None gets ignored by __in filter

[Martin Chase]

from django.db import models

class A(models.Model):
  b = models.ForeignKey(B, null=True)

class B(models.Model):
  pass

for _ in range(0,10):
  A.objects.create()

b = B.objects.create()
A.objects.create(b=b)

A.objects.filter(b=None).count()  # => 10
A.objects.filter(b__in=[None,]).count() # => 0 (should be 10)
A.objects.filter(b__in=[None,b]).count() # => 1 (should be 11)
A.objects.filter(b__in=[b,None]).count() # => 1 (should be 11)

I haven't tested with the above code exactly, but this is the failing behavior I'm seeing.

[Russell Keith-Magee]

This is an edge case caused by the fact that x = NULL and x IS NULL are not equivalent in SQL. The filter clause "bin=[None, value]" needs to be rolled out as "b IS NULL OR b IN (value)"

[Filip Gruszczyński]

patch + simple test

[Filip Gruszczyński]

I don't know, whether this is a way to go, so I would appreciate any advice on the patch and would be happy to work more on it, if required.

[dmclain]

Looks like the implementation for __in has changed since the patch was made. This is more than just the patch not applying cleanly, there are more code paths that need to be considered now.

[Jacob]

Duplicate of #13768.