pythonpython-typing

from __future__ import annotations


Python doc __future__

In the python docs about __future__ there is a table where it shows that annotations are "optional in" 3.7.0b1 and "mandatory in" 4.0 but I am still able to use annotations in 3.8.2 without importing annotations. Given that, what is the use of it?

>>> def add_int(a:int, b:int) -> int:
...     return a + b
>>> add_int.__annotations__
{'a': <class 'int'>, 'b': <class 'int'>, 'return': <class 'int'>}

I doubt I clearly understand the meaning of "optional in" and "mandatory in" here


Solution

  • Mandatory is an interesting word choice. I guess it means that it's by default in the language. You don't have to enable it with from __future__ import annotations

    The annotations feature are referring to the PEP 563: Postponed evaluation of annotations. It's an enhancement to the existing annotations feature which was initially introduced in python 3.0 and redefined as type hints in python 3.5, that's why your code works under python 3.8.

    Here's what optional from __future__ import annotations changes in python 3.7+:

    class A:
        def f(self) -> A: # NameError: name 'A' is not defined
            pass
    

    but this works

    from __future__ import annotations
    
    class A:
        def f(self) -> A:
            pass
    

    See this chapter in python 3.7 what's new about postponed annotations:

    Since this change breaks compatibility, the new behavior needs to be enabled on a per-module basis in Python 3.7 using a __future__ import:

    from __future__ import annotations

    It will become the default in Python 3.10*.

    * it was announced to be default in 3.10 (when python3.7 was released), but it was now moved to a later release