Question

How efficient is sqlalchemy if I use several query objects, lines of code, or functions to build my final query? Is there a performance hit as it executes each step, or is it smart enough to figure out what’s going on and only execute the final query?

e.g. say I have code that does this:

get_shiniest_robots(self, polish_date):
    """Get robot_ids for robots with recently polished spikes."""
    return session.query(Robots.robot_id).filter_by(spike_polish_date > polish_date)

Then I have another function that does

get_dullest_robots(self, polish_date):
    """Get robot_ids for robots that have not been polished lately."""
    return session.query(Robots.robot_id).filter_by(~Robots.robot_id.in_(get_shiniest_robots(polish_date))).values(Robots.robot_id)

Will get_dullest_robots() send two queries to the database instead of one?

Answer 1

Is there a performance hit as it executes each step, or is it smart enough to figure out what’s going on and only execute the final query?

Query has well defined points at which SQL is emitted:

1. when you iterate over it, that is, “for item in query:”

2. When you call .all(), .one(). .first(), .scalar(), .values(), .count(), and sometimes .get().

3. most other methods are what we call “generative”, which is also known as “method chaining”. The methods listed out at http://docs.sqlalchemy.org/en/latest/orm/query.html should give a hint, if not explicitly, if you’re getting back a new Query with additional state, or if you’re getting SQL emitted results back.