Error when referencing a class within its __init__ method

[shreydesai]

I have a class called Node which represents a general node for a linked list. As defined, a node has two attributes: the node's data and a memory reference to the following node.

from typing import TypeVar, Generic

T = TypeVar('T')

class Node(Generic[T]):
  def __init__(self, data: T, next: Node[T]) -> None:
    self.data = data
    self.next = next

When I run this module with the mypy checker (mypy node.py), it has no issue with this. However, python3 node.py yields the following error: NameError: name 'Node' is not defined.

I tried changing next: Node[T] to next: object, but that generated a mypy error (error: Incompatible types in assignment (expression has type "object", variable has type Node[Any])).

How do I reconcile the differences in behavior when running this module with python3 and mypy while also maintaining the structure of this class?

[Michael0x2a]

What's happening here is that when the Python interpreter hits the __init__ method, the Node class is still in the process of being defined, and so Node hasn't been bound to the environment yet. This explains why Python claims Node[T] isn't defined.

The workaround is to use forward references and write Node[T] as a string, which mypy will also understand:

from typing import TypeVar, Generic

T = TypeVar('T')

class Node(Generic[T]):
  def __init__(self, data: T, next: 'Node[T]') -> None:
    self.data = data
    self.next = next

This will make your code work nicely with both mypy and Python.

That said, it would probably be a good idea for mypy to report an error in cases like this, since it seems to be a common tripup. I found a pre-existing issue for it here, but it doesn't seem like anybody has started working on this yet.

[gvanrossum]

(Since this question was answered and there's no bug here, shouldn't we just close it?)