Dataclasses

Why this matters

Many classes are mostly a bag of fields: __init__ copies arguments onto self, __repr__ prints them, __eq__ compares them. A dataclass writes all of that for you from a short list of fields.

The idea

from dataclasses import dataclass, field

@dataclass
class Player:
    name: str
    score: int = 0                       # a default value
    items: list = field(default_factory=list)   # a NEW list for each object

p = Player("Ada")
print(p)                    # Player(name='Ada', score=0, items=[])
print(p == Player("Ada"))   # True: compares all fields

Useful options:

• @dataclass(frozen=True): instances cannot be changed after creation, and can be dictionary keys.
• @dataclass(order=True): adds <, >, <=, >= using the fields in order.
• field(compare=False): leave a field out of comparisons.
• __post_init__(self): a method run right after the generated __init__, handy for validation.

The type hints are required, but Python does not enforce them.

Try it

Make a Player dataclass and print one. Then add frozen=True and try p.score = 5. What error do you get?

Your turn

Write three dataclasses:

• Product with fields name, price and tags (a list, new for each product, empty by default). A negative price raises ValueError("price cannot be negative"). Add a method price_with_tax(rate) returning the price plus tax, rounded to 2 decimals.
• Coordinate with lat and lon, frozen, so it can be used as a dictionary key.
• Task with priority (an int) and title (a string that is not used for comparing), ordered so that sorted() puts the lowest priority number first.

Recap

• @dataclass generates __init__, __repr__ and __eq__ from the annotated fields.
• Use field(default_factory=list) for mutable defaults.
• frozen=True and order=True add immutability and sorting.

Common mistakes

• tags: list = [] raises ValueError: mutable default <class 'list'> for field tags is not allowed: use default_factory.
• A field with a default before one without: TypeError: non-default argument 'x' follows default argument.
• Forgetting the type hint, so the name is a plain class attribute and not a field.