On Wed, Dec 27, 2017 at 9:41 PM, Kirill Balunov <kirillbalu...@gmail.com> wrote:
> Will there be any implications (or is it possible) if all variables will
> have an attribute *something like* `__type__` which by default will be
> initialized to *something like* `AnyType`. So in the case `x = 12` will be
> equivalent to `x: AnyType = 12`.
>
> x: int
> x = 12
> x.__type__ # int
>
> a, b = 11, 12
> a.__type__ # AnyType
> b.__type__ # AnyType
This won't work. When you say "a.__type__", it means "take the *value*
referenced by a, and look up its __type__ attribute". So it's
equivalent to writing:
(11).__type__
(12).__type__
To query something about the *variable*, you have to look at its
enclosing namespace. If these are at top level, you'd be looking at
the module-level __annotations__ dictionary. That's also where
x.__type__ would be - it's actually stored in __annotations__["x"].
> And also in this case:
>
> class Dummy:
> a: int
> b: float
> c = []
>
> Dummy.__annotations__ # will be {'a': int, 'b': float, 'c': AnyType}
Now, this situation might be of some interest to the dataclass
discussions. Currently, there's a limitation in that no single source
has information about all three attributes; and that means that the
order ["a", "b", "c"] is actually not stored anywhere. (Annotations
are ordered within themselves; values are ordered within themselves;
but annotations and values are separate.)
There is definitely room to ask the question "can we get default
annotations for any global or class-level name that gets assigned to
without an annotation?". I fully expect that thread to be one of those
gigantic ones, but have fun :)
> While I ask purely out of curiosity, I think this is not an off-topic for
> python-list.
Absolutely, it's not off-topic! Though at some point it might need to
migrate to python-ideas, if there's a proposal to actually change the
language. But there's nothing wrong with asking the question here on
python-list.
ChrisA
--
https://mail.python.org/mailman/listinfo/python-list
