Filtering __in a sliced queryset with a 0 limit raises an unexpected error

[Bertrand Bordage]

AnyModel.objects.filter(pk__in=AnyModel.objects.all()[i:i]) (where i is a positive integer) raises a ProgrammingError at least on Django 1.6 and 1.7. It most certainly affects all other Django versions too.

That’s because the ORM avoids converting AnyModel.objects.all()[:0] to a SQL query, while .filter(in___=…) tries to include that inexistant SQL query. The resulting SQL looks like SELECT * FROM any_model WHERE id IN (), which is invalid.

A working fix is to change ​line 91 of compiler.py to raise EmptyResultSet instead of return '', ().
WhereNode.as_sql then ​catches this exception and notes that it should select nothing. However, I’m not sure that it doesn’t break something else. I can make a pull request with a unit test if you want.

In the meantime, a workaround is to detect whether the inner QuerySet verifies queryset.query.low_mark == queryset.query.high_mark, and do something relevant if it does.

[Tim Graham]

I could reproduce this on PostgreSQL (but not SQLite) with the attached test.

[Tim Graham]

Duplicate of #19263