Annotations in GeoDjango on many-to-many tables - django

Question

short
I am trying to query a model Foo which has a many-to-many relationship to Address where addresses will be within a specified distance from a given point and sort results by ascending distance. Seems like annotation would be able to do this however I can't figure out how to do that in GeoDjango since it does not support geo annotations.
longer
Here is the basic model structure I have:
# app name is bar
from django.contrib.gis.db import models
class Location(models.Model):
latlon = Models.PointFields(spatial_index=True)
# other fields ommitted
objects = models.GeoManager()
class Address(models.Model):
latlon = models.PointField(spatial_index=True)
# other fields omitted
objects = models.GeoManager()
class Foo(models.Model):
addresses = models.ManyToManyField(Address)
# other fields omitted
objects = models.GeoManager()
Using the above models I am able to construct a query which selects all Foo objects which have addresses within a specific distance from a specific point. For example:
from django.contrib.gis.geos import Point
from django.contrib.gis.measure import Distance
new_york = Point(-73.98497, 40.75813) # == Location.latlon
Foo.objects.filter(addresses__latlon__distance_lte=(new_york, Distance(mi=20)))
That generates a query something like:
SELECT
"bar_foo"."id",
...
FROM "bar_foo"
INNER JOIN "bar_foo_address"
ON ("bar_foo"."id" = "bar_foo_address"."foo_id")
INNER JOIN "bar_address"
ON ("bar_foo_address"."address_id" = "bar_address"."id")
WHERE (ST_distance_sphere("bar_address"."latlon",ST_GeomFromEWKB(
'\x0101000020e6100000aaf1d24d628052c096218e75715b4440' :: BYTEA)) <= 32186.88)
That works very well except I run into trouble if I want to sort the all foos by their distance from the given point. I tried something like:
(Foo
.objects
.filter(addresses__latlon__distance_lte=(new_york, Distance(mi=20)))
.distance(Location.latlon)
.order_by('distance'))
# produces
TypeError: ST_Distance output only available on GeometryFields.
When I read some source code I tried to modify the query and yet still getting errors:
(Foo
.objects
.filter(addresses__latlon__distance_lte=(new_york, Distance(mi=20)))
.distance(Location.latlon)
.order_by('distance', field_name='addresses_latlon'))
# produces
ValueError: <django.contrib.gis.db.models.fields.PointField: latlon> not in self.query.related_select_cols
I guess this is related to a fact that Address and Foo have many-to-many relationship. Unfortunately regular annotations are not supported in GeoDjango so I cant do something like:
# hypothetical syntax
(Foo
.objects
.annotate(distance=DistanceAnnotation('addresses__latlon', new_york, unit='mi'))
.filter(distance__lte=20)
.order_by('distance'))
# which would generate
SELECT
"bar_foo"."id",
(ST_distance_sphere("bar_address"."latlon",ST_GeomFromEWKB(
'\x0101000020e6100000aaf1d24d628052c096218e75715b4440' :: BYTEA)) as distance,
...
FROM "bar_foo"
INNER JOIN "bar_foo_address"
ON ("bar_foo"."id" = "bar_foo_address"."foo_id")
INNER JOIN "bar_address"
ON ("bar_foo_address"."address_id" = "bar_address"."id")
WHERE distance <= 32186.88)
ORDER BY distance ASC
So the question is how can I do do regular annotation using existing API? Or maybe some other way I can accomplish the desired result?

Answer 1

Distance annotation was implemented in 1.9
https://docs.djangoproject.com/en/dev/ref/contrib/gis/functions/#distance
Distance
class Distance(expr1, expr2, spheroid=None, **extra) Availability:
MySQL, PostGIS, Oracle, SpatiaLite
Accepts two geographic fields or expressions and returns the distance
between them, as a Distance object. On MySQL, a raw float value is
returned when the coordinates are geodetic.
(...)
In the following example, the distance from the city of Hobart to
every other PointField in the AustraliaCity queryset is calculated:
>>> from django.contrib.gis.db.models.functions import Distance
>>> pnt = AustraliaCity.objects.get(name='Hobart').point
>>> for city in AustraliaCity.objects.annotate(distance=Distance('point', pnt)):
... print(city.name, city.distance)
Wollongong 990071.220408 m
Shellharbour 972804.613941 m
Thirroul 1002334.36351 m
...