Inheritance from base class with property in class makes them non-instantiatable

[Germandrummer92]

BPO	47237
Nosy	@gvanrossum, @ericvsmith, @JelleZijlstra, @Fidget-Spinner, @AlexWaygood, @Germandrummer92

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = None
created_at = <Date 2022-04-06.08:41:16.184>
labels = ['type-bug', 'library', '3.10']
title = 'Inheritance from base class with property in class makes them non-instantiatable'
updated_at = <Date 2022-04-06.15:54:34.248>
user = 'https://github.com/Germandrummer92'

bugs.python.org fields:

activity = <Date 2022-04-06.15:54:34.248>
actor = 'eric.smith'
assignee = 'none'
closed = False
closed_date = None
closer = None
components = ['Library (Lib)']
creation = <Date 2022-04-06.08:41:16.184>
creator = 'Germandrummer92'
dependencies = []
files = []
hgrepos = []
issue_num = 47237
keywords = []
message_count = 5.0
messages = ['416846', '416853', '416880', '416881', '416882']
nosy_count = 6.0
nosy_names = ['gvanrossum', 'eric.smith', 'JelleZijlstra', 'kj', 'AlexWaygood', 'Germandrummer92']
pr_nums = []
priority = 'normal'
resolution = None
stage = None
status = 'open'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue47237'
versions = ['Python 3.10']

[Germandrummer92]

Hi,

According to https://peps.python.org/pep-0544/#explicitly-declaring-implementation it should be possible to explicitly inherit from Protocols. This however breaks the dataclass constructor when using the @Property decorator in the protocol, see this example:

from typing import Protocol
from dataclasses import dataclass

class SomeProtocol(Protocol):
    @property
    def some_value(self) -> str: ...

@dataclass
class SomeDataclasss(SomeProtocol):
    some_value: str

if __name__ == '__main__':
    a = SomeDataclasss(some_value="value") # this crashes with AttributeError: can't set attribute 'some_value'

The pattern of @Property in the protocol is one taken from the mypy docs (see https://mypy.readthedocs.io/en/stable/protocols.html#recursive-protocols for example).

When removing the explicit inheritiance mypy also correctly typechecks the dataclass implementation when doing something like, only the explicit inheritance seems to fail in python

a: SomeProtocol = SomeDataclass()

[ericvsmith]

Here's the error without dataclasses:

------

from typing import Protocol

class SomeProtocol(Protocol):
    @property
    def some_value(self) -> str: ...

class SomeClass(SomeProtocol):
    def __init__(self, some_value):
        self.some_value = some_value

if __name__ == '__main__':
    a = SomeClass(some_value="value")

---

Traceback (most recent call last):
  File "foo.py", line 12, in <module>
    a = SomeClass(some_value="value")
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "foo.py", line 9, in __init__
    self.some_value = some_value
    ^^^^^^^^^^^^^^^
AttributeError: property 'some_value' of 'SomeClass' object has no setter

And here it is without Protocol:
--------

class SomeProperty:
    @property
    def some_value(self) -> str: ...

class SomeClass(SomeProperty):
    def __init__(self, some_value):
        self.some_value = some_value

if __name__ == '__main__':
    a = SomeClass(some_value="value")

---

Traceback (most recent call last):
  File "foo.py", line 10, in <module>
    a = SomeClass(some_value="value")
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "foo.py", line 7, in __init__
    self.some_value = some_value
    ^^^^^^^^^^^^^^^
AttributeError: property 'some_value' of 'SomeClass' object has no setter

[gvanrossum]

So is the conclusion that this should be closed as "not a bug"?

[JelleZijlstra]

I think the behavior with regular classes is expected (that's just how inheritance works), but a case could be made that dataclasses should handle this case specially.

[ericvsmith]

What would dataclasses do that's different from a regular class?