add_entity and before_compile event

[Jared]

When adding an entity to a query "before_compile" the query is correctly executed but the row processor(s) used in orm.loading.instances still refer to the original query object - specifically the _entities list.

So if you had say related User and Address models and some query,

q = sess.query(User).filter_by(first_name='John')

and a listener,

@event.listens_for(SomeQuery, 'before_compile', retval=True)
def change_query(query):
    return query.add_entity(Address).join(Address)

You would expect the query to yield KeyedTuple but instead User entity is used.

A potential solution would be to change Query._execute_and_instances to return,

loading.instances(querycontext.query, result, query context)

instead of,

loading.instances(self, result, querycontext)

But this change may have other side-effects?

[Mike Bayer]

kudos for using the new feature so quickly! I think your solution is correct:

diff --git a/lib/sqlalchemy/orm/query.py b/lib/sqlalchemy/orm/query.py
index 2bd6889..c3638f6 100644
--- a/lib/sqlalchemy/orm/query.py
+++ b/lib/sqlalchemy/orm/query.py
@@ -2528,7 +2528,7 @@ class Query(object):
             close_with_result=True)

         result = conn.execute(querycontext.statement, self._params)
-        return loading.instances(self, result, querycontext)
+        return loading.instances(querycontext.query, result, querycontext)

     @property
     def column_descriptions(self):

all tests pass with that and the change seems like it should be appropriate.

tagging this for 1.0.3 and I'll see if i can test more fully. the feature is very new so only has one test so far :)

[Mike Bayer]

• Fixed issue in new :meth:.QueryEvents.before_compile event where changes made to the :class:.Query object's collection of entities to load within the event would render in the SQL, but would not be reflected during the loading process. fixes #3387

→ <<cset d48acff23bc0>>

[Mike Bayer]

thanks for reporting!