Convert a complex SQL query to SQLAlchemy

Question

I run out of ideas. I googled now more then one day and I still can\'t find any useful answer to my question.

What I did until now, I tried to use raw SQL but withou

Answer 1

Your HAVING is handled correctly, but you're passing it the wrong expression. It seems that you're using Python 2, since the relational comparison between a string and an integer

'distance' < 25

does not raise an exception, but evaluates to False instead. In other words your query is equal to

locations = db.session.query(...).having(False).all()

which explains why you get zero results: all rows are explicitly filtered out by the HAVING clause, as seen in the printed version:

...
HAVING false = 1  -- remove all rows

A solution is to use a suitable construct, such as column(), to produce the expression:

locations = db.session.query(...).having(column('distance') < 25).all()

You shouldn't wrap the complex select list item expression in a select(), which represents a SELECT statement. Either label the text() as is:

text('( 6371 * acos( cos( radians("53.6209798282177") ) * '
     'cos( radians( lat ) ) * cos( radians( lng ) - radians("13.96948162900808") ) + '
     'sin( radians("53.6209798282177") ) * sin( radians( lat ) ) ) ) '
     'AS distance')

or build the expression using the model:

(6371 *
 func.acos(func.cos(func.radians(53.6209798282177)) *
           func.cos(func.radians(Location.lat)) *
           func.cos(func.radians(Location.lng) - func.radians(13.96948162900808)) +
           func.sin(func.radians(53.6209798282177)) *
           func.sin(func.radians(Location.lat)))).label('distance')

You could improve the readability of your query construction by making a function for the great-circle distance, and with a little bit of work you could implement a hybrid method on Location:

import math

def gc_distance(lat1, lng1, lat2, lng2, math=math):
    ang = math.acos(math.cos(math.radians(lat1)) *
                    math.cos(math.radians(lat2)) *
                    math.cos(math.radians(lng2) -
                             math.radians(lng1)) +
                    math.sin(math.radians(lat1)) *
                    math.sin(math.radians(lat2)))

    return 6371 * ang

class Location(db.Model):
    ...
    @hybrid_method
    def distance(self, lat, lng):
        return gc_distance(lat, lng, self.lat, self.lng)

    @distance.expression
    def distance(cls, lat, lng):
        return gc_distance(lat, lng, cls.lat, cls.lng, math=func)

locations = db.session.query(
        Location,
        Location.distance(53.6209798282177,
                          13.96948162900808).label('distance')).\
    having(column('distance') < 25).\
    order_by('distance').\
    all()

---

Note that the way you use HAVING to eliminate non-group rows is not portable. For example in Postgresql the presence of HAVING clause turns a query in to a grouped query, even without a GROUP BY clause. You could use a subquery instead:

stmt = db.session.query(
        Location,
        Location.distance(53.6209798282177,
                          13.96948162900808).label('distance')).\
    subquery()

location_alias = db.aliased(Location, stmt)

locations = db.session.query(location_alias).\
    filter(stmt.c.distance < 25).\
    order_by(stmt.c.distance).\
    all()