Django ORM：筛选外键属性的元组

Question

I'm writing a matching service that will match you with profiles based on overlapping expertise. Eg if you are REQUESTING python expertise, it will match you with folks OFFERING python expertise.

I'm having trouble determining how to structure that query with the Django ORM.

Setup

I have models:

class Profile(models.Model):
    pass

class Expertise(models.Model):
    profile = db.ForeignKey('profile', on_delete=models.CASCADE)
    name = db.CharField(choices=[(1, 'python'), (2, 'javascript'), (3, 'golang')], max_length =255)
    direction = db.CharField(choices=[('offered', 'offered'), ('requested', 'requested')], max_length = 255)

Data

I've got data that essentially looks like the following:

# setup some fields
profile = Profile()
profile.save()

Expertise(profile=profile, name=1, direction='requested').save()
Expertise(profile=profile, name=2, direction='offered').save()

Obviously Wrong Attempt

Profile.objects.filter(
   expertise__name__in = [e.name for e in profile.expertise_set()]
   expertise__direction__in = [e.direction for e in profile.expertise_set()]
)

I'm essentially looking for the ability to combine boolean AND and boolean OR in the query.

In a different technology, I'd...

In SQL I'd do something along the lines of:

SELECT * FROM app_profiles JOIN app_expertise on app_profiles.id = app_expertise.app_profiles_id 
WHERE 
   (app_expertise.direction = 'offered' AND app_expertise.name = 1) OR
   (app_expertise.direction = 'requested' AND app_expertise.name = 2)

Expected behavior

>>> maus = Profile.objects.get_or_create(name='Maus')[0]
>>> kungphu = Profile.objects.get_or_create(name='kungphu')[0]
>>> strange = Profile.objects.get_or_create(name='Dr. Strange')[0]
>>> thanos = Profile.objects.get_or_create(name='thanos')[0]
>>> 
>>> # maus offers golang and requests python
>>> Expertise.objects.get_or_create(
...     profile=maus, 
...     name=Expertise.NAME_PYTHON, 
...     direction=Expertise.DIRECTION_REQUESTED,
... )[0]
<Expertise: Python (Requested by Maus)>
>>> Expertise.objects.get_or_create(
...     profile=maus, 
...     name=Expertise.NAME_GOLANG, 
...     direction=Expertise.DIRECTION_OFFERED,
... )[0]
<Expertise: Golang (Offered by Maus)>
>>> # kungphu offers python and requests golang, they will match 
>>> # with mous both because of golang and python compatibility
>>> Expertise.objects.get_or_create(
...     profile=kungphu, 
...     name=Expertise.NAME_PYTHON, 
...     direction=Expertise.DIRECTION_OFFERED,
... )[0]
<Expertise: Python (Requested by kungphu)>
>>> Expertise.objects.get_or_create(
...     profile=kungphu, 
...     name=Expertise.NAME_GOLANG, 
...     direction=Expertise.DIRECTION_REQUESTED,
... )[0]
<Expertise:Golang (Requested by kungphu)>
>>> # Doctor Strange is trying to learn golang, he will match with maus as 
>>> # a result.
>>> Expertise.objects.get_or_create(
...     profile=kungphu, 
...     name=Expertise.NAME_GOLANG, 
...     direction=Expertise.DIRECTION_REQUESTED,
... )[0]
<Expertise:Golang (Requested by strange)>
>>> 
>>> # Thanos both offers and requests Python, because balance, I guess.
>>> Expertise.objects.get_or_create(
...     profile=thanos, 
...     name=Expertise.NAME_PYTHON, 
...     direction=Expertise.DIRECTION_REQUESTED,
... )[0]
<Expertise: Python (Requested by thanos)>
>>> Expertise.objects.get_or_create(
...     profile=thanos, 
...     name=Expertise.NAME_PYTHON, 
...     direction=Expertise.DIRECTION_OFFERED,
... )[0]
<Expertise: Python (Offered by thanos)>
>>> 
>>> # maus has requested Python, thanos and kungphu have offered it
>>> # maus has offered golang, dr. strange and kungphu have requested it.
>>> maus.find_matches()
[<Profile: kungphu>, <Profile: thanos>, <Profile: strange>]

Answer 1

For complex queries Django offers Q objects

So you might query that table with Q objects like;

Expertise.objects.select_related('profile').filter(
    (Q(direction='offered') & Q(name=1)) | (Q(direction='requested') & Q(name=2))
)

The | is the OR and the & is self explanatory. You can also do ~Q for NOT style queries.

select_related , if you've not come across it, makes a more efficient query for related objects. You can read more about that here; https://medium.com/@lucasmagnum/djangotip-select-prefetch-related-e76b683aa457