This code:
class A:
    a = 1
    b = {str(x): x for x in range(5) if x != a}

Produces following error:
Traceback (most recent call last):
  File "<input>", line 1, in <module>
  File "<input>", line 3, in A
  File "<input>", line 3, in <dictcomp>
NameError: name 'a' is not defined


Which i think it shouldn't produce. Issue occurs only in dict comprehensions inside class definitions referencing class attributes(which are in local scope during definition).

Update: any class definition comprehension referencing local variables inside filtering block produces this error.

This is a duplicate of issue3692.

[I wrote this before Serhiy posted his reply; sending it anyway, in case the doc links are useful.]

This is documented, by-design behaviour. Essentially name resolution skips class scopes.

See https://docs.python.org/3/reference/executionmodel.html#resolution-of-names, and particularly the example at the end of that section, which is very similar to your example.

See also: https://docs.python.org/3/reference/expressions.html#displays-for-lists-sets-and-dictionaries, and in particular this part:

> However, aside from the iterable expression in the leftmost for clause,
> the comprehension is executed in a separate implicitly nested scope.
> This ensures that names assigned to in the target list don’t “leak” into
> the enclosing scope.

That "leftmost for clause" bit explains why, in the following example, the first comprehension is evaluated successfully, but the second raises a `NameError`.

>>> class A:
...     a = 3
...     b = [x+y for x in range(a) for y in range(3)]  # okay
...     c = [x+y for x in range(3) for y in range(a)]  # raises