• cattrs version: 22.1.dev0 - 3ee4b40
• Python version: 3.10.1
• Operating System: irrelevant

Description

Unstructuring tuples work differently if they are annotated as tuple and typing.Tuple.

Also, unstructuring differs if the tuple is variable-length or precisely known.

What I Did

Consider these 6 classes:

import typing as t
import attrs
import cattrs

@attrs.define
class T1:
    xs: tuple

@attrs.define
class T2:
    xs: t.Tuple

@attrs.define
class T3:
    xs: tuple[int, str, float]

@attrs.define
class T4:
    xs: t.Tuple[int, str, float]

@attrs.define
class T5:
    xs: tuple[float, ...]

@attrs.define
class T6:
    xs: t.Tuple[float, ...]

Now, if you unstructure each of them, you get:

>>> converter.unstructure(T1(xs=(1, "2", 3.4)))
{'xs': [1, '2', 3.4]}
>>> converter.unstructure(T2(xs=(1, "2", 3.4)))
{'xs': (1, '2', 3.4)}


>>> converter.unstructure(T3(xs=(1, "2", 3.4)))
{'xs': (1, '2', 3.4)}
>>> converter.unstructure(T4(xs=(1, "2", 3.4)))
{'xs': (1, '2', 3.4)}


>>> converter.unstructure(T5(xs=(1, "2", 3.4)))
{'xs': [1, '2', 3.4]}
>>> converter.unstructure(T6(xs=(1, "2", 3.4)))
{'xs': [1, '2', 3.4]}

What I expected

I think you mentioned in some other issue, that unstructuring tuples into lists is more natural, since lists are supported by more languages and I agree with that, therefore I expected all the unstructurings to produce a list.

As an aside, structuring works correctly for all the cases, so a dictionary {"xs": [1, '2', 3.4]} always produces a tuple, and, in the case of T5/6, all the values are converted to float.