalchemy.orm.strategies.LazyLoader:Asset.Related lazy loading
> clause %(param_1)s = relation.src_asset
> INFO:sqlalchemy.orm.strategies.LazyLoader:Asset.Relatee lazy loading
> clause %(param_1)s = relation.target_asset
> INFO:sqlalchemy.orm.strategies.LazyLoader:AssetRelation.Target lazy
> loading clause asset.path = %(param_1)s
> INFO:sqlalchemy.orm.strategies.LazyLoader:AssetRelation.Target will use
> query.get() to optimize instance loads
> testshow/eps/201/s01/t01
> testshow/chr/test/test
> testshow/eps/201/s01/t01
> dgard...@cssun32 ~/assetdb/one-time:$ python eager_loadtest.py text
> using Text
> INFO:sqlalchemy.orm.strategies.LazyLoader:AssetRelation.Source lazy
> loading clause asset.path = %(param_1)s
> INFO:sqlalchemy.orm.strategies.LazyLoader:Asset.Related lazy loading
> clause %(param_1)s = relation.src_asset
> INFO:sqlalchemy.orm.strategies.LazyLoader:Asset.Relatee lazy loading
> clause %(param_1)s = relation.target_asset
> INFO:sqlalchemy.orm.strategies.LazyLoader:AssetRelation.Target lazy
> loading clause asset.path = %(param_1)s
> testshow/eps/201/s01/t01
> testshow/chr/test/test
> testshow/eps/201/s01/t01
> dgard...@cssun32 ~/assetdb/one-time:$ python eager_loadtest.py auto
> using auto
> INFO:sqlalchemy.orm.strategies.LazyLoader:AssetRelation.Source lazy
> loading clause asset.path = %(param_1)s
> INFO:sqlalchemy.orm.strategies.LazyLoader:AssetRelation.Source will use
> query.get() to optimize instance loads
> INFO:sqlalchemy.orm.strategies.LazyLoader:Asset.Related lazy loading
> clause %(param_1)s = relation.src_asset
> INFO:sqlalchemy.orm.strategies.LazyLoader:Asset.Relatee lazy loading
> clause %(param_1)s = relation.target_asset
> INFO:sqlalchemy.orm.strategies.LazyLoader:AssetRelation.Target lazy
> loading clause asset.path = %(param_1)s
> INFO:sqlalchemy.orm.strategies.LazyLoader:AssetRelation.Target will use
> query.get() to optimize instance loads
> testshow/eps/201/s01/t01
> testshow/chr/test/test
> testshow/eps/201/s01/t01
>
>
> Michael Bayer wrote:
>> On Sep 24, 2009, at 8:21 PM, David Gardner wrote:
>>
>>
>>> Ran across something that I suspect might be a bug. If I define my
>>> table like:
>>>
>>> asset_table = Table('asset', metadata,
>>> Column('path', Text, primary_key=True,
>>> server_default=FetchedValue(),
>>> server_onupdate=FetchedValue()),
>>> autoload=True)
>>>
>>> Then anytime I query for an asset and eagerload a related table the
>>> backref on the related table isn't populated, causing a second query
>>> to
>>> the DB.
>>> If instead I define that column of type PGText then the backrefs are
>>> populated properly. I attached a test which is a simplified version of
>>> my table mappings.
>>>
>>
>> what does the logging output say if you turn on logging.INFO for the
>> "sqlalchemy.orm" logger ? that would illustrate some things about the
>> join conditions.
>>
>> For me to test this I'd have to build up some table names and guess
>> what you have for those defaults....can you share your table
>> definitions ?
>>
>>
>>
>>
>>
>>> Attached is a test of this behavior. The output when the column is
>>> defined as Text or String looks like:
>>> testshow/eps/201/s01/t01
>>> testshow/chr/test/test
>>> 2009-09-24 17:17:03,214 INFO sqlalchemy.engine.base.Engine.0x...1f10
>>> SELECT asset.updated AS asset_updated, asset.name AS asset_name,
>>> asset.type AS asset_type, asset.path AS asset_path, asset.parent AS
>>> asset_parent, asset.is_file AS asset_is_file, asset.created_by AS
>>> asset_created_by
>>> FROM asset
>>> WHERE asset.path = %(param_1)s
>>> 2009-09-24 17:17:03,214 INFO sqlalchemy.engine.base.Engine.0x...1f10
>>> {'param_1': 'testshow/eps/201/s01/t01'}
>>> testshow/eps/201/s01/t01
>>>
>>>
>>> When defined as PGText the output is:
>>> testshow/eps/201/s01/t01
>>> testshow/chr/test/test
>>> testshow/eps/201/s01/t01
>>>
>>>
>>> --
>>> David Gardner
>>> Pipeline Tools Programmer
>>> Jim Henson Creature Shop
>>> [email protected]
>>>
>>>
>>>
>>> import sys
>>> from sqlalchemy import *
>>> from sqlalchemy.orm import *
>>> from sqlalchemy.types import *
>>> from sqlalchemy.databases.postgres import PGText
>>>
>>> DB_HOST = 'localhost'
>>> DB_NAME = 'test_db'
>>> DB_USER = 'testuser'
>>> DB_PASS = 'testpass'
>>> db_uri = 'postgres://%s:%...@%s/%s' % (DB_USER,DB_PASS,DB_HOST,DB_NAME)
>>>
>>> db = create_engine (db_uri)
>>> metadata = MetaData(db)
>>>
>>> class Asset(object):
>>> pass
>>>
>>> class AssetRelation(object):
>>> pass
>>>
>>> #asset_table = Table('asset', metadata,autoload=True)
>>>
>>> #asset_table = Table('asset', metadata,
>>> # Column('path', Text, primary_key=True,
>>> # server_default=FetchedValue(),
>>> # server_onupdate=FetchedValue()),
>>> # autoload=True)
>>>
>>>
>>> asset_table = Table('asset', metadata,
>>> Column('path', PGText, primary_key=True,
>>> server_default=FetchedValue(),
>>> server_onupdate=FetchedValue()),
>>> autoload=True)
>>>
>>> relation_table = Table('relation',metadata, autoload=True)
>>>
>>> asset_mapper = mapper(Asset, asset_table,
>>> properties = {
>>> 'Related' : relation(AssetRelation, backref='Source',
>>> primaryjoin
>>> =
>>> asset_table.c.path
>>> =
>>> =
>>> relation_table.c.src_asset
>>> ,order_by=relation_table.c.target_asset,lazy=True)
>>> })
>>>
>>> mapper(AssetRelation, relation_table, properties = {
>>> 'Target' : relation(Asset, backref='Relatee',
>>> primaryjoin=asset_table.c.path==relation_table.c.target_asset,
>>> viewonly=True,lazy=False)
>>> })
>>>
>>> session=create_session()
>>> a=session.query(Asset).options(eagerload(Asset.Related)).get
>>> ('testshow/eps/201/s01/t01')
>>> db.echo=True
>>> print a.path
>>> r=a.Related[0]
>>> print r.target_asset
>>> b=r.Source
>>> print b.path
>>> session.close()
>>>
>>> sys.exit(0)
>>>
>>
>>
>> >
>>
>>
>
>
> --
> David Gardner
> Pipeline Tools Programmer
> Jim Henson Creature Shop
> [email protected]
>
>
> >
>
--~--~---------~--~----~------------~-------~--~----~
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
-~----------~----~----~----~------~----~------~--~---
