从SQLAlchemy中的子查询生成布尔表达式

Question

I have the following SQLAlchemy models:

PENDING_STATE = 'pending'
COMPLETE_STATE = 'success'
ERROR_STATE = 'error'

class Assessment(db.Model):
    __tablename__ = 'assessments'

    id = db.Column(db.Integer, primary_key=True)
    state = db.Column(
        db.Enum(PENDING_STATE, COMPLETE_STATE, ERROR_STATE,
                name='assessment_state'),
        default=PENDING_STATE,
        nullable=False,
        index=True)

    test_results = db.relationship("TestResult")


class TestResult(db.Model):
    __tablename__ = 'test_results'

    name = db.Column(db.String, primary_key=True)
    state = db.Column(
        db.Enum(PENDING_STATE, COMPLETE_STATE, ERROR_STATE,
                name='test_result_state_state'),
        default=PENDING_STATE,
        nullable=False,
        index=True)

    assessment_id = db.Column(
        db.Integer,
        db.ForeignKey(
            'assessments.id', onupdate='CASCADE', ondelete='CASCADE'),
        primary_key=True)

And I am trying to implement logic to update an assessment to the error state if any of its test results are in the error state and update the assessment to the success state if all of its test results are in the success state.

I can write raw SQL like this:

SELECT 'error'
FROM assessments
WHERE assessments.state = 'error' OR 'error' IN (
    SELECT test_results.state
    FROM test_results
    WHERE test_results.assessment_id = 1);

But I don't know how to translate that into SQLAlchemy. I'd think that subquery would be something like:

(select([test_results.state]).where(test_results.assessment_id == 1)).in_('error')

but I can't find any way to compare query results against literals like I'm doing in the raw SQL. I swear I must be missing something, but I'm just not seeing a way to write queries which return boolean expressions, which I think is fundamentally what I'm butting up against. Just something as simple as:

SELECT 'a' = 'b'

Seems to be absent from the documentation.

Any ideas on how to express this state change in SQLAlchemy? I'd also be perfectly open to rethinking my schemas if it looks like I'm going about this in a silly way.

Thanks!

Answer 1

Query below should do it for error check. Keep in mind that no rows will be returned in case it is not an eror.

q = (db.session.query(literal_column("'error'"))
     .select_from(Assessment)
     .filter(Assessment.id == sid)
     .filter(or_(
         Assessment.state == ERROR_STATE,
         Assessment.test_results.any(TestResult.state == ERROR_STATE),
     )))

If you wish to do similar check for success , you could find if there is any TestResult which is not a success and negate boolean result.

Answer 2

I actually ended up doing this with postgres triggers, which is probably the better way to handle state updates. So for the error case, I've got:

sqlalchemy.event.listen(TestResult.__table__, 'after_create', sqlalchemy.DDL("""
CREATE OR REPLACE FUNCTION set_assessment_failure() RETURNS trigger AS $$
BEGIN
    UPDATE assessments
    SET state='error'
    WHERE id=NEW.assessment_id;
    RETURN NEW;
END;
$$ LANGUAGE 'plpgsql';

CREATE TRIGGER assessment_failure
    AFTER INSERT OR UPDATE OF state ON test_results
    FOR EACH ROW
    WHEN (NEW.state = 'error')
    EXECUTE PROCEDURE set_assessment_failure();"""))

And something similar for the 'success' case where I count the number of test results vs the number of successful test results.

Credit to van for answering my question as I asked it, though! Thanks, I hadn't bumped into relationship.any before.