SQLAlchemy select_from a single table

In trying to replicate a MySQL query in SQL Alchemy, I've hit a snag in specifying which tables to select from.

The query that works is

SELECT c.* FROM attacks AS a INNER JOIN hosts h ON a.host_id = h.id INNER JOIN cities c ON h.city_id = c.id GROUP BY c.id; 

I try to accomplish this in SQLAlchemy using the following function

def all_cities(): session = connection.globe.get_session() destination_city = aliased(City, name='destination_city') query = session.query(City). \ select_from(Attack).\ join((Host, Attack.host_id == Host.id)).\ join((destination_city, Host.city_id == destination_city.id)).\ group_by(destination_city.id) print query results = [result.serialize() for result in query] session.close() file(os.path.join(os.path.dirname(__file__), "servers.geojson"), 'a').write(geojson.feature_collection(results)) 

When printing the query, I end up with ALMOST the right query

SELECT cities.id AS cities_id, cities.country_id AS cities_country_id, cities.province AS cities_province, cities.latitude AS cities_latitude, cities.longitude AS cities_longitude, cities.name AS cities_name FROM cities, attacks INNER JOIN hosts ON attacks.host_id = hosts.id INNER JOIN cities AS destination_city ON hosts.city_id = destination_city.id GROUP BY destination_city.id 

However, you will note that it is selecting from cities, attacks...

How can I get it to select only from the attacks table?