Inheritance and Subclassing

Lost me at the end

@user45755 Feeling lost is a good sign. It means that you’re learning 😉 Which part of the lesson did you find problematic, specifically?

If you set kw_only field to True in the @dataclass decorator in both parent and subclass, then you don’t have to make sure that all new fields have default value. kw_only=True not only makes sure you use keyword arguments when setting instances of the class, but also marks all instance fields as keyword arguments whether they have default values or not.

from dataclasses import dataclass

@dataclass(slots=True, kw_only=True)
class Position:
    name: str
    lon: float = field(default=0.0, metadata={"unit": "degrees"})
    lat: float = field(default=0.0, metadata={"unit": "degrees"})

    def distance(self, other):
        earth_radius = 6371
        lam_1, lam_2 = radians(self.lon), radians(other.lon)
        phi_1, phi_2 = radians(self.lat), radians(other.lat)
        h = (
            sin((phi_2 - phi_1) / 2) ** 2
            + cos(phi_1) * cos(phi_2) * sin((lam_2 - lam_1) / 2) ** 2
        )
        return 2 * earth_radius * asin(sqrt(h))


@dataclass(slots=True, kw_only=True)
class Capital(Position):
    country: str


oslo = Position(name="Oslo", lon=10.8, lat=59.9)
vancouver = Position(name="Vancouver", lon=-123.1, lat=49.3)
print(oslo.distance(vancouver))

@DimaG True, and it’s a clean way to sidestep the ordering rule the lesson runs into. Since kw_only=True makes every field keyword-only, the “non-default argument follows default argument” restriction no longer applies, so Capital can add a required country with no default of its own. That was added in Python 3.10, so it’s a nice modern alternative to giving every subclass field a default.

A couple of small things on the snippet, in case someone copies it to try out:

• The paste only imports dataclass, but it also uses field() (and radians(), sin(), cos(), asin(), sqrt() inside .distance()), so you’d need something like from dataclasses import dataclass, field and from math import radians, sin, cos, asin, sqrt for it to run.
• The example builds a Position, so it doesn’t actually exercise the subclass part. To see the payoff, you’d instantiate Capital with the extra required field:

oslo = Capital(name="Oslo", lon=10.8, lat=59.9, country="Norway")

Without kw_only=True, that same Capital definition raises TypeError: non-default argument 'country' follows default argument, which is exactly the problem the lesson describes.

One trade-off worth flagging: kw_only=True means callers have to pass everything by keyword, so Position("Oslo", 10.8, 59.9) stops working and you have to name the arguments. That’s often what you want with a wide dataclass, but it is a behavior change rather than a free win. The lesson’s approach of giving the new fields default values still works too, so it comes down to whether you’d rather invent defaults or require keywords.

Thanks for adding this 🙂